US 3872503 A
In a system wherein a continuous signal is divided into a plurality of discrete segments each consisting of a plurality of signal samples, a predetermined number of the last signal samples of one segment are repeated as the first signal samples of the next segment. The predetermined number is selected to correspond to a time duration greater than at least a major portion of the transients introduced as a result of dividing the continuous signal into segments. The continuous signal can thus be reconstructed from the segments without discontinuities and without the unwanted transients.
Claims available in
Description (OCR text may contain errors)
United States Patent Shutterly 1 Mar. 18, 1975 ELIMINATION OF TRANSIENTS IN PROCESSING SEGMENTS OF AUDIO INFORMATION EXTERNAL REFERENCE ZOAUDIO PLAYBACK 24 ANALOG ig IN PUT SAMPLE DIG AUDIO g'i ei g lnecolw 'lf I: CONVE R i ER 'RECIRCULATE SWITCH 3,789,137 l/l974 Newell 360/8 Primary Examiner--.lames W. Moffitt Attorney, Agent, or Firm--M. P. Lynch  ABSTRACT In a system wherein a continuous signal is divided into a plurality of discrete segments each consisting of a plurality of signal samples, a predetermined number of the last signal samples of one segment are repeated as the first signal samples of the next segment. The pre- I determined number is selected to correspond to a time duration greater than at least a major portion of the transients introduced as a result of dividing the continuous signal into segments. The continuous signal can thus be reconstructed from the segments without discontinuities and without the unwanted transients.
5 Claims, 10 Drawing Figures HIF RE Gl S'l' ER BUFFER CONTINUOUS SLOW CLOCK TIMING IR; l
RECIRCULATE FAST CLOCK XE MULTIPLEXED C LOCK IOO UN n' IPLAYBACK l FAST CLOCK RECORD 32 VI DEO BANDPASS Fl LTER l l l l PLAYBACK RECORD so, 52 54 BAiPD PA ss $318 Fl LTE R PATENTEI] MAR 1 8 I975 SHEET 1 BF 3 PATEMEB 3,872,503
sumsq s RECIRCULATED I S TA SSI COMATJRDEIOSED FIGA I I i i TIME START OF FAST END OF FAST CLOCK BURST .CLOCK BURST DURING PLAYBACK DURING PLAYBACK SLIgDLIjVLgIEgCK I I I I I I I I I I I I I I I I I I 'FIG.50
pA c l cK IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII FIG-5b AFJIEs n II 5I2I/3JL\I I\ SES Ag g L o k IIIII IIIII -FlG.5d
FAST CLOCK BURST g gg g ggg I I .I I IIIIIII I I... I I IIIIIIII I, .T|MEF' SIGNALS I I 495 SLOW CLOCK PULSES ---TIME SIGNAL I ELIMINATION OF TRANSIENTS' IN PROCESSING SEGMENTS OF AUDIO INFORMATION BACKGROUND OF THE INVENTION A technique for time-compressing audio signals 5 through the use of a time buffer store for recording in a format similar to video signals and the subsequent time expansion of the signals to produce the original audio signals is disclosed in detail in the copending application Ser. No. 241,944 filed Apr. 7, i972 and now format for the time-compressed audio information is that corresponding to a standard television line which would permit the transmission of audio and video signals in a time multiplex mode and permit the use of conventional TV channels and equipment for processing the time-compressed audio signals. The capability 2 of time multiplexing audio and video signals facilitates the transmission of audio accompanied by video stills in a time significantly shorter than the time required for presentation of the original audio information.
The invention described in the above-identified co- 3 pending patent application permits time-compressed audio and video waveforms to be combined in time multiplex form, recorded, replayed and transmitted over a conventional television network without modification to the network equipment. The transmitted like waveforms with synchronizingsignals added at conventional TV sweep intervals in order to make the audio information appear as a video composite waveform. The time-compressed audio samples are formed in segments comparable to the active portion of a television line, i.e., 52 microseconds. The conversion of the time-compressed segments of audio information into conventional audio bandwidth information for subsequent reproduction has however suffered due to transients and periodic noise occurring at the interface of 5 adjacent segments of the audio information.
