In a response to Rene Debets dated 97-03-09, Duane Marcroft writes:

<< I do not like the local loop simulators supplied by TAS or PTT.  In my
 opinion they are a poor representation of real world local loops. They are
 sourced at 600 ohms and should be sourced at 900 ohms.  I don't use them
 unless a client insists they must be used.  In my opinion TSB-37A testing
 makes poor and good modems look good.  >>


Duane and Rene:

I've been following (a bit delayed) the dialog that you and Rene have been
having regarding return loss and trans-hybrid loss.  While it is true that
the TAS simulator uses a 600 ohm termination, I'm not sure that it makes as
much of a difference as you may think.  

On short loops (0 kft to 3 kft), switching the termination at the central
office between 600 ohms and 900 ohms does make a noticeable difference in the
impedance seen by the TE.  However, on longer loops, the impedance of the
loop dominates what the TE sees, and changing the termination at the central
office has very little effect.  At least, this has been my observation in
laboratory tests on trans-hybrid loss.

So, unless the differences that you see between "real world" performance and
performance on a TAS are on very short loops (e.g., loop 1 in TSB 37A), I
suspect that the 600 ohm termination in the TAS is not the culprit.  

In my experience, most modems tend to have more trouble on the longer loops
in TSB-37A (loop 4 and loop 5).  On these loops, the receive signal level is
lower, there are larger loop-related impairments, and the impedance mis-match
is usually greater.  When these loops are used, the termination at the
central office end (be it 600 ohms or 900 ohms) has very little effect on
what the modem sees.

So much for my opinions on TAS vs. the "real world."  After reviewing the
correspondence that has taken place on the issue of modem performance, I can
summarize my views as follows:

1) There is no compelling case for modems to have high return loss, whether
it be referenced to 600 ohms, 900 ohms, or a complex reference.

2) Trans-hybrid loss is much more likely to affect modem performance than
return loss.  While return loss and trans-hybrid loss are related, they are
NOT measures of the same parameter.  

3) As Duane has pointed out, the main reason that return loss becomes an
issue for modem performance is that inexperienced DAA designers sometimes
make arbitrary changes to improve return loss (a parameter that is subject to
regulatory requirements), without checking to see what effect these changes
have on trans-hybrid loss (a parameter that is not subject to regulatory
requirements).  Sometimes, the effect is catastrophic.

4) As both Rene and I have noted, a good near echo canceler in the modem can
make up for poor trans-hybrid loss.  In theory, with a perfect D/A, perfect
A/D, and infinite precision in the near canceler, it is not necessary for the
DAA to provide any trans-hybrid cancelation at all.  The extent to which
cancelation is required in the DAA relates to how non-ideal these resources
are.

5) My impression is that in many commercial chip sets, these resources have
improved in recent years.  I am familiar with one commercial chip set that
seems to perform well as long as the trans-hybrid loss (THL) is greater than
6 dB.  However, the "THL tolerance" will always be implementation-dependent,
so greater THL may be required for certain vendors or certain implementations
(Duane: can you venture an opinion about the required THL for Rockwell
modems?).



Joe Randolph
Telecom Design Consultant
Randolph Telecom, Inc.






Reply via email to