CA2110182C - Electronic signal encoding and decoding - Google Patents

Abstract

An electronic method and apparatus for signal encoding and decoding to provide ultra low distortion reproduction of analog signals, while remaining compatible with industry standardized signal playback apparatus not incorporating the decoding features of the invention, and wherein the improved system provides an interplay of gain, slew rate and wave synthesis operations to reduce signal distortions and improve apparent resolution, all under the control of concealed control codes for triggering appropriate decoding signal reconstruction compensation complementing the signal analysis made during encoding. In addition, signals lacking the encoding process features of the invention are likewise compatible with playback decoders which do embody the invention, to provide some overall restorations enhancement.

Description

2~~01.8~
'~G~'~~2'i?6~ - pCT/US92/~4629 I~3PFCOVED SIG1VAI, IJI~t'ODE,~DECODE SYSTEM
~ACItGItOIJI~D OF THE IN'~~ENTIC~1V
This invention relates generally to improvements in signal encoding/decoding meEhods and apparatus and, more particularly, to a new and improved digital encodang and decoding system for lower distortion, higher resolution, and increased dynamic range rcproduexion of analog signals while remaining compatible with industry standardized sy~nal playback apparatus and standards not ineorparating the deending feaeurcs of the prrseat invention. In addition, recordings lacking the encoding process ffeatuses of the fiavenxion are likewise eompatkle with playback decoders which do embody the invention, and ase provided some enhancement ~.0 Quatc often a rr~rdiag or communications system is standardized and its format cannot be readily attemd wi&hout a~'e~ing a substantial quantity of ecyuipment already in c~dsienoc. Hence, adding information wrath supplemental codes may act always be practical unless provisions have been standardised ffor such insertions. Unfortuaately, modern dig'~tal systems are not very expandable since data bammdwldth, resolutson, error corntdioa, syarhronization, andllary data and other "2~ousekoeping' information essentially occupy the entire digital capacity of the storage or transmission mctiium.
However, tlectronic equipment manufacharers and users of such devices continue to seek enhanced perforananoe and more ftaturrs from such standardized systems. An iaupoetant example is the need to make a a~ompatible recording well suited snnultaneously for portable, 211 automotive, television and audiophile markets. Today, many recordings are made for the most profitable market whale other users suffer compromised routes. The obvious conflicting performance requu-ements of diifarent listening environments and the need for sonic improvement should desirably be implemented by a acw system which is compatible with older systems and recordings.
Automobile and portable equipment arc usually low cost and must operate 2S an noisy envimameats. I-Ient~, is such situations, a slightly restaic~ed dynamic range playback is beneficial. Audiophile syseems require utanost acotuacy, dy~amie range, and resolution beyond that which is available as the current standards. Thus, is any new oompshbk system, as pc~ovided by the pres~at imreatioa, encoded dynamics and slew rate modidicatioas which achieve lowest distortion and beset resolution for the audiophile when decoded, should also provide improved sonics for portable and 30 automotive playback when not decoded.
Compact Disc pulse code modulation and other digital audio encoding schemes are good examples of highly developed and. standardized systems which push signal conditioning and digual information limits. )Most such digital systems originally evolved around then practical 25 to 35 mHz rotary head video recorder bandwidths. In such standards, the data bits with 35 error oomecdon and housekeep'utg entirely fill the available bandwidth.
Accordingly, the need for a SI~B~'d'ITIIT~ ~h9E~T

~~1~~.82 ,VO ~2~220E~ - P~CT/U892/04629 "smart" optimization technique, which does not rely upon increased bandwidth for its implementation, becomes apparent.
By way of background, let us consider a typical digital audio record-play system, its most frequently encountered ~mponents, operation, and difficulties. In its simplest form, the recorder includes a sampling switch and an analog to digital converter.
The switch breaks the contanuous analog signal into a sexier of voltage saps, each of which is converted to number groups or digital words. Digital level meters and simple communication systems often operate with just these I'umctaons an a single IC chip. pracxic~l high performance remrd and playback systems require many added operations to prevent undesired internal and external analog-digital signal inttractions, as well as brats and non-linear feedthrough between digital and analog frequensaes.
't3~o11-known technologies to deal with these problems include sharp atst-off or 'brick wall" low-pass filters, fast sample and hold circuits, and high common mode rejection amplifiers. Unfortunately, although these ~omponencs and subsystems solve many problems, they also create others.
Briefly, in typical digital recording systems, low-pass filters ring, and if of iS analog constru~ion, have pre-echo, are subject to sudden phase shifts near band edge, and hat~c cape 'e~rors whisk often cause troublesome dielactsic hysteresis effesxs.
Sample and hold arcuits have unpredictable timing and capture exxors for different signal slew rafts and also suffer from capacitor problems. Fast digital signals and the high Speed amplifiers needed to handle them often create and arc sensitive to gxound cusreats which can caux sudiblc strobe-beat effects.
Digital reproducing systems Gave similar problems, along with spike or glitch generation caused by digital to analog oon~siou, and digital filter word length round off problems. 1',Isually the recorder is designed to have state of eke art performance while that of the reproducer degrades depending on the economies of "consumer" construction. These and other problems continue to plague modern high performance digital audio systems.
Unfortunately, such technical difficulties usually create jarring non-harmonic distortions, typically centered in the most sensitive and percepdive human hear7ng range,. Dften these distortions are caused by the highest, almost inaudible frequencies contaaaed within the program material. Taking the ratio of high and low frequency hearing acuity into account, and the fact that souads.usrelated to the program material stand out, the presraa~ of wen an eactraordinarily small 3~ amount of these distoriioas can be quite objectionable to the listener.
Fortunately, often only very small ~orrections are needed to minimize some of these distortions. piowever, left as is, these distortion errors can combine to yield the equivalent of 13 to 14 bit performance accuracy from systems originally designed fns 16 bit resolution. In pracxicc, while some feel the advantages of current digital recordings outweigh the disadvantages of their distortion errors, many sophisticated listeners and audiophiles are not so tolerant.
Accordingly, those concerned with the development and use of digital signal encoding and dct:odimg systems for analog signals have long recognized the need for a higher qualiy, '~VC~~'~!lZzi?6n ~ hCT/US92/04629 lower distortion digital systean for reproducxion of such analog signals, which for all practical purposes is also compatible with existing uluipment standards. The present invention ffu1611s all of these needs.
,~~,1MMAItY OF THE ~,~IYENT'lON
l3ric.fly, and in general terms, the present invention provides new and improved digital encoding/
decoding methods and apparatus ffor,'ultra low distortion reproduction of analog signaL~s whisk are also compatible with industry siandard~od s>gmal playback apparatus not incorporating the decoding features of the present invention. In addition, s>gnals lacking the encodang process features of the invention are likewise compatible with playback deeodCrs whisk do embody the iwxutaon, and are provided some overall enhancement.
l3asisaliy, the present invention is diresded to various aspects of an improved encode/decode system for providing a predetermined balance or interplay of gain structures, filter characteristics, ~rarious slew sate modibcations, and wave synthesis operations to seduce signal distortions and aanprove apparemi resolution. During the ending pros, an analysis of the sisal to be encoded is made over time and the rersutts of this analysis are subsequently atilized in the eneodang LS and deatding process to more a~rattly aeeonstrud the original waveform upon playback. This is accomplished while minimising the deleterious effects noranally encountered in sampling and converting analog signals to digital signals and subsequently reconverting the digital Signals back to an accurate simulation of the original analog waveform.
In accordance with the invention, control information developed during the ~0 aforedescribed waveform analysis is concealed within a standard digital code and this deformation is subsequently used to dynamically change and ~ntrol the reproduction proacss for best performance.
These concealed control codes trigger appropriate decoding signal reconstruction compensation complementing the encoding process selected as a result of the aforemeneioned signal analysis. Since the eoatrol code is silent and the overall distal informaeion sate is nnsmally ~Ced, the process can 25 operate compatibly with existiag equipment and industry standards. In addition, and as previously indicated, sigaals lacking the one pt~osxss features of the invention are likewise compai~'ble with playback decoders which do embody the invention, to provide some beneficial enhancement.
To achieve higher performance with a fined information eats, an on-going trade-off is made between dynamic range, to achieve improved small signal seaolution, and peak level ~0 and/or slew rate, to achieve fast signal response ascusaty. These small change and fast change aspects of a signal, ss well as large and small amplitude asports, cash have their own dipltal distortion or System compromise mechammisms. Since both large and small aspects will not occur at the same time, an optimum encoding proctors or mix of pr.s favoring sash signal condition ran be chosen dynamically, in accordance with the invention, t~ achieve an improved signal seproducdon within a fixed ~ digital information rate. A silent or hidden ~ntrol rode documents these changes from time to time 2~.:1~~.~~
1~'CD 1." ~?Ob0 PCT/US92/04ti29 in the signal encoding process and is used to create the complementary level, slew rate, filter character, and waveform synthesis necessary to restore the original signal during the decoding process.
In a presently preferred embodiment of the 9nvention, the encoder system has much higher resolution and speed than the industry standard or encoded product, and is set up as an acquisition system with sufficient look forward and look behind memory to compute the opi'unum processing of the signal and its corresponding reconstruction control code. As previously noted, the processing of the signal is determined baset~ on a consideration of which trade-tiffs of resolution, speed, and level are most appropriate for the signal conditions over time and how the reproducer can best be programmed to allow the most acear~ate reproduction of the original analog signal.
To be inaudible, the computed reconstruction taatrol signal is encoded or ended to a random number sequence which may be inserted ~ntinuousty or dynamically when needed into the least sigrtificaat digital bit or bits. The processed audio or signal becomes encoded to the remaining bits.
Caaveatioaal decoding by a maple digital to analog converter of all bits of a recording encoded in accosdaace with the invention, yields a signal with slightly less dynamic range and only slightly higher bacD~r~ound noise. h3owtvrr, the signal will have lowtr quantization and slew induced distoxiions and, heave, the processed eaaoded product, when reproduced on non-decoding standard equipment, wdl sound equal to or better than an unenooded producx.
A fully decoding player, in acxardaace with the invention, retrieves the control signal and uses ft to set tip, operate and dynamically change a complementary process to recover the pre-sompueed high accuracy information and provide low distortion reproduction of the original analog signal. operations to do this include fast pealk expansion, averaged low level gain reductions, selecting complementary interpolation friters, wav~efarm synthesis, arid others. When these are selected a«:ording to ongoing trade-tiffs, optimum for a particular set of signal conditions, an appartnt increase of 2S bandwidth and resolution occurs.
An improved digital system, in accordance with the invention, uses groups of dynamac;ally changang pre-detet~atined pt;rforananc;e trado-ofl's made ovhen signal conditions of the recorded program would create distortion. Since digital distortions occtv at extremes of high level, slew rate, and high frequca~cies, on one hand, and with quiet signals and short small transients on the other hand, a best ~noode/deo~e strategy is chosen for that extreme without the process compromise hurting the opposite aspects of the program. To achieve this, the program is delayed long enough so chat a most likely distortion mechanism is adentr4ied prior to its ~emergen~ from the time delay, thereby allowing a best encoding strategy sad complementary decoding method to be determined and encoded.
I'erformanae as ~~proved because any distortion compromise made occurs for opposite signal conditions, which are essentially nonexistent at that t9me.
In the simplest form of the system, an encoded dynamic range compression and extmplcmentary reproduce expansion will improve per~'ormance. Furthermore, improvements are had by using averaged levels of small signals independent of their lower frequency and near supersonic ~~.~i~~.o~
fVg ~ ~Z/2Sd160 Pi.°l'/US92/(14629 frequency specwal components to control processing providing improved complementary restored resolution. In a similar manner, the staoctgest signals aeoeive processing having DC to ruaximum bandwidth for instantaneous peak conditions, which also yields best complementary restoration. ~aly one correction need operate at a time and, hence, digital information is saved, or conversely, more apparent performance is obtained from an unchanged digital information rate.
3n addition, a further reduction of known and predictable digital distortions occurs by selecting s brat low pass filter with the least compaomise for program conditions during tnooding and using a complementaryc finterpalation or low pass filter during reproduction. Also, other 'smprovtmenis art had from the n of known recurrent distorts, such as transient ereors, by synthes~ng chest components from lookup table etuves of these distortions or missing information, and then scaling these to the signal at hand.
All of the aforedescrilxcd improvements can also operate with varying degrees of success in a defauie or "open loop" mode at the reproducer by dcteixing information about the encoded signal and thin varying chest proocs,Ses from the detected s>gnal.
Dagital systems typically hay a very high signal to noise ratio, but hae~ a restaicted working dynamac range of levtl5 and a~raaed frtquenry response. Tht improved system of the present invention reduax distortions and, as such, uses signal character dependent gain changes, filter optim:~aation, slew raft processing, and waveform rtconstrudnon or synthesis to do this. 'fhe improved system computes, within memory and proa~ss time limits, a continuously changing best compromise strategy of available processes to give the best signal reconstruction. This obviously complex task yields a restoration control signal silently encrypted or noise disguised is a least significant bit code. By comparison, the reproducer system is simple, since its decoding and complementary signal restoration can occur with convtntional multiplying converters, digital signal processors and other analog and digital devices similar to or already used in consumer electronics.
A conventional recording and rtproduring digital system appears relatively simple and potentially accurate for all the data bits encoded, do practice, however, using a very near to theoretical minirattm sampling rate a>md the least acceptable number of data bits substantially agg~ravaees spud and accuracy limitations from even the best state-of the-art circuits and components.
In this regard, the worst offenders are items such as filters, sample and.hold circuits, analog-to-digital converters, digital-to-aualog converters, and system grounding, timing and various process interactions and crosstalk.
~ht aforede~'btd practical tcchnolog'tcal dlllicialilts and their potential distottaous can be greatly minimized by usang higher sampling rates and more data bits than current standards allow In fact, current Lechxtolog'tcal capability permits the reduction of cross-talk, time fitter and other ttoisc interatxlon problems which, along with digital bandwidth limitatiotu, prevented the practical implementation of higher data rates when currtmt digital standards were fu~t cnvis9oned and sstablishcd. With today's high speed couvcrters operating touch faster with snore data bits, filters can become less severe and the greater diffenaaoe between highest audio fretluencies and the digital Sl9~STl'TlJ'fE SI-iEE1' 2~.:~~9~.8~
f!'~ 97 "~206U PCT/US92/04629 sampling a-ate then reduces beats, sideband foldoveax, abasing, as well as loss of small signal information. The present invention uses chess capabilities by employ'mg a high speed conversion process. The dig;~tal information rate, though now much higher, can be computed, as an ongoing acquisition process, to an "error free" mathematically fileered lower sampling rate 16 bit code ~mpatiblc with current standards. fi~fost decimation oversampling encoders work like this. However, in addition, the invention anticipates alias, aperture, ineerpolation and amplitude resolution distortions from am "ideal" standard reproducer and computes them during the e$coding proctss for correction during reproduction. ii7hen the full process of the invention is used, even certain frequencies above the audio range o~r Idyquist limit of industry standard equipment can be sent through the system without creating sub-harmonic or foldover distortions. I-Ienee, a closer to perfect reoord/playback system is provided with minianal problems from filters, converters, and other components or subsystems while remaining compatible with industry standards.
For a Compact IDisc system, "perfecx" reproduction to 16 bit industry standards will have a ma~amum of 65,53b aucll defined equally spaced resolution cps, each about 150 microvolts 1S in amplitude when scaled to normal professional audio levels (10 molts peak to-peak ma7amum). 'This number, when stepped consceattwely at the industry standard 44.1 kHz sampling rate, provides a slew of less than 7 volts per second. Faster rates will skip numbers until, for a 10 kHa triangle segment, only 2.2 sample points remain to define that waveshape as it would be faltered to its 20 kliz bandwidth. In this regard, more than a 1 gaga Hertz sample rata would be r~equu~ed to include all 65,536 resolution points !o create that wave segment. Fortunately, an ideal interpolation filter will fill in all of these points prow<dcd the 2.2 samples have been timed accurately enough. To do this to achieve a one half bit RMS
averaged socusate sample of a fast changing Signal the sample timing must occur within:
~~p~Q~S~ond ~tMSI
x0 Volt ppL50 p Volt / 4 sqrt(2) X = 375 x 10'~ / 2 sqrt(2) = 133 pico sec.
This sample, accurate in time and amplitude, must beheld long enough for conversion eo digital code. Usually, a charge on a capaator represents this information. However, most dielectrics and insulators used to fabricate capacitors hava complex losses as well as past history memory which create a complex delayed voltage change, held re-disisibution errors and leakago. llVhen abrupt changes in level from sample to sample occur, as they do with sampled high lGrequency audio signals, these errors often become much greater than when signal levels don't change. To have less than a half ISB of RIviS averaged error the hold accuracy becomes:
X50 ,~ Volt / LSB x 4a.1 k Hz= 2.3 volt / sec 2 x s~rt(2) or about 23 p Volt per p, sec.

~~.:~~~.8~
wa~T-~Ziaio~o ~crius9aio~a~
Such performance is well beyond simple applications of most modern electrical passive components, much less integrated circuits. Obviously, practical consumer playback equipment will not do better, and the resulting errors can produce slew rate related transient intermodulation distortion ~mponents, which are among the most audibly objectionable. Spe 'cifacally, these result from acquisition time uncertainty or fitter, slew ;rate related non-linear switching offsets, various types of dielectric hysteresis causing previous went related errors, polariey dependent sample discrepancies, and unpredie~able hysttresis within converters as well as other factors. 'Thus, practical systems often have ~mplex signal related errors as hid as twenty times more than theoretical resolution limits of the cur ent ~6 bit standard. Z~Iencc, a process providing more sample points per second with the least voltage change per sample will yield a signal with lower transient intermodulation distortion.
A second distortion mechanism occurs with very small signal amplitude changes of about 5 to a0 millivolts represented by digital activities of less than about 8 bits in a typical 16 bit system. These levels seldom occu$ by themselves yet can still be a small but audible part of a larger low frequency dominated signal )rienct, these small signals can occwir averaged at many different voltage levels or digital numbers of a larger slow waveform. A psacxical example of this would be midband Ball reverberation decay and bass sounds oomb9ned. ~'he roverb<xstion signal attenuates and sometimes completely disappeatx as it becomes chopped or broken st,~tent parts of the bass wavcform.
As previously indicated, these breaYs aepreseat ehe 150 pV acsolution limits of a "perfect' 16 bit reproducer. In practico, very small signal changes can becrome stepped oneputs, or more often distort a0 to irregular step to step changes with an nn~ainty or hysteresis which orxurs due to errors within converters and from external interference and exosstalk. This produces a collapse of the sense of space in a recording and generates impulsive grainy noise effects which are usually made less objectionable by adding a random noise voltage to the signal prior to encoding so that the step errors become randomised from the uncertain samples cheated. '1'lius, the steppod or quaneizcd distortion becomes a less objectionable noise modulation and the least bit signal cut-off levels are now smoothed to a gradual gear loss with progFessevely smaller signal changes. A better form of distortion reduction occurs by iacreasia,g the samplo points per unit voltage change. ~Jnfortuoately, like the process to increase slew accuracy, s much higher digital information rate than that off the current standard is needed to accomplish this.
il.ow signal level digital errors produce distortions such as quantization noise and resolution loss. Whereas, high signal level high frequency and slew rate related errors produce distortions such as sporadic beats and fast signal change envelope related snbharmonics, referred to as transient intermodulation distortion or ~'iPvl. ~ne is easily misled by test signals with a continuous tnvelope nature, in that they lend to average over time and ean~l many of these distoations and therefore ineorrecily indicate only very small resolution and converter iaac~curacy distortions.
Unfortunately, waveforms like those in music continually change and, as noted, may provide much higher and far more objecxionable non-harmonic TI1~4 and resolution problems.
SUE3STiTUTE ~HE~T

~'V~'9,"-~Z20fit) ~ ~ ~ ~ ~ F'CT/US92104629 Digual distortions occur with high slew rate and small amplitude signal change conditions and, as previously indicated, both are not likely to occur at the same time. Hence, in accordance with the invention, the system identifies either a fast slew or a small change character of the signal wavcform and implements the appropriate corrective process. During encoding" the nature of program signal changes can then determine which corrective process is used as well as a best reprodt:ce conjugate or process at any time during decoding. Dne process can borrow information sate from a less needed performance capability when potentially severe distortion conditions in the signal call for it. In this manner, a dccasioa to provide more points per fast voltage change yields an equivalent higher sampling race at the expense of less important low level resolution. Comvrrsely, a x0 sm~lle3 voltage change per sample antomatic~lly anduces the momentarily unneeded speed capability.
hush interplay and eompnonai~se can be managed and/or eompute<1 to maintain a substantially constant digital information rate. IIJnder these clrcttmstanccs, the processed, decoded, analog output may have an apparent inaxease of bandwidth and resolution and, as noted earlier, when these improvements occur, one or the other as needed, the fundamental causes of digtal distortions as well as their effect on imperfect reproducers can be reduced.
A similar correction strategy is applied to reduce fitter tradeoff ~mpromise errors between transient response, phase accuracy, settling lima, group delay, and oihrx distortions inherent with filderiutg methods. Such errors may not be non-linear, and hence, will not appear as harmonic distortion; howevrx human hearing is sensieive to manipulations of wavcform shape and to the settling limo of complcat signals. Typically, the smallest amplitude high frequency signals are likely to have tacocssive transient ringing and process noisr~s from aggressive filtering, whereas sub-harmonic brats and other filtering noises may occur with intense high frequency signals. The anstautancous versus non-instantaneous character of complex signals is reproduced differently from one filter type to another.
As before, the same large si~al/small signal selection criteria bold, allowing a best encode and decode filtea choice, without having to compromise for the opposits~., essentially non-coexistent, program conditions.
gience, the method and apparatus of the present invention utilize a pre-c~tlcuiated optimal interplay of gain, shw, filter sehc~ion, and waveform synthesis operations done individually or as a oomposiee all inclusive process which becomes ended and decoded in a ~J complementary manner to reduce distoations and improve resolution. Included in such a system is a record c~r~press - play expand system with some features similar in ways to those used in noise rrducBion systems. teiost such noise reduction systems use either peak or 12MS
deteaxoas to examine the incoming signal and convert its icvel to either fast or slowly changing internal DC control signals vrhich ttttimately drive a transient free switching element or art :analog variable gain device. dNhen set up for gain reduction, with increased input signal level, the output signal is compressed so that tiny signals are amplified and stroatg distortion prone signals arc attenuated.
~Clpon playback or decoding, a sunilar circuit set up for gain expansion, detects level changes and restores the signal to an approximation of its original dynamics.

