[comp.dcom.modems] What all does S120=16 do?

gandrews@netcom.COM (Greg Andrews) (03/17/91)

In article <1991Mar16.160708.152892@ramona.Cary.NC.US> andrew@ramona.Cary.NC.US (Andrew Ernest) writes:
>In article <9908@discus.technion.ac.il> devil@techunix.BITNET (Gil Tene) writes:
>>To detail : S120=12 means no use only LONG packets, no micro
>
>What are the side-effects of S120=16?  This is the value Telebit
>says turns off the short burst of V.25 tone at the beginning of
>each PEP answer tone.  This tone, which appeared in the BC5.10A
>firmware upgrade for my TB+, caused my neighbors with older firmware
>to no longer be able to connect to my modem.  S120=16 solved the
>problem but I am now curious if that register value (16) has any other
>effect I should know about (like packet size, etc).
>

It turns off all of the mods put into the PEP modulation to combat trouble
with echo cancellation devices on fiberoptic long distance carriers.
One of the changes was a 3 second V.25 answer tone when the modem first
answers the phone, even if PEP tones are first.  Previous versions of
firmware didn't put out the V.25 answer tone before the PEP tones.

>
>Can anyone guess why the V.25 tone was added?  Any reason why turning it
>off is a bad thing?  (Not that I have any intention of turning it back
>on...compatibility with my PEP neighbors is of the most importance to me.)
>

It was added to better conform to the V.25 recommendation, which says
an automatically answering modem transmits a 2100 Hz tone when it goes
off hook.  Part of the reason for the V.25 answer tone is linked to the
reason PEP was modified for the fiberoptic carriers - the tone helps to
disable the telco's echo cancelling and/or echo suppressing equipment.

PEP was having trouble over those types of lines, so it was modified to
compensate.  The V.25 answer tone was only part of those mods.  Setting
S120=16 disables ALL the mods - not just the V.25 answer tone.

The V.25 answer tone probably isn't the problem.  It's more likely to
be due to how the mods work when you have PEP tones last.  If PEP tones
are first, only the oldest TrailBlazers have trouble with the V.25 answer
tone before PEP, and many of those can set S50=255 to overcome it.  When
PEP tones are last, it's a different story.


-- 
.-------------------------------------------.
| Greg Andrews      |   gandrews@netcom.COM |
`-------------------------------------------'

root@minixug.mugnet.org (MINIXUG-ONLINE System Manager) (03/23/91)

gandrews@netcom.COM (Greg Andrews) wrote:
> In article <1991Mar16.160708.152892@ramona.Cary.NC.US> andrew@ramona.Cary.NC.US (Andrew Ernest) writes:
>>In article <9908@discus.technion.ac.il> devil@techunix.BITNET (Gil Tene) writes:
>>>To detail : S120=12 means no use only LONG packets, no micro
>>
>>What are the side-effects of S120=16?  This is the value Telebit
>>says turns off the short burst of V.25 tone at the beginning of
>>each PEP answer tone.  This tone, which appeared in the BC5.10A
>>firmware upgrade for my TB+, caused my neighbors with older firmware
>>to no longer be able to connect to my modem.  S120=16 solved the
>>problem but I am now curious if that register value (16) has any other
>>effect I should know about (like packet size, etc).

S120=16 causes the firmware to ONLY send "long" packets, like in file
transfer ("spoofed") mode.  This is nice for links that have clicks in
them (like my link from The Netherlands to Japan), since the firmware
spends less time re-synching.

With micro-packets and short-packets (interactive resp. normal operation),
the firmware sometimes gets completely out sync, after which it has to
do a resynch ("retrain")..  this is annoying, and sometimes this can be
prevented by just simply increasing S120 to 12 (only short + long pkts)
or even 16 (only long pkts).

The disadvantage of 12 and 16 is, of course, that interactive operation
(login and such) gets really bad, since for each _character_ typed in,
the TB sends a _long_ packet, instead of a micropacket.  This, of course,
results in very long delay times.

So, on my link to Japan, login is very inefficient, as is file switching
during transfers.  So: we only send _ONE_ large (4-10Mbyte) file per
session, which makes up for login time loss...

> It turns off all of the mods put into the PEP modulation to combat trouble
> with echo cancellation devices on fiberoptic long distance carriers.
> One of the changes was a 3 second V.25 answer tone when the modem first
> answers the phone, even if PEP tones are first.  Previous versions of
> firmware didn't put out the V.25 answer tone before the PEP tones.
No, that is S121=1 .
S120 is for setting the low-level packet protocol specs (detail: packet
sizes).  With J6 (enter maintenance mode) you can also issue several
other register commands to further optimize the PEP protocol for a given
link situation...

All this "low level PEP optimization settings" stuff is documented in a
text file available from Telebit Corporation...

I could post it here, but I am sure that is not what people want...

