Index Home About Blog
From: (Floyd Davidson)
Subject: Re: V.90 Server Symmetrical 56/64K mode (was Re: Modem's can't connect 
Date: 24 Dec 1998 09:00:49 GMT

Rhodes <> wrote:
>Floyd Davidson wrote in message <75s07r$>...
>>Jeffrey Rhodes  <> wrote:
>>>Well, I wonder? If Robbed Bit Signaling is the only impairment, and
>>>since RBS, on average, only compromises one of 96 bits, I wonder if
>>I'm not sure I see the average of 1 in 96.  More like a minimum of
>>of 1 in 48 when RBS is used?
>>  Floyd
>>Floyd L. Davidson                      
>>Ukpeagvik (Barrow, Alaska)
>Hi Floyd,
>I am counting every other robbed bit to have a 50-50 chance of being correct
>in the first place, hence only every other robbed bit is not really robbed
>and all the others remain intact ;-)

Won't fly.  If we test the receive end to determine which ones
were good or bad (using CRC or parity for example), what do we
do with the bad ones????  Resend all 8 bits!  Ooops, that means
instead of being able to use half the bits, instead we actually
are encumbering 4 times as many bits.  There goes the efficiency
of that method...  :-(

Instead it is more useful to test frames to determine which ones
are losing bits to RBS, and then just using 7 bit symbols on those

>I guess you are correct about the minimum, since there is no practical
>method to effectively take advantage of an uncompromised robbed bit?

Well, the minimum is because each facility link robs its own
bits and has no idea which frame other facilities are robbing
from.  With a little bad luck that mean 6 different frames on
the first 6 different facilities... which is every single frame
experiencing a lost 8th bit. (Instant 56K circuit instead of
64K!)  On short haul connections like most PCM modem users are
probably getting, there might not even be a total of 6 different
facilities, so the likelihood of seeing that happen is not so
bad.  But variations in robbed bit signaling frame selection is
one (of many) reasons that some users get such a variety of
connection speeds for no apparent reason.


Floyd L. Davidson                      
Ukpeagvik (Barrow, Alaska)

From: (Floyd Davidson)
Subject: Re: V.90/91 Server Symmetrical 56/64K mode
Date: 4 Jan 1999 15:33:58 GMT

Rhodes <> wrote:
>As Laurence points out, whatever the robbed bits are, they will
>remain the same value whilst the call is answered,
>eg. "off-hook".

That is not a good assumption.

There are, and will be for many years, a goodly number of
facilities using equipment which cannot be configured to disable
robbed bit signaling.  When those facilities are used on
circuits that have common channel signaling the signaling inputs
can end up floating, and will "chatter".  On leased line
circuits which connect to equipment with relays, the noise
generally causes someone to notice and find a way to force the
signaling bits on or off.  But on switched trunks there is
nothing to require that that be done.  It is not uncommon at all
to plug a test set into DSX jacks and see the supervision
indicators randomly pulsing.

Hence, on some connections it is very likely that robbed bit
signaling will indeed be detected, but it won't be actually in
use for signaling and therefore will not necessarily remain
steady state.  Two different considerations are of concern.
One is that a simple one time sampling of a given frame to
verify that cycling bits on and off follows input to output is
not enough to determine if robbed bit signaling is used.
Several cycles need to be monitored.  And of course it also
means that once robbed bit signaling is in fact detected, that
bit must be forced on or off if it can make any difference


Floyd L. Davidson                      
Ukpeagvik (Barrow, Alaska)             
  Pictures of the North Slope at  <>

Index Home About Blog