~r~ -.,Z~(Pg0 ~ ~ ~ ~ 'PfrTlCJS9Z/(14G29 In contrast to traditional noise reduction, the system of the present invention ~rrraxs distortion. It dies this by altering gain structure, as well as amplitude and stew rate linearity, for extreme low and high level signal conditions. Low level, small changing parts of the signal are detecxed and used to control the gain of the whole signal which then includes more encoded bits. This S gain eon~ol is derived from a broad middle specxrum of the signal and is active at signal levels representing the lowest levels perceived by human hearing. It is not activated by low frequencies, near supersonic frequencies, or when higher level mid-band signals are paeseni. In this manner, the gain struexure increase maintains s minimum L,~l~ dither-like ae~iry independent of inaudtblc sounds and maintains ambient and background information as well as masking quantization and monotonistic error 19 distortions previously described, Infrequent peak levels are instantly compressed with a transfer function having very low distortion for signals meat maximum 9evcd and producing m~ininaunt upper harnaommics once the limit threshold as Iaavcrsed. This type of operation does a~an occasional higher d'utortion on peaks, ~how~r it prevents catastrophic overloading during recording and allows a higher recording level 15 with overall lower distortion.
Infrequent fast slaw portions of the waveform can be expanded symmetrically in time, and/or in samples, to onc~ompass roots encoded bits, and, as before, other parts of the waveform may be unaltered. This op~a~ation tnay be a dispersion process where time delay is altered, or it can be a graphical waveform synthesLS. It takes can instantaneous event and spreads it in time, and 20 like the peak limner, it creates distortion in tuxdecoded playback.
Gain change, peak limit, and slew rate compression operations and their complements or restorative operations are practical with analog or digital techniques. 'doitage controlled amplifiers, diodes, delay lines, and chirp filters, and mnltipliera are typical analog building blocks which can be assembled to exeate these ftanctions. Equiv~l~nt digital sub-routines and dedicated 25 process algorithms and c~mpotuats are also available. Distortion ff~e digital processing is complex;
for example, rounding off' errors may have to be dithered seed interpolated over time. However, once implcmentod, digital operations are any sfnble and ~proas~ ~eomparetl to the ~rariables subjecx to tolerances and adjustments aeqttixed for the analog control of gain, disperxioa, bandwidth and time constants.
30 The atoredescribed level and slow pisses of the present invention correct distortions ocsxtaxing from opposite signal oondicsons which are not likely to occatr at the same time.
Hlence, these can anttaphty and at maximum oorrec~iott caper=ity pan borrow from an opposite less needed pesfformance capability to maintain constant digital 'information rates. The wave gvnthesis process of the present invention operates with known distortion waveshapes which, when encountered ~.S daring encoding, are subsequently called out of memory by code for complementary correction during reproduction.
bevel and slew correaxion worlcs for known signal conditions having unpredictable distortions and synthesis works for known distortions occurring from signal conditions SIIBSTITIIT~ al-1EE'~' WO 9~ 9~?tl6G ~ ~ ~ ~ ~ ~ P~T'/US921fD~b629 .
unpredictable at the reproducer. fJnlike even state-of the-art noise reduction processes, this system's processing is under intelligent control and given sufficaene computation, trial and error, or successive appro.~mation time, the best correeqion scheme and its encoding for reproducer process control is readily determined and optimized.

5 Wave synthesis, in accordance with ehe invention, is a keyed operation used to recall from memory a number of predictable and/or recurrent distortions known to occur ae the reproduces. gmall waveform segments falling outside of the ldyquist sampling limits, repeated quantization distortions, and inteapoiation filter parameters can be recalled from a look-up table in memory or s~rniliesiz~ from inf~mation sent in the hidden ~, and need for improved playback. ~'hc l!0 synthesis memory can s~rty several interpolation waveshapes which best connect points ai and betv~cn samples. These larger waveforms will maintain their characteristic shape independent of level, just as the a~roducsd signal would do. ~nce the ~nnesting waveshape has been recalled from ~DM, it must be sealed to bt the a>. ~inoe only very slowly chaagxng waveforms wall have samples without bit resolution levels in between, a form of heel detection is necessary to make synthesized segments scaltd to the signal. What would have been level detecdors and gain controlled devises in an analog system are replayed by equtvaleset digital steal processing functions in a digital system. ~nce this has been accomplished, the a~e~n~ucic;d waveform has more equivalent data points in time and level and, when pre~omputed properly, a lower distortion results from the curve fitting.
l(n light of the foregoing, a practical system, in accordance with the invention, may have many times beater signal eesolution and much better fast transient signal accuracy. A much greater digital information rate would normally be necessary to achieve these results. Data is saved by processing only distortion producing conditions. As noted, resolution is selectierely and adaptively traded off for slew accxuacy and slew rate or maximum lave! is bornow~ed for higher resolution. Information rate is conserved by toggling back and forth or fading from process to process when needed.
It should also be apparent that implementation of various subsystem desigtu may be in oithea analog or digital form, monitoring and analysis of the waveform may be accomplished at varying Ancations in the system including the reproducer acrd in either analog or digital form, other parameters of the wavefoa~tns may be selected for compensation, and control codes or other waveform corrtetive message information may be inserted and extracted in a variety of different ways, without departing from the basic Lroncxpts of the present invention.
Hcnee, the method and apparatus of the present invention for eatcoding/dccoding Vitals with mmimimal .distortion saeisfies a long lasting need for a compatible system which provides ate adaptive interplay of gain, slew rate, filter action and wave synthesis processes to substantially reduce signal distortions and improve apparent resolution, 'li4ee above and other objects and advantages of the invention will become apparent from the following more detailed descxiption, when taken in conjunction with the ~ceompanyng drawings of illustrative embodiments.

~:~~.~~8~
~'... ..,/22060 1'tr l'/US92/04629 ~F."SCT21P'I°d~ld ~F "I'fdE g9RA5~i'IP1GS
FIG. 1 is an overall block diagram of an analog to digital encoding system in accordance with the invention;
F"IG. 2 is an overall block diagram of a distal to analog decoding and rcproducang system in accordance with the invention;
FIG. 3 is an more detailed block diagram of an example of an analog to digital encoding system in accordance with the irn~niion;
FIG. 4 is an more detailed block diagram of an example of a dig'ual to analog deccsding and reproducing system in accordance with the ~~~ent:on;
FIGS.Sa through Sd and FIGS. Sel through Se5 graphically depict wavefozms ~lustrating sampling and encoding errors eneotmtered wsth low levtl and rapidly changing waveforms;
FIGS. 5a through 6f graphically depict various signal waveforms during the limiting and reoonstntction of a 8xiangle wave, in one embodiment of the navention;
FIGg. 7a through 7d graphically depict waveforms illustrating various types of distortion encountered with different types of filters;
IFdG. 8 is a block diagram of a processing system in accordance with the invention, using analog processing tcchttology;
k~G. ~1 is a blockk diagram illustrating filter selection control in one embodiment of the invention;
2iD ~ ff~IG.10 is a block diagram of a process switcher utilized in one embodiment of the invention;
FIGS.11 and 12 graphically depict waveforms and distortion plots illustrating system respotose before and after procxss for two 9'alter types, in accordance with the invention;
lrlG. 13 is a block diagram of an analog implementation of a slew rate eompression and expansion system;
~UB~TI~'~~T~ ShIE;:

~~.lfl~~~
i~V~ 93 ~ __'ii60 PCd'JUS51~/~4629 IFdGS. 14a through 14e sholv wavefforms illustrating the operation of a slew rate compression and expansion system ~presouce off high frequency ranging ffrom a low pass filter);
1~IG. 15a is a schematic diagram of an analog implementation of a variable slew rate compression and restoration circuit and FiG. lib shows wavtforms for input and outputs;
FIGS.16a and b fform a simplified schematic diagram of a variable slew rate amplifier; ;
~F1G. 17 is a schematic diagram of an analog implementation of a slew rate expansion circuit and FIGS. i7a through 17e show key waveforms;
~1G. 1$ is a bloclt diagram of a more advanced, presently preferred digital embodiment of the encode system, in accordance with the invention; and FdG. 19 is a block diayam of a more advan~xd, presently preferred digital embodiment of the decode system, ire aocoedance with the iavention;
F1G. 20 is a schematic d'aagram of a circuit used to pseudo-random noise encode control signals and insert them into the least significant bit of the data stream in one embodiment of the invention;
~G. 21 is a schematic diagram of a circuit used to recover and decode control signals inserted into the least s'igni(icant bit of the data stream ~bn one embodiment of the invention.
,~~sc~IP~(DN ~F TllETfl<E 1' )<~ttltE~ 1si19fl~D1P11uT~
The present invention is directed to a system far an electronic method and ~0 apparatus far signal encoding and decoding to provide ultra low distortion reproduction of analog %tgttals, while remaining compatible with industry standardized signal playback apparatus not necessarily incorporating the dccodirtg features of the invention. The improved system provides a selective interplay of gain, filter selection, slew rate and wave synthesis operations to reduce signal distortions and improve apparent resolution from a recorded product, under the control of concealed or silent ?.5 control codes when necessary for triggering appropriate decoding signal r~onstruction compensation based upon a previous signal waveforn3 analysis made during the encoding process for the recorded product. la addition, signals laa~eking the encoding process features of the envention are likewise t~ompatible with playbaedc decoders which do embody the invetttion, anti ace pcovided the benefits of sornc overall enhancement based upon a signal waveform analysis made during playback.

~Vf -._'2/220!0 ~ ~ ~ ~ ~ ~ ~ 1'C'1'/U5~2/fb4629 Rcfemng now to the drawiaigs, and more particularly to Fig. 1 thereof, there is shown, 'sn general terms, the analog to digital conversion and encoding subsystem of a typical recording system cmbodymg features of the present invention.
As shown in Fig. 1, an analog signal 99 is directed as input to a processing subsystem 100 which converts the analag signal into digital form, including such tasks as filtering, sample and Gold, analog to digital conversion and the like. The digieal output 100a from the subsystem 100 is directed to two subsystems, a memory subsystem 101 and an analysis and computation subsystem 102. In memory subsystem 101, the dig~al signal is delayed or stored foe further use and manipulation.
The dual signal output of memory 101 is sent to subsystem 102 at input 102E.
iJsing the output of subsystem 100 at input 102x1, the ~rav~efoexn analysis and eorarclion computation subsystem 102 continuously monitors and evaluates the digital format waveform as it as being stored in the memory subsystem 101 fn order to determine the physical characteristics of the stored waveform ultimately to be retsu~ed and the required corrections necessary for ao~nrate reoonstrtt~ion and restoration of the orfiginal analog waveform $9. This evaluatnoa relates to reoonstractivt level, slew, and wavcform synthesis ~quia~eme,nts ultimately to be provaded by complementary compensation in an appropriate decoding and signal s~eproduction system (Fig. 2). The evaluation may also predict alias ~mponents for subsequent ~njugate mmeutt~lication. Some aspects of the sisal evaluation may be performed on the analog signal by attbsystem 100, and the results sent to subsystem 102 at input 1024.
The ~rreaxive procedux~s are applitd to the distal signal from the memory 2D subsystem 101 by subsystem 102 under the control of signals resulting from the analysis. The procrss controller 102 also generates control codes for use by the decoder which are converted to proper format and appropriately encrypted into the digital signal sa that the control codes can silently ride along wieh the digital a~prcsentation of the original analog waveform 99 and be provided as an encoded digital output 103. Some off these corrective procedures will relate, not just to distortion charadcristia occurring as a result of the basic conversion of the analog wavcform itself, but also to procedures deliberately introduced by the encoder ffor subsequent complementary decoding, such as peak lunit/atabaequently exgand for high !cruel s~nals and averaged ~mptuss/subsequGntly expand for low atv~et signals.
eAs beat observed in Fig. 2, there is shown, again in general terms to illustrate some of the basic overall concepts embodied in the ptnsunt 9nveation, a digital to analog inversion and decoding subsystem ef a typical reproducing system embodying various features of the present imrention foe reconststtexing the original analog waveform.
In Fig. 2, the encoded distal signal 103, recaptured from any appropriaee recording medium (not shown) such as tape or disc, is directed as input to a digital signal analysis and processing subsystem 104 and to memory subsystem 107, which delays flee digital signal. Si~ta1 analysis subsystem 104 extracxs control onde anfocmatiott inserted in the signal at the encoder and may also atuityie the signal itself to determine its characteristics. Tlstose operations include appropriate mesas for control code detection, signal ftlteting, level detection, specteal analysis and the like. !'he detected SU~~''a1'11'U°~'E SWEET

4Vfl 9;' <'2~~ ~ ~ ~ ~ ~ 1'CT/f.JS92/04ti29 montrol codes and signal analysis in the processing subsystem 104 are used to generate control signals directed to a seeonsiruction compensation subsystem LOS which interacts math the processing subsystem 104 and operaees on the delayed digital input signal L08. Subsystem i05 includes digital to analog convcrsaoa, and may include farther memory, such as one or more ltd~yYI's or look-up tables, for various types of rcoonstruciion compensation used, iu a~ordauee with the invention, to correct the digital signal 203.
The compensation subsystem 105 typieally will respond to the various arntrol codes, or the absence thereof, to generate a variety of ~rrec~ve compensations sack as slew rate, le cl, filler aelcction, and waveform synthesis wharh, through appropriate interaction with the processing LO subsystem 104, yields a aeeonstructad analog signal L~ with minimal distortion and enhanced apparent resolution, all without the need for ancreasietg industry standardized digual bandwidth.
It will be appreciated by those of ordinary skill in the art that the systems of Figs. 3 and 2 arc merely illustradv~e of simplified general approaches for practicing axtain basin aspecxs of the present invention, and implementation of the systems of Figs. L and 2 may take a wide rraricty of spctnfic forms without in any way departing firom the spirit sad seeps of trine anven6on.
It should also be appateatt that implementation of various subsystem designs stay be in either analog or dlg'ttttl forma, m and analyses of the waveform may be accomplished at vaeying locations in the system and in ae3iher analog or digital form, outer parameters of the wavefotms may be selected for compensation, and ~net~ol ides or other waveform corrective message information may be inserted and extracted in a variety of different ways, without departing from the basic concepts of she present invention.
By way of example, one possible implementation of the general structure above is presented in more detail in Figs. 3 and ~. These drawings correspond to Figs. 1 and 2, and illustrate more internal detail.
Refersing now mores spec~f"tcally to F'tg. 3 of the drawings, there is shown an mualog to digital encoding system in aeoordance with the invention. Analog input signal 99 is applied Go a buffca amplifier, the first element of the analog to digital a'ubsystem 300. The output of alts buffer amplifier drives an analog low pass and-alias filter, which removes any high frequency eompoaents of the input signal falling above the l~lyquixt limit of half the sampling frequency. The ouepue of the low pass filter has an analog dither signal added to it and then it is applied to the input of a sampling easing to digital converter. In the converter, the signal amplitude is sampled at regular intervals and flee amplitude of oath sample is converted into a number or dig'tial vNOrd.
~'he series of digital avords Wrom the converter make up the digital signal, which is sent to the analog to digittal conversion process e~ontroller. This process controller has generated the dither signal which vans added to the analog signal 33 before conversion, and, typically, the controller subtracts the dither from the digital signal, giving a vrrnier enhancement to the oonvcrsion accuracy as well as spreading any converter nonllnearities unto a noise-like signal. Trite ABC process ~ontroller may also make other corrections or additions to the conversion pros, such as noise shaping. 'The output of this module is a high resolution digital signal SU~S°f'I'fU'flw SH~~-'f 'fi'f' '2/2x4160 P~1'/US92J04629 100a which is sent to subsyseems 101 and 102 It should be noted that this digital signal has boeh higher amplitude resolution and greater sampling rate or 4ime domain resolution than the industry standard digital signal which is the final output of the encoding system.
PvIIemory subsystem 101 is used to delay the high resolution digital signal 100a 5 before sending it to 102e. ~ his time delay gives subsystem 102 time to analyze the signal and choose appropriate corrective procedures to be applied during encoding.
The high resolution digital sagnal from subsystem 100 is also sent to the signal analysis process ~ntrollor unit of subsystem 302 at input 102x. 'this unit analyzes the characteristics of the signal as ii is being stored in the delay memory 101 and makes decbsions about employing 10 corm procedures sash as ins3antancous peak amplitude lamiting, low level gain compression, choice of best 'brick wall" low pass ~lter9 transient reconstruction and so forth.
The unit then sends commands 102b to the units which pay the delayed digital signal to carry out the corauxive proceduees. 'fhe signal analysis process ~ntroller also generates a condrol ~ode i02c which ie sends to the Mode encryption unit for addition to the output signal. 'his ~ntrol coda tells flee decode system what has 15 been done and how to recover an aecusate representation of Lbe original input signal.
The delayed high resolution digital sagnal from the anemory subsystem 101 is sent to the deamation filter trait at 102e. )Eierey the orrcrsampled i~ut signal is dceimafed down to the industry staadard sampling rate. ~'he choice of optimal filter characioristics is dependent on the nature of the program signal at the limo. such facxom as transient content of the signal, presence of large amounts of alias producing high frequencies, otc. are taken into account by the signal analysis proecss controller, and a filter control signal 102b tells the decimation filter which parameters to use. The output of the decimatnon filter has the industry standard sampling rate and very high amplitude resolution. It is sent to the level control processing unit.
~'he level control processing unit uses such operations as instantaneous peak ~5 level compression and low level average gain compression to squi"ezo the high amplitude resolueion of the signal into the industry standard resolution (such as 16 bits for gyp).
Whose operations arc done under the control of the signal analysts pros oontrollor. fhe level control unit array also include other techniques such as the addition of digital dither to allow rosoludon below the Icast significant bit level and transient time domain or slew rate compression. 'f6e output of this u~t is sent to the silent code encryption unit.
'fhe silent code encryption unit Hakes the condrol codes 102c from the signal analysis process controller, which arc commands and information ffor the deader system, and adds them to the digital signal. One method of doing this involvrs encrypting them into a pseudo-random noise-like signal and inserting it as aneoded into the least significant bit of the digital signal. ~iher methods include the use of "useei" bits in standard code or unused bit combinations which may appear to be errors to a normal deader. The ~mmon characterastic of those methods is that they provide a silent silo channel for control informae4on which rides along with the program digital signal.
aU~~TITIJT~ ~H~ET

