|Publication number||USRE40497 E1|
|Application number||US 09/771,010|
|Publication date||Sep 9, 2008|
|Filing date||Jan 26, 2001|
|Priority date||Apr 16, 1996|
|Also published as||DE19781701T0, DE19781701T1, EP0954793A1, EP0954793A4, US5864714, US6108720, WO1997039409A1|
|Publication number||09771010, 771010, US RE40497 E1, US RE40497E1, US-E1-RE40497, USRE40497 E1, USRE40497E1|
|Inventors||Nir Tal, Ron Cohen, Zeev Collin|
|Original Assignee||Silicon Laboratories Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (16), Non-Patent Citations (1), Classifications (13), Legal Events (2)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present invention relates to communication systems and in particular to modems utilizing native signal processing.
Traditional full duplex communications systems art typically implemented using dedicated hardware, such the prior at modem 12 illustrated in FIG. 1. Modem 12 includes a digital signal processor (DSP) integrated circuit (IC) 18 at its core, a random access memory (RAMS) 20, digital to analog converter (D/A) 24, analog to digital converter (A/D) 26 and data access arrangement (DAA) 28. Modem 12 is shown coupled to a host computer bus through bus interface circuitry 16. Host computer central processing unit (CPU) 14 generates the data to be transmitted through modem 12 and receives the data received by modem 12.
DSP based moderns such as the type illustrated in
Although modems such as the one illustrated in
Although the interrupt mechanism is designed to switch the currently running task to a task that handles the interrupt in as little time as possible, in practice it may take a substantial amount of time before an interrupt is actually handled. The time delay may be caused by hardware delays, multiple sources of interrupts in the computer system or priority given to some interrupts over others. The time between the actual occurrence of the interrupt trigger and the begs of execution of the interrupt hailer routine is defined as ‘interrupt latency.’ A typical interrupt latency in the PC environment is in the range of 0 to 5 milliseconds. However, In the PC environment, there is no guarantee that a running program will give up control within a predefined amount of time. Well behaved programs can be expected not to dominate the CPU for an unreasonable time period, however some tasks may hold the CPU resource for a relatively long period of time.
Therefore a straight forward translation of a traditional DSP based modem to a native processing environment (NSP) is very problematic, since it requires the execution of the modem task each sample (or symbol) and the completion of its execution before the next sample (symbol) arrives in order to meet the ‘real time’ operation requirement. In order to minimize the time between a sample (or symbol) arrival and the modem activation, a straight forward implementation would be to generate an interrupt upon the arrival of each sample (i.e., symbol). In such an implementation, the real time constraint may be too difficult to overcome and the modem routine may not be executed on time due to long interrupt latencies, which may result in data loss. This problem can be overcome by designing the modem routine to operate on a buffer of samples rather than on one sample only. A buffer of samples means a longer time period between consecutive calls to the modem routine. The real time requirement in this case is that the time to process m input buffer of samples and to generate an output buffer for transmission is smaller than the time it takes to receive/transmit a buffer.
The buffer operation scheme, however, poses a new problem. It suffers from an inherent delay disadvantage, since a sample received at the beginning of a buffer is processed only after a whole buffer is received. This disadvantage conflicts with some high rate data pumping modem standards, such as the ITU V.32bis 14,400 bps modem standard, which impose strict time constraints for processing the samples and responding to certain signals from the modem located on the other end of the connection. For example, the V,32bis stank contains a ranging stage at the beginning of the modem connection. During this stage, the specification requires a maximum response time of 26.6 ms to respond to the other modem's ranging signal. The minimum turn around time from the signal detection to response transmission must take into account sample acquisltion, processing, interrupt latency and buffer transmission.
The present invention has been developed to alleviate the problems discussed above in the implementation of an NSP based full duplex communication system. These problems include the necessity to be resistant to interrupt operating system and other task related latencies, in addition to opting and being able to operate with to implementation overhead, etc. As discussed above, severe time constraints exist in implementing modems for the public switched telephone network (PSTN) as the transmission bit rate is pushed higher and higher. In most communication systems, the time constraints are typically present during the initial startup of the communication link. In particular, during the ranging phase of the modem connection.
One solution to designing NSP modems capable of implementing higher bit rate standards, e.g., V.32 and V.34, taught by the present invention, is to utilize within the NSP modem buffers of non-fixed size. Allowing the buffer size to vary, allows the NSP software to adapt to the constraints of the standard. Small buffers provide the communication system with short and accurate response times. On the other hand, increasing the buffer size would make the modem processing more robust, reducing the implementation overhead making it more resistant to operating system latencies and giving it greater flexibility in the exact time within the time slice, i.e. between interrupts, that processing can occur without disturbing the data flow. A more robust task is less susceptible to the tasks that are not well behaved, i.e., they hold the CPU for a relatively long period of time.