In the system described in the above-identified application, the time compression is achieved by sampling an audio signal and storing the samples in a buffer store. The samples in the buffer are then read out at a much higher rate, producing a segment of the compressed audio signal. Synchronizing signals, if required, are added in front of or following the audio segment, and the resultant combination is stored in a recirculating store, such as a track of a magnetic disc. A second group of samples is then collected in the buffer and recorded on the same track in an adjacent position, and this process is continued until the track is completely recorded. The buffer sequence is reversed during playback. A segment of compressed audio information from a recirculating store is sampled at a high rate into a buffer store. The samples are then read out of the buffer at a relatively low rate to produce the original audio signal. When the buffer has beenemptied, a new compressed audio segment is read in at a high speed and the process is continued.
Due to the elapsed time and the intervening synchronizing signals between successive compressed audio segments, transient interference occurs at the beginning of each new segment of compressed audio. The
transient interference is due in part to the energy storage elements, i.e., inductors and capacitors, the filters, modulators, and demodulators used to process the time-compressed audio signal. There are essentially two types of interefercnce at the beginning of an audio segment:
I. transients due to the absence of audio samples which would normally be present if there was no time separation between successive audio segments, and
2. transients caused by the synchronizing signals occurring between successive audio segments.
Since transient interference occurs at the same periodic rate at which the time-compressed audio segments are converted into normal audio, the effect of the interference is to introduce an undesirable audible noise, i.e., buzzing, into the recovered audio information.
Another source of audio noise is inaccurate sampling during playback when the time-compressed audio is converted for playback. More specifically, the first sampling pulse must sample the first signal sample; the second sampling pulse must sample the second signal sample, etc. Otherwise, one or more of the signal samples of each segment of time-compressed audio information will be omitted and a number of samples from the non-audio, synchronizing portion of the signal will be added, thus producing a very objectionable audible noise. Such exact sampling is difficult to achieve and maintain due to the many delays encountered in the signal processing circuitry.
SUMMARY OF THE INVENTION There is described below with reference to the accompanying drawings an application of the invention described in the Abstract for record and playback of time compressed audio segments as developed in the referenced copending application such that the first few samples in each segment are identical tothe last few samples from the previous compressed audio segment; that is, the last few segments of each compressed audio segment are repeated as the first few samples of the next segment. If, for example, each compressed audio segment consists of 512 signal samples, then an arbitrary number of samples, i.e., 12, at the end of each time-compressed audio segment would be repeated as the first 12 samples of the next time-compressed audio segment. .During the conversion of the timecompressed audio signals for reproduction, the sampling of each segment begins at an audio sample in the group of repeated audio samples located at the beginning of each segment.
The transient errors are eliminated since the sampling of each segment is begun after transients due to the non-audio (sync) portion of the signal have decreased to a negligible value and transients due to the initial samples, those being repeated, have reestablished the signal conditions that would have been present had there been no break between the adjacent audio segments. If, for example, the overall transient decay time is equal to 4 sample periods, the playback sampling in each compressed audio segment could start with the th sample.
The repeatedsamples also eliminates the need for extreme accuracy for initiating the sampling operation. If, for example, one playback unit starts the sampling at sample 5 in each segment and a second unit starts at sample in each segment, both recover a continuous audio signal if 12 samples have been repeated. One unit would sample from sample 5 to sample 504, assuming 512 samples per segment with 12 repeated samples, and a second unit would sample from sample l0 to sample 509. Since samples 505 through 509 are repeated as samples 5 through 9 in the next segment, both playback units will have selected the proper samples'for recovering the audio information.
DESCRIPTION OF THE DRAWINGS The invention will become more readily apparent from the following exemplary description in connection with the accompanying drawings:
FIG. 1 is a schematic illustration-of an audio record playback system incorporating the invention;
FIG. 2 is a detailed schematic illustration of the timing circuit of the embodiment of FIG. 1;
FIGS. 3 and 4 are waveform illustrations of the timecompressed audio information processed in the embodiment of FIG. 1;
FIGS. 5a, 5b, 5c, 5d, 5e, and Sfare timing waveforms illustrating the operation of the timing circuit of FIG. 2.
