[H-GEN] 24/7 56k modems

[Bruce Campbell]

On Wed, 21 Oct 1998, Rob Kearey wrote:

> David Shackleton wrote:
> 
> > Andrew's comments reflect a number of responces I've received regarding
> > X2's superiority as a protocol - what I'm concerned at, is if I buy the
> > v.90/x2 Sportster, can I still dial into a Kflex bank due to the v.90?  I

*provided* that the Flex bank supports V.90 as well.  Otherwise you'll
drop back to V.34 (33.6K)

Food for thought:  (this is from le big troubleshooting guide for ours and
UQ's 56K equipment.)

x2 Issues

There are several issues that one MUST be aware of when dealing with x2
calls.  The first deals with connection quality.  The x2 protocol is
extremely optimistic in making a PCM mode connection.  This is, it will
train up into a high speed connection under conditions that are far from
favorable for this modulation.  This results in a high level of extremely
poor x2 connections that should have remained V.34.  These poor x2 calls
can experience a high level of retransmissions, renegotiations, and
retrains.  It is common for these calls to underperform a V.34 call under
the same conditions.  In addition, many of these calls end with unexpected
disconnects including retrain timeouts, carrier lost, and telephone
disconnects.  In these situations, the user is advised to disable x2 on
their modem.  Consult the modem's user's manual for instructions on how to
do this. 

A second issue deals with connect speeds.  In any x2 call, the modem
ALWAYS initially trains at the lowest possible x2 speed of 33.3333 Kbps
and then ramps up from there.  Since it would be meaningless for the
client modem to report this speed in its connect message, the modem makes
a ``guess'' at the final speed based on channel probing information and it
is this speed that is reported.  This guess is often extremely optimistic
and is often high by 3 or 4 Kbps.  The Annex reports the initial speed as
33.333 Kbps and the actual or final speed, depending if it is reported by
the Actcall or Histcall statistic. 

The actual speed of the modem is further modified by the ``spectral
shaping bits'' as reported by the Annex.  These bits, as described in
Annex 1, reduce the actual connect speed by 1.333 Kbps / bit.  As a
result, the actual transmission speed of an x2 connections be up to 10
Kbps less than what is reported by the client modem. 

K56flex Issues

Unlike x2, K56flex modems are quite conservative in making a PCM
connection.  So conservative that users who experience high quality V.34
connections may find themselves unable to achieve a K56flex connection. 

In addition, several versions of K56flex exist, each incompatible with the
other.  The BayDSP platform supports V1.1 of the K56flex specification. 
Modems that support V1.0 and older protocols have been shipped in great
numbers and will all be unable to make K56flex connections to a BayDSP
platform.  In these cases, the user should be encouraged to upgrade their
modem through its manufacturer's web page.  Note:  The manufacturer's
software version number will NOT correlate to the K56flex version number. 

As in x2, the speed reported by the modem will differ from that reported
by the Annex.  As is shown in Appendix 1, this difference is based on the
number of robbed bits distributed throughout the network.  The client
modem reports the actual final throughput speed, or ``Actual Data Rate.''
The Annex reports the ``Nominal'' data rate as well as the number of
robbed bits/RBS frame.  From this, the Actual data rate can be calculated.