When the system is in a steady state and can got by with longer response times, it should be able to operate with larger buffers. In such a system, a point in time is reached where the buffer size can be increased without data overruns/underruns or other errors occurring.
In addition, the buffer switching taught by the present invention does not incur any data loss. Switching occurs smoothly and coherently without the loss of any data. Coherency is hereby defied as not losing or towing away any input or output samples.
Accordingly, it is an object of the present invention to provide a system for enabling high bite rate NSP based communications thereby overcoming the problems associated with the prior art.
It is another object of the present invention to provide a system that can vary the size of its data buffers in accordance with the desired latency time period.
Yet another object of the present invention to provide a system that can vary the size of its data buffers without incurring any data overruns or other data errors.
Another object of the present invention is to provide a system which optimizes system resources such as CPU usage, by reducing the relative implementation overhead of a call to the communication signal processing process.
It is yet another object of the present invention to provide a system which optimizes processor usage by using block optimized signal processing techniques which are more efficient for long data blocks.
The present invention discloses as full duplex communication system that utilizes a buffer interface for both signal reception and transmittal. The buffer interface includes a mechanism of switching between buffers of different sizes without losing any input or output samples. The buffer switching based communication system of the present invention collects a group of input samples and places them in a buffer. The system then processes the samples contained in the buffer and generates a buffer of samples to be transmitted. The reception and transmission buffers are typically of the same length. The length of the sample buffer determines the memory that must be allocated to store the buffer and it also determines the latency time. The latency is defined as the minimum communication system response time between the occurrence of an event at an input port to the generation of another event on an output port in response to the event at the input port. The worst case latency is shown to be exactly twice the time it takes to fill a buffer.
During the initial startup sequence a relatively short and accurate response time is required. Thus, during this phase, short buffers are used. However, later on in the long run (i.e., in the steady state), long buffers become more efficient. Therefore, there exists some point in time when the modem switches to a buffer with a different size. The buffer switching occurs without losing any coherency (i.e. without losing any data at the input or output ports).
There is thus provided, in accordance with a preferred embodiment, a method, in a communications system of achieving a balance between processing response time, on the one hand, and robustness to interrupt latency and processor implementation overhead, on the other hand, the method including of the steps of utilizing data buffers having a first buffer size when it is desired to optimize the communication system so as to have quick processing response times, and utilizing data buffers having a second buffer size when it is desired to optimize the communication system so as to be robust to interrupt latency and to have low processor implementation overhead.
In addition, the method other includes the step of providing switching means enabling the communication system to switch between using the buffers having a first buffer size and the buffers having a second buffer size. Also, the size of the buffers is coherently switched without any loss of data and the second buffer size is greater than the first buffer size. In addition, the second buffer size is switched back to a smaller size when modem connection is reinitialized or restarted. The second buffer size can be switched back to a smaller size when a retrain sequence has been initialized, wherein the communication system implements an International Telecommunication Union standard chosen from the group of V.32, V,32bis an V.34.
There is also provided, in accordance with a preferred embodiment of the present invention, a method of implementing a communications system the communications system comprising a transmitter, echo canceler and a receiver, the method including the steps of performing echo cancellation, utilizing the echo canceler, on delayed output samples transmitted during time slice K−1, performing receive processing, utilizing the receiver, on the difference between input samples received from time slice K−1 and samples generated by the echo canceler during time slice K, an performing transmit processing, utilizing the transmitter, to generate the output samples to be transmitted during time slice K+1.
In addition, the transmitter, the echo canceler and the receiver are implemented using a central processing unit (CPU) of a computer and the size of the buffers is coherently switches without any loss of data.
The invention is herein described, by way of example only, with reference to the accompanying drawing;, wherein:
To better illustrate the operation and utility of the buffer switching system of the present invention, the system Is described in the framework of a full duplex voiceband modem. However, it is understood that the example presented throughout this disclosure in no way limits the scope of the present invention, One skilled in the art may take the principles of the system and methods of the present invention disclosed herein and apply them to many other types of full duplex communication systems, those of which that are well known in the art.
