Rod, at 08:08 on Apr 18 1996, you wrote to Tom Bulum...

RS> Its possible tho, there clearly is a major problem with the USR line
RS> probe stats when the modem on the other end isnt a USR, the entire curve
RS> shifts the best part of 7db and that may well be confusing the USR when
RS> it attempts to set the Tx level to the optimum level. In other words it
RS> gets confused about the line loss in that case and boosts the Tx level
RS> in the auto level setting with a non USR modem at the other end. That
RS> could explain why the effect with USR/USR is different to USR/non USR
RS> on the manual override of the Tx level.

The following stats were generated by my USR V.34+ Sportster during
connects with Paul's M34F, David's test-bed M34F, and David's Courier
respectively, and although they'd already been posted in Locuser a couple
of weeks ago, they're highly relevant to this current thread as well, and
at V.34 frequencies, the entire curve shift to Rockwells is indeed around
the 6dB mark.

---------------------------------       Legend: PE = Paul Edwards
Freq | PE M34F | DD M34F | DD USR               DD = David Drummond 
---------------------------------
 150      39        42        41
 300      28        29        25
 450      26        27        23
 600      25        26        22
 750      25        26        22
 900      25        26        22
1050      26        26        22
1200      27        27        22
1350      27        27        22
1500      28        28        23
1650      29        28        23
1800      30        28        24
1950      31        29        24
2100      31        29        24
2250      32        30        24
2400      33        30        25
2550      34        31        25
2700      34        31        26
2850      35        32        26
3000      36        33        27
3150      38        33        27
3300      39        35        28
3450      43        38        30
3600      48        43        33
3750      57        52        40

----------------------------------------------------------------------------
Connect Statistics       Paul's M34F     David's M34F    David's Courier
----------------------------------------------------------------------------
Modulation               V.34            V.34            V.34+
Carrier Freq (Hz)        1829/1829       1959/1959       1959/1959
Symbol Rate              3200/3200       3429/3429       3429/3429
Trellis Code             64S-4D/16S-4D   64S-4D/16S-4D   64S-4D/64S-4D
Nonlinear Encoding       ON/ON           ON/ON           ON/ON
Precoding                ON/ON           ON/ON           OFF/OFF
Shaping                  OFF/ON          ON/ON           OFF/OFF
Preemphasis (-dB)        6/8             6/6             4/4
Recv/Xmit Level (-dB)    35/11           30/13           25/14
Roundtrip Delay (msec)   20              4               3
Data Compression         V42BIS 2048/32  V42BIS 2048/32  V42BIS 2048/32
Equalization             Long            Long            Long
Fallback                 Enabled         Enabled         Enabled
Protocol                 LAPM 244/6      LAPM 244/6      LAPM 244/15
Speed                    26400/26400     28800/28800     33600/33600
-----------------------------------------------------------------------

It can be clearly seen from these stats that as the available bandwidth
decreases, the Tx level increases automatically to compensate, but why the
entire curve shifts ~6dB is beyond me, I'm afraid.

Regards, Bill

--- Msgedsq/2 3.20

Regards, Bill

--- Msgedsq/2 3.20

--- Squish/386 v1.11