3'V092~'~.2UbJ ~ ~ ~ ~ ~ a ~ PC~'/U~92/04b29 The final task of the code encryption unit as to encode the composite digital signal into an industry standard format for recording, etc. The output of this unit is a standard digital sigaal 103, which, ffor instance, could be sent to a recorder. This a:ompleers the description of the encoding system.
Referricag now to g'ig. ~ of the drawings, there is shown an example of a digital to analog decode/reproduce system in a~ccordan~ with the invention. The input digital signal 103, from a tape recorder, Cue, etc., is applied to the signal and ode analysis subsystem 104 and to memory subsystem 107. ;
Memory subsystem 107 provides a time delay for the digdtal input signal in order to allow subsyseem~104 time to do ate analysis. The delayed digital signal output 108 of the memory subsystem is sent to the level control unit of subsystem 105.
'The digital input signal 103 is also applied to the signal analysis, code analysis and process control subsystem 104. This subsystem separates from the signal the silent control code inserted by the encoder. This control code ~niains information about what proccssaaag choices were made by the encoder and what ~mpllmentary corrections should be applied to reconstntct the most acaarate reproduction of the original analog input signal. The subsystem may also analy~ the signal itself to deterrmdne the best recxuutrudion strategy, measuring such parameters as the signal amplitude, spectral content, etc. The subsystem then generates a series of caatrol signals to control the various units within the reconsteurtion processor 105, each of which performs a specific type of operation on the program signal.
The reconctrucdion compensation and digital to analog conversion subsystem 105 is made up of a number of proctascing units which operate on the program signal under the control of the analysis and process control subsystem 104. Each of these units has a control connection to 104.
The first of these is the level control unit, which receives the delayed digital input signal 108. The level control unit pcrfarms amplitude scaling complementary to that performed by the encoder, such as instantaneous peak expansion and signal averaging based low level cvpansion.
The output of this unit is a digital signal at the input sautpling salt, but with higher amplitatde resolution than the input. This output is sent to the interpolation filter unit.
The interpolation filter unit creates an oversampled digital signal by ~0 interpolating between the paints represented by the input signal. The best filter parameters for this interpolation arc chosen dynamically based on the control codes, and possibly also signal analysis, so that they complement the parameters of the des~axion filter in the encoder.
Other prot~ing such as noise shaping and transient reconstruction may also be done by this unie.
The output stgaal of this unit is a high resolution ovrasaanplcd digital signal which is sort to the digteal to analog converter unit.
The digital to analog converter dL~AC) unit converts the high resolution digital v signal into an analog signal. It may be a standard converter or a multiplying converter which is used to further affect level changes in the signal. The output of this unit is an analog signal which is sent to the analog processing unit.
SUB~TIT~JT~ SHI~E'.f 1Wg' ~?/22~b0 ~f.'T/L1S92/04629 ~'!ne analog processing unit contains an analog anterpolation filter and buffer amplifier. It may also contain other prorxssing, such as level conxrol, under ihc control of the analysis and process coatrol unit. Since it is operating on an analog signal, the control signal may be converted do analog form in the ~ntrol d~AC before being applied here.
The output of the analog processing unit is an extended range analog signal lOb, which is a close replica of the original analog input signal 99. The overall system of the invention nxakes possible a morn accurate reconstruction of the arigonal analog signal than would have been possible from conventional systems wising the same digital rsoording standards.
The ensuing analyses and discussion are intended to provide further ba.~kground for a proper understanding of the prac~ticc of the invention and io further illustrate and descrfbe a variety of anaiog/digital modes presently contemplated as feasible for carrying out the invention.
Analog to digital enaoditig, in accordance with one aspect of the present invention, works as a sample sate down converter which allows resolution enhancement and reduced filter artefacts. Analog to digital conversion is made at a high sample rate and extended bit g~solution, both well above that rt;gui~d for the final ptv~duct encoded format. This high density code is then arithmetically processed back to the desiaed end use number of bits and sample rate. ~l'atl this arrangement" many advantages occur. Analog'~rick walB" low pass filtering as unnecessary as the very high angoing Sample rate allows much more gentle and phase - time domain controlled audio band cut off before lVyqutst distortions occur. The sample and hold - analog to digital converter subassembly can be supersonically dithered in a knowrf and controlled way to create duty cycle modulated low level code parts in which any monotonirity or code eraors are spread nut as noise sidebands around the dither signal. These will be arery high frequency and, hence, alnnost inaudible, unlike signal related noises of standard systems.
~.n important advantage is that the "brick wall" low-pass Falter required to prevent hlyquist - Alias errors can be implemented as a digital filter, which has higkity repmdusx'ble characteristics free from phase distortions. The eharacxerisdes of this filcea can be dioscn dynamically based upon an analysis of the high scsolution signal to mlnimi~ disiorbion.
~lenoe major Filter and analog to digital encode system prxiblems such as pry-echo, transient ranging, group delay anomalies, missing code errors, alias distonrion and beats are greatly reduced or elimanatcd.
A very powerful arithmetic "engine" and operating program, as well as digitally operated feedback and feedfoaward parts are utilised to make the wldad audio to selected format digital conversion. ~lowever, such a system also easily performs inseantaneous high level limiting said averaged low level expand operations. In pracEice, approximately an crises 4 bits dynamic range ran be 3S had faom systems having complementary playback and cautly improved conies will occur from standard unprocessed compatible playback systems. This is becattse the recording engineer ran raise levels without overload problems, thereby simplifying recording sessions and very low level ambient information will always maintain least bit actavity to prevtaat monotonistic errors. Both processes on ;IJ~STl1-l9-T~ Sh~E~f w~ ~-~ -~~os~ ~ ~- ~ ~ ~ ~ ~ lamv~9zeo~sz9 ig a °perfeci" system would be inaudible. However, on real digital systems the sonics will improve, as the slightly higher levels reproduce with lower distortions through digital systems. Both the instantaneous peak limit/expand and the averaged compand/decompand functions are controlled so that the degree of pm~ssing can be computed and automatically Controlled as needed for best prograan reproduction.
This configuration allows very fast corrective action to be varied by a low bandwidth control signal.
This control signal can lx hidden within error correction codes, placed on other audio channels within the system, or random noise encrypted and inserted as needed into the least significant bit or hits. In this regard, the benefits far exceed any added error which is below the practical resolution limits of most equipment.
~ ~ Basically, one aspect of the system of the prCSent invtntion addresses and partially Corrects several distortions lmown to occaar with A to D and D to A
sonvergiotrs of ~mplcx signals. Some of these errors are hardware related and are oorredable with more exacting methodology. ~ther distortions are the result of the bit depth and sampling sate fixed by industry standards and are m'animiaed ax~ativeiy by the system, by varying dynamic optimisation beawsen performance aspects. Determining the best form of optimization cats be very complex, as many such dlistortions do not acxur with the steady state tgpe sagnaLt ttscd for distortion tests, and must be minimized to subjective cxiccaia. R~i~st are transient anicrmodulation dis8ortions ) of which sertain types are objectionable 50 to 50 dB below the psogram material. As will be shown, the hardware mechanisms for producing these distortions are non-linear switching and hystenccis in capacitors in sample and hold circatits, digital to analog crosstalk, slew rate asymmetry from large numbers of parts in the signal path and ~sstalk between analog and digital signals.
The encoding process of the invention makes program signal alterations with the least audible coascquenoe for reproduction from standard equipment. R'hese changes reduce certain types a~f distortions and increase s'agnal resolution, thus providing improved playback, e.g., better spatial sense ~5 and less brittleness. Using the decoding process of the invention, the reproducer can be made to track and inversely compensate these signal alterations thereby allowing substantially "exact" playback, with greatly reduced distortion.
In order to provide a further understanding of the problems with digital systems and how they are resolved by practice of the invention, the difficult tectmology areas and distortion anechanisms occurring from ideal as well as practical implementations are presented as follows:
L Resolution limitations with small signals:
Distortions in digital systems increase with decreasing signal levels, and the smallest signals become broken and tend to disappear. ,~1 good analog system and a 16 bit digital system can both handle signals with an 85 dB dynamic range.
Typical 3~ analog systems have infrequent high distortion at signal peaks whereas digital systems have continuous distortions at low levels. lratperien~ has shown these distortions to be audible, hence some form of noise dithering is often used to smooth out SIJ~STITt9lf~ ShI~~T

.
V1'd' ;'2/2B~6t) PCI"/'US92J04629 quantization steps and allow information fill-in. Th9s technique rxeates new distortion from beats between dither, sampling and signal frequency differences. Nearly inaudible tow level upper harmonics then vxeate more perceptible low level sub-harmonic interferencxes.
Fig. Sa,bd Sampled, encoded, and derided low level low ffrequency signal.
Figs. SeJ1 through SrS. above dithered with high pass Tiered random noise. See fable I for notes.
.a1 unique solution to the aforlbed problem is provided by the system of the present invention.
1!D In this regard, a minimum low level signal activity is maintained at adl tiaaes by using a ~ devir; or ate equivalent digital pry to increase the gain of the system only when the average lev~ei of the signal is low. This low level Rain riding is programmed to maintain a minimum dither-like activity which will tend to mask quanrizatnon effects or other least sig~uQt bit azonotonistic discontinuities.
. The reproduced program wil! sound litter on less expensive players net incarporafmg the invention, which often have li~h distortions from these Itinds of problems. For exact decoding, the invention provides an opposite gain-structured average level compression device or equivalent digual process which restores the full dynamics of the original signal. When this complementary process is used at the reproducer, low ?B level gain reduction occurs and quantiaadan noise is reduced.
Best operation of the process occurs when gain control is based upon a broad middle audio spectrum, as this restriction prevents gain pumping from noise and bass fundamentals. An average level detector~oonte~s:~ls a variable gain function as follows:
An D3~IIS detector ~r other averaging type devicx retxives these signals. Gain boost is determined from the band restricted pt~ram and its average Boost level is controlled by attack, sustain and decay parameters much like those in synthesizers.
To prevent ~verload from sudden signal changes, the unmodified full Bandwidth program is delayed icing enough to allow time ecinstants needed for the averaging pcot~ respond and antncipate events. The delayed program is then gain cantrolled either by adding the signal to itself or by multiplying. These methods provide the benefit of undeooded reproduction with the leev~sse audible artifacts.
.~ decoding reproducer can operate in the aforedescribed manner except that it performs a gain seduction in response to its own determination of average signal level, using a tmntrol signal detected from a reduced bandwidth version of the input ~5 signal, having attack, sustain, and decay time constant averaging to operate a gain a'°1.9E3~'fl'~'d3ll'~ w!"i~

bil'~ 9"'2200 F~'/U~92/0~4529 control process passing the delayed program. This default or non-controlled mode can be made snfficicntly accurate since the recorder has the same building blocks and can test cbe reproducer response for a program event and make touch up corrections prior to the ffitll band signal reaching the variable gain device.
5 This system operates like many compress/expand type systems. Various anternal operations like filtering, detection, gain ~ntrol, integration for tame constants, time delays, !og conversions, and curve generation can be made operable ffrom functional modules or from &nown dlgstal prosx,~ programs. An analog system can be constricted from such building block ffimctional units. )each unit is fully buffered, 10 gain structured, and instrumented allowing many different types of systems to be set up easaly A more detailed descaiption is found below as pare of the circuit destription.
2 d:atascrophic overload fiom large signals:
large signals ran suddenly overload a digital system to state severe 15 unmusical ssackling and other breakup sounds. Mgt analog recorders gradually overload with program related hasinonits, are nose fotwing, and therefore work closer to their maximum capability. Ite~rding engineers wising digrtal equipment typically will choose more sonsetvative leincls and rislt the resulting lower resolution and higher distortion.
20 Fig. 6a-f Iaigital/Analog overload on triangle wave.
A unique solution to this problem is also provided by ilie system of the present tat ~ntion.
In this re~trd, a peak litniter is used to imitate analog overload. ! Iigher peak distortions occur infrequently, but now the average psograni level can be higher, and the average percent distortion as usually lower. lacact program reproduction is restored by a peak expansion Having a conjugate transfer function to that used to limit the program. 'This peak limiting a~tt be applacd to the signet either in the analog domain, before it as converted to digital, or in the distal domain, and the conjugate expansion ban likewise be applied in cnttter domain. tic preferred xuethod is to do both operatioias in the digital domain, since the expansion a~n then be made to track the ~otapruscion exactly, shad the shape of the limiter curve can be controlled for least distortion on undecoded playback. In order to make the scheme work effectively in the digiwl domain, however, one must leave an analog..to-digjtal and digital-to-analog converters with suffiaent amplitude resolution to Handle the dynamic rang.
SlJ'BST1'fU"~'E ~~H~'f w~ 92i22oso ~t.-ri~~oxit~a~z~>

If the peak limiting operation is applied to analog signals at the encoder input it can be digitally conjugated to yield a linear signal of more data bits.
This effectively creates an A-to-D converter with higher resolution which has higher errors for signal peaks where the diode compression occurs than for the majority of its dynamic range.
Building block or functional module circuits can be hocked up to perform the input and output analog operations. Voltage controlled amplifiers, four quadrant multipliers, log-exponential converters, and multiplying DAC systems are available.
Most of these devices make a more logical and direct implementation, but they also have temperature tracking problems and most of them create higher noise and distortions than the method we ttse.
In the digital domain, DSP programs can work from lookup tables, arithmetic sub-routines, and process combinations. Just like low level averaging this operation can test decoder response, determine a fix, and tlien make the correction to a delayed data signal. Certain types of transient ringing occurring after decimation will create 1~ some re-construct distortion to very large signals. These errors are similar to those in other dynamic range enhancement systems. They appear to be an accepted compromise of dynamic range enhancement and for now we have not dealt with the problem.
3. Limitations of low pass anti-alias filters:
The industry standard low sampling rates force a very narrow transition region between pass and reject responses of anti-alias falters. This very greatly increases complexity of either analog or digital filter implementations and prevents having all aspects of filter performance optimal in one design. A compromise results.
Best resolution for extended high freqttcncy response and Nyquist rejection necessitate poor dransient and time settling response. Less aggressive filtering gives less high frequency extension with improved transient settling or gives improved high frequencies and poor alias rejection. Filter sonics are different from one to another and each works best with certain types of program material. The filters shown below are symmetrical, fixed group delay types made from large numbers of sections representative of good finite impulse design practice.
Fig. 76 Low Alias filter: extended maximally flat high frequency response, maximum alias rejection, steep transition region.
susswrs-ru~r~ s~o~' s."~,,,~'T"~,'N'aN:.--_. __..~,.__....~.y..~,..<.uaeuae.~,:....~__. .
~ -1' 92/22060 PC°f/US92/~4629 Electrical: Rapid changing high frequencies have unsymmetrical sidebands which create vestigial amplitude ripple effects. Long settling time for transients.
Sonics: Inner detail sound is compromised to get low alias distortion at high signal levels.
Fig.7c Compromise filter: Reduced high frequency response, larger eransition region, reasonable Nyquist rejection.
Electrical: slower transient response with faster settling, less flat frequency response.
ltd Sonics: Dull soft dynamics, moderate inner detail, moderately clean high frequencies - best for simple low level signals.
Fig. ~d High resolution filter; good transients, but peaked high frequencies and poor alias rejection.
lrlectrical: High frequency response has a dip and then a peak, 15 producing steep transients and fast transient settling time.
Complementary decoding yields very fast transition speed to infrequent transient edges.
Sonics: Best for complex sigyals when alias distortion does not create problems. Poor cymbal sound.
2~ Near ideal digital or analog implementation off the above filters is theoretically possible. Both are characterized by similar equations. No one of the above works best fox all situaeions; each has its compromises.
The aforedescribed problem is likewise resolved by the system of the present invention.
Observe the program data and dynamically choose a best encoding filter automatically, based on the program content. Send to the reproducer coded control signals indicating these filter choices. This makes it possible for the reproducer to c~34,3~~'fA'6't,-I"'I'6~ ~E~~'E' PVC' ~~2t~b0 ~ ~ ~ ~ ~ ~ ~ PC.'f/d:S9Z/04629 initiate its own complementary or conjugate response to the encodzng filter characteristics.
Falter corrections require an operational system consisting of encode filters, decode falters, selection log'ac, a means of switching from one filter to another, time delays, and a means of encodiaig control signals for the reprodaacer. Each of these can be performed in the analog or digital domain and quite often easy processes in one are quite difficult tai the other. A brief description of each subsystem follows:
Digital lw'alters:
Compleae filters are best created by specialized DSP chips. A typical DSP chip tt) is ~nfigurcd with ?.4 bit mtaltipliers and 56 bit acr~mulatotx in a functional configuration which is vary c~t~ent at performing the multiply and add operations required for digidal falter algoagthms. 17SP chaps can be used to make symmetaacal Emits impulse response falters, convolution networks, spectrum analysers etc.
In most instances, a filter equation is des'agncd using a ~mputer to smaulate its response and the resulting coe~cients are then incorporated ~to a program for insertion into the DSP chip's memory. PI~DIvI chips are programmed with these numbers and connected to the I7SP, or for volume production an eeltiwalent mask programmed yt~M anay be used, which may even be resident in the DSP chip. I-lencc, different falser respoaases ,can be achieved by changing to different program coefficients or ~0 different programs, all stored in ItOIvf, or by using several separaee DSP
processors, each running a siuagle filter program.
"fhis approach applies to decianadon, which reduces oveasampled data to an "alias free" lower sampling rate and is used during encoding, or to interpolation, which pnoduoes an oversampled signal sad may be used during the decoding process.
Eoth 23 imvolvc the ttse of low pass filters and both rely on multiple or oversamples of a base or industry standard sample rate.
Decimation lilacs a stepped "suave" of multiple samples for each one needed and fords the best number for each sample at she des9sed rate and discards the rest.
Since wav~eform reveals within each final sample period do not reproduce and only 30 waste distoation because of the ldyquisi limit, a low pass filter operation muse remove the alias frequencies of such events, eliminating the audio conscquen~s. In one ' system, ~ue use a 16 hit conveeter operating at an 8 times oversample rate, and in another, an 1S bit converter mperating at 4 times oversaanple ante. In theory we gain ~-3 bits from subtractive dither with oversampling and gain another two from limiting.
wring any one sample, hundreds of DSP operations on these bits will have been in S~E3 ifl"1'49'i'llr ~hl~~"6' ~~ ~:.~a~~
~~.~0~.8~ .
~~iu~gai~asa9 process and accumulated to produce an enonded low pass filtered number ~i bits long.
For our system, 20 - a2 of these bits can have useful information.
Interpolation takes each sample and creates a stepped "curve" of many computed intermediate values or oversamples. here the intent is to reconstruct a S waveform like the original signal. As before, hundreds of DSP operations arc needed and again fully processed points between samples as well as any computed signal restorations become added bits to the DAC. With the system of the present invention, we then get 2.4 bits at an f3 times oversample rate. Of these 24 bits, appro:dmately ~0 bits contain useful inforaaation, and digital to analog coaverters with 1!) this resolution arc just becoming commercially available.
I3ccause ~f DAC performance lamiiations, it may be ne~ssaay to handle the added bits from Limit rc~oration and low level averaging during or after digital to analog conversion. In one ~cssion of the system, an 18 bit 13?C sampling DAC
and noise draping is used to achieve a theoretical ~-3 bit resolution improvement to the 1S seeppcd curve interpolation.
Filter Selection Logic Best filter choices are made when program conditions reveal a compromise problem. Strong high frequencies, isolated fast transients, and continuous low lcvds encode best with specialized filters. Fortunately, each condition as easy to identify and 20 each is most likely to occur by itself: Conditions such as Loud or soft, eontinuoru or bro&en, aced strong treble are xepresenLatave of those which are identifiable and cause problems. Program materials arc not predictable and solo voiceJinstruments, synthesizzr, percussion, and so forth array present rapidly changing requirements.
1'Tnforiunately, filter lengths, cuhan~d 'transients, fitter merges, and ideneifying x5 program conditions all operate with time constraints. Therefore, filter changes may have to be resixicted when program a~nditions ~!~ll for choices which alternate too rapidly.
A best compromise over time is made by memory enhanced variable threshold smart logic. Normally a filter "call" as initiated in rcssto a compromise 30 situation. The call rcpnesents the intensity of the demand for a particular biter, caused by the compromise situation, integrated over a weaghted time window. If the program doesn't change much and the call is atot continuous and doesn't happen again, the previous best filter remains. 3Fach sussive call, its length of time and intensity increases the response sensitivity to engage that filter. If this choice is made, then the 35 response sensitivity to engage any other filter is reduced ours a preset time and the process like above now can repeat for a new filter. This selection method eliminaecs ~tJB~'TITIJT~ SHI~~' ~'(''-'~?/~2f160 ~ ~ 1 ~ ~ ~ ~ PC"T/IJ~92/i9~lb~~
haphazard filter toggling and still allows qtuck filter changes when a strong filter compromise situation occurs.
f~slthouglt this operation is suited for digital implementation, a hybrid analog and digital circuit is less complex and has allowed easy experimentation with S programs. Diodes, resistors, capacitors, comparators, and current sources make up anost of the adaptive decision making elements. Calls arc voltages through resistors charging integrating capacitors. Cal) urgency translates to higher, valtage for a longer time thereby increasing the capacitor charge rate and amount. Freguency of calls is the number of times this charging takes place. A filter seleri threshold is initiated d0 when the capacitor voltage triggers a voltage comparator. To prevent indefinite erdtaneemeni, a time slot for decision making is creaeed by a negative going current source which eventually discharges the capacitor to reset conditions.
Each &iter is engaged from a comparator as above. I-Iowever, when triggering and a ~Hter change ocettrs, ail the antegrattng capacitors for the other filter 15 slaoit~;s are ° sod held inactive for a pre-determined setup and run time.
Gnu completed the pros resets and starts again without prior memory.
Program ~mmditions are auowath but~ding block type analog circuits.
Fligh pass filters and peak detectors sort out alias causing high level upper frequencies. Dealt and average level detectors are ~mpared and the difference ~l response integrated to identify transients. To reduce the influence of program level variations, voltage ~antrolled amplillers sexvoed from RMS detectors are used to scale the peak-average operation. Each of these circuit groups then produces a positive going averaged output voltage whose amplitude and time duration is related to the degree of filter compromise, or the desirability of choosing a particular filter type.
2,5 Strong high fa~,guencies, fast peaks. at various program levels, and high overall average intensity all convert to similar filter call voltages, each proportional to the magnitude and repetitive nature of the program event.
Filter Switching and hrferging:
Filters leave di~'ererat lengi6s, instantaneous phase shifts, time delays, responses eta SimQle brattal switching fiom oae filter to another would create serious glitches, and other very audible disturbances. Some form of tone alignment, fading and merging or a parametric change within the filter from one type to another is nu~ssary. All of these techniques have been done in both analog and digital falter cltattgtng. Early light bulb photo-sell and VCA type fader-switchers are common analog methods. Ivfany digital synthesizers use mix, merge and parametric falter changes in various eombinaeions to produce inaudible transitions. Good examples are ~U~S°('I°I'l9'8'~ ~W~~' ~'O 32050' Pt.'T/iJ~92/04629 Fairlight and Synclavier machines which have elaborate digital tracking filters which work on these prlnciplrs.
A simple implementation would use LED photocells and time delay correexion to switch filters. For more advanced versions, the filters arc implemented using DSP programs, and DSP pro~ams are used to perform the mix, merge, and coefficient changing functions.
4. Frequency response ljmitations imposed by industry standard sampling rates:
The frequency respaase of digital systems is fundamentally limited to half of the sampling frequency, in accordance with the Nyquist theorem. For a current digital recording media, Compact Disc, this means that one cannot record anything above 221t91oHerta. This limit was chosen based ~on the assumption that the human ear cannot hear sounds above about 20 kiloHcrtx. Rest research has shown, however, that humans use transient information in sounds with frequencies much higher than that to detcsmine the direction fiom which the sound has come, and that eliminating those very high frequency companeats impairs ones ability to locate the source of the sound. The inner ear actually lass nerve acceptors for frequencies up to about 80 ka~oHertz. Therefore, if the "prick walC' low-pass filter, which is a neocssary part of all digteal recording, r~emov~ frequencies above about 2~ kiloPIertx in transients, ii reduces the level of realism in the sonic image.
Tn accoa~danae with the invention, the waveshape of critical transients is reconstructed at the reproducer based on information sent from the encoder over time. The steady state bandwidth of the digital channel is set by the industry standards, but, for ooarsnonal transient events, additional information on the shape of the waveform cart be spread out in time and sent along for use by the decoder.
There:
2i arc a number of different methods which can be t!sed to accomplish this, all of which make use of a control siglaai or "side channel" of irtformacion sent along with the main signal, described in more detail lader. They are all non-linear processes and therefore should be used sparingly.
The methods of daansient reconstruction employed fall into three categories:
a. Waveform synthesis b. In-Between sample generation Slew rate compression All of these methods rely on starting with an accurate waveform of the transient resulting from an ovcrsampled original signal which leas the higher frequency ~UB~'u'rll~lJ'T~ ~0'~~~