A high level functional block diagram illustrating a general realization of a full duplex voiceband modem 30 utilized in the communication system of the present invention is shown in FIG. 2. Modem 30 contains a transmitter 32, a receiver 34, an echo canceler unit 36, a summer 37, a digital to analog (D/A) convener 38, an analog to digital (A/D) convener 40 and a digital access arrangement (DAA) or hybrid 42. Echo canceler unit 36 comprises delay register stack 35 and echo canceler circuitry 41. Transmitter 32 received data from the transmitter data in port and outputs transmits (Tx) samples to echo canceler 36 and D/A 38. DAA 42 functions to match the impedances between the telephone line and the transmitter and receiver. It transforms balanced analog voltage On the two-wire pair from the central office (CO) to two two-wire unbalanced pairs, one for the transmitter and one for the receiver. Echo canceler 36 functions to remove echoes from the received signal by applying standard echo canceling techniques, which are well known in the art, to the transmitted signal. The output echo canceler 36 is summed with the receive signal using summer 37. Receiver 34 outputs a digital receive (Rx) data out signal.
A high level block diagram illustrating the native signal processing (NSP) modem, generally referenced 10, coupled to a host computer is shown In FIG. 3. NSP modem 10 generally comprises a hardware portion and a software portion. The software portion runs on host computer CPU 54.
A DAA 64 forms the physical line interface to the 2-wire pair from the CO (e.g. RJ-11, RJ-45 or any other suitable connection method). The host CPU 54 communicates to NSP modem 10 through bus interface circuitry 56. Two first in first out (FIFO) buffers are used to buffer samples to and from the host computer CPU. A transmit FIFO 58 buffers outbound samples and a receive FIFO 60 buffers inbound samples. A coder/decoder (CODEC) 62 couples a transmit FIFO 58 and receive FIFO 60 to DAA 64. CODEC 62 performs the D/A and A/D functions of D/A 24 and A/D 26 (FIG. 2).
Since the majority of personal computers today run some type of multi-tasking operating system, It will be assumed that host computer CPU 54 is executing some form of mult-tasking operating system. In this case, the system of using buffers of varying size taught by the present invention is used by NSP modem 10 to conform to the strict time tolerances imposed by the modem standards, e.g., V.32 and V.34 standards. To aid in understanding the mechanism of using buffers to varying sizes as taught by the present invention, the V.32 bis standard is used as an example. In particular, the V.32 bis start-up procedure which occurs before data can be exchanged, is described. Illustrated in
To respond within the allotted time frame, short response times are needed. Thus, during this initial phase, short buffers are used. After the training stage ‘TRN’ onward, short and accurate event handling in not an absolute necessity and long buffers may be utilized, thus reducing the CPU context switching task overhead. Experiments undertaken by the inventor, using a PC equipped with a Pentium 100 MHz processor, have shown hat CPU utilization during the ranging period is relatively low, on the order of less than five percent During the training period CPU utilization increases add con exceed thirty percent. During the steady state data portion CPU utilization falls to less than thirty percent.
As previously discussed, during the ranging period of the start-up sequence (
At some point in time, the decision to switch to large buffers is made. In the case of V.32 bis, the switch to large buffers con occur after the ranging period has concluded, preferably before the training period has begun. The system of the present invention is oblivious to the samples contained in the sample buffers. It makes no difference whatsoever what the samples within the buffer are. The coherency characteristic of the present invention is described in more detail below.
The operation of the buffer switching mechanism around the point of switching will now be described in more detail. Assume time slice N-1 is current and a buffer full of samples is received during this slice. Assume also that small buffers are no longer needed and the decision to switch to larger buffers has been made. What entity makes the decision to switch is not relevant to the present invention. During time slice N, the sample received during slice N-1 are processed. However, the processor knows that from the next time slice forward, large buffers are to be used, Thus, the processor generates a buffer of samples to be transmitted that has a length L2 greater than L1. For example purposes L2 is equal to 256. The transmit process is independent of the other components of the system and therefore can produce buffers having any arbitrary number of samples. Thus, a buffer of size 256 samples is generated by the processor during slice N. These samples will be transmitted during the following time slice N+1. Also during slice N, a small buffer of samples is received that are to be processed during slice N+1.
During the next slice N+1, a large buffer of samples is acquired and the small buffer of samples acquired during slice N is now processed. A large buffer of samples is generated for transmission during slice N+2. In addition, the large buffer of samples generated during slice N are transmitted during this slice N+1. Similarly, during slice N+2, the large buffer of samples received during slice N+1 are processed and a large buffer of samples is generated for transmission during slice N+3. The following table shows the buffer sizes used before and after the switching transition.
N + 1
>N + 1
With reference to the table above, the echo canceler processes the received buffer of slice K−1 during slice K, K being any arbitrary time slice. Thus, processing L1 samples in buffers less than or equal to N+1 and outputting buffers of exactly L1 samples. During time slice K, the receive process processes the difference between the buffer received in slice K−1 and the output of the echo canceler in slice K. Thus, in time slice N, the receiver will process the difference between the samples received during slice N−1, having a length L1, and the output of the echo canceler during slice N, also having a length L1. During time slice N+1, the receive process will process the difference between the received samples in time slice N, having a length L1, and the output of the echo canceler during slice N+1, also having a length L1.
