SCOTT, THE FOLLOWING POSTS SHOULD BE VERY INTERSTING TO YOU ON POSSIBLY
WHY NONE OF OUR  CALLERS  HAVE BEEN ABLE TO ACHIEVE FULL 28.8 RATES...

JUST THE INFO ON BANDWIDTH TRANSMISSION OF DATA IN PHONE LINES SHOULD BE
WORTH THE LOOK SEE.....

   LOOK FOR THREE LINES  OF  ********* 's TO SEE BEGINNING OF POSTS...
I PUT THE MODEM ONES FIRST !   BY THE SYSOP OF MOG-UR'S BBS:
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 08-11-94 (08:16)             Number: 229
From: TOM TCIMPIDIS                Refer#: 39683
  To: ALL                           Recvd: NO  
Subj: 28.8k Connects?                Conf: (2) TechInf-LO
---------------------------------------------------------------------------
MP> I would like to ask if the 28.8.k connection is as rare as it seams
MP> me.  I have seen a 28.8k connect only once, and during a transfer, i
MP> seamed to drop to maybe 26.4k.  As on a compressed file, I reciived
MP> little more than 2900 cps.  I was wondering what you might suggest t
MP> improve line conditions in my house, I have 2 data lines, that have
MP> the past gave good troughput with 16.8k connects, and 21.6k connects
MP> Also, I would like to know if v.34 would improve the line condition
MP> needs for a 28.8k connect?  Is the v.FC protocol more demanding?
MP> Thank you for your time...

A 28,800 bps V.FC/V.34 link requires 3200 Hz bandwidth, from 320 - 3520
A 26,400 bps V.FC/V.34 link requires 3000 Hz bandwidth, from 375 - 3375
A 24,000 bps V.FC/V.34 link requires 2800 Hz bandwidth, from 467 - 3267

by contrast..
A 21,600/19,200/16,800 link needs a usable bandwidth of only 2400 Hz,
from 600 - 3000 Hz. This is the same amount required for 9600/V32 &
14,400/V32-bis.

Alas, while many domestic phone lines can easily support the requirement
of V.32, V.32bis and V.32-terbo, some may not have sufficient bandwidth
to support V.FC and V.34 at *FULL* SPEED.  This, in a nutshell, is what
you (and others) are experiencing.

Compared to V.32/bis/terbo, 28,800 bps requires 33% more bandwidth,
26,400 bps requires 25% more bandwidth and 24,000 bps needs 17% more
bandwidth. This additional bandwidth has to be there from end to end,
from one modem to the other.  Either you have it or you don't - your
modem is telling you.

As the domestic telephone companies race to install fiber optic cables,
the bandwidth situation should gradually improve.

The only suggestions for 'at-home' modeming is to try disconnecting
*ALL* telephonic devices attached to the phone line.  This includes:
extension phones, answering machines, fax machines, caller-id boxes,
line-in-use indicators, cordless phone base units, demon dialers,
lightning spike protectors or line filters like those commonly found in
PC Desktop master-switch power directors. If any of this helps, then
start plugging things back in one-by-one until the culprit is found.

Your mileage will vary.
---
  QMPro 1.52  MOG-UR'S EMS  Internet: sysop@mogur.com 
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 08-11-94 (19:09)             Number: 230
From: TOM TCIMPIDIS                Refer#: NONE
  To: ALL                           Recvd: NO  
Subj: Modem statistics screen        Conf: (2) TechInf-LO
---------------------------------------------------------------------------
    For those of you without a V.FC or V.34 modem, this is what the
Modem Stats (Door 13) screen looks like on such a connect.  This screen
was captured on a call from a line in my home to a node of the BBS and
represents a good, clean connection and line.

 Protocol                     LAPM  Modulation                   V.34
 Compression        V42BIS 2048/32  Speed                       28800
 Link Timeouts                   0  Carrier Freq (Hz)       2000/2000
 Link Naks                       0  Symbol Rate             3200/3200
 Blers                           0  Retrains Requested              0
 Blocks resent                   0  Retrains Granted                0
 Chars lost                      0  Fallback                  Enabled
 Chars sent                   2560  HST Line Reversals              0
 Chars Received                  0  HST Equalization             Long
 Blocks sent                   152  Trellis Code        32S-2D/32S-2D
 Blocks Received                33  Nonlinear Encoding            OFF
 Octets sent                  2407  Precoding                      ON
 Octets Received                37  Shaping                       OFF
 Preemphasis Index             5/4
   Frequency (Hz)  Level (-dB)      Rx/Tx Level (-dB)           30/18
        1025           25           Roundtrip Delay mS              2
        3025           29           Current Call             00:00:39
        3325           31           SV: 08/02/94        DSP: 08/04/94
        3375           33
        3475           36           08-05-94 20:44 Caller 5
        3625           42                USRSTATS  Version 2.34B

For the those of you who care about such things, it should be noted that
modems essentially become deaf at a level of -43 dB...  Thus, a
frequency response for a given frequency at a level of -43 dB or less
(higher number is less) is unusable by a modem.