wt~ ~zrzio~o ~ - ~ ~ ~ ~ '~ '~ ~ ~c: r>v~9zr~z9 information intact. In the wavcfform synthesis method, a transient to be rerconstrucxed is identified at the cnco~dcr, sad it's wave shape is matched to one of a number of predetermined "standard" isansient shapes, which are known to both the encoder and decoder. ~. command erode idemmtif3nng the shape is sent through the control channel to the decoder, which regenerates the shape, either by seeding it out of a lookup table or algorithmically generating it, and scales it to the amplitude of the bandlimited transient arriving in the main signal. 33ie decoder then uses the synthesized swaveshapc to norteM the shape of the transient and approximate the original.
~'he c~c~on can be 'sn faun of a difference ~iween the band limited transient and the ltl original, which is added to the band limited signal at the decoder.
obviously, only a limited number different ax~xxa:etions ran be used, lance one enact be chorea an a aseasonable tame at the cneoder, all of them anis$t be remeanbered at both ends, tokens must be assigned to to the choices, and time is rr~tdred to s~mthcsi~cc and scale the correction at 4he xleon3er. Idevertheicss, nt is possible to achieve an apparent ILS increase in available bandwidth wath only a few shapes. °I'his method has no oomprosnisc to the shape of the bandl~ited signal ta~ept the prtsenre of the montrol command, and therefore is not aucL'ble on non-decoded playback.
fee in-between sample generation method is very similar to the above, except that instead of sending a token representing a tremcmbered ~rrecsion, the eaooder ?UI sends the actual waveform correction over the eoatrol channel, spread out in time to accommodate the low bandwidth of the side eL 1n it's simplest form, this correction can be the value of a snngle 'in between" sample point falling between the noa~nal samples of the band limited signal. 1"he decoder can use this point to correct its interpolation of the signal as it generates an ovcrsamplcd signal prior to conversion 25 back to analog form. .As above, the only effect on non-decoded playback is the presence of the oantrol channel.
'The slew rate compression method is different from the above two in that the additional infotmatioa reeluired to eanstrtid the transient is spread out in time and sent as part of the main dal. 'T'he ~ontrol claaanel is used simply to activate the 30 pt. ~'bis method is con~ptatally similar to a technique used to enhanex the apparent bandwidth ~f a video monitor during transicats by slowing down the assn rate during the traatsacnt and speeding it up a~ to make up the lost tinic.
'~6lhen the slew rate, or sate eaf change, ~f the waar~'ot~ e:ds a thrcsliold, it is limited to .a value vvhiiclt can be seps~ented aerurately in the band limited sisal. 'The degree 35 to wvhich the speed is slowed down is soled to the speed of the original transient so that the decoder rain anfcr the original slcvv rate from the slow oats which it can observe in the recorded signal and speed up again. Since the transient is spread out ~tJ~~'TdTtJT~ ~HE~' i1'~ 9Z~~2fi6~ ' .
I'CI'lUS9~/Odb29 2.g nn time, the time maest be made up somewhere, normally aftcswards. do order work properly, the transient must be isolated so that information near it in time is not lost.
"I2lis method definitely does have sonic consequences for undecoded playback, but analog tests indicate that a surprising amount of slew limitatatg can be done without being objectionable.
Dynamically controlled system:
.All the above improvements arc mast effective when the reproduce decoder change to complement ~nditions of the record~l progeam. The record encode proeess~can generate a hidden control code soneealed in noise as one method of lttl controlling these aciivitacs, dtandom modulation of forbidden numbers in au error ~ort~taon code ar user code is another way that the control codes can be included with floe data. These caa be continuous or initiated when needed. Whey the code hides in &hc psog<am, digital c~Pies from one format to anoeher will preserve the code whey analog copies will not. These features can identify unauthosiacd copies, 15 as well as convey production process information that anight be used for motion picture work, ete.
Algorithms and lookup tables in the decoder provide cusvattue shapes, bane constants, l6evcl thresholds, muirsplieas, filter oocfficients and ocher useful data also "known" by the encoder. Without continuous control iaformation, the system can run 20 default where the recorder/encader is set up to anticipate the reproducer response.
Either feedback firom an internal test for best encoding or feedforward of a previously worked out response will do dais. Ivdost control activity is needed to access and change a particular complementary reproduce ffunction or correction. Hence the improvement is much greater than the information bandwidth loss necessary to make ~5 the improvement.
S. Digital to analog and analog to digital crosstalk:
The smallest mmnalog signals, 'sn the iQa microvolt range and under, are easily contaminated or interfered with by digital data streams having millions of times greater energy. Faster prormmg nncrcases the energy per bit as well as the number 30 of interferences per second. The same situation osxurs with larger numbers of bits.
interconnects, cables, and enclosures pick np this eacrgy, stare it, and create delayed coaupound antcrac~lons. Ifiigher speeds require smaller packa~g which increases such crosstalk, unless wires and ~ther parts are also made smaller.
Fig. a'b Analog waveform changes occurring from digital interaction.
St~L3~'~aT~JI ~~ ~~~~' ~'~ °'x/22~60 ~ ~ ~ ~ ~ ~ ~ ~G 1'/LJS92/~4629 F"ag.5c ;sample and encoding errors from rapidly changing waveforxns.
F"~g. Sd Sample and encoding errors from low level waveforms.
gn accordance with the invention, a silent conveasion system is used to resolve the above problem.
Pdosmally a;digital system operates with a continuous clock which times its inter~al options. ~llaons of timed oo~ur each second. ~iencx the system state may or be an a state of change at any moment, particularly the sxitical sample einae high accuracy is needed most. Sample tune fitter and rhgital to analog crosstalk may a~asult.
The system of the pit anv~cneion stops all operations except the sampling clock long enough before sampling to allow the energy stored on cables and other energy storage parts to dissipate. One pulse inieiates sampling which then occurs during eiesxxical silence. ~nce dal cagiure is complete, other processes resume.
1S ~. numlxr of prior art approaches have been developed td r~lu~ some of the above distortions and are d~cr~'bcd as follows:
1. Group delay of low pass Falter arxomplisha;d by using an all pass phase shift network.
2. 4~uantia~tio~n noise reduction Pram use of 1 or 2 least significant bits "keep alive" ultrasonic dither.
3. Reduced granularity noise from Case of balanced or push pull circuits, achieving high oammon mode rejection of noise spikes and e~,gital-analog stalk.
Further reducBions made with optical isolation of logcc and oom~ertes systems, as well as high impedance isolated power supplies.
4. Slew induced errors reduced by super fast symmetrical analog circuits.
'lxaese improvements help reduce the harsh, congested sonics, and may slighely expand spatial sense in good lTowever, inner detail and correct space perspective are still lost even with such prior art approaches.
~efexring now more particularly to Fig. 8 of the drawings, there is shown a ~0 prooes$ing system, in accordance wills the invention. °~fhe system modifies signals to achieve Z8 bit perfozmancx fram a standard 16 bit converter syseem. As mentioned peeviottsly, the average level of ~l9~ar''T~IT~~I'1'~ .cuk~~~°T

very small signals is expanded while the occasional instantaneous peaks of very large signals are soft limited. When both operations are carefully done and are digitally encoded and then reproduced in a standard manner without decoding, the sound is improved. Ambience and articulation more like the original program does occur, even though the process would be aud~le without the intervening analog 5 to digital and digital to analog conversion. By the time this expand Iimit process bccomea clearly audible on standard equipment, the dynamic range for fully decoded reproduction in accordance with the invention has increased almost 20 dB. Average resolution is substantially more than 18 bits.
Each numbered subsystem element in Fig. 8 is an important stand alone operation typically performed by an independent arcuit card or module. The corresponding schematic 7l0 circuits for implementing the system of .Eig. 8, while deemed to be within the design purview of one of ordinary skill in the art, are included for convenience in Appendices A and C
included with the U.S.
Priority Application, serial number 707,073 filed May 29, 1991, and issued to United States Patent 5,479,168 on December 26, 1995. These circuits are numbered to their respective block functions in Fig. 8.
15 The small signal average expand subsystem 61 operates in the following manner. The incoming audio signal is band restricted to SHz to SOOkHz, thereby preventing DC level ~hift~c supersonics, and radio frcqueary components from ovcrslcwing amplifiers and influencing process control parameters. Components Cl, Ll, Rl, G2 and R2 and buBer follower components J211 and J271 in Appendix A of the U.S. Priority Document perform this inside-outside world isolation. Two active 211 signal paths are provided, one passing through a voltage controlled amplifier (VCA and ICl) and the other origuiating from a buffer. Both signals arc in phase, however, a control signal marked 'compensation input" can set the VCA output from -40 dB to well above the buffered output.
Since VCA devices are well-known distortion producers, this configuration allows the clean buffered signal to pass uncontaminated when the VCA is shut-off at it's -40 dB gain.
2'.~ Only during very small signal conditions" when distortions are less important, is the VCA gain made large, allowing its substantial output signal to be added to the buffered signal. A higher overall output resvus. Other circuit components are needed to perform housekeeping functions necessary to prevent crosstalk betvveea control signal sad VCA output and to adjust for lowest distortions.
The phase shift network subsystem 62 corrects analog filter group delay 30 problems and operates as follows. The buSered and VCA signals summate or combine in phase at an atteauator Rl R2 R3 in Appendix A of the US. Priority Document. With average program signals in the 0.05 to OS volt (0 to -20 dB range), the VCA gain is set at -40 dB, making its output and distortions inconsequential to the accurate bu~'er signal. VCA gain increases substantially for small signals in the 0.005 volt and under (-40 to -$0 dB range) where monotonidty and discxete step-by step type quantization errors from the AD_DA process are becoming increasingly large.
This added VCA output maintains a "keep alive' status or minimum bit number change rate at the A to D converter. Here the signal and its background noise become dither-like and a minimum useful amount of it is maintained by the variable gain VCA independent of signal conditions. Sometimes noise dither is added to digital SUBSTITUTE SHEET

PCfJU~92J04b29 ~A'~ "22060 systems and in pratxice this noise is quasi-audible. The active-dynamic dither has similar properties except that when needed, the origuial program dynamics can he restored with a controlled playback VCA system.
Analog filter compromises are then corrected by all pass phase shift correcaion circuits placed in the signal path. These stages marked A, ~, are twice iterated. Each section of J557, 1211 and J271 is connected to make a unity gain buffer - inverter giving in phase signals at RS and 180 de~ecs out of phase ai Iz;4. Foth signals combine through Ra and Ca making a flat response all pass system havi~ near 0 degrees ~t low frequencies and 380 degree phase at high frequencies. The four eonabined scions still has flat frr~uency response yet exln'bit an abrnpt ?s0 degree phase shcft an the 5 DcEIz to 30 Ic~Ea r~~on. Tinis eorresponds to a ,saSe~ group delay change which partially cantxLs a suddeaz group delay shift occtersing with mangy analog cllipti~l low pass filters. PTathout comgetuation, serious transient ringing would ripple an~lulate on and o~ the subsequent peak level limner and would muse excesave gain modulation to oot:ur at the reproducer when restoring or expanding the signal pta&s. The group delay corrected filter lass very little ringing and allows much more predictable peak limit and expand operations.
The low pass falter subsystem 63 X701 filter) operates in the following manner.
'This is an-essential sand very troublesome part of all A to D converter systems. It stops or rej~ats frequencies near and above the ldyquist limit or 1/2 sampling rate. For ideal systems, its 'stopp'utg:
action aqausi be better tlaaa the least bit resolution, while its pass action must be ripple fret and in proper phase alignment in the IS Hz to 201d-lz range. At 44.1 kl3z sampling and 16 bit encoding for Compact Risk formats, the filter snarl drop of least 85 dB between the 2D 1d-Iz audio limie and the 22 kHz hdyquist limit. Mathematics of traditional analog filter designs require compromise decisions related to numbers of parts and them signal degradation, and alias, group delay, and ripple compromises. For the system of Fig. g, a eompromise of more pasts for better transient and group delay is taken to allow bettea peals limit ~ expand operation. For example a delayed high ffrequcncy pan of a revraae sweep, which happens fast, ran add or subtract fr~t~m lower frequency parts of the sweep happening Later. Thus, the instantaneous frequency aespoase can change with fast shanging_signals.
ltipplybeats throughout envelope and 'tail" occur when such a sweep repeats back to highest frequency.
A shortened envelope also occurs as delayed high frequencies of an uncompensated filter occur within the sw~e~p envelope. In this regard the delayed high frequencies continue propagating and internally combine with other s~gnaLc'withiu the falter after the sweep envelope has completed and started the next cycae.
1'he high level peak limner subsyseem 6h (limner, expander, dither generator) operates as follows. IC101 in Appendix A of the U.S. Priority lJocument receives the filter output and completes transient response ripple ~ompensation previously ataentioned. (31 and f22 with IC 1028 perform the peak limit function while Q3 and Q4 with IC 103 perform the restorative peals expand function at the reprodutxr. Added parts IC10?A allow one to observe the signal waveforan peals that SUB~1'1TUTE ~FiEE~°

'WO 92J22ti6f~

has been limited aaxd IC104 generated supersonic and near supersonic noise to dittaer the A to D
conversion.
One low pass altar compensation marked "B" in Fig. 8 inciudes a combined notch - peak circuit around OP amp ICIUCi. A tuning of 18.5 kh3z high O
partaal notch and a 2L5 kHz sharp peak become added ~Iter sections which help smooth and reduce singang. A
rougher, but acceptable, frequency xesponse is made and the peak transient ripple is less than 5% with completion the same as the square way risetime.
Divider Iti and R2 sets gain structure and source impedance to a peals limner oxslng idealized diodes made from transistors Q1 and 02. These parts sin be °supermatched" pairs i0 having many devices random connected on an IC substsaRe yielding near ideal jundxon performance.
These behave close to the ideal logarithmic junction relationship of:
vt - d2 ~ .~T 1°gn .~1 9 Ha giving change of forward voltage to ~opcrating current ratio. Boltzman canstant, temperature, and electron charge arc considered constant. Instantaneous resistance of dV/dl (rate of change of voltage to curacnt) becomes related to 1/currertt once limiting action begins. This relationship tracks over a ~tl dB (100 times) range for reasonably good transistors thereby allowing easy accord - play peak signal tracking in the 10 dB one to two bit process range. A practical setting is a 24 volt peak to peak triangle wave compressed to i.?. v peak through (.~i, O2 yielding a 3S v pp output at IC10?B. This 1 bit (6dB) compression can be monitored at °dest° output showing the clipped portion which for set-up adjustment 9s made symmetaical with the SOk ohm control. Itrstorative operation is demonstrated by connecting IC102 out to IC106 in. Observations at IC103 "protect" output shows tracking ranges in excess of 20 dB whoa needed Operation of the analog to digital conversion subsystem b5 and digital to analog subsystem 66 is as follows: Output reconstruction from the 17 to A
signal must occur prior to the low pass filter subsystem G8, otherwise phase shifts would alter the signal waveshape and consequent expand tDaseshold points. Consequent sampling fetdthrough, and inttrferences as well as step sampled data require very fast circuits. Amplifiers have additional stabilization and speed enhancement.
Components marked Rs Cs perform these operations and are specific to the amplifier types used. 9~Jith decoded stop waveforxns, the limner function must quickly settle to each level and associated amplifiers must not overshoot, x'mg, or have unsymmetrical rise and fall times while doing this. As noted, limit - catpand fiunctions must occur at direct ~upled or DC pass circuits not having phase shift. A to I~
inputs and I) to A outputs satisfy these requirements when the low pass &itess are not included in the path. In pracxice, limiting will create upper harmonics an the Nyquist range which could create alias excise which would confuse reproduce reconstruction and add considerable distortion. Fortunately, practical operation allows modest compression and expansion of occasional peaks which happen in music and speech program material. Unlike instrumentation signals of instant amplitude maximum energy to band edge character, upper music frequencies are usually harmonics of less energy than WC -'!22060 ~ ~ ~ ~ ~ ~ PC,'T/U592/m4629 fundamental tones. Alias foldover is then infrequent and occurs only at peaks which best mask these problems. Figures 6a-6f show various signal waveforms during the limiting and reconstruction of an illustrative triangle wave.
The low pass filter subsystem 68 operates in the following manner. The peak reconstructed sampled A7 to A signal output of It~°103 .is routed to a simple low pass filter. The 14.1 kHz and up step components arc removed and the waveform is rounded and smoothed inherent to the filter characteristic. filoise and transient spikes are reduced to tolerable levels to prevent overslewing the ~/CA portions of the law level srgnal campr~ess circuits to follow.
Operation of tire small signal a~rage mompress buBer and Vt:.A subsystem 69 and Line .Amp 'hfrer Subsystem 70 (small signal compress, line drivt) is as follows: Both subsystems serve similar functions eo their subsystems 61 and 62 ~unteaparts in the record sccrion. )3ulfercd and VC:A output voltages are similar. However, this trine the 1'C:A output as subtracted. As before, lower distortion is achieved by g the Vd:A at -30 to -~0 dB level relative to normal level srgnaLs.
Increased V~, gains rrxlu~ sagnal outputs until at +10 dB, a null or ~ srgnal maximum compression occurs. With this arrangement, any reasonable signal expansion can he compensated and the system distortion is lowest for the most probable average level srgnal conditions.
<:ontrol srgnal generation is aecomplishcd by the limitrx-buffer subsystem 71, barrdpass filter 72a and Ii.P~IS detector subsystem 72b and gain subsystem ?3.
High level peak limit -expand thresholds and low level average gain set aooncrols are needed. O~rcuit settings can allow both types of control to be tested independently of each other. Best operation occurs when tire control sig~ral anticipates the program waveform to be proetssed and, hence, an audio pre-delay is used to allow contml stabilization prior to control circuit action. Low control bandwidth is nettled to m'mamize non-program least si~niCcant bit activity. One method of doing this as to have an aMive - inacxive control status. Since high level - low level program signals do not oocw simultaneously, the reproducer can make its own decisions as to where the control is applied. The unused operation then xcturru to its inactive or nominal process state. High level, normal program type signals, whore cxgand - compress functions are urutecesaary, then have inactive control scales. As signal levels de~rase, an internal Dimit diodc/cla~mp rtleases and the VCA gain rapidly increases to create summation signals.
Further program level reduriions modulate the FICA gain in a controllable, ~0 predictable manner needed to maintain digital "keep alive" hSB (least significant bit) activity. For most eondidons, these lonarest level will be mid-band acoustic nurses and arumerous types of electrical noises. The latter may include ItF antesfcrenoe, light dimmer pulses, security system signals, and high frequency peaked electronic noises. These often have low audibility compared to midband acoustic sounds. Hence, a sharp cut-off bandpass ftltcr and wend wide pack to average level capability 115 detector a~ used to assure that the e~ontrol signal tracks audio sounds and riot inaudible interference.
In the small signal average expand subsystem t51, to handle the entire large dynamic range of modern program materials would require very low noise falttrs and detectors of quite difficult electronic design. Since the avrrage level circuits are low program level active only, the process ~IJBSTITUI"~ ~uH~~