To ensure that the receiver gets samples that have had echoes properly removed from them, the transmitted and received samples must be suitably aligned. In other words, the samples used by the echo canceler must be synchronized in time with the samples transmitted and received. This is achieved by placing one buffer delay register 35 (
In order to meet the time constraints of the modem standards, the processing of the data must occur in a pertain order. Illustrated in
Therefore, echo canceler processing on delayed transmitter data is performed first (step 70). Then receiver processing is performed on the difference between the received samples from time slice K−1 and the samples generated by the echo canceler process in slice K (step 72). Finally, transmit processing is performed, generating a buffer full of samples to be transmitted during time slice K+1 (step 74) and stored in delay registers 35.
Thus, utilizing the buffer switching mechanism of the present invention, resistance to interrupt latency can be maximized. In the example provided above, buffers of size 256 at 8000 samples/sec yields 32 ms buffer times. Assuming thirty percent processing period gives a maximum interrupt latency of 22.4 ms, which provides a very large time margin. In addition, a side benefit of using long buffers is lowered CPU utilization in terms of lowered overhead enabling the CPU to perform other functions. The main benefit provided by the present invention is the ability of an NSP modem to conform to the strict time constraints of the higher bit rate modem standards (e.g., V.32 9600 bps) V.32bis 14,400 bps and V.34 28,800 bps).
In addition, if for any reason one of the modems requests a retrain process during a connection, the buffer size can be changed back to a short buffer size so that a retain process can occur At at later time, the buffer size Is switched back to a larger buffer size.
While the invention has been described with respect to a limited number of embodiments, it will be appreciated that many variations, modifications and other applications of the invention may be made.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4823312||Oct 30, 1986||Apr 18, 1989||National Semiconductor Corp.||Asynchronous communications element|
|US4841526||Nov 19, 1986||Jun 20, 1989||Wilson Jon C||Data communications system|
|US5077550||Sep 19, 1990||Dec 31, 1991||Allen-Bradley Company, Inc.||Burner flame sensing system and method|
|US5327428||Oct 7, 1991||Jul 5, 1994||International Business Machines Corporation||Collision-free insertion and removal of circuit-switched channels in a packet-switched transmission structure|
|US5425098||Jun 23, 1993||Jun 13, 1995||Yamaha Corporation||Data transmitter-receiver having echo canceler with data delay|
|US5440740||Oct 13, 1992||Aug 8, 1995||Chen; Fetchi||System and method for managing devices on multiple digital signal processors|
|US5585803||Aug 29, 1995||Dec 17, 1996||Atr Optical And Radio Communications Research Labs||Apparatus and method for controlling array antenna comprising a plurality of antenna elements with improved incoming beam tracking|
|US5864714||Apr 16, 1996||Jan 26, 1999||Comsys Communication & Signal Processing Ltd.||Communication system which dynamically switches sizes of sample buffer between first size for quick response time and second size for robustness to interrupt latency|
|DE3705176A1||Feb 18, 1987||Sep 1, 1988||Siemens Ag||Digital signal converter with nonlinear conversion characteristic|
|EP0340613A1||Apr 26, 1989||Nov 8, 1989||National Semiconductor Corporation||Integrated Modem which operates without a dedicated controller|
|EP0577115A2||Jun 30, 1993||Jan 5, 1994||3Com Corporation||Programmed I/O Ethernet adapter with early interrupts for accelerating data transfer|
|EP0657824A1||Oct 19, 1994||Jun 14, 1995||Advanced Micro Devices Inc.||Apparatus for Ethernet packet reception|
|JPH066387A||Title not available|
|JPH088977A||Title not available|
|JPH06216970A||Title not available|
|JPH06326748A||Title not available|
|1||Kenworthy, "Modems. (Optimizing modem data flow)(World-Class Tune-Up Tips: Hardware Heaven)", CMP Publications Inc., Windows Magazine, v6,n7, pp. 150-152, Jun. 20, 1995.|
|U.S. Classification||713/1, 710/57, 710/260, 710/56|
|International Classification||H04M11/00, H04L13/08, H04L29/10, G06F13/00, H04M11/06, G06F13/14|
|Cooperative Classification||H04M11/06, H04L7/005|
|May 23, 2005||AS||Assignment|
Owner name: SILICON LABORATORIES, INC., TEXAS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:PCTEL, INC.;REEL/FRAME:016580/0323
Effective date: 20050429
|Jun 28, 2010||FPAY||Fee payment|
Year of fee payment: 12