David: I saw your original posting on trans-hybrid loss, and read a few of the replies. I have some comments to add:
Thans-hybrid loss is poorly understood by many modem designers (both hardware and software). I have seen a lot of effort expended to optimize this parameter against an unrealistic "reference termination." This effort is largely wasted. The range of possible line impedances that can be encountered in the field is truly amazing. Scatter plots of actual measurements on a sample of loops were published in Bellcore ST-TSY-000041, which summarized the results of the 1983 loop survey. Another useful reference is TIA/EIA TSB37-A, where tables A.1 to A.7 show the complex impedance of each of the seven test loops, as a function of frequency. I believe that the only realistic way to set up the trans-hybrid loss a modem DAA is to test it against the full range of actual loop conditions that are likely to be encountered. The seven test loops in TSB-37A could certainly be used as one representation of the range of likely terminations. I use a simplified procedure with five terminations: 1) 600 ohms 2) 900 ohms 3) 900 ohms plus 3 kft of 26 gage loop 4) 900 ohms plus 9 kft of 26 gage loop 5) 900 ohms plus 15 kft of 26 gage loop I adjust my fixed-compromise solution to achieve the best overall trans-hybrid loss over these five conditions. Using this method, I find that exceeding 10 dB THL over all five conditions is about all one can expect to achieve with most transformers. (I define THL as the transmit level at the receive input, referenced to the transmit level on tip/ring, with the receive gain normalized to 0 dB from tip/ring. Some other designers use different definitions). For the remaining 60 to 70 dB of THL required by a V.34 modem, you should rely on a good implementation of the near echo canceler in the DSP. If the near canceler is done well, it will provide 60 to 70 dB of cancellation, and will be very tolerant of the frequency slope that you mention in your posting. That's the beauty of DSP adaptive echo cancelers. In summary, I think that the design goal for your DAA should be to achieve 6 to 10 dB THL over all loop terminations. The DSP should do the rest. If your software developer feels that this is too much to ask of the DSP, try taking a look at some of the existing commercial modem chip sets. I won't name names here, but I know that one of the leading vendors of V.34 modem chip sets has designed their chip set to operate well with only 6 dB of THL in the DAA. I am under the impression that other leading vendors have similar performance. I hope this helps you to resolve your design goals for the DAA. If anyone on treg has other opinions/experiences related to this subject, I would be interested to hear them. Joe Randolph Telecom Design Consultant Randolph Telecom, Inc.
