Pulse or digital communications – Cable systems and components
Reexamination Certificate
1998-09-30
2002-06-18
Deppe, Betsy L. (Department: 2634)
Pulse or digital communications
Cable systems and components
C375S258000, C375S295000, C375S296000, C327S307000
Reexamination Certificate
active
06408032
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to the field of digital data transmission, and particularly to a method of compensating for baseline wander in a transformer coupled transmission link.
BACKGROUND TO THE INVENTION
When transmitting data from a transmitter to a receiver, bandpass limitations can cause a phenomenon called baseline wander, which is the shifting of a central e.g. neutral, voltage or current value about which data signal excursions occur, positively or negatively from the proper neutral or central value. As will be described below, this can cause jitter or data loss.
With reference to
FIG. 1
, a schematic diagram is shown as a Norton equivalent circuit, which includes a transmitter
1
comprising a current driver, which may, depending on the network standard, be as simple as an on/off switch or as complex as a linear driver or a Digital-to-Analog Converter (DAC). A receiver
3
(shown as an amplifier) may be as simple as an error amplifier or as complex as a full analog adaptive equalizer. In general, both the transmitter and receiver are terminated by resistances that match the characteristic impedance of twisted-pair cabling
5
which is connected between transformers
7
and
9
which couple to the transmitter
1
and receiver
3
. The transformers are included both to block DC currents and to minimize common-mode coupling from the transmitter onto the cabling, thus minimizing EMI radiation from the system.
In an AC-coupled data transmission systems such as that of
FIG. 1
, there often is the possibility of a phenomenon called Baseline Wander, also called BLW. It is caused by the high-pass nature of the AC coupling, and appears as a data-dependent drift of the theoretical zero-crossing point of the data over time.
FIG. 2
shows the effects of BLW for Non-Return to Zero (NRZ) data.
The top trace in
FIG. 2
is the ideal differential signal that would be expected by the receiver without AC coupling. The second trace shows how this is actually encoded as a pair of differential signals arriving at the receiver. The third trace shows the effects of AC coupling on these differential receiver inputs; as the data pattern density moves from balanced data on the left to unbalanced data (more −1's than +1's), the signals will start to collapse toward one another. As a result, the overall differential signal's baseline will wander, as shown in the final trace.
For non-return to zero (NRZ) data, where there is only a single threshold required to determine whether the signal transmitted was a +1 or a −1, the effects of BLW are comparatively minor. If the receiver slices the data using a slightly wrong threshold, the only effects appear as edge jitter. For multi-level (Pulse Amplitude Modulated or PAM) codes, however, where there are multiple thresholds required to resolve the transmitted symbol, the effects can be catastrophic and result in data errors. As a result, compensation for BLW becomes important.
BLW may be somewhat controlled by maximizing the open-circuit-inductance (OCL) of the transformers used. However, transformers act as band-pass filters. The low-frequency cutoff is controlled by an induction/resistance L/R time constant, where L is the OCL. A larger L provides a lower frequency cutoff.
The high-frequency cutoff is controlled by the coupling coefficient of the transformer and the same L/R time constant. An ideal transformer has a coupling coefficient K of 1, while real transformers have a coupling coefficient somewhat smaller than, but still close to, 1. The high-frequency corner of the bandpass response has a proportional 1/(1−K
2
) term in it, so the closer K is to 1, the higher frequency this corner is. It is difficult to manufacture transformers with wide frequency responses (i.e. K close to 1), so when attempting to transmit high data rates, designers are forced to use transformers with the minimum OCL that they can. As a result, designers are competing against two mutually exclusive requirements: high speed requires low OCL's, while BLW immunity requires larger OCL's or more expensive transformers. By compensating for BLW somewhere other than in the transformer, the designer is free to use less expensive low-inductance transformers to achieve high speeds.
The simplest way to compensate for BLW is to choose a transmission coding scheme to remove all spectral components from the transmitted data that are below the high-pass limit of the transformer coupling. Among others, transmit line coding techniques known as Return-to-Zero Alternate Mark Inversion (RZ-AMI), Modified Frequency Modulation (MFM, also known as Manchester Coding), and the technique described in U.S. Pat. No. 5,200,979 all do this at the expense of increased bandwidth or required signal to noise ratio (SNR). The MLT
3
A technique described in U.S. Pat. No. 5,655,078 also removes low-frequency spectral components, but re-uses the increased SNR required by MLT
3
and so is not quite as expensive as the other techniques.
Another way to compensate for BLW is to scramble the data with a known Pseudo-Random Bit Sequence (PRBS) before transmitting it, then descramble it at the receiver. This technique is based entirely on statistical arguments based on the length of the PRBS sequence, since there is a small but finite probability that transmitted data will exactly match up to the inverse of the PRBS sequence over a long enough period of time to induce a BLW event.
In the Fiber Distributed Data Interface (FDDI) standard, there is a known valid “killer packet”, that, if transmitted, will induce a BLW event that will break the network in the absence of BLW correction. Other standards, such as 155Mbit/s ATM and 100 BaseT Ethernet may also suffer from BLW “killer packets”, but this is a matter of some debate.
Yet another way to compensate for BLW is to include BLW correction circuitry in the receiver. This requires that there be circuitry built to detect the BLW event and add an offset to compensate. Depending on the complexity of the rest of the equalizer, this may be easy or difficult. As an example of the trade-offs required,
FIGS. 3A
,
3
B and
3
C show three possible BLW correction loop configurations with an adaptive equalizer
11
receiver. The first, in
FIG. 3A
, shows a BLW correction loop
13
applied before equalization. This is, in theory, the simplest to conceive; however the BLW correction algorithm in the loop needs to extract information from the signal before the equalizer, so any correction algorithm that assumes a known decision out of the equalizer is impossible.
The second configuration, as shown in
FIG. 3B
, wraps the BLW correction loop
13
around the adaptive equalizer
11
, in effect embedding the adaptive equalization loop as an inner loop of the BLW loop. It has the advantage over the first configuration in that now the BLW correction algorithm may use the information coming out of the equalizer as its input. However embedding the adaptive algorithm loop as part of the BLW loop makes stability analysis of the BLW loop challenging.
The third configuration shown in
FIG. 3C
puts the BLW correction
13
after the adaptive equalizer
11
. This has none of the difficulties of the first or second configuration, but requires that the entire equalizer have an increased dynamic range (as much as twice if the transformer voltages completely decay) to accommodate the extremes of a BLW event. As supply voltages are reduced with smaller-geometry microelectronic processes, this linearity becomes harder and harder to accomplish.
Yet another technique i$ to include BLW correction circuitry into the transmitter. Because in this technique the BLW correction circuitry is a single loop, not wrapped around an adaptive equalizer, it is much simpler to analyze and design than the receiver BLW correction circuitry of FIG.
3
B. At the same time, because the transmitter comes before the cable
5
, the BLW correction algorithm already has information about what signal is being transmitted and doesn't have the difficulties
Candage Anthony B.
Lye William
Baker Harold C.
Deppe Betsy L.
Hendry Robert G.
PMC-Sierra Ltd.
Wilkes Robert A.
LandOfFree
Transmit baseline wander correction technique does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Transmit baseline wander correction technique, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Transmit baseline wander correction technique will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2979906