Fred van Kempen
waltje@uwalt.nl.mugnet.org

gandrews@netcom.COM (Greg Andrews) (03/24/91)

 Fred posted an article where he quotes three other articles.
 I've added pointers to clarify who posted what.

In article <910322577@minixug.mugnet.org> root@minixug.mugnet.org (MINIXUG-ONLINE System Manager) writes:
>> In article <1991Mar16.160708.152892@ramona.Cary.NC.US> 
>>andrew@ramona.Cary.NC.US (Andrew Ernest) writes:
>>>In article <9908@discus.technion.ac.il> devil@techunix.BITNET 
>>>(Gil Tene) writes:
       [Gil wrote]
>>>>To detail : S120=12 means no use only LONG packets, no micro
>>>
       [Andrew asked]
>>>What are the side-effects of S120=16?  This is the value Telebit
>>>says turns off the short burst of V.25 tone at the beginning of
>>>each PEP answer tone.  This tone, which appeared in the BC5.10A
>>>firmware upgrade for my TB+, caused my neighbors with older firmware
>>>to no longer be able to connect to my modem.  S120=16 solved the
>>>problem but I am now curious if that register value (16) has any other
>>>effect I should know about (like packet size, etc).
>
       [Fred answers]
>S120=16 causes the firmware to ONLY send "long" packets, like in file
>transfer ("spoofed") mode.  This is nice for links that have clicks in
>them (like my link from The Netherlands to Japan), since the firmware
>spends less time re-synching.
>

       [I respond]
No, that is S120=2.

       [Fred]
>With micro-packets and short-packets (interactive resp. normal operation),
>the firmware sometimes gets completely out sync, after which it has to
>do a resynch ("retrain")..  this is annoying, and sometimes this can be
>prevented by just simply increasing S120 to 12 (only short + long pkts)
>or even 16 (only long pkts).
>

       [I respond]
The firmware does NOT get "completely out of synch".  Retrains are caused
by getting too many data errors within a short period of time.  The modems
track the error profile and if it looks bad, they decide the modulation
is not optimal for the line conditions, and perform a retrain in order to
adapt to the present conditions.

Micro packets don't work very well over satellite links because the link
has rather long delays.  The modems expect a response to a micro packet
within a certain period of time.  Same for short packets.  The timing
constraints are very tight for micros, looser for shorts, and loose for
long packets.  The delays introduced by satellite links can cause the
micro response to exceed the timing limit and generate an error.  Since
micros are extremely brief blips of sound, echo suppressors may not be 
able to respond in time, cutting off part or all of the packet.  Using
shorts and/or longs can help eliminate these kinds of errors, reducing
the need to retrain.

>      [Fred quotes my answer to Andrew]
>gandrews@netcom.COM (Greg Andrews) wrote:
>> It turns off all of the mods put into the PEP modulation to combat trouble
>> with echo cancellation devices on fiberoptic long distance carriers.
>> One of the changes was a 3 second V.25 answer tone when the modem first
>> answers the phone, even if PEP tones are first.  Previous versions of
>> firmware didn't put out the V.25 answer tone before the PEP tones.
>
>      [Fred responds]
>No, that is S121=1.

       [I answer]
That's not correct.  The S121 register deals with echo **suppressors**
of conventional design, not echo **cancellers** used on fiberoptic long
distance lines.  They are entirely different beasts, and setting S121=1
does not fix the trouble PEP was having over Sprint and MCI lines.

As I said, PEP was modified to handle the echo cancellers used for that
new technology.  Those modifications are completely different than the
calibration comb controlled by S121.  S120=16 disables the modifications
to PEP introduced in version 7 to solve the problems that existed over
fiberoptic lines.

        [Fred continues]
>S120 is for setting the low-level packet protocol specs (detail: packet
>sizes).  With J6 (enter maintenance mode) you can also issue several
>other register commands to further optimize the PEP protocol for a given
>link situation...

        [I answer]
That's right.  And that's why S120 contains the bit to disable the
fiberoptic mods.  That bit is bit 4, S120=16.

        [Fred continues]
>All this "low level PEP optimization settings" stuff is documented in a
>text file available from Telebit Corporation...

        [I answer]
Yes, and it gives only three S120 settings:  12, 2, and 3.

-- 
.------------------------------------------------------------------------.
|  Greg Andrews   |       UUCP: {apple,amdahl,claris}!netcom!gandrews    |
|                 |   Internet: gandrews@netcom.COM                      |
`------------------------------------------------------------------------'

root@Topsail.ORG (Chuck Murcko) (03/27/91)

Please post the Telebits text on adverse links. modems@telebit.com tells me
they are no longer mailing it out due to problems with users misapplying the
fixes. So, let those who read the post, beware!
-- 
Chuck Murcko   The Topsail Group   538 E. Church Rd., Elkins Park, PA 19117
Internet: cmurcko@topsail.Topsail.ORG
UUCP: ...!uunet!lgnp1!gvlv2!topsail!cmurcko