In a message dated 97-02-18, Vic Boersma writes: << Joe, after all this, what does it take to provide the user with a good grade of service ??? >>
Vic: I hesitate to venture a response to your question, because the debate about the "required" amount of return loss, and the "correct" reference termination sometimes reaches the proportions of a religious debate. Having said that, I will now make the following remarks, and then stand back to see what happens. 1) Note that FCC Part 68 contains no requirement for return loss of a simple one-port TE. As a result, there are a lot of TEs connected to the USA network that have pretty poor return loss. Nevertheless, my observation has been that the USA network works pretty well. 2) From the perspective of network harms, the principal goal is to ensure that oscillation and singing do not erupt on a given connection. In theory, this is ensured as long as the return loss is less than 0 dB (no gain in the reflected signal). 3) From the perspective of the (voice) user, the principal goal is to ensure that the connection does not sound "hollow." The "hollow" effect is a combination of reflection and delay, so it is difficult to pin a requirement on return loss alone. Connections will noticeable delay *usually* have higher than average loss, so the need for high return loss is mitigated to some extent on such connections. 4) From the perspective of an echo-canceling modem (V.32, V.34, etc.) poor return loss of the far end modem can result in larger far end echo. Most modems can handle far echo pretty well. In my experience, modems have more trouble with the near echo (trans-hybrid loss), which is only indirectly related to return loss. In summary, I do not think that return loss is especially critical for modem performance. 5) Regarding the "correct" reference impedance, most telecom engineers recognize that 600 ohms resistive is a pretty poor approximation of what a TE typically sees looking into the phone line. Real phone lines exhibit a capacitive element, and the various complex references (such as the one in NTR 3 and TBR 21) attempt to account for this. However, if you look at scatter plots of impedance vs. loop length, and impedance vs. frequency on real phone lines, you will see that real world impedances are literally all over the map. Thus, while the complex reference impedances are perhaps a better approximation than 600 ohms resistive, they are not THAT much better. So, in summary, I think the whole return loss thing is frequently overblown. My feeling is that a return loss of 6 dB or so is probably fine for the vast majority of applications. As for the "correct" reference impedance? I'm not sure it matters whether we use 600 ohms or a suitable complex reference. In one sense, I think the requirement proposed in TBR 21 was brilliant. TBR 21 called out a complex reference, which helped to satisfy the purists who insisted that the reference impedance must be complex. However, TBR 21 also set the required return loss at only 6 dB, so that even a 600 ohm resistive impedance would comply when tested against the complex reference. That way, the people who felt that TE should present 600 ohms could continue to use 600 ohms. Everybody won. Of course, not everyone liked the proposed spec in TBR 21. One of the reasons that the Germans cited for voting "no" on TBR 21 was that they felt the return loss requirement was too lax. They wanted at least 14 dB in the DTMF band, presumably to improve DTMF tone detection at the central office. I do not know the technical basis for their argument. Well, Vic, there's my two cents on this subject. I suspect that we will hear from others out on treg who disagree with my perspective on this. Joe Randolph Telecom Design Consultant Randolph Telecom, Inc.