A poor signal to noise ratio or excessive harmonic distortion can also
cause a reduction in connect speed even if you have adequate frequency
response.

         -Tom
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 08-17-94 (18:47)             Number: 236
From: TOM TCIMPIDIS                Refer#: NONE
  To: ALL                           Recvd: NO  
Subj: About my cps rates...          Conf: (2) TechInf-LO
---------------------------------------------------------------------------
   I am often asked what the optimum cps rates are for different bps
rates so here is a list.  These are the maximum theoretical rates for
each bps using Zmodem and transferring a ziped (uncompressable) file.
Actual rates will generally be up to about 5% less due to modem and
computer system overhead.  Connections not using MNP or V.42
error correction (typical of older 1200 and 2400 baud modems) will
achieve cps rates 20% lower than those listed.

Modem protocol:  V.FC  V.34  V.32bis  V.32terbo  V.32   HST

BPS      CPS

1200      144     X     X      X         X        X      X
2400      288     X     X      X         X        X      X
9600     1152     X     X      X         X        X      X
12000    1440     X     X      X         X               X
14400    1755     X     X      X         X               X
16800    2016     X     X                X               X
19200    2304     X     X                X
21600    2592     X     X                X
24000    2880     X     X
26400    3168     X     X
28800    3510     X     X

All the protocols and rates shown above are presently supported by
MOG-UR'S on all nodes using USR Dual Standard V.Everything modems.

However, experience nationwide during beta testing on a variety of
systems by a variety of persons has shown that only about 10% achieve a
28.8 connect, 40% achieve a 26.4 connect and 40% achieve a 24.0
connect. The remaining 10% receive a 21.6 connect or lower. See earlier
messages in this conference for a discussion of why this is so and how
the telephone company enters in to the picture.
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 09-09-94 (10:09)             Number: 245
From: TOM TCIMPIDIS                Refer#: 12731
  To: ALL                           Recvd: YES 
Subj: 8/25 V.34 code                 Conf: (2) TechInf-LO
---------------------------------------------------------------------------
BB> Yes, but my available bandwidth here (using -43 as the threshold
BB> benchmark) should not allow beyond a 24K or maybe at best a very
BB> occasional 26.4K connect, which is exactly what I used to get. Howev
BB> I now get solid 28.8 connects everytime which tells me that they mus
BB> now be able to get a usuable signal from levels well below -43...
BB> Whatever the reason, it is a most impressive performance.

With the USR Courier Flashrom 08/23 code (and later) they've enabled the
1829 Hz carrier frequency, thus shifting the entire envelope down 91 Hz
lower when conditions warrant, for an additional 182 Hertz of bandwidth.
This may be where you're achieving those 3200 Hz bandwidth calls. Now
if the modem can't find the bandwidth in the attic it looks for it in
the basement. The -43 level threshold is still there.

Not 100% sure, but I think the modems -may- be able to do this
envelope shifting independently, transmit versus receive.

Isn't V34 neat?

         -Tom
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 09-09-94 (10:39)             Number: 246
From: TOM TCIMPIDIS                Refer#: 12748
  To: ALL                           Recvd: NO  
Subj: 8/25 V.34 CODE                 Conf: (2) TechInf-LO
---------------------------------------------------------------------------
RH|I am seeing that the V.34 code is not reporting correct connect
  |rates.. I have seen my comm. program reporting one thing but the bbs
  |says something else... sometime my comm. program will say 21.6 and
  |the bbs says 24.0 sometime it is reversed.. Today, I called myself
  |and I said 19.2 on the comm program and PCBoard reporte 26.4... The
  |transfer rates seem to point out the higher rates, no matter what it
  |says it connected at...  Which I guess is good, (Grin)...

There is nothing incorrect about what the V.34 modems are reporting.
V.34 incorporates ASL (adaptive speed leveling) which allows one side
(transmission or receive) of the connection to be higher than the
other. All you have to do is examine an ati6 after (or during) a call
(or use Door 13 online here) to see this. When I call from work to my
bbs, my comm program reports a 28800 connect while the bbs reports
26400. What that means is *my* receive is 28800 and the bbs's receive
is 26400. Both speeds MAY be the same but they do not HAVE to be the
same. This is one of the many big advantages of V.34 over V.FC.

