[comp.dcom.lans] DESPR performance problems

crosson@cam.nbs.gov (Bob Crosson) (03/28/89)

Your problem may be with circuitry in both boxes that test 'signal
quality', or SQ.  SQ circuitry causes a JAM signal for about 0.6
microseconds after each packet that is transmitted by the box (your
DELNI).  This usually doesn't hurt anything because the Ethernet
standard requires that everybody wait during that 0.6 microsecond
before attempting to transmit.  The jam is used to test the
collision detect circuitry in the transceiver.

SQ circuitry has to be disabled in regular transceivers that
connect repeaters to Ethernet cables.  This is because the repeater
is not really the transmitting device, it's just repeating what
it saw on the other cable.  But its transceiver doesn't know that
and will try to jam at the end of the packet if its circuitry is
enabled.  If the circuitry is enabled in the repeater's transceiver,
then when the real transmitting transceiver jams, the repeater
repeats that jam, and then adds a jam of its own.  The second jam
can collide with the packet being transmitted next no matter who is
sending it.

Of course, the SQ circuitry can be disabled on all transceivers
without hurting anything.  It can also be disabled on some with no
effects that I've seen.  But it must be disabled on transceivers
(your DELNI in your case) that connect repeaters to Ethernets.

I don't know whether you can disable SQ circuitry on DELNIs.  Try
putting the DESPR on its own transceiver, or a transceiver that has
the SQ circuitry disabled.  Your local DEC rep. might know how to do
this.

smb@ulysses.homer.nj.att.com (Steven M. Bellovin) (03/29/89)

In article <176@fs1.cam.nbs.gov>, crosson@cam.nbs.gov (Bob Crosson) writes:
> Of course, the SQ circuitry can be disabled on all transceivers
> without hurting anything.

Be warned that some devices -- for example, the DEUNA -- get upset
if they don't see SQ.