DESCRIPTION OF THE PREFERRED EMBODIMENT There is illustrated in FIG. 1 a magnetic disc recorder D typically including a clock track and a plurality of additional tracks for recording information. The magnetic disc recorder D is illustrated as including a clock head CH, a playback head PH and record head RH. For the purpose of discussion it will be assumed that 525 equally spaced square wave pulses had been recorded on the clock track of magnetic disc recorder D. Inasmuch as a television frame consists of 525 lines, a servo drive system is implemented through the use of a phase comparator having a first input consisting of horizontal drive signal, i.e., l5,734.26 Hz, for television and a second input from the clock head CH associated with the clock track of the magnetic disc recorder D. The output of the phase comparator 12 is subsequentially amplified by motor drive amplifier l4 and applied to the magnetic disc drive motor 16 to establish a disc rotation equivalent to one revolution of the disc per TV frame i.e., 30 revolutions per second. The schematic illustration of the servo control system for the magnetic disc D is merely representative of numerous servo drive techniques suitable for establishing a disc rotation at 30 revolutions per second.
Having established the desired revolution rate of the magnetic disc recorder, the following discusion will relate to a technique for time-compressing audio signals into segments of compressed audio information equivalent to television line information with the additional improvement of providing for elimination of transient and periodic noise during the record and playback of the discrete segments of time-compressed audio information. While the description will relate to FM recording it will be apparent that the technique for eliminating periodic noise will be fully applicable to other recording modes including AM recording. It will be apparent to those skilled in the art that the inventive concept is applicable to recording media other than a magnetic disc as well as to direct transmission of compressed segments without intermediate recording:
Audio information is supplied through an audio band pass filter 20 which band limits the signal for subsequent application through a record-playback switch 22, comprised of switch sections 22a, 22b and 220, when positioned in the record position to the sample and hold circuit 24. The sample and hold circuit 24 responds to clock pulses received from the timing circuit by sampling the audio input information. The sample and hold circuit 24 retains the sampled audio information until a subsequent clock pulse is received at which time a new sample of audio information is stored by the sample and hold circuit 24. The analog to digital converter 26, like the sample and hold circuit 24, is activated by fast clock pulses i.e., typically 10.74 megahertz, from the timing circuit 100 with the analog to digital converter producing a digital output representation of the analog signal transmitted by the sample and hold circuit 24.
For the purpose of discussing the operation of a typical embodiment of the invention, it will be assumed that the digital output signal developed by the analog to digital converter 26 is in the form of an 8 bit binary coded word. The 8 bit digital output signal of the analog to digital converter 26 is transmitted through a recirculating switch 28 to a shift register buffer store 30. Binary coded words supplied to the shift register buffer store 30 are supplied to the shift register buffer store 30 by clock pulses originating from the timing circuit 100. During the record mode the clock pulses supplied to the shift register buffer store 30 correspond to slow clock pulses, i.e., l4.8 kilohertz, which are transmitted from the timing circuit 100 through the logic OR gate 32. In the embodiment described herein, wherein recording of time-compressed audio segments consisting of 5 12 signal samples is desired, the timing circuit 100 is designed, as will be described with reference to FIG. 2, to supply a number of slow pulses which is less than 512, herein chosen to be 495, to the shift register buffer store 30 for entering 495 of the 8 bit binary code words. After 495 binary code words have been entered into the shift register buffer store 30 in response to 495 slow clock pulses, the timing circuit I00 supplies a burst of 512 fast clock pulses i.e., at a rate of l0.74 megahertz, to the shift register buffer store 30. Simultaneously the timing circuit 100 transmits a control pulse to the recirculating switch 28 to transfer it to the recirculating mode. In this mode the digital signals contained within the buffer store, 30 which includes the 495 just entered plus 17 remaining from the information previously stored in the buffer store, are simultaneously recirculated through the recirculating switch 28 back into the buffer store 30 as well as being transmitted to the digital to analog converter 34. The recirculating switch 28 can be satisfied through the use of a simple two position digital switch. Commercially available Motorola chip MC 3023 will satisfy the function of recirculating switch 28.