         -Tom

***************************************************************************
***************************************************************************
***************************************************************************
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 06-20-94 (09:17)             Number: 208
From: TOM DEVLIN                   Refer#: 10650
  To: DOUG HAIRE                    Recvd: NO  
Subj: New 28.8 Modem                 Conf: (2) TechInf-LO
---------------------------------------------------------------------------
DH>Yes, the need for extra capacity at a low price is the basic reason for
DH>the SLC usage. My area is growing quickly and is in a very rual area of
DH>the county. There is insufficient outside plant (cables) to support the
DH>amount of growth we have. Since we are also in a transition to fiber,
DH>there is a problem with expanding the existing metallic wiring to handle
DH>the growth. Converting it to SLC systems is a good method until the
DH>fiber comes in...

Then you see the SLC's as a temporary expediant, right?  There was (as
I'm sure you recall) a lot of talk in the Field Trial conference about
the effect SLC's were having on connect speeds, you seem convinced
that they *should* allow a 28.8K connect but other folks noticed a
direct correlation between connect rates and the presence (or absence)
of a SLC in the path.

DH>I already have two V.everythings (one on the board and one personal). I
DH>get some interesting results when calling from home and office. When
DH>calling from my home voice line into the BBS (call is routed to the CO
DH>and back), I get 26400 bps connects.

Is it possible to have a SLC in the circuit on an in-and-out connect
like that?  You say they are used for inter-office expansion but would
something similar ever be used to expand line capacity in or out of a
CO?  Possible at a site remote from the actual Central Office?

DH>When I call from my office (about 16 miles as the crow flies), I
DH>get only a 24000 bps connect. Dumps of the various displays show no
DH>real difference in conditions.

Does that include the ATY7 screen?  I'm convinced that there's some
valuable information there if we could understand it.

DH>However, from the office, I go through a PBX to a local CO (in the
DH>same building) to my home CO so there's more room for problems.

Yeah, any time you add *anything* to the link you increase the
possiblity of problems. I frequently see better LD connects than to
"local" boards that are an office or two away. This was also true with
V.32Bis, HST and terbo, I saw fewer Blers on LD calls than calls to a
couple of local boards.

DH>I have passed the data on to a friend who works in Southern Bell

Keep us posted...

DH>It's real hard to get any sympathy for the telcos when your complaint
DH>goes:

DH> I can't seem to get any better than 24000 bps connections on my
DH>voice line.

Once V.34 is "officially" approved I'm going to try the "I'm using a
protocol the was approved internationally for use on voice grade
lines, why aren't *your* lines up to snuff?" line. Be interesting to
see what they say.

DH> In fact, it's hard to get any sympathy from my friends with 14.4's ...
DH> [g]

Sadly true... <g>

DH>  SLMR 2.1a #40  Why does 14400 bps seem so slow now?

ROTFL!  I still remember my first 14400 HST connect, I just couldn't
believe it. Now, if I call a board that's still "stuck" at 14400 it
seems like it takes an enternity for anything to happen. How soon we
get spoiled... <g>

---
 * BCSUTI Version 1.3
 * The Warm Boot  Waterford, MI  (810) 683-8032
 * PostLink(tm) v1.20 WARMBOOT (#1003) : RelayNet (tm)
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 06-20-94 (19:40)             Number: 209
From: BILL GARFIELD                Refer#: 10827
  To: DOUG HAIRE                    Recvd: NO  
Subj: New 28.8 Modem                 Conf: (2) TechInf-LO
---------------------------------------------------------------------------
TD> I've heard a fair amount about SLC's lately, I get the feeling that
TD> we're going to be seeing a lot more of these as the phone companies
TD> scramble to add capacity, true?

DH>Yes, the need for extra capacity at a low price is the basic reason for
DH>the SLC usage. My area is growing quickly and is in a very rual area of
DH>the county. There is insufficient outside plant (cables) to support the
DH>amount of growth we have. Since we are also in a transition to fiber,

But fiber or copper, the SLCs are here to stay.  The old SLC96 is giving
way to the new SLC5 and SLC2000. (higher capacity)

TD> Do you think that a properly set up SLC should allow 28.8K connects?

No. Contrary to what some will tell you, an SLC-96 operating in either
Mode 1 or Mode 2 provides only 56-kilobit voice channels, not 64k bps.
They are 64k bps capable, but only in Mode 3. Mode 3 is not used for
switched voice applications.  SLC = Subscriber Loop Concentrator.

SLC-96 "Mode 1" is for high capacity voice applications, normally in
commercial settings. A Mode 1 SLC-96 is provisioned with four (4) 1.544
megabit T1 lines (and sometimes a 5th for failsafe). It can also be
provisioned with a fiber DS-2. But whether fiber or copper, the Mode 1
SLC-96 is capable of providing 96 -simultaneous- 56 kilobit voice paths.

SLC-96 "Mode 2" is used typically in residential areas where the
likelihood that more than 50% of the households would be using their
phones at any one moment in time is very remote. In Mode 2 the SLC-96
is provisioned with only two (2) T1 lines, but no compression takes
place. It still serves 96 subscribers, though only 48 can be off hook
(talking) at any one time. Again, these are 56-kilobit channels, not
64k.

