RS> Yes, that approach has some theoretical validity, BUT we ALREADY know that
RS> the Rx level the M34F claims to have seen with one particular AT command
RS> is dud coz it says 50dbm all the time. Clearly the M34F has problems too.

PE> And we can bin the 50dbm result too.  We can recognize it and bin it.

RS> Yes, but YOU want to just bin ALL the USR data, even when there is
RS> another USR on the other end of the line. For some reason you only
RS> bin SOME of the M34F data, the data with the obvious problem.

PE> Well it is obvious which bit of the M34F data is stuffed.

Well, its certainly obvious that an Rx level which never varys from
call to call must be stuffed. Its less clear if the OTHER reports of
the Rx level by M34F modems make sense with different modems on the
other end of the line, particularly with the same line in use.

PE> What is not obvious is which bit of the USR data is stuffed.

Thats going too far. It is clear that USR/USR stats, particularly the
full graph of the line characteristics, does make lots of sense. When
we know that particular lines like Poes have changed in particular ways,
by increasing the length of line to the exchange, and the graph does show
the increased loss at higher frequencys that is supposed to happen on the
longer lines, it is clear that the numbers do make considerable sense.

The only real problem with them is that you get a pronounced shift of the
entire curve by about 6dBm when you change the modem at the other end from a
USR to a Rockwell. Its currently not clear if the USR is at fault, having a
brain fart with a non USR at the other end for some reason, or whether the
Rockwell is putting out the standardised test thats used to produce that
graph with transmit levels that are a full 6dBm too low.

One obvious test worth doing is to call a Motorola, coz one that doesnt use
the Rockwell datapump shouldnt show the same effect if its a Rockwell quirk.

PE> I can see what you mean about using
PE> the relativity only, on dual-USRs though.

Yeah, the data does make lots of sense as a measure of the line in that config.

And the other thing about the USR graph is that there is a LOT more
information in it that the M34F provides, even if the M34F figures
are accurate, the Rx level and EQM are very minimal figures to use
to characterise the line itself. Particularly when EQM suffers from
the problem that the number varys with the symbol and carrier rate.
@EOT:

---