The waveform comprised of a plurality of timecompressed audio segments produced in response to this recirculation of a portion of each segment for inclusion with the subsequent time-compressed segment is illustrated in FIG. 3. The net result of this operation is such that after the burst of 512 fast clock pulses, the contents in the shift register buffer store correspond identically to the contents that existed prior to the burst of 512 fast clock pulses. The digital to analog converter 34 in response to clocking by the fast clock pulses from the timing circuit 100 converts each of the 512. binary code words into corresponding analog samples. These analog samples are subsequently supplied to a video band pass filter 38 through the record-playback switch 22b when positioned in the record position. The video band pass filter 38 functions to pass the base band portion of the analog information while eliminating all the higher harmonics present in the analog sample signals. This base band portion of the audio analog signals is subsequently processed through FM modulator for subsequent recording on the magnetic disc D by the record head RH. The recording of the 512 analog samples takes place in approximately 47 microseconds which is less than the 52 microseconds of the active portion of a conventional television line.
The cycle thus described is again repeated with 495 slow clock pulses being applied to enter 495 new digital 8 bit binary words into the buffer store 30 such that the 512 digital code words stored in the buffer store 30 correspond to the 495 new digital code words and 17 digital code words remaining from the previous 512 digital words stored in the buffer store 30. Once again 512 fast clock pulses are subsequently applied to the shift register buffer store 30 to recirculate the 512 pulses through the recirculating switch 28 and back to the buffer store 30 while simultaneously converting the 5 l2 digital words into analog samples by the digital to analog converter 34. These samples are processed as described above and recorded on the magnetic disc D. As noted above the speed of the magnetic disc is regulated relative to the fast clock burst such that the rotation of the magnetic disc will advance the disc by one television line for each burst of fast clock pulses. This assures that each subsequent segment of time-compressed audio information comprised of 512 analog samples is recorded adjacent to the previously recorded segment of timecompressed audio samples. This recording process continues until all audio information has been recorded allowing, if necessary, for space to include television horizontal and vertical sync information between adjacent segments of time-compressed audio information to ultimately produce an audio information waveform comparable to a video composite waveform for processing on video apparatus. The time-compressed audio information in segments eorrespponding to the active TV portion of a television line in combination with the appropriate television synchronizing information is illustrated in the waveform of FIG. 3.
For operation in the playback mode, the switch 22 is transferred from a record position to a playback position and the playback head PH supplies the time compressed analog information recorded on the tracks of the magnetic disc D through an amplifier 40 and FM discriminator 42 asinput information to the sample and hold circuit 24. As described above the sample and hold circuit 24 responds to fast clock pulses by sampling the time-compressed analog information supplied to it and transmitting the sampled analog information to the analog to digital converter in response to a subsequent clock pulse. A sample thus transmitted to the analog to digital converter is converted into an 8 bit binary digital word in response to the fast clock pulses and subsequently applied through the recirculating switch 28 to the shift register buffer store 30. The recirculating switch 28 is set in a non-recirculation position during the playback mode of operation by the timing circuit 100. The burst of fast clock pulses supplied to the shift register buffer store during the playback mode of operation is delayed slightly with respect to the segment of time-compressed audio information being read from the magnetic disc. This delay is developed by the timing circuit 100, as will be explained in detail later, and as illustrated in the waveform of FIG. 4. The delay corresponds to a number of analog samples which is less than the number of digital words repeated from one segment of time-compressed audio information to a subsequent segment of time-compressed audio information during record. In the embodiment described thus far wherein 17 digital binary coded words from the end of a segment is repeated at the beginning of the next segment. The delay in clocking the timecompressed analog information from the disc D into the buffer store 30 might typically be represented by a time corresponding to between 10 and 12 analog samples. Since the first portion of the time-compressed analog audio information coming from the disc D corresponds to the 17 samples repeated from the previous segment of time-compressed information the time lapse corresponding to this repeated portion gives time for any foreign transients present to decay to an acceptable level before the burst of fast clock pulses is initiated in the playback mode. The repeated samples reestablish the transient conditions that would have been present if there had been no break between the audio segments, while substantially eliminating the transients developed as a result of the break between audio segments.