i3r0 92;!22060 ~ ~ ~ ~ ~_ ~ ~ PCT/US92/Od629 gain can be very high. This allows reasonable circuit voltages to occur during quiet passages. Strong signals normally creaeing overload are smooth Limited with a semi-logarithmic curve to create minimal compression harmonics. Noise, transients, and other uncontrolled overload behavior are then prevented from crosstalking to the signal path. Campanent ICl is configured as approximately 100 times small signal unity ~ in Su~ecsive diode llmiters provide cx3nduction with incxeasing signal voltage to give smooth "overload" behavior.
The bandpass filter subsystem ?2a prevents low level inaudible electronic noises from modulating the average low lGVel process. The an log breadboard part covers a 200 Hz to S kHz range and is made ffrom two sections of active combined low pass high pass feedback type filters.
These have a slight rise at band edge frequencies followed by a near 24 dB per actavc cut-off. Note, that a front end buffer (J: ll and 1271) is used to prevent blur impedance loading from 'snteracting with input signals.
The filter output drives an I~dS detector module, a DBX-type component having an averaged DC logarithmic output aelative to AC input matching the logarithimic VCA
IS charaaer. As configured, a 1~ mV output change occurs for each 20 dB input signal change. This gain structure from the limiter and filter through the converter 100 mV
control rangt e~ieth very liule noise for 86e miliivolt type sdgnals ooctuaing at least sigxtibc~nt bit resolution limits. Control levels on either side of this range reprcsen4 front end electranics noise and normal signal operation.
Normal compressor - expander compromises are employed to assure minimal VCA gain modulation (distortion) from ~.C components in the control signal.
Components Cl, Rl perform this acsponse averaging for the turn-off time constant. A much shorter turn-on time constant from the IiJktS module internal impedance and Cl occurs to allow fastest rrecpommse to sudden program level increases. The short - long time constant action is typical of many compand systems and because of low frequency distortion requirements, is set very long. (20 m~ec on, S00 mSec. off for ?dl dB gain change) This very slow response necessitates an 9nput signal delay to allow control signal buildup before sudden signals ooau when VCA gain ns maximum at low level signals. In practice, thc.analog delay adds considerable distortion, and would not be user? in its analog form ffor high quality systems.
Digital proccssoss can perform all of the above level limits, band restriction and detection. The needed time delays to get best pexformancc arc simple, brst in first out type, opuations. Two advantages of long constant operations occur. Low frequency distortion is reduced gad control signal bandwidth is much less, thereby reducing the amount of bit borrowing needed to pass the ooutsol through the audio encoding.
The DC offset and gain adjust subsystem 73 is the control signal amplitude, offset, and limit nerve center. It adjusts for t~lerances between amplifiers, delay lines, VCA's and the ItMS detector and, hence, performs general circuit housekcepaug funtxious. It also is a limner to give maximum and minimum VCfI, gams needed to implement the control signal inactive and minimum input Signal presets.
SI~E'~~T'ITI~~°E S~BE~' 9'Vt7 ~ "f2205~ . ~ ~ Pf.'1'IL1~921~h29 For syseem considerations, both practical Vt:A's and the ItMS detector operate with logarithmic control to signal and signal to control relationships. Hence, changing an offset changes a bxed program gain in d)3. This makes possible large dynamic range gain control and yet still maintains low level sagnal control. Both can occur with reasonable conerol signal limits. In addition, 5 coairol systerai gain changes, such as limiting, gave direct dB to voltage gain ratios making very simple compress - expand ratios and assu~rlag input - output tracking by simple polarity reversal of the control The o~'set din adjust cit inverts the control signal for input output tracking, adjusts I~ offset of each to match gains at a pro-determined ~onttxtl signal level, and has 10 control gain adjusts to make level inch match reproduoe~r gain decrease for varying control levels. Diodes CRi and C~i2 perform level limits to preset a masdmum practical expand redo and a maximum compress ratio under psk. As constructed and e~n~gmed, this circuit has compress expand ratio adjustment int~rac~ive on a single control and the threshold of when the process starts on a second control. Sing these am I,C level of dB operators, process control ~tgnals to these points will 15 creaee dynamic prop changes. A8 presef'at, this ~s a manual adjustment.
However, clearly the process start level can be dynamietHly depending on program a~vaty and other considerations to rtdua audibility with normal non-process playback.
The delay line subsystem 94a~cb can be a balanced 1D0 coupled self-cloc&ing variable delay line. C~rge coupled devices arse operated push pull v~ath staggered clocking to make 20 the lowest posst'ble l~ drift, distortion, and clock noise from relatively poor performance devices. OC
to 25 kHz, minimum overshoot, » dB dynamics are achieved.
One delay 7~a is need to allow control signal stab'alizatioa to prevent VCA
overload from sudden signal changes. 'his delay also allows anticipatory pros control signal strategy to be computed. A second line 79b is for test purposes when tas;ng a system wriehout the hidden code 23 subsyst~t to match the A to D and 1D to A encode-decode process time to make output signals track inputs. 13n borrowing, in which the cxtntrol signal is noise cncry;~ted and hidden in the least sigaificaat bit or bits of the digit~ml signal, ire the normal mode of operation of the system.
lamner dynamic control via VCA 'anpue/output tracking can be had at high signal levels by removing the limiter-buffer subsystem 7i and operating the system as a straight ~0 compressor-expander. Since flee peak limner-eacpander is within dais system, its operation is changed along with the gain variation programmed with the offset gain adjust snbsyscem ?3. As noted before, a process control input can be operated so that all times when signals are loud, a certain percentage of limiting takes place. This is peogram dependent as seine classical music will have infrequent peaks vrhile studio processed rock and loll is snore likely hard ltnxited and will leave many small peaks 35 occurring neatly. Although this threshold control may be accomplished manually, ntmaerous computed variations vvtlt bvor& more effectively to keep the process least audible wtaen reproduced on a standard non-restoring player, just as in the case of tine lose level pto~ss dynanuc control. As with the average level expand-mompand, the limit ettpand threshold control need only be very low bandwidth.
SU~~'f'1"1'IJTE :yHE~'f W~ 5~~~.Z2060 ~ ~ ~ ~ ~ ~ ~ g"~'/U592/0~629 A 10 Hz ~onuol bandwidth is adequate and since only one operation occurs at a time, only one control for both operations is needed. The player can dciesmine program levtl and switch functions. In the a7lustrative system described, the control is manual since each operation is a different set-up. However, them is no foreseeable diffiaaUty in malting this automatic if duplicate iICA's were set up for limner gain structuring.
The following further describes tire theory, design concepts, and early development, construction and operation of an encode-decode system demonstrating the basic principles of the prc~ent invention. Its operation,is analogous to the rcoonstruct proeess based on choosing optimal curve fitting tuhniclt~s to get best waveform re~ns~nrtion. As desrn'bed, the process changes for different signal conditions and the nnmbes of such proctss optimizations per unit time can range from a few per program to many times per second. Even faster operation changes are pc~iblc.
Hawc~er, the Col signal nWrded to ache proper opeeati~al program bceomes more aoomplex and bandwidth consenting.
The basic system contains two record proxssors and two decode processors, each of which is eomplemeacary to and matching as a system. Either system is selocxed automatically by variable resistance photoc~ductive cells working as slow faders in the s>gnal path. Iaght emitting diodes illuminate these stns and are driven by variable Level signals emaaating from a segaal analysis logic Cisatit. During operation, the logic cho~cs the least distortion process based on signal conditions. Similar switching and soutang rsn be aeeomplished with voltage controlled amplifiers, digital aetenuators, and faeld effect devices or other ~aomponents acxiag on the analog signal. Similar routing, mixing, or merging operations can be made from digual processors operating on numbers representing the signal. Such operations like those with the photoconducxive cells can be transient disturbance free by virtue of their slow switching action. Each record proaxs, and its complementary reproduce pros resembles a ~Iter-equalizer operation which is made optimtm~ for the pmgram regret. Both parts operate as a system so that the encoder can anticipate reproduce ewers and can create complementary corrections. In this manner, the record and reproduce circuits rice not working as an individual, :tend-alone theoretical ideal, rather as an optimal system. The breadboard dcxigm has two such systems, one for best articttlstion and transient response and the other for lowest distortion or most accurate instrument timbre.
Filters, equalizers and catrve fitting operations are accomplished as follows.
vOne can define a filter mathematically by coe~'tcient strings in polynomial sequences. In addiction, the same Ether can be defined by how it responds to a given waveform stimulus.
Essentially, extrve Patting in tinge, frequency domain, and amplitude is axeated from numbers which can be stored in tables. For analog systems, sash operations are performed by circuit elements connected to band sesirict, equalise, time correct, and to perform dispersion operations on a signal. These circuits can also be analyzed back to sunilar polynomial dents which can run as digital process progeams. As can be seen, very awkward circuit ~oonstruotion problrans~oocur when one mu.~t alter these numbers from time to time, as would happen with a dynamically shangang process. lviultiple component values of induceors, SUBSI'I'fU'fE SI'~~~d' ~,~, ,/zi~o ~ ~ ~- ~ -~ ~ ~ PGT/US9~/0~4629 resistors and capacitors, as well as gain stages, would be all simultaneously changiag to produce such a merge operation. This ran, however, be accomplished by digital programs.
Dynamically changing digital filters have become practical only recently as the necessary processing power has btcome ~c~nomically available. Of course, simple networks such voltage controlled parametsic equalizers and variable RC section tone controls have been available for a long time.
However, complex variable filters arc still rare. As one can sec, the dynamically changing curve fitting operations can be handled most directly by digital pro~sing. In an analog system complete filetrs must be changed, whereas the digital process merely changes filtering. Both have s'unalar potential cum fitting capability, however, they differ during the transition region from one process to another.
'' In theory, only one set of fetter coefficients is needed to make a near ideal analog to aiegital conversion and its reciprocal, provided all signal asiivity in the frequency domain is greatly removed from the Nyquist sample limit and there are ample numbers to characterize the signal.
Commercial digital standards do nest allow either of these conditions and, constqucntly, some practical stato-of the-art compromise of time, transient, alias, quantization, and flatness of response must be made. The best of each performance aspect cannot t~cttr simultaneously and one of ordinary skill muu choose a compromise based on knowledge and subjecxevc expcrieatoe with audio programs.
As previously indicated, the signal to noise ratio of digital protxsses can be excellent while camplcx distortions at average signal level can be higher than with good analog systems.
For high quality work, there is a need for improved sesolution, time and eransient accuracy as well as redu~;d high order distortion. This aspect of the encode-decode system, in accordance with the present invention, addresses such a need.
At this point, a further understanding of digital distortions will prove useful.
Typical digital systems have between 0.01 and 0.05°!o total harmonic distortion ('f'H~) at high signal levels and about 10~~c cumulative errors in the time-transient domain. lWost analog systems have opposite problems to these, as they often operate at above 1% THD but seldom have more than 0.1%
transient time error. Under low to average signal condetions, dig'rlal THD
increases while analog THD
decreases. As noted, digital distortions tetsd to bt uppea~ order and non-gtaranonie and, therefore, stand out due to their non-musical nature. Analog distortions occur less frequently and are less objectionable, even at higher levels, as they tend to blend in or musically merge with tire signal. Similar problems occur with transient limo domain type distortions. At first, it was thought such problems were inaudible, since simple square wave tests would show few sonic consequences from such distortions.
Today9 we can show serious deterioration of spatial sense, as well as lost inner detail perception as a result. As digital time domain distortion is much more complex than the simple stng'tng measured in early tests, resolution performance of industry standard 16 bit encoding is also inadequate. A system which can produce 10 volt peak to peak signal will !have approximately 150 u~l best possible resolution from one least significant number step to the neat. Practical systems have ~tgnal discontinuity of 4 to g times greater than this, as the state-of Phe-art has not yet allowt;d near theoretical perfot~tnanoe. A

W~'92~2~60 ~ ~ ~ ~ ~ ~ PCl'/CJ~92/04629 0 to SO ullolt disoondinuity limit is typically considered just audible.
1'saclical systems have distortions often ton timres lidglmr than this.
As noted earlier, the peak signal lianit-expand and low level average racpand-compress operations deal with resolution problems. Other distortions from time shift, alias, and S quantization, which are inherent with even ideal encode decode operating to industry standards, still remain.
Distortion seduction may be accomplished in the following manner. Most digital distortions can be paedicxcd, as they are strongly related to sagnal'sonditions which are easy to identify. It follows that, for a given sagnal, one ran choose a best snrode-decode proxss having the i0 Itast audible or sonira~ly damagutg distortion. if one must opcsate to industry standards where the Nyquist limit is just outside she audio range, then a trasuicnt response versus alias compromise exists.
This compromise occurs when requiring flat passband response and a very narrow pass to reject uansition bandwidth. As the signa9 changes, one can choose ahe best process.
In practect, phase and time response arc not equal from one complex 1S ~Iter-equaliser network to another and a slow fade or merge is neodul to prevent inevitable switdting transients wish process shifting. Sunihir problems are dealt evith for analog nonce rrdudion processors.
With digital processing, these operations of meaging from one optimum filter or curve fit to another ran be lookup table ooeffecients which are ace~scd as a sequrrt~ to merge from one filter type to another. While phase anomalies still occur, the deoodetl signal can, however, be free of beats or 2D cancellations which plagttc analog iader type systems. The mix or filter change merge occurs just fast enough to prevent audible transients or other paramctrically generated phase disturbance.
Since digital process timing is almost always crystal controlled" the record -play transitions can be made to track each other by pratimcd sequence programming which can be initiated by a single command. This eliminates the aced for continuous numeric control and higher 2.S bandwidth for the control signal al,ll process types, transition speuic, and intermediaee cocfCtcients can be stored and run as a program from a single, one time oommami and the recorder-reproducer are effectively locked to each other.
The basic analogsystem uses resistor-capacitor time constants within threshold sensing logic to simulate pre-determined transition speeds and the rcsuitant tracking of intermediate 30 filter mix states. In addition" other time constants also serve as internal memory to add hysteresis or hold back to decision making operations. These allow a first time quick process chaange decision and a redueacd sensitivity to further changes thereafter until the time constant assets. This pr~snts unnecessary process changes during grey area or uncertain signal conditions.
like a digital system, the analog system has the ability to operate with simple switch on switch off control where the output tracks ~5 the input and where the output or reproduce subsystem does not have to detect signal sxanditions to do so.
Plormal analog systems are not DC or direct coupled as is the case wish digital, and these would require an additional data channel with a linear control signal or an internal ~'~."'~r22(D60 ~ ~ ~ ~ ~ ~ ~ 1PC.°T/US92/~1629 analog signal conditions detector to make such a system operational. For practical operation, the basic analog system has been tested without buried or Hidden code control and a third control channel with appropriate time delays has peen included. .
B'here are numerous ways to sneak through and hide the control signal within digital systems. .~s previously pointed out, one can random nouc encode-decode a control signal in the least $ignifacane bit(s). 'l~his operation has no counterpart in the analog domain as it is nearly amposs~le to locate minuscule portion of a complex waveform carryang this information. In the dig'Ual realm, the least significant bit air always known. Hence, this bit can actually be borrowed for xoaatrol puss. Other ways to hide a control signal include using forbidden numbers or unused data wards ar number strings within a digital system which the same or a ddli'erent system considers an error or nonexistent data. itJhen the forbidden numbers are carefully chosen during encoding, the reproducer will z~gnize the error, but still decade its data signal ~tre~ty. ~f course, the forbidden number is data which can be ~traa~td and used for the eontaol function. lFather method of concealment has enough bandwidth within C;ompad Hisc standards to allow ample bandwidth ar information carrynng capacity to pass complex contxol signets. .flny signal degradation from performing this operation is very small when compared to the improvements resulting fmm the added process power and re-construction capability being controlled.
)fre5ent industry standards are hugely based on providing good perforrman~
in terms of flat response, low harmonic distortion, and high signal to noise ratio. 'f"mie, transient, alias, and resolution are compromised, bue their problems ar de6dency occur predictably with signal conditions. Consequently, a control logic roust analyse the incoming program material and determine a best process.
'The relatively simple basic analog system carcuit makes quite accurate decisions off quantization versus alias distortion based on high frequency intensity and its ratio to average program level. 'g'his fellows from the flat frequency r~ponsc-sharp cut off compromise of the low pass filter design. In practice, the filter xuust be SS to 90 d~8 down at the end of its 2 kHz trarasicion region. Just prier to this, it must be flat to 20 btFix. Unfortunate and serious transient ringing must occrtr, as can be demonstrated from the analysis of a square v;~v~ with its upper harmonics sharply removed. 'flte filter having good transient response will not rt;mm~e enough alias causing upper frequencies.
3.ive program spectral energy, in the transition region and above, is unpredictable and ranges from bursts caused by mic~ophonn clement peaks, instrument overtones, amplifier distortions, etc. ~3ence, a simple high frequency level detector can determine whether added filtering for alias reduction is needed ar not. Since these distortions can be oo4~ereel up by program ~S sonics, an added v~t~hting factor of reduced rejection duriu~g high program levels can be used.
'flaerufore, the detector ltso&s for ratios of high frequency HVyqufist energy to average program levels to determine when more aggressive filtering is needed.
SU~STiTl9TE ~'~EET

i'V~ 9~''~.2050 I'C'f1U592/~29 40 ' F-sscntially, a reduced second derivative ground the cut-off slope yields improved lime and transient performance. It is assumed that symmetrical t'rlterx taaving coe~stane group delay arc used, as they are practical analog and digjtal process types. These have symmuxeral pulse behavior and can be made to mix/mergc from one curve fit shape to another without altering group delay and creating eaccess phase interference during transit time. ~racticai systems can have as much as 2tD0 uSec. time shifts near ~ttt-otf when full 90 dB alias rejection occurs. These numbers relate to about 0.15 inch rapid displacemeru or doppler shift of high frequencies which can occur very rapidly with mtuic type wavreforms, t:ertain types of transient intermodulatiori distortion (TIM) ran occur under aimaDar conditions. lVhen correcxed to 3 uSec/kHz change of upper passband conditions, a filter may have kss than SO dB rejection. I-towcvcr, as can be seen, a best choice compromise switchable system is practical.
A xoond group of compromises t~elates to quantiaation distortions and the smallest signals which can be proxssed. Ids noted, level change operations reduce these problems in a compatible manner. Some further improvements can be made by antasipatoev forced resolution I5 cuhartcement. Like the alias/transient opcratioru, these are also curve fitting in nature and can be accomplished by record-play circuit systems resembling equalizing filterx. In this case,, a fortxd high frequency extension during record is made when signals have small amounts of high frequeacy iatosmation. When normal signal lemels with high frequency conceit arse present, the frequency response of the system is flat, but when the signat level is low sad them ~
lietae energy in the high part of the spectrum, the frequency response in the record half of the system is $toosted. The play circuit does the inverse operation. The overall least significant bit activity is substantially increased and more information becomes ended via duty cycle modulation and increased dithering.
When the record equalization (F..Q) contour rises very sharply, most of this added information is just at and slightly beyond audio range. It has little effect on standard players or on hearing, because hearing acuity is low for these small, low level signals. Frssentially, one has traded non-harmonic distortion for a similar amount of harmonically rtdated program distoetion. To a degro;:e, the less accurate the player, the better this protxss works to disguise grainy noise as upper music: harmonics.
CDI course, a decoder can be instructed to perform fiat response reoonstrttction and there would then be more data bits making the complete signal. Hence, quantization noisy. is ~0 reduced. This is another curve fitting operation which might be called dynamic dither, as it must be removed in the presence of strong signals. If left continuously on, alias or beat frequencies will oecur from strong signal harmonics interacting with such ess snergy high frequency dither. CBearly, the process must shut-off under intense high frequency ~nditions where resolution benefits become minor.
Control signals for resolution enhanewnent and distortion reduction processes 33 can be derived by Docking for critical energy in the alias frequency range, 1"%igh ratios of these high frequencies to average signal conditions ave indicative; of possible foldover distortions made audible= as they are not masked by program maeerial.
~v~Sl-aYl..l'YE 5~~~~i' 1'S'(;~ 9~/.~2iDG0 ~ ~ ~ ~ ~ ~ ~ PCIf'/iJ~q2/dl~lb29 Ivtost significant are complex high frequencies by themselves such as those encountered anth cymbals since low frequency problems are completely unmasked and are audible 50 to 80 dB below where hearing aaaity is strong. Such signals and how fast they .change can be sensed to determine a best process. wick high ~equency bursts above average level conditions suggest halt aaltcring and best trammsicni response, provided some midband energy in the predicted alias range is present. hw levels of high ffrequency energy suggest quantiaafion or dyaamic dither correction.
Sance some proc~css/hlter/ egualixcr sient change operations Can be more audible than othces, some maximum number of changes become part of the deasaon making process.
' dithering and resolution cnhaacement (l~rQj arc simple high frequ~ay operations which can be turned off and on rapidly without sonic consequences from sudden phase shifts, beats, etc. Transient alias switching is much slower as time-phase changes do oxur. fuse of ihes~
p~sble process to process time change constraints, h is xee~sary to look athead to observe the before and after signal conditions stu-risunding the dec~ssion point. In addition, the ocxttrr~n~
frequency of these Changes, past and present, is important to ps~at process hunting or dr~sion in52abiliiy resulciag in unnecessary process changes.
f~r~ctaits to pdrform derision ma~ng are deceptively simple ~mparad to what one might eapecR front the afoa~edescrib~d funtxianal descriptions and the same holds true when the earcuit e~nivalents operate from dagctal systems programming. The basic analog system uses analog circuit subsystems to perfo~ran these operations. 3 here inc3ude, dolays, voltage somparat~rs, speraral analysers, multipliers and st~tal detectors with time constant memory. Nigh frequencies are detected and routed to three voltage eomparatoss. ~ne is set to detect minimum ~
thereby switching on dynamic dither. The second is set to masdmum allowable Hl: to switch on the large anti-alias filter.
The third has a variable threshold dependent on progam level. IEanch comparator has its own time constant or hold-ofd so that, onct brad, or its on-off state is cleansed, then a certain time must lapse before the circuit will .ar~nd again. In pratxice., these time Constants art:
performed by diodes sharg<ng resistor capacitor networks and, as configured, the rha;uge to dischaage time can be easily made itnsymmctrical. R'his behavior allows ~uask doeisioms of a °on~;
ghot° nature without having the quit jump from state to state from near threshold conditions. 3.amp/L'~ sources lllumiaatc sigpal steering photocells to give quick fades ~ one pra~ss to another. As with the comparators, the lamp drivers far each process type have di~'eaeut on-o~'timc constants to acxo~nmodate tune-phase differences from one process to another. In addition, several time delays are used to allow logic action to happen prior to the signal conditions requiring the Change.
Unlike analog noise processors which require record-play tracking and very carefully wtarked out signal thrtsholds, the basic analog system process decisions Can are very rough.
.Accuracy ~ aennecxssaay as the s~epr~itt~ar process is always tracking. Since the operations are industry standard compata'ble, no major disaster w911 occur from a wrong decision.
Uenc~ the analog circuits in the basic analog system have worked °'as is" without rehnemcnt.
ilJ~.cii"I'I"~.91'1~: ~H~~'1'°

