Pulse or digital communications – Pulse code modulation
Reexamination Certificate
2000-09-08
2002-10-08
Corrielus, Jean (Department: 2631)
Pulse or digital communications
Pulse code modulation
Reexamination Certificate
active
06463103
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to data modems, and more particularly to pulse coded modulation modems which operate at rates as high as 56K bits-per-second.
DESCRIPTION OF THE PRIOR ART
The next generation in modem technology operates at a 56K bit-per-second (KBPS) data rate. Computer users are expected to aggressively seek upgrades that provide ever faster access to the Internet. Pulse code modulation (PCM) 56K BPS modems are now expected to flood the personal computer (PC) market faster than any previous generation.
Historically, telephone networks were intended to carry analog voice traffic. Therefore, equipment was designed to create, transmit, and process analog signals. As computer and digital transmission technologies have advanced, nearly all equipment installed now in new facilities is digital.
Almost every long distance telephone communication now uses digital transmission on the majority of their lines, because it offers better transmission quality. But since voice in its natural form is analog, it is necessary to convert these. In order to transmit analog signals over digital facilities, to capitalize on its numerous advantages, analog signals are converted to digital signals.
PCM is a conversion process typically used by coder/decoder (codec) devices. PCM is a method of converting analog signals into binary ones and zeroes suitable for digital transmission. At the receiving end of such digital transmissions, the digitally coded signals are reconverted into the kind of analog signals that sound like voices to the listener.
PCM uses a three-step process that begins with sampling. The analog voltage levels in the voice signals are sampled at discrete time intervals, according to Nyist's Theorem. If an analog signal is sampled at twice the rate of the highest frequency it attains, the reproduced signal will be a highly accurate reproduction of the original.
The highest frequency used in voice communications is 4000 Hz, so a signal sampled 8000 times per second, will preserve the voice quality of the speaker such that a listener will be unaware of the digital sampling in between.
The next step is quantizing. Since analog signals are continuous, an infinite number of voltage levels are needed to perfectly describe any analog signal. In practice, each analog sample is rounded to the nearest of 256 predetermined voltage levels by quantizing. Quantizing assigns one of 256 amplitude levels to each sample. The amplitudes of samples do not all exactly match their original, but are close enough so only a small quantizing error occurs that can create an audible noise over the transmission line.
Low-amplitude signals are more affected by quantizing error than are high-amplitude signals, so a method called companding is used. Companding reduces the effects of quantizing errors on the lower amplitude signals where the effects are greatest. This is paid for by increasing the error on high amplitude signals where the quantizing errors effect is minimal. But even with companding, the sampling rate remains the same at about 8K samples per second.
Two common companding formulas are used in different parts of the world. The United States and Japan use a companding formula called Mu-law (&mgr;-law). In Europe, and other areas of the world, the so-called A-law is used. Although the two appear to differ only slightly, they are nevertheless incompatible, e.g., &mgr;-law hardware cannot be mixed in use with A-law hardware.
Encoding is the third step and converts each of the 256 possible numeric amplitude voltage levels into a corresponding binary 8-bit digital code, e.g., the final PCM ready for digital transmission.
At the receiver, the transmitted digital bit stream must be converted back to an analog waveform, in order to be recognized as a voice to the listener. Such digital-to-analog conversion (DAC) is essentially the reverse of PCM, and involves decoding, reconstruction, and filtering.
Decoding converts the 8-bit PCM code into PAM voltage levels. Reconstruction reads the converted voltage level and reproduces the original analog signal. The decoding process creates unwanted audible high frequency noise in the 4000-8000 Hz range. So a low-pass filter is needed to block the frequencies higher than one-half the sampling rate, e.g., 4000 Hz.
A PCM modem transmitter transmits user data bits by encoding them into PCM code symbols. These PCM code symbols are fed into a digital network channel. The digital network channel carries these PCM code symbols to a site near the end user where they are converted into analog waveforms as if they are voice samples. These analog waveforms reach a user's site through a physical communication channel. At the user's site, a PCM modem receiver processes these waveforms to recover the PCM code symbols transmitted from the PCM modem transmitter. These recovered PCM code symbols are then decoded to recover user data bits.
A physical communication channel usually does not transfer signals at different frequencies equally well. For example, some communication channels have a DC block capacitor or transformer, therefore they cannot pass the DC component of the signals. Yet, some other communication channels have band limitation, therefore they cannot pass a signal component above a certain frequency. For all these communication channels, some kind of spectral shaping is normally required on the transmitted modem signal to use a better part of the channel in order to improve the modem performance.
A conventional way to do spectral shaping is through modulation and pulse shaping filtering, e.g., using a modulating carrier to shift the signal spectrum to a desired frequency band and using a filter to define the bandwidth and shape of the transmitted signal spectrum.
The modulation and filter approach may not always be feasible or desirable, such as in the case of a PCM modem. For a PCM data stream sampled at 8000 Hz, a minimum of 4000 Hz bandwidth (Nyquist rate) is needed. Since the telephone channel bandwidth itself is only 4000 Hz, no bandwidth margin can exist.
In almost every modem it is advantageous to avoid sending low frequency signals near DC, because such signals tend to suffer more from distortion and interference because of the non-linear way that transformers operate and power lines interfere. Such low frequency signals also cannot pass through transformers in the hybrid or the subscriber line card at the central office.
One way to measure the amount of the DC component in the transmitted sample stream is through the use of a running-digital-sum (“RDS”). For a given sample in a PCM sample sequence, the running-digital-sum is defined as the algebraic sum of all samples up to and including that sample.
Several methods recently suggested for DC-suppression in a PCM modem all use sample inversion. Such methods first group PCM code stream samples into segments of equal length, and then inverts selected ones of the PCM codes to balance the DC component on-the-fly. Information about which particular PCM codes are inverted is appended to the user data for the receiver use during decoding. The receiver uses the inversion flag information to recover the original polarity of each PCM code. Including such overhead data for the inversion information erodes the maximum user data rates possible.
In one prior art method, a fixed sample in each segment, such as the first sample in each segment, is selectively inverted to balance out DC. In another prior art method, a best sample in a segment is selected, and selectively inverted to balance out DC. In yet another prior art method, the sample with maximum magnitude is selectively inverted to balance out DC. All prior art methods allow trade off between the amount of redundancy and spectral shaping capability.
In such prior art sample inversion methods, at least one redundant bit is required to flag the inversion of each sample in a segment. More redundant bits are needed when the inverted-sample locations in a segment must be pointed to. So it is highly desirable to use schem
Dong Ping
Lin Kuo-Yu
Bayard Emmanuel
Corrielus Jean
ESS Technology, Inc.
Stevens & Westberg LLP
LandOfFree
Frame-based sign inversion method and system for spectral... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Frame-based sign inversion method and system for spectral..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Frame-based sign inversion method and system for spectral... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2991378