In the playback mode the burst of fast clock pulses precedes the slow clock pulses such that following the delay established by the timing circuit 100, a burst of 512 fast clock functions to enter the recorded audio information into the shift register buffer store 30. Following the conclusion of the 512 fast clock pulses, a burst of 495 slow clock pulses are applied to the shift register buffer store 30 to discharge 495 of the 8 bit binary words from the buffer store 30 at the slow clock rate of 14.8 kilohertz to the digital to analog converter 34. The digital to analog converter 34, is clocked also by the slow clock pulses as indicated by the positionof the record-playback switch 220 so as to discharge analog representations of the 8 bit binary digital words at the slow clock rate through the record-playback switch 22!: to the audio band pass filter 50. The audio band pass filter 50 functions to transmit the audio base band spectrum without harmonics to the audio amplifier 52 for amplification and subsequent transmission to the audio speaker 54.
The pulse rates illustrated in FIGS. 5a, 5b, 5c, 5d, 5e, and 5f illustrate the operation of the system in response to the output pulses developed by the-timing circuit 100.
Typical implementation of the timing circuit to achieve the necessary clock drive signals is illustrated schematically in FIG. 2. There is illustrated in FIG. 2 an oscillator circuit 1 10 for supplying pulses at the rate of 42.95454 megahertz and a horizontal television drive source 113 for supplying pulses at a rate of 15,734.26 hertz which corresponds to the same drive signal applied to the phase comparator of FIG. 1. The pulse output of the oscillator is supplied simultaneously to divider circuits 112 and 114. Divider 112 functions to divide this frequency rate by a factor of 4 to produce the 10.74 megahertz fast clock pulses. The combination of the divide by 3 circuit 114 and the divide by 967 circuit 116 functions to produceslow clock pulses at a rate of l4,806.8 hertz from the divide circuit 116. The slow clock pulses developed at the output of the divider circuit 116 are transmitted through the pulse monostable circuit 118 and OR gate 120 in the absence of the application of a reset pulse to the divider circuit 116 by the output of the flip-flop circuit 122. The pulse monostable circuit 118 determines the width of the slow clock pulses. It connects a square wave signal into a train of narrow pulses. The state of the flipflop circuit 122 is controlled by a reset signal derived from the output of the horizontal drive oscillator 113 while the set input signal to the flip-flop circuit 122 corresponds to the output of the flip-flop circuit 124. The output pulses developed by the horizontal drive oscillator 113 are supplied to divider circuit 126 which divides the output pulse rate of the horizontal drive oscillator 113 by 526. Since there are 525 lines in a TV frame, the operation of divider circuit 126 function to select for example line n in one television frame, line n+l in the next television frame, line n+2 in the next television frame and so on. Since the magnetic disc D is rotating at a rate corresponding to the television frame rate and is phase locked to the horizontal drive oscillator reference signal, the initiation of each burst of fast clock pulses is offset by one line period from the previous burst of fast clock pulses for each revolution of the magnetic disc. This offset provides for the accurate positioning of the adjacent segments of timecompressed audio information on the tracks of the magnetic discs D.