A~'O 92!22060 ~ ~ ~ ~ ~ ~ ~ PCl'1US92129 It is clear that when a digital signal and process is used, the encodiag must have greater accuracy and resolution than the final industry standard product.
One method to assure this is to encode with added bits at a high sampling rate and then perform successive decimation and arithmetic roundoff or truncation to make the final format (~4.1 kl3z, 16 bit) Processing becomes multiple stages of delays, filtering,, equavang, instanaaneous gala changing and averaged gain changing.
The signal is analyzed and the results are used to intccaogate a process "rule book". Several procxsses and their reproduce conjugates are available to be chosen based on predicted error and best signal reprodttctiwn. Once determined and initiated, transition parameters are ao~sed and the process starts chan~tng. I~uaing this decision making time, the music has been delayed to allow the proctss transition to complete prior to the signal being matched. A control word is generated and encoded for inclusion in the recording, so chat the reproducer can acoas from Its memory the complcm~tary process and its synchronous transition parameters. D3oth operations commence relative to the'u timed sequences and to their stored data. Since the reorder has already simulated the pre-programmed reprodueer atxion or the rnrrcctions for consequences of its action, both processes change synchronously within the tiaae accurary limits to the ~nooder-decoder crystals or a:locks. ~'he system then changes itself without having major Hransition abcrxations and then operates with the best performance for the signal conditions.
Referring now more partaaalarly to Figs. 9 and 10 of the drawln$s, there is shown a systeEn which can choose an optimum recording pmecss and its reproduce conjugate to achieve low alias or quickest changing, fastest settling transient response. Phase interferences during transition time are r.,outrolled by "Eider" time instants and signal delays. The logic circuit has one oomparator set up to change state when alias distortion would be ~reater than an approximate 40 dB below peak program level.
This a:ircuat, shown in Appendix C included with the U.S. Priority Document, contains four sets of back-to-back LSD-Photoresutive axll switrhcrs and driver circuits. Signals are delayed to allow process decision times and transition time ~~~taants aro adjustable to allow smooth fading bettveen two process signal path.. Controls are derived from an analysis &lter and detector made sensitive to highest alias frequencies. A peak lever detector sets a voltage threshold from which a comparator can be referenced. Alias companenis above this level setting wilt switch processes. A
second program delay is used to synchronic reoosd-play tracking and effectively matches variable time constants of the LIrD driveax. Diffferent control settings allow this circuit to operate either alias/transient as a dynamic operation or quatitization/distoxiion as an independent operation.
These circuits are set up eo be compatible to industry Compact Disc standards.
The encoded product having these variable-dyaiamic prop will play back with equal to or better 35. sonics than without procesuug, wen on a standard home player without the decoding featut5es of the invention. Circuit subsystem blocks correspond to those used in previously discussed embodiments of the invention.
~UBSTITt9TE ~~HEET

~'~j'-' ~~2460 i'C.'T/iJS92/tb4529 4~
As observed in 1~ig. ll, "process A", there are waveforms and distortion plots starling with input signal, output of uncompensated encoding, the conjugate restorative response, and the overall correaxed system response. ~d'est signals enc9udc slow sweep forward from 20 Hz to 30 kHz, 3kHz square wave, and distortion naeasur-Jd from 20 Hz to 3t) kHz frequency sweep at near operating level.
Fig.12 illustrates "process f3" and uses the same format as Fg.12 for °proaess A", except the plots arc for the fast transient process.
~ne anetbod of over~ming the frequency resfDonse limitation imposed on digital recording systcens by industry standards and its effect on transient response is the use of slew rate compression,~as discussed earlier. Slew rate limieing and expansion operate in a similar manner to the peak amplitude methods previously described. As before, a nonlinear clement is introduced in the signal path to perform the desired limiting, and the expansion or rruciion method involves placing the same devaa or circuit in the fecdbac9c loop of an operational amplifier. Variable eonduaion of diodes with eaecreased voltage is for peak amplitude limiting whereas increased current through 1<S a capacitor with incar.ased sigaad speed is used for slew rate limiting and expansion. Slew rate limiting takes an event and spreads il in time and, hens, its use must tse limited 80 occasional events like those occurring in musical progams.
~'he basic system using an analog implementation of slew rate impression is shown in Fig. .13. ~'ypical wavtforms associated with its operation are illuserated in F'tg. l4a~c.
Schematic diagrams of the key modules are included in Figs. 15a, 15b, 16a, 16b, and 17 and are discussed below.
Wigs. x5a and b show an example real time slew limieer having a circuit configuration somewhat analogous to the diode limiter types previously described. Here a representative wideband square wave signal with transitions of many volts per microsecond is shown applied to an amplifu, marked Af, which is constructed to have a restricted slew rate performance of much lass than one mitxovolt per mnicrosecond. its square wane output now has a well defined rise and fall character. When the input and output of this amplifier art compared and the gain structure is appropriately set to cancel slowly changing signals, the slew rate limited part of the signal becomes available. A bridgo-like circuit of the amplifier A$ and the resistors R~
through R6 perform this cask ~0 ated its output is the distorted portion of the signal occurring during slew limiting. ~Vlten 8his correction signal is appropriately amplified and added to the slew limited signet, the original input square wave is restored.
A vary high performance slew limiting amplifier is needed for this task and a specialized circuit configuration must be carefully worked out to prevent sub-harmonic, recovery, and overload distortions. In addition, the degree of slew rate limiting with respect to signal speed must be prulictable so that acceptable reproduction can be rrdonstructed when the correction signal is not present, as might be the case in a simple reproducer. A standard operational ttrraplificr will not work adequately for this task. pigs. 16a and 16b show a sttrtplified conceptual variable slew rate amplifier SUE~STITUTE SHEET

1W4~ X2122060 ~ ~ ~ Q .~ ~ ~ IPCd'/U~9~/~4629 a~
where all parts operate in linear class A so that conduction occurs sander ail sagmai and limiting conditions. Voltage controlled variable current sources marked I+ and I- are used to achieve slew limiting. q'wo of these circuit groups marked A and B oppose sacks other and the balance betavcen them is modulated by the input signal through FE'T devicxs marked C and D. tit limits rather side of balaatce are restricted by diodes E and F which are in turn controlled from a balan~;d phase inverter FIrT' marked G. Slow changing signals create small currents through the capadtor marked Hl and have inconsequential effesx. Large fast changing signals demand more rurrcnt and the limiter r~stri~ions then restrict slew rate in an ever increasing manner following the dio~o saonducRion versus voltage curves. 1-lena a tow distortion predicable a$d controllable symmetrical slew D'tmit occurs. A more 1~ detailed schematic is shown in Fig.. D3 of the U.S. Priority Dotxtment.
In pracxice, it would be desirable to be able to reproduce an occasional fast signal such as pcacusuve transients. 'lfhtsc may have large fast transition waveshapes which are quicker than filter and sampling limitations allow and the above circuets arranged like the diode mcpander will perform this operation without requiring an external correction ssgnaL Fig.
l'7 shows this arrangement.
1S The variable slew ampliiaer is now made siow~er than the anticipated input sigaal from the repr~u~r so that the diffuence between the recorded stgJtal limitations and the reproducer amplifier performanea now becomes a synthesized correction signal. As before, 121 -R6 and amplil"ters A1 and AZ are like a bridge which ranecls :.~nlimitcd srgnals and presents the slew rate diif'ereatoe betw~n input and output.
Prev'rously, the limitul and ~rrccted signals were added to restore the input.
Now alt averrorrection 20 is made to antiapate the signal that would have been at the encoder input prior to band limit filtering.
This operation then uses an overeorrection signal which will vary from one signal condition to another, hens a controllexl variable gain device, VCA, replaces the tuted'R~ of the previous circuit. ~Vhcn the control signal has bean properly set up for this event, an error correction signal can be added to the input signal to yield a much faster transient reproduction which now more closely resembles waveshapes 25 of wider bandwidth input signals. As can be seen, a transition shift approximating slew restoration occurs and if time integrity is needed, the signs! must be advancxc9 back in time by a variable delay so that during this reconstrucxion, the edge transition occurs where it would have in the original program material.
Both slew rate and correction signnl gains are controlled. 'These arc analogous 30 to curve segment shapes which might be stored in and recalled from lookup tables and size scalings which can be determined from examination of the signal. Gtpacitors and diodes from analog circuits create predictable slew dependent curve shapes and voltage controlled amplifiers respond to size information. hate of change of numbers compared to some averaged number scale and multiplier coeffesict»ts in memory simulating curvature or a non-linear fun~ion ran do the same operations in the 35 di~ttal domain. Either operation depends on having first tested the serord said predicted reproduce synthesis during encoding and then generaeing a control signal which sets up the reproducer to track the best tested results. To do this with the analog circuit, the input signal is lowpass filtered, possibly slew limited, and then compared with a yet lower slew rate limit circuit to get a correction signal. 'FICA
__ _ .. , ..-.--a.

w~ ~zrzzo6o ~ ~. ~ ~ ~. ~ ~ pcrit~s9zvoa62~
~s gain is then adjusted to get a best match brxween band restricted and unrestricxed signals. Slew limit, VC.A srttings" and engage time become the control information Since the reproducer has the same circuits as those used for enoodirlg, the output waveform will track. Clearly many other analog and digital methods to determine and synthesize slew limit and expand can be used.
However, a unique aspect is that these operations can synthesize many high definition points of a wavrshape portion from pre-ceded curve shaape;s which are access,~d from a low information content control signal or can perform a fast order appv~nmation of the waves6ape in the abseaoe of control.
A presently preferred embodiment, in accordance with the invention, will operate primarily in the digital domain and has the same entry basic overall system design, as shown in ?<0 Figs. 3 and 4 of the drawings. As in the case of the analog system (e.g.
Fg. 8), each subsystem is a subtantialfy self-contained circuit or functional module which performs a unique operation. Input and output signals of these modules are quite often simelar i'rom one design or product to another. Hence, if the component or subsystem added to improve performance does not signiG~ntly change these intermediate signals or the circuit configuration, then compatibility to standard equipment and Li recordings fis mush more likely. In our case the ~~SI'" or digital stgna) processing subsystems in 10(1-102 and 104/IQS art the unique elements while the remaining ~mponents of the system have few changes and are left as they nomnally appear in products.
As best observ~ul in Fig. 18, a more detailed diagram of the presently preferred digital embodiment of the eaacode system, highly specialized operations are performed by 20 fun~ional groups of electronic components. Quite often, each performs a unique task which can be examined, ew~luated, and descxibcd independently without involving other portions of the digital system.
Htnce, each clement is a functional building block much like a sound system component which can be specified and compared to others.
'fhc analog enput signal is applied to a balanced input amplifier 201 followed 23 by a supersonic low-pass analog filter subsystem 202, which first isolattes both signal and processor grounds and then removes frequcnaes above the D~lyquist lirr~it. In this manner, crosstalk bcteveen digital send analog signals is reduced. ~etdio sigtaals must be isolated from digital circuits in order to prevent interaction and crosstalk noises. If not done effectively, those problems propagate throughout the audio component chain as well as within the encoder electronicx. 'fhe supersonic filter is needed 3(l to eliminate high ft~equency components in the incoming signal, Including radio frequency leakage and other noises, which would otherwise create alias and foldover distortions or beats when the signal is sampled. The output of the filter is applied to an oversampling A to b converter. In the embodiment shown in Fig. 18, the signal is sampled at 4 times the final 44.9 kilof-leria frequency. In another embodiment, we used a converter running at 8 times 44.1 kl3z ~1s pari off the transient analysis 35 describeal below, we art interested in frequencies up to at least 40 kilol3ettz, and therefore, the filter response begins to role off above this region. In both eases, the cutoff frequency of the filter is well above thtr normal audio range;, so that the filter can have a gradual roll off and not introducx audible phase distortions. The "brick wall" alias filter for the encoded signal is implemented as a digital filter ~IJB~ 8'FU'~°~ ~hi~~T°

w0 ~zrzzoso ~ ~ -~ ~ ~. 8 ~ ~c°rru~9zro~~z~

in the decimation proorss dcscritxd below. It is essential, however, that the response of the analog Otter ~ down below the resolution limits for frequencies which would alias into the audible range (i.e.
input above 232 kHz for 4-tirues oversampiing), since these alias products cannot be distinguished from the program material or fettered out later.
State of the art filter designs attempt to keep al'~as and foldover no'ues well below the resolution limits of the d'sgital code. The well known Compact Disc encoding yields 16 bits of data sampled at 44.1 kHz Input frequencies above 2205 kI la txteed the half sampling rate Idyquist limit and simply wiU not play back. Instead, one gets lower frequency difference components which, to bt inaudible, should be at past 85 d0 down for a standard CD. Iiowcvtr, the inrtntion needs a digital signal wish higteer resolution, which means proportionately more suiatgent filter characteristics.
Since we are dialing with a signal with appro~naattty 20 bit resolution, as described below, wt need to keep input signals which would reuse alias producxs down by at least 108 d~. dust of their similar ion-musical cW ractee, crosstalk intexftrtn~ t~xween analog and digital processes must be at least as low.
While analog filtering and isolating operations arc functionally separate operations, the rtquires9 circuits are related and often work best when constaru~td together from one group of components. Good designs may have fully balanced push-pull signal paths, as vvcll as separate power, grounding, and shielding.
The output of the supersonic analog filter ~2 is applied to the sample and z0 hold and analog to digital conversion subsystem 203 through a summing junction in which dither is added. The continuous analog ~yals are sampled at regales ineervals and the satnplc voltage held unchanging long enough to be converted to a binary number or word which represents the amplitude of the sample. As has been discussed previously, faster sampling ratty give more points to dtiinc the signal wavcshapc and longer digital codes or more bits give Cner re.colution for each sample. Accurate conversion is very, diC(icult and many clever etchniquts to achieve it are represented in commerdal products. Wt art currently using a commercial hybrid integrated sample and hold and A to D
converttr which sea operate at a 176.4 kiloHcrtz or d times oversample raft and produce 18 bit d'~tal words representing the sample amplitudes. This unit is at eht limits of the current stage of the art en commtreially available converters. Prior to the availability of these converters, wt used another commercial converter wvith 16 bit accuracy at an 8 times ovtrsampling rats.
In order to get resolution in an A to D system which matches the capability of modern eonvtrtcrs, great care must be exercised to minemize tao'tse and analog-digital interaction.
One of the techniques which wt use is called silent conversion. In order to prevent digital interfercnces to analog signals and conversion timing, the entire logic and conversion system except the sampling ~5 stock shuts down prior to the critical sampling operation. Noise from cables, ICs and other parts btsomes ten to one hundred times less and a signal sample to accurate to millionth°s of a volt occurs.
Once the analog signal is sampled and softly held, the conversion process resumes and the dip~tal rode Sl9B~-f9Tll'T~ S1HE~'T

~:~.1 ~~_~~
a~~ ~~ixi~so ~t°r>us~zit>a~29 is sent to the digital signal processors. Otbrs systems do not work Pike this and are severely hiadered by noose, crosstalk, or glitches.
Another aspect of A to D aced D to ~. conversion whieh is very important, as dis~tsstd earlier, is the minimization of sampling time fitter. F'mding$
re.~ntly reported in the audio trade press indicate that a fitter of 7L00 picoseconds in the sampling time is Clearly audible. In order to keep ibis fitter to a minimum, we place the .system clock 209 in the A to D
a~nvert~r $iodule. We use a clock circuie designed to have very low phase noise, and use a short path to Ihc ~averter. The clock is also buffered and used to p~vide the auaster taming to the rest of the system.
~veraampling, in addition to the advantages regarding analog filter design previously discusxd, allows a given converter to achieve a h°~gher amplitude resolution, or more bits to represent signal levels, when dccamateal. ~aCh additional bit doubles the enc~dad resolution to yield an almost d d13 greater dynamic range. In 4 times ovCrrsam~pl'mg, for example, 4 samples are taken for each one present in the final format, and those extra samples contain more information about the oaiginal signal. Some decimating converters simply discard ehis additional information, but we convert it into amplitude resolution by using subtractive dither. ~ne of the $uns~ions of the first DSl' subsysctm 205 is to generate a dither slgna~ which cam take one of several forms, including a saaROOth, a sine wart, and a pseudo-random noise. A process within the DSl' generates small seemingly random numbers, which are scaled to fractional bat levels. Thcsc numbers are appticd to a digital to analog oonvertu 307 whose output is smoothed and scaled or attenuated in 208 to achieve fra~ional bit levels when added to the incoming analog signal. The voltage is addul to the audio signal thereby creating a vtrnicr effect. Within the DSP system 2D5, the dither atumbers are delayed to match the system delay for samples Coming from the A to D ~nvt.rter 203 and the dither is subtracted out again. When the signal is averaged by the low pass filter process in 205 as part of decimation, the smallest signal components can bt detcrmintd to fractions of a least significant bit of the converter. Thcst operations must occur at the incoming sampling frequency and, in the present scheme of 8 times ovtrsampling, up to an extra two bits of resolution is possible.
The digital output of the analog to digital converter is direcacal to the signal analysis subsystem 211 and through delay subsyseem zU4 to iht first digital signal procGSSing subsystem 205. The delay provided by 204 allows analysis of the signal to be trade and a process control decision 3~0 taken before iht signal reaches the DSl' system. In this way, the DSl' is never "surprised" oe caught off guard by changing signal conditions.
In a presently preferred embodiment, the digital signal pnnoessing subsystem 205 is implemented using two Commercial DSP processors with 24 bit avord length and 56 bit accumulators, It performs a variety of functions, including: generation, delay and subtraction of the dither signal, described about; low pass ;filtering the signal using a variety ~f filters9 decimation of the signal to the industry standard sampling rate; and handling the transitions from one filter to another under the mmmand of the process control subsystem X11. First, a delayed Copy of the dither which was added to the analog signal prior to conversion is subtracted from the incoming digital signal. The war ~~razoso ~ ~ ~ ~ ~ g ~ ~~riu~~zm6z9 srgnal then undergoes decimation, which involves low pass Filtering followed by repeatedly discarding three samples and keeping the fourth. It is this dig~ially implemented low pass falter which performs the .anti-alias function for the signal at its final sampling frequency, and, as discussed previously, no single filter implementation can be ideal tender all program monditions because of the steep transition between the passband and stopband sequarvd. m~dhilc a symmetrical finite impalse response digital filter is free of the variable group delay and phase distortion effects which plague analog filters, there are still tradeoffs between alias reje~lon, transient response, and passband frequency response. The invention solves this problem by using different ~'xlter characteeistics for different signal conditions, and making a smooth transition or merge from one filter to another. The implementation of the falters i.4 a standard one for FI126Cr~ters using multiply and a~cumulate functions. The result of decimation is a signal having appro~amately 20 bit resolution at one-times sampling rate. This 20 bit accuracy necessitates s falter siopband rejection of at least 1(~ dI3 to keep alias products below the resolution of ihc signal.
The output of DSP subsystem 205 is a digital signal at ehe industry standard sampling rate (44.1 kliz, for aCD's~ having 20 bits of information. This signal is passed to the second digital signal processing subsystem 210, which packs the 20 bit rtrsolution into 16 bit words snatching the industry standard and adds control information for tale by the reproducer.
These operations are carried out under the oomrnand of the prorrcss control subsystem 211. The information pacing is acoomplishad using a dia"r'tal implementation of the analog syseem described earlier. For lave! peaks in the program, an instantaneous soft limit transfer function is used. Sincx it is implemented in an axed mathematical way, the transfer function can be chosen to have minimum audible effect for undetected playback and can be exactly reconstructed in the roproduce decoder. It is also possible for the process control subsystem 211 to alter the limit parametees, such as changing the limit threshold in response to the degree of limiting which may already have been applied to the signal before it reached the encoder. In doing so, the controller can also send the parameter information to the reproducer using the control codes hidden in the signal.
For very low level signals, an average gain compression is used to increase the system gain. 'This gain increase ra'rst's the level of those small signals further into the upper 16 bits of the digital word, and then the ~0 bit word is rounded off io 16 bits, matching the industry standard format. The gain is controlled by subsystem 211, which looks ahead in time from the point of view of the DSl' system 210. 211 sets an undclayed srgnal, while the DSl' system gets one delayed by 204 and ?O5. The control subsystem also inserts control ides which tell the reproducer what it has done with the gain. The second DSl' subsystem is also used to apply "dynamic dither" or noise draping as discussed above in the analog description.
The Penal task of the DSl' system 210 is to encrypt and insert the control codes into the least significant bit of the digital words. The details of this prootss arc discussed later. These are the Code.S WhlZl1 tell the decoder what has been done to the signal, so that it can carry out complementary processes.
SUBSTI"flJ'1E SWEET