The output of the divider circuit 126 functions to set flip-flop circuit 124 producing an input to AND gate 128. When this input is present, fast clock input pulses from the divider circuit 1 12 are gated through the AND gate 128 to the lO-stage counter 130. The AND gate 128 functions to gate 532 of the l0.74 megahertz pulses. When the flip-flop circuit 124 is set by the output of the divider circuit 126 the output of the flip-flop circuit 124 functions to set flip-flop circuit 122 such that the output of the flip-flop circuit 122 resets divider 116 preventing the generation of additional slow clock pulses. However, one television line period later, the flip-flop circuit 122 is reset by the output of the horizontal drive oscillator circuit 113 thus causing a change in the state of the flip-flop circuit 122 so as to permit the generation of additional slow clock pulses by divider circuit 116 while at the same time triggering the back edge monostable circuit 132 which produces one slow clock pulse. The back edge monostable circuit 132 converts the negative going edge of the output from flip-flop circuit 122, which is generated when the flip-flop is reset, into a narrow slow clock pulse.
With the record-playback switch 22 of the embodiment of FIG. 1 positioned on the record mode, the count detector circuits 134 and 136 respond to the fast clock count in the lO-stage counter 130 to establish the fast clock burst timing. The count detector circuit 134 responds to a fast clock pulse count of 10 by setting flip-flop circuit 138 thereby actuating AND gate 140 which is enabled by the positioning of the switch 141 in the record mode. The count of 10 is arbitrary; it determines the position of the fast clock burst relative to the external l5,734.26 Hz reference. The count detector 136 is set to respond to the 522nd fast clock pulse entered into the lO-stage counter during the record mode to reset flip-flop circuit 138 and terminate the output of the AND gate 140. The AND gate actuated by the record mode function and the output of the flip-flop circuit 138 develops an output pulse of a duration corresponding to the burst of 512 fast clock pulses. The output pulse of the AND gate 140 is supplied to control the recirculating switch 28 of FIG. 1 as well as being supplied through the OR gate 142 to the AND gate 144. The application of the pulse developed by ANd gate 140 to the AND gate 144 functions to gate a burst of fast clock pulses from divider circuit 112 of a duration corresponding to the duration of the enabling pulse. This produces a burst of 512 fast clock pulses for application through the OR gate 32 to the shift register 30 and to the digital to analog converter 34 in the record mode.
With the record-playback switch 22 of FIG. 1 positioned in the playback mode, thus enabling AND gate 150, the count detector circuit 152 functions to set flipflop circuit 154 in response to occurrence of the 20th fast clock pulse entered into the lO-stage counter 130. Flip-flop circuit 154 provides a second input of the AND gate 150. Detector circuit 156 functions to reset the flip-flop circuit 154 at the occurrence of the 532nd fast clock pulse thereby terminating the activating inut to the AND gate 150. The selection of a lO-sample delay is solely for the purpose of illustration and it is apparent that other delay periods would also be appropriate. As noted earlier the delay time in the playback fast clock burst must be less than the time of the repeated samples in a given segment of time-compressed audio information and yet must be long enough to allow for decay of unacceptable transients.
The AND gate functions similarly to the AND gate 140 in that it develops an output pulse of a duration corresponding to 512 fast clock input pulses which output pulse is transmitted through OR gate 142 to enable AND gate 144 to gate a fast clock burst of 512 pulses to the OR gate 32in the embodiment of FIG. 1. The control signal for the recirculating switch 28 as generated by AND gate 140 is maintained for the entire fast clock burst generated during the record mode.
What we claim is:
1. In a method of segmenting a continuous signal to permit the subsequent reconstruction of the continuous signal without introducing transients developed during processing of the segmented signals, the steps of:
dividing a continuous signal into discrete segments,
each of said segments being comprised of the plurality of signal samples, and repeating one or more of the last signal samples of a signal segment as the first signal samples of the next signal segment, the number of signal samples repeated corresponding to a time duration greater than the time duration of transients introduced during the processing of the segmented signals.
2. A method as claimed in claim 1 further including the step of time compressing said signal segments for storage or transmission.
3. A method as claimed in claim 1 further including the step of reconstructing said continuous signal by reproducing said segments in a serial, continuous manner without repeating the portion of each segment corresponding to said transients.
4. A method as claimed in claim 1 wherein said conof audio information from audio bandwidth to video tinuous signal consists of audio information. bandwidth for transmission and storage on video appa- 5. A method as claimed in claim 4 further including ratus. the step of converting the bandwidth of the segments