;~~ ~Z,Z~~a . 2 ~. ~. o ~ ~ z pC'1'/US92/(141629 4g Both DSP subsystems receive commands froua the signal analysis and process control subsystem 2711. This module receives the oversampled digital signal directly from the A to D
converter, conditions it, analyzes it, and based on the analysis, makes process control decisions and sends commands to the I~SP modules. It also generates the conerol codes for the reproducer which are included in the encoder output. The module uses digital versions of ehe analog algorithms discussed curlier.
It uses ratios of high frequency intent to total amplitude along with dceecced isolated transecnts to select filter programs for the decimaeion filter.
It measures the average signal level of the broad middle frequency spectrum antl uses a0 the resulas to control the gain of the low Icvel ~mpressor. It also generates reproducer control codes eo Correrxly ~compleratent the encode gain structure.
It rxacasures thz average level of low level high frequency signals, and invokes dynamic dither insertion of extra high fscquencies when approprlatc.
It analyzes the distribution of peak amplitudes to determine if the an~miarg signal leas been limited prior to the encoder. If so, it can raise the threshold of the encoder's soft limit function, or eum it off altogether.
It can Compare the decimated signal to the oversampled one delayed to match the decimation to look for isolated bursts of high frequency information which represent trancicnts which would not fit within the normal x2 kilol-lerta bandwideh.
These 2(1 difference signals can be sent to the reproducer ice the Control Channel, spread in time, so that the reproducer can correct the transient on playback.
It Can also use the transient analysis to atntrol slew rate limiting of the train signal as an alternate approach to increasing the apparent bandwidth of the system, as previously diseusscd.
a5 It controls the insertion of hidden Codes is the (Cast significant bit of the encoded signal, putting them in when needed and letting the 1SB be used for the main signal ' when ie is not needed for Control.
°fhe process conerol subsystem is the nerve veneer of the encoder, making the decisions and controlling the functions of the ~SP units. It is not necessary for a given implementation 30 to incorporate all of the features above. For economic reasons, it may be desu~able to only include a S~3BSTITt3TE S~i~~'T' WO X2122050 ~ ~ ~ ~ ~ ~ 1PCT/IJ~92/0~5~9 SO
particluar subset. Since the encoder arses control codes to tell the reproducer what it is doing, a more capable reproducer will not be confused, and a less capable one will ignore those functions that it cannot compltmcnt.
The digieal dais output from the second DSP module 210 goes to the format converter and then to the recorder. Compact Dasc, digital audio tape, etc.
operate on similar encoding principles. I-iowever, these systems have different recorded formats and electronic signals for the same bits of encoded program data. In the andustey standard format ~av~crsion subsystem 212, spccaafrced IC chips are configured to add progress track informatics and other housekoeping information to the data and combine the tvro channels of 36 bit progam digital data into a single data stream which has 10 been ~nfigurcd to andus'try seandard formats. The end result is a combined data and operational code made compatible with the input of a standard digital recorder. This module performs functions common to ail digital recording systems, and uses. commercially available special function integrated circuits to pcr6orm the format conversions.
As bast observed in Fg. 19, a more detailed diagram of the presently 15 preferred digital embodiment of tho decode ~syseem, highly specialized operations are performed by functional groups of elcaronic components. In the playback subsystem, the fleet element of the reproduce chaip could be a video player, GD player, receiver or other equipment. These components usually have servos, monversions from specialized standards, buffer memories, and oocasionaily phase or frequency locked timing systems to achieve stable continuous playback signals. for example, such a system could be a CD transport. Each type of digital system requires its unique unscrambling, parching, and fixing operations to eventually extras "error free" program digita9 data and this is accomplished by standard circuitry within the player or other device. The output of the player is a stream of digital data in one of several industry standard formats, and this stream of data forms the input to our decode system.
Iteftrring to lFig. 19, the data from the reproducer is applied to a format converter 220, in which one of the industry standard serial digital data formats is converted ineo a form suitable for use internally within the decoder. The data is normally split into right and left channels at this point for separate processing. This format conversion is carried cut using commercially available integrated circuits designed for this function. This subsystem also may provide servo feedback control to the transport to control the incoming data rate, and it provides timing information 4o the decoder system clock.
The data output of the format converter goes to the control decode module 221. This subsystem is complementary to the process control subsystem X11 in the encoder. Its functions include ~deteeting and decoding the hidden control codes inserted by the encoder, possible code stripping or removal of the code from thG signal, signal analysis of the data signal, and generation of process control signals to control the DS1P modules based on the nature of the signal and the hidden rides.
SUIBSTlTUTIE S~WET

2~~~t~.8~
w~-~-~i2ioso I'f_'T/L1S92/o4629 The data signal then goes to delay module 222, which gives the control decode module 221 time to figure cue what to do with the signal before it gets to the first DSP subsystem 223.
The first DSP module 223 is the complement to module 210 in the encoder. It does a peak expansion which restores the peaks limited in 210. It dots a low level gain expansion, restoring the low level dynamics compressed in 210. It can complement the low level forcing of high frequencies in the dynamic dither operation, restoring a flat frequency response and lowering quanttaation noise. It performs some housekeeping functions, and its signal output has 18 to 20 bits of real information at the one-times sampling talc (44.1 ki~~z for ~IC1~.
This more socrrrate digital signal at the media samprng rate is routed to the second digital signer! proc~or subsystem 224, which is complementary to encode module 205. In this subsystetrr, the signal is interpolated to a higher sampling rate using a variety of smoothing falters which arc chosen to complement the decimation filters in 205.
All d3 to A conversion systems involve a smoothing operation to convert the disaxeet sampled signal back to a continuous analog waveform. Digital interpolation is frequently used to increase the sampling rate by calculating a larger number of seeps representing the continuous wavefomr. A larger srumber of smaller amplitude steps reducers the burdea placed on the analog smoothing filter 227 following the conversnon back to analog form. Ivlost player circuits employ some version of this technique. Again an "oversampTmg' has o~ttrrcd. I~Iowever, in a normal player, the information ooatent between input and output from interpolation has nee changtd since the filter cannot create new information from its curve fttrarg computations. By contrast, our interpolation subsystem has knowledge of the signal resulting from an analysis made in the encoder prior to the bandwidth limiting decimation. This information has been sent to it through the control channel in the form of filter selection oontrnl, and transient correction or enhancement data, and thus this interpolator can restore some of the information deleted by the deamator.
The decode system can also provide some improvement to playback of standard unenooded signals by analyzing the incoming signal charaueristics in module 221 and using the results to pick a smoothing ~lt~er which is probably best. This single ended operation results in an improvement over a conventional player, but it cannot achievt the perfomnanco of ehe full system.
The smoothing or interpolation filters in DSP subsystem 224 arc standard finite impulse response or FIR typos, which arc made symmetrical to avoid phase distortions. The subsystem mast implement smooth tratr5itions or merge operations from one tnterpolatron filter to another in the same manner as the decimator does. It may also include transient synthesis and slew rate modification, similar to the analog implementation discussed earlier.
In summary, the first decoder DSP module 223 restores amplitude resolution, and the second DSP module 221 restores frequency or transient resolution. Both of these operations are the complements of operations in the encoder.
The high resolution oversarnpled signal goes to the digital to analog converter subsystem 225. As in the encoder, we use a commercially available D to A
convtrier module which AV~ ~Z12Z~0 ~ ~ ~ ~ ~ ~ PCTfUS~32/~52q represents the current state of the.art. °Ilrc current embodiment uses 20 bit converters operating at 4 eimes oversampling. We have also used 18 bit converters ae 8 times oversampling. As in the encoder, great care must be taken t~ isolate the anaEog signal from the digital noise, and sample timing fitter is minimized by using a low noise master clock tightly coupled to the converter module. The aaalog output goes to the supersonic filter.
In the analog smoothing filter subsystem 227 and output buffer amplifier subsystem 228, final rounding and removal of supersonic signals occurs with an analog low pass filter along with ampiit'rcation to standard line levels and output 'smpedance. A
sophisiieated design such as ours treats flee analog and digrtal Crltcring as a whole system to achieve the benefits both methods offer.
fps in the encoder, tht: isohrtion of digital and analog processes is achieved through fully balanced digital and analog systems, floating power supplies, and isolated grounding schemes which prevent ineeraction with cables and other external components. The result is a line lzvtl analog output signal. This completes the description of the signet path from the analog input of the encoder to the analog output of the decoder.
1S The following is a description of the control channel in the presenQy preferred embodiment which allows tts to send control commands and su~dlaary signal informateon from the cn~der to the decoder in the same signal as the main program.
The coanmand codes and other auxiliary data are encrypted with a pseudo-random noix and instated into the last sigaiGcant bit of tire main signal digital words in a 24 serial fashion, one bit per word. The 1.,SB of the audeo is replaced by a "random" noise for the duration of the control insertion. (Of course, more than one bit could be 'borrowed"
for this purpose, but more of the main program would be lost.) The system is set up so that when ihc ~ntrol channel is not needed, the 1,SB carries the normal audio signal. Sine the digital to analog converters in most of the current generation of digital audio producxs arc not accurate to 16 bits, the loss of the 16th bit will not 25 be audible during undecoded playback, as long as the information inserted there has noise-like properties. Even in high quality systems tvhich do rrsolve all 16 bits, the inserHion is not normally audible because the 1SB of most programs already has arery noise-like properties. The low level gain compression and dynarnic dither dcssribed previously raise the level of the progam during very quiet periods and help hide the code insertions clueing those program conditions under which they might be 3(l noticeable. In typical classical music programming, the control signal would be inserted for intervals of about a millisecond each occurring several trines per second at most. The loss of full program resolution for these brief intervals is not noticeable.
Circuits to create random noise, modulate a control signal, insert it in the LSB
of the data stream and then aetrlcve and decode it have been assembled and made io initiaee alter 35 selection from a !ridden ~ntrol signal. These circuits are included in Wigs. ?.if, a Pseudo Random a.Jnscrambier/Dccoder, and 21, a Pseudo Random Scrambler-Encoder.
The process control signal is hidden in the least signiCacane bit of the digital audio channel by modulating it with a n~rse signal. Our circuit consists of a pseudo-random noise SllBSTITd,9TE ~i~-$E~' ~~.1~1~2 ~uvQ 92maoso PCT/iJS921f14629 generator based on a shift register with feedback which implements a maximal length sequence. This type of generator peoduces a deterministic sequence of bits which sounds very random, and yet is a reprodud-ble sequence. The oueput of the noise generator is added to the control signal module-two (exclusive-or'ed), modulating the signal with noise, or scrambling it. ~'he result fis then inserted into the least significant bit of the record serial data stream. On the play side, the (east significant bit is extracted from the ser'sal digital stream and the output of a matching shift register is subtracted from it, module-two gexrlusive-or again). The result is tire process coaitrol signal, unscrambled again.
There are two basic variations of this scheme. The farst version uses two noise generators, one on the record side iirad one on the play side. The record generator noise is added to the signal, and the play generator noise is suburacted. If the two generators produce the same bit sequence, the original signal is recovered. The problem is that the play generator must be synchronized with the rao~.se sequence added during record. While there are many well-known approaches to solvang this problem which arc covered in the literature on spread specxrum communications, it is still a non-trivial problem. Although this approach is feasible, a presently preferred embodiment of the system employs the following technology.
In the preferred embodiment, ehe stun of the process control signal and the generator output is fed back to the generator input. This effecxivcly "folds"
the signal into the generator sequence so that the scrambled signal depends only on the reocnt history of the bits, and the play side contains a matching shift register with no feedback. Because the play side only uses "feed-forward"
addition of the bits in the shift register, it is guaranteed to become synchronized as soon as N+ 1 bits have arrived, where N is the length of the shift register. The disadvantage of this approach is that it is possible for the noise generator to become stuck temporarily, depending on the characteristics of the process control signal. The probability of this happening can be made arbitrarily small by increasing the length of the shift registers. In the implementateon shown in Fags. 20 and 21, which sues a 17 bit shift register, the probability is on the order of 1 in 100,000 that a bit sequence might occur which could cattle the generator to stick. I3y going to a 31 bit shift register, the probability drops to about 1 in 2 billion, whieh corresponds to once every lL2.b hours for a CD. If the process control signal is changing rapidly, the anifart of a "stuck" noise generator will be of short enough duration to be noise-like and inaudible. The problem of a stuck noise generator is only relevant if the control sequence is inserted continuously. In the preferred embodiment, in which the control .is only inserted for brief intervals, it is not a problem for two reasons. First, since the ISB is returned to the program xignal most of the time, the stuck generator output is riot inserted into ehe signal Secondly, dynamic insertion requires the use of a synchronizing sequence, as described below, which can be designed to guarantee that the generator dace not get stuck.
3.S 9~ynamic insertion of the control signal into the L.SB or sharing the LS8 with the main program data means that the reproducer has to be able.to identify the commands embedded within the stream of arbitrary main program data. Tlits is accomplished by preceding a command code with a synchronizing sequence of bits which the decoder looks for in the data strram. The sequence ~i~B~~T~1~~ SW~~

WO 32/22(lfs0 ~ ~ ~ ~ ~ ~ ~ pCT/US92/f14629 sa can be made long enough that the probability of its occurrence in the program data is extremely small, assuming that the program data is essentially random. !Of course, one must avoid patterns that might appear with more than random frequency, such as long strings of ones or zeros which could occur during silent periods in the program. False triggering of the reproducer on program data can be completely eliminated on recordings incorporating the invention by having the encoder monitor the program data stream during recording and after the least significant bit in one word if the synchronizing sequence is about to occur, thereby preventing it. This results in a bit error probability on ehe same order as the.prevented false trigger probability, which can be made much less disruptive than the insertion of control codes, and so inconsequential.
~ It should be noted, that the aforedescribed technique can be used to hide arbitrary digital data in a digital audio signal or other digital signal representing analog data in which accuracy to the least significant bit is not continuously necessary. Such data inserted in place of or in addition do our process control signal could be used to control a multi-media presentation or for some other completely unrelated purpose.
S It will be apparent from the foregoing description that those of ordinary skill in the audio, digital and data processing arcs should be able to utilize a wide variety of computer and other electronic implementations in both hardware and software to practice many of the analysis, evaluation, encoding, decoding and compensation techniques embodied within the methods and apparatus of the present invention.
ZO The aforedescribcd systems of the present invention satisfy a long existing need in the art by provdlng new and improved digital encoding/decoding methods and apparatus for ultra low distortion reproduction of analog signals and which arc also compatible with industry uandardized signal playback apparatus not incorporating the decoding features of the present invention.
!n addition, signals lacking the encoding process features of the invention era likewise compatible with 23 playback decoders which do embody dhe invention and are provided some overall enhancement.
The present invention provides an improved encode/dccode system .enabling a predetermined balance or interplay of gain, slew and wave synthesis operations to reduce signal distortions and improve apparent resolution. Analysis is made of wavcform .characteristics during the encoding proeec.~, and the results of this analysis are subsequently utilized in the decoding process to :3(1 more accurately reconstruct the original waveform, while; minimizing the deleterious effects normal)v encountered in sampling and converting analog signals to digital signals and subsequently reconverting the digita9 signals back to an accurate simulation of the original analog waveforan.
In accordance with the invention, control infornaation developed during the aforedcscribed waveform analysis is concealed within a standard digital code and this information is 35 subsequently used to dynamically change and control the reproduction process for best performance:.
Thetc concealed control cods trigger appropriate decoding signal reconstruction compensation complementing the encoding process resulting from the signal analysis. Since the control code is silent and Ihc overall difiital information rate is normally fixed, the proccse can operate compatibly wth SUBS'TITl9T'~ SH~~~T' ~'~ 9Z/220511 IPCIf/US92/04629 54/~
existing equipment and to manufacturer's spe~cations and standards. In addition, and as previously indicated, signals lacking the encoding process features of the invention are likewise compatible with playback decoders which do erabody the invention and are affordtd some beneficial enhancement.
It will be apparent from the foregoing that, while particular forms and several aspects of the invention have been illustrated and described. various modifications can be made without departing from the spirit and scope of the invention. Accordingly, it is not intended that Ihc invention be limited, except as by the appended claims.

2~:~~~8~
~r~ ~ataaobo p~ 1'lus9~~04629 TAB1,E I
Notes to Figures 5a-Sd Fag Sa: Test signal = ~ 5 volt ramp + ~Om volt sinewave Fig Sb: Section AA of figure Sa magnified.
Signal Discontinuity From:
1. LSB roundofty,A-D, D ~1 crosstalk 2 Resolution limit, no code change 3. '. °Missing oodr,~ code error 4. Sample Bold or DAC glitch (spurious signal or energy) F'ig Sc: Sections at beginning and :,nds ~f sawtooth transition (magnified) 1. Solid = idea waveform 2. Shaded = best possible filter 3. A-B correct analog to digital conversion 4. C-D Slew rate error For low distortion:
-Area of transitions must be symmetrical to the signal wavetorm, since if they arc not, a residual offset will have o~urred after the signal.
The test signal is made symmetrical, therefore Area of overshoot A~B should equal area C--D
In drawing: C-~D is larger than A-B =
1. Unsymmctt~ical Slew Rate at Sample Hold 2. Feedback delay in A~D converter ~~es = 3. Data related noise glitches 4. Crosstalk lxtween data and timing (capture uncertainty) .5. Hysteresis/memory of previous sample.
Fig Sd: Section BB of Figure Sa (magnified) ,~.iUB~1'I~TIIT~ .'t'''~9-a~~'1"

Claims (159)

The embodiments of the invention in which an exclusive property or privilege is claimed are defined as follows:

1. A method for converting and encoding analog signals to a standardized digital format, comprising the steps of:
monitoring the physical characteristics of an analog waveform to be converted to a standardized digital format;
converting said analog waveform to said digital format; and encoding, within said standardized digital format, information indicative of the physical characteristics of said analog waveform, said information facilitating subsequent more accurate reconstruction of said analog waveform from said standardized digital format without altering the ability to also recover said analog waveform independent of said information.

2. A method as set forth in claim 1, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

3. A method as set forth in claim 1, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

4. A method as set forth in any of claims 1, 2 or 3, wherein said information is encrypted to the least significant bits of said digital format.

5. A method as set forth in any one of claims 1 to 4, wherein said information provides control codes.

6. A method as set forth in any one of claims 1 to 5, wherein said information is selectively inserted in the least significant bits of said digital format as control codes for prescribed time periods, and said least significant bits represent said analog waveform during time periods other than said prescribed time periods.

7. A method as set forth in any one of claims 1 to 6, and further including the step of dispersing within said information, over a period of time, additional analog waveform data as hidden code, whereby the apparent signal spectrum is expanded.

8. A method as set forth in any one of claims 1 to 7, and further including the steps of:
interrupting all system operations to create a period of electrical silence;
performing digital sampling of said analog waveform during said period of electrical silence; and resuming system operations after said digital sampling has been completed.

9. A method as set forth in any one of claims 1 to 8, wherein said digital format represents a compressed signal to be complementary expanded by a subsequent decoder.

10. A method as set forth in any one of claims 1 to 9, wherein said information relates to slew correction.

11. A method as set forth in any one of claims 1 to 10, wherein said information relates to level correction.

12. A method as set forth in any one of claims 1 to 11, wherein said information relates to waveform synthesis.

13. A method as set forth in any one of claims 1 to 12, wherein said information provides control codes encoded to a random-number sequence.

14. A method as set forth in claim 13, wherein said random-number sequence modulates the least significant bits of said digital format.

15. A method as set forth in claim 13 or 14, and further including the step of processing small signal changes independent of lower frequency average level.

16. A method as set forth in any one of claims 1 to 15, including the step of managing said information for optimum encoding processes favoring each signal condition for gain, level, slew rate and waveform synthesis.

17. A system for converting and encoding analog signals to a standardized digital format, comprising:
means for monitoring the physical characteristics of an analog waveform to be converted to a standardized digital format;
means for converting said analog waveform to said digital format; and means for encoding, within said digital format, information indicative of the physical characteristics of said analog waveform, said information facilitating subsequent more accurate reconstruction of said analog waveform from said standardized digital format without altering the ability also to recover said analog waveform independent of said information.

18. A system as set forth in claim 17, and further including limiting means for peak limiting said digital format, and means for selectively activating said limiting means.

19. A system as set forth in claim 17 or 18, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

20. A system as set forth in claim 17 or 18, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

21. A system as set forth in any one of claims 17 to 20, wherein said information is encrypted to the least significant bits of said digital format.

22. A system as set forth in any one of claims 17 to 21, wherein said information provides control codes.

23. A system as set forth in any one of claims 17 to 22, wherein said digital format represents a compressed signal to be complementary expanded by a subsequent decoder.

24. A system as set forth in any one of claims 17 to 23, wherein said information relates to slew correction.

25. A system as set forth in any one of claims 17 to 24, wherein said information relates to level correction.

26. A system as set forth in any one of claims 17 to 25, wherein said information relates to waveform synthesis.

27. A system as set forth in any one of claims 17 to 26, wherein said information is in the form of control codes encoded to a random-number sequence.

28. A system as set forth in claim 27, wherein said random-number sequence modulates the least significant bits of said digital format.

29. A system as set forth in any one of claims 17 to 28, and further including means for processing small signal changes independent of lower frequency average level.

30. A method for converting and encoding analog signals to a digital format, comprising the steps of:
monitoring the physical characteristics of an analog waveform to be converted to a digital format;
converting said analog waveform to said digital format;
providing frequency extension by slew-limiting for enhanced playback; and encoding, within said digital format, information indicative of the physical characteristics of said analog waveform, whereby said analog waveform subsequently can be more accurately reconstructed from said digital format.

31. A method as set forth in claim 30, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

32. A method as set forth in claim 30, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

33. A method as set forth in claim 30, 31 or 32, wherein said information is encrypted to the least significant bits of said digital format.

34. A method as set forth in any one of claims 30 to 33, wherein said information is in the form of control codes.

35. A method as set forth in claim 30, 31 or 32, wherein said information is selectively inserted in the least significant bits of said digital format as control codes for prescribed time periods, and said least significant bits represent said analog waveform during time periods other than said prescribed time periods.

36. A method as set forth in any one of claims 30 to 35, and further including the step of dispersing within said information, over a period of time, additional analog waveform data as hidden code, whereby the apparent signal spectrum is expanded.

37. A method as set forth in any one of claims 30 to 36, wherein said slew limiting is in band transient frequency extension.

38. A method as set forth in any one of claims 30 to 37, wherein said slew limiting includes warping the time base temporarily to represent a transient.

39. A method for converting and encoding analog signals to a standardized digital format, comprising the steps of:
monitoring the physical characteristics of an analog waveform to be converted to a standardized digital format;
converting said analog waveform to said digital format; and encoding, as compatible hidden code inserted asynchronously within said standardized digital format, information indicative of said physical characteristics of said analog waveform, said information facilitating subsequent more accurate reconstruction of said analog waveform from said digital format.

40. A method as set forth in claim 39, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

41. A method as set forth in claim 39, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

42. A method as set forth in claim 39, 40 or 41, wherein said information is encrypted to the least significant bits of said digital format.

43. A method as set forth in any one of claims 39 to 42, wherein said information is in the form of control codes.

44. A method as set forth in any one of claims 39 to 43, wherein said information is selectively inserted in the least significant bits of said digital format as control codes for prescribed time periods, and said least significant bits represent said analog waveform during, time periods other than said prescribed time periods.

45. A method as set forth in any one of claims 39 to 44, and further including the step of dispersing within said information, over a period of time, additional analog waveform data as hidden code, whereby the apparent signal spectrum is expanded.

46. A method as set forth in any one of claims 39 to 45, wherein side channel bandwidth is borrowed from amplitude resolution.

47. A method as set forth in any one of claims 39 to 46, wherein side channel bandwidth is borrowed on an as-needed basis.

48. A method as set forth in any one of claims 39 to 47, wherein a side channel signal is encrypted to a noise-like characteristic.

49. A method as set forth in any one of claims 39 to 48, wherein encrypted data acts as dither for the remaining signal.

50. A method for converting and encoding analog signals to a digital format, comprising the steps of:
monitoring the physical characteristics of an analog waveform to be converted to a digital format;

applying frequency extension asynchronously to enhance playback;
converting said analog waveform to said digital format; and encoding, within said digital format, information indicative of the physical characteristics of said analog waveform, said information facilitating subsequent more accurate reconstruction of said analog waveform from said digital format.

51. A method as set forth in claim 50, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

52. A method as set forth in claim 50, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

53. A method as set forth in claim 50, 51 or 52, wherein said information is encrypted to the least significant bits of said digital format.

54. A method as set forth in any one of claims 50 to 53, wherein said information is in the form of control codes.

55. A method as set forth in claim 50, 51 or 52, wherein said information is selectively inserted in the least significant bits of said digital format as control codes, and said least significant bits represent said analog waveform during time periods other than the time periods of code insertion.

56. A method as set forth in any one of claims 50 to 55, and further including the step of dispersing within said information, over a period of time, additional analog waveform data as hidden code, whereby the apparent signal spectrum is expanded.

57. A method as set forth in any one of claims 50 to 56, wherein said frequency extension is based upon sending information about transient events.

58. A method as set forth in claim 57, wherein said frequency extension is based upon waveform synthesis.

59. A method as set forth in claim 58, including coding an actual difference signal in a side channel.

60. A method as set forth in any one of claims 50 to 56, wherein said frequency extension is based upon an analysis to determine transient waveshape.

61. A method as set forth in any one of claims 50 to 56, wherein said frequency extension is based upon a best fit of said analog waveform in a lookup table.

62. A method as set forth in claim 61, wherein said frequency extension is based upon scaling information.

63. A method as set forth in any one of claims 50 to 56, wherein said frequency extension is based upon a token for waveform shape transmitted in a side channel.

64. A method as set forth in any one of claims 50 to 56, wherein said frequency extension is based upon adding a scaled waveshape to a main signal.

65. A method as set forth in any one of claims 50 to 56, wherein said frequency extension is based upon sending of actual out-of-band information.

66. A method for decoding and converting an encoded digital-format signal to an analog waveform, comprising the steps of:
decoding, from said digital-format signal, control information indicative of certain specified physical characteristics of said analog waveform;
converting said digital-format signal to said analog waveform; and selectively introducing signal-reconstruction compensation, in accordance with said control information during said converting process, whereby said analog waveform subsequently can be more accurately reconstructed from said digital format by utilization of said control information, while preserving the ability to recover said analog waveform independent of said control information.

67. A method as set forth in claim 66, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

68. A method as set forth in claim 66, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

69. A method as set forth in claim 66, 67 or 68, wherein said compensation involves slew correction.

70. A method as set forth in any one of claims 66 to 69, wherein said compensation involves level correction.

71. A method as set forth in any one of claims 66 to 70, wherein said compensation involves waveform synthesis.

72. A method as set forth in any one of claims 66 to 71, wherein said control information is in the form of control codes encoded to a random-number sequence.

73. A method as set forth in claim 72, wherein said random-number sequence modulates the least significant bits of said digital format.

74. A method for converting a digital-format signal to an analog waveform, comprising the steps of:
converting said digital-format signal to said analog waveform; and introducing signal-reconstruction compensation, in accordance with said digital-format signal, during said converting process to facilitate more accurate reconstruction of said analog waveform from said digital-format signal, without altering the ability to also recover said analog waveform independent of said signal-reconstruction compensation.

75. A system for decoding and converting an encoded digital-format signal to an analog waveform, comprising:
means for extracting, from said digital format, control information indicative of certain specified physical characteristics of said analog waveform;

means for converting said digital-format signal to said analog waveform; and means for introducing signal-reconstruction compensation, in accordance with said control information during said converting process, whereby said analog waveform subsequently is more accurately reconstructed from said digital formal without altering the ability also to recover said analog waveform independent of said control information.

76. A system as set forth in claim 75, wherein said control information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

77. A system as set forth in claim 75, wherein said control information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

78. A system as set forth in claim 75, 76 or 77, wherein said compensation involves slew correction.

79. A system as set forth in any one of claims 75 to 78, wherein said compensation involves level correction.

80. A system as set forth in any one of claims 75 to 79, wherein said compensation involves waveform synthesis.

81. A system as set forth in any one of claims 75 to 80, wherein said control information relates to level, slew and corrective waveform synthesis.

82. A system as set forth in any one of claims 75 to 81, wherein said control information is in the form of control codes encoded to a random-number sequence.

83. A system as set forth in claim 82, wherein said random-number sequence modulates the least significant bits of said digital format.

84. A method for converting and encoding an analog signal to a standardized digital format and subsequently decoding and converting the digital format to recover the analog signal, the method comprising the steps of:
monitoring the physical characteristics of an analog waveform to be converted to a standardized digital format;
converting said analog waveform to said digital format;
encoding, within said digital format, control information indicative of the physical characteristics of said analog waveform, to facilitate subsequent more accurate reconstruction of said analog waveform from said digital format;
decoding from said digital format said control information indicative of certain specified physical characteristics of said analog waveform;
converting said standardized digital-format signal to said analog waveform;
and introducing signal-reconstruction compensation, in accordance with said control information during said converting process, whereby said analog waveform subsequently is more accurately reconstructed from said digital format without altering the ability to also recover said analog waveform independent of said control information.

85. A method as set forth in claim 84, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

86. A method as set forth in claim 84, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

87. A method as set forth in claim 84, 85 or 86, wherein said control information is encrypted to the least significant bits of said digital format.

88. A method as set forth in claim 87, wherein said control information provides control codes.

89. A method as set forth in any one of claims 84 to 88, wherein said digital format represents a compressed signal to be complementary expanded by a subsequent decoding.

90. A method as set forth in any one of claims 84 to 89, wherein said control information relates to slew correction.

91. A method as set forth in any one of claims 84 to 90, wherein said control information relates to level correction.

92. A method as set forth in any one of claims 84 to 91, wherein said control information relates to waveform synthesis.

93. A method as set forth in any one of claims 84 to 92, wherein said control information provides control codes encoded to a random-number sequence.

94. A method as set forth in claim 93, wherein said random-number sequence modulates the least significant bits of said digital format.

95. A method as set forth in any one of claims 84 to 94, and further including the step of processing small signal changes independent of lower frequency average level.

96. A method as set forth in any one of claims 84 to 95, wherein said compensation involves slew correction.

97. A method as set forth in any one of claims 84 to 96, wherein said compensation involves level correction.

98. A method as set forth in any one of claims 84 to 97, wherein said compensation involves waveform synthesis.

99. A system for converting and encoding an analog signal to a standardized digital format and subsequently decoding and converting said digital formal to recover the analog signal, the system comprising:
means for monitoring the physical characteristics of an analog waveform to be converted to a standardized digital format;
means for converting said analog waveform to said digital format;

means for encoding, within said digital format, control information indicative of the physical characteristics of said analog waveform, to facilitate subsequent more accurate reconstruction of said analog waveform from said digital format;
means for decoding, from said digital format, said control information indicative of certain specified physical characteristics of said analog waveform;
means for converting said digital-format signal to said analog waveform; and means for introducing signal-reconstruction compensation, in accordance with said control information during said converting process, whereby said analog waveform subsequently is more accurately reconstructed from said digital format without altering the ability also to recover said analog waveform independent of said control information.

100. A system as set forth in claim 99, wherein said control information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

101. A system as set forth in claim 99, wherein said control information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

102. A system as set forth in claim 99, 100 or 101, wherein said signal-reconstruction compensation involves slew correction.

103. A system as set forth in any one of claims 99 to 102, wherein said signal-reconstruction compensation involves level correction.

104. A system as set forth in any one of claims 99 to 103, wherein said signal-reconstruction compensation involves waveform synthesis.

105. A system as set forth in any one of claims 99 to 104, wherein said control information relates to level, slew and corrective waveform synthesis.

106. A system as set forth in any one of claims 99 to 105, wherein said control information provides control codes encoded to a random-number sequence.

107. A system as set forth in claim 106, wherein said random-number sequence modulates the least significant bits of said digital format.

108. A system for converting and encoding an analog signal to a digital format and subsequently decoding and converting said digital format to recover the analog signal, the system comprising:
means for monitoring the physical characteristics of an analog waveform to be converted to a digital format;
means for frequency-resolution enhancement by selection of decimation and interpolation filters in response to said monitored physical characteristics to facilitate enhanced playback;
means for converting said analog waveform to said digital format;
means for encoding, within said digital format, control information indicative of the physical characteristics of said analog waveform, whereby said analog waveform subsequently can be more accurately reconstructed from said digital format;

means for decoding from said digital format said control information indicative of certain specified physical characteristics of said analog waveform;
means for converting said digital-format signal to said analog waveform; and means for introducing compatible signal-reconstruction compensation, in accordance with said control information during said converting process, said information facilitating subsequent more accurate reconstruction of said analog waveform from said digital format.

109. A system as set forth in claim 108, wherein said control information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

110. A system as set forth in claim 108 or 109, wherein said means for frequency-resolution enhancement includes a choice of different filters for different signal conditions.

111. A system as set forth in claim 110, wherein each of said filters is switched on and off dynamically.

112. A system as set forth in claim 111, wherein said filters are switched dynamically by fading.

113. A system as set forth in claim 111, wherein said filters are switched dynamically by merging.

114. A system as set forth in any one of claims 108 to 113, wherein said means for frequency-resolution enhancement includes choosing filters based on analysis of ratios of energy in different frequency bands.

115. A system as set forth in any one of claims 108 to 113, wherein said means for frequency resolution enhancement includes choosing filters based on analysis of presence of isolated transients.

116. A system as set forth in any one of claims 108 to 113, wherein said means for frequency resolution enhancement includes choosing filters based on analysis of overall signal level.

117. A system as set forth in any one of claims 108 to 113, wherein said means for frequency-resolution enhancement includes using complementary filter pairs.

118. A system as set forth in claim 117, wherein said filter pair includes one filter for decimation during encoding, and another filter for interpolation during decoding/playback.

119. A system as set forth in any one of claims 108 to 118, wherein said decimation filter is applied single-ended for analog-to-digital conversion.

120. A system as set forth in any one of claims 108 to 119, wherein said interpolation filter is applied single-ended for digital-to-analog conversion.

121. A system for converting and encoding an analog signal to a standardized digital format and subsequently decoding and converting said digital format to recover the analog signal, the system comprising:
means for monitoring the physical characteristics of an analog waveform to be converted to a digital format;
means for compatible amplitude-resolution enhancement;
means for converting said analog waveform to said digital format;
means for encoding, as compatible hidden code within said standardized digital format, control information indicative of the physical characteristics of said analog waveform, to facilitate said analog waveform being subsequently more accurately reconstructed from said digital format;
means for decoding from said digital format said control information indicative of certain specified physical characteristics of said analog waveform;
means for converting said digital-format signal to said analog waveform; and means for introducing compatible signal-reconstruction compensation, in accordance with said control information during said converting process, said information facilitating subsequent more accurate reconstruction of said analog waveform from said standardized digital format.

122. A system as set forth in claim 121, wherein said control information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

123. A system as set forth in claim 121 or 122, wherein said compensation involves slew correction.

124. A system as set forth in claim 121, 122 or 123, wherein said compensation involves level correction.

125. A system as set forth in any one of claims 121 to 124, wherein said compensation involves waveform synthesis.

126. A system as set forth in any one of claims 121 to 125, wherein said means for compatible amplitude-resolution enhancement includes low-level control.

127. A system as set forth in any one of claims 121 to 126, wherein said compatible amplitude-resolution enhancement is done in complementary fashions controlled by a side channel.

128. A system as set forth in any one of claims 121 to 127, wherein said means for compatible amplitude-resolution enhancement includes control of decoding to complement encoding.

129. A system as set forth in any one of claims 121 to 128, wherein said compatible amplitude-resolution enhancement is done only at very low levels.

130. A method for converting and encoding an analog signal to a digital format and subsequently decoding and converting said digital format to recover the analog signal, the method comprising the steps of:

monitoring the physical characteristics of an analog waveform to be converted to a digital format;
converting said analog waveform to said digital format;
encoding, within said digital format, control information indicative of the physical characteristics of said analog waveform, whereby said analog waveform subsequently can be more accurately reconstructed from said digital format;
decoding from said digital format said control information indicative of certain specified physical characteristics of said analog waveform;
converting said digital-format signal to said analog waveform;
peak limiting and restoring; and introducing compatible signal-reconstruction compensation, in accordance with said control information during said converting process, said information facilitating subsequent more accurate reconstruction of said analog waveform from said digital format.

131. A method as set forth in claim 130, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

132. A method as set forth in claim 130, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

133. A method as set forth in claim 130, 131 or 132, wherein said control information is encrypted to the least significant bits of said digital format.

134. A method as set forth in claim 133, wherein said control information is in the form of control codes.

135. A method as set forth in any one of claims 130 to 134, wherein said peak limiting and restoring includes an encoder-selected algorithm for decoding.

136. A method as set forth in any one of claims 130 to 134, wherein said peak limiting and restoring includes one-to-one mapping for reconstruction by a decoder.

137. A method as set forth in any one of claims 130 to 134, wherein said peak limiting and restoring includes use only at highest peak levels.

138. A method as set forth in any one of claims 130 to 134, wherein said peak limiting and restoring includes mimicking analog tape saturation curve shape for low distortion in un-decoded playback.

139. A system for converting and encoding analog signals to a digital format, comprising:
means for monitoring the physical characteristics of an analog waveform to be converted to a digital format;
means for converting said analog waveform to said digital format;
limiting means for peak limiting said digital format;
means for selectively activating said limiting means; and means for encoding, within said digital format, information indicative of the physical characteristics of said analog waveform, to facilitate subsequent more accurate reconstruction of said analog waveform from said digital format.

140. A system as set forth in claim 139, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

141. A system as set forth in claim 139, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

142. A system as set forth in claim 139, 140 or 141, wherein said information is encrypted to the least significant bits of said digital format.

143. A system as set forth in any one of claims 139 to 142, wherein said information is in the form of control codes.

144. A system as set forth in any one of claims 139 to 143, wherein said digital format represents a compressed signal to be complementary-expanded by a subsequent decoder.

145. A system as set forth in any one of claims 139 to 144, wherein said information relates to slew correction.

146. A system as set forth in any one of claims 139 to 145, wherein said information relates to level correction.

147. A system as set forth in any one of claims 139 to 146, wherein said information relates to waveform synthesis.

148. A recording product having recorded thereon an analog signal waveform encoded in a standardized digital format, the product comprising:
a recording medium having recorded thereon said analog waveform converted to said digital format, and further including, within said digital format, encoded information indicative of the physical characteristics of said analog waveform, said information facilitating subsequent more accurate reconstruction of said analog waveform from said standardized digital format without altering the ability also to recover said analog waveform independent of said information.

149. A recording product as set forth an claim 148, wherein said information indicative of the physical characteristics of said analog waveform is encrypted within said digital format.

150. A recording product as set forth in claim 148, wherein said information indicative of the physical characteristics of said analog waveform is concealed within said digital format.

151. A recording product as set forth in claim 148, 149 or 150, wherein said information is encrypted to the least significant bits of said digital format.

152. A recording product as set forth in any one of claims 148 to 151, wherein said information provides process control codes.

153. A recording product as set forth in any one of claims 148 to 152, wherein said digital format represents a compressed signal to be complementary expanded by a subsequent decoder.

154. A recording product as set forth in any one of claims 148 to 152, wherein said digital format represents a peak limited signal for high signal levels and an expanded signal for low signal levels.

155. A recording product as set forth in any one of claims 148 to 154, wherein said information relates to slew correction.

156. A recording product as set forth in any one of claims 148 to 155, wherein said information relates to level correction.

157. A recording product as set forth in any one of claims 148 to 156, wherein said information relates to waveform synthesis.

158. A recording product as set forth in any one of claims 148 to 157, wherein said information provides process control codes encoded to a random-number sequence.

159. A recording product as set forth in claim 158, wherein said random-number sequence is recorded as modulation of the least significant bits of said digital format.