In either Mode 1 or Mode 2 an SLC-96 uses "robbed bit signalling" to
carry in band all status information such as E&M signalling states.

Before we put this to bed I think we need to have a crash course in T1
digital transmission techniques.  A T-1 line or Span as they are
sometimes called, runs at a composite data rate of 1.544 megabits. For a
SLC-96 loop concentrator, the line format is AMI (alternate mark
inversion). It uses an 8-bit character interleaved Time Division
Multiplexing (TDM) pulse code modulated (PCM) format.

There are 24 time slots (channels) on a standard T1 line. Each time slot
(channel) is divided into 8-bits and sampled at an 8-kilobit rate to
provide twenty-four 64k bps channels plus a single-frame synchronization
bit per frame (193 bits per frame total times 8k bps sample rate = 1.544
megabits). - It's easier to follow drawn out on a white board, but follow
along, we're almost there...

One of the rules (laws) of this signalling format is the zeroes density
rule. We can have no more than 15 consecutive "zeroes" before we -must-
output a "one".  To prevent violation of the zero density rule, the 7th
bit of any 8-bit byte must be altered. This is of no consequence for
voice transmission but means data is effectively restricted to 7/8ths of
the maximum 64k bps, or 56 kilobits max.

It is possible to achieve "clear channel" 64k bps speed but doing so
requires that we use binary 8-zero substitution (B8ZS) instead of the
altername mark inversion (AMI) format.  Use of B8ZS in a T1 multiplexor
requires that all channel service units, central ofice equipment and
test equipment be strapped for this option. It is not commonly used
in SLC-96 equipment.

In theory, if I understand all this smoke & mirror magic, the max
speed one can expect running data over a Mode-1 or Mode-2 voice channel
on an SLC-96, under the most favorable of conditions, is 26,400 bps (due
to the theoretical restrictions of Shannon's Law as applied to the
reduced bandwidth of a 56 kilobit voice channel.

Not what you wanted to hear, I know...

---
  OLX 3.0  Friends feel sorry for friends who are on an SLC-96!
 * USRobotics Customer Support, U.S.A., 708-982-5092 (28.8k)
 * PostLink(tm) v1.11  USRUSA (#174) : RelayNet(tm)
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 06-21-94 (23:33)             Number: 210
From: DOUG HAIRE                   Refer#: 10842
  To: TOM DEVLIN                    Recvd: NO  
Subj: New 28.8 Modem                 Conf: (2) TechInf-LO
---------------------------------------------------------------------------
DH>Yes, the need for extra capacity at a low price is the basic reason for
DH>the SLC usage. My area is growing quickly and is in a very rual area of
DH>the county. There is insufficient outside plant (cables) to support the
DH>amount of growth we have. Since we are also in a transition to fiber,

TD|Then you see the SLC's as a temporary expediant, right?  There was (as

  Yes, that's what its purpose has become. It was envisioned as a way to
  handle growth on a permanent basis but it's apparent that nothing is
  that permanent in the telecommunications field :).

TD|I'm sure you recall) a lot of talk in the Field Trial conference about
  |the effect SLC's were having on connect speeds, you seem convinced
  |that they *should* allow a 28.8K connect but other folks noticed a
  |direct correlation between connect rates and the presence (or absence)
  |of a SLC in the path.

And I still feel that way (since I have managed 28.8k connects on rare
occasions though they have been generally poor and unstable) in spite of
Bill's explanation about the SLC's.

DH>I already have two V.everythings (one on the board and one personal). I
DH>get some interesting results when calling from home and office. When

TD|Is it possible to have a SLC in the circuit on an in-and-out connect
  |like that?  You say they are used for inter-office expansion but would
  |something similar ever be used to expand line capacity in or out of a
  |CO?  Possible at a site remote from the actual Central Office?

The SLC is not between offices, it's from CO to house/business areas
(though more to rural/suburban housing than any major business area).
Straight T-cxr systems are used between offices (usually in T-3 "pipes"
or 1.7gbps fiber routes).

DH>When I call from my office (about 16 miles as the crow flies), I
DH>get only a 24000 bps connect. Dumps of the various displays show no
DH>real difference in conditions.

TD|Does that include the ATY7 screen?  I'm convinced that there's some
  |valuable information there if we could understand it.

I'm sure the aty7 screen would be quite useful if we had *any* idea of
what it was reporting. Since we don't, it's totally useless.

TD|Yeah, any time you add *anything* to the link you increase the
  |possiblity of problems. I frequently see better LD connects than to
  |"local" boards that are an office or two away. This was also true with
  |V.32Bis, HST and terbo, I saw fewer Blers on LD calls than calls to a
  |couple of local boards.

When you make an LD call, the path is usually like this:

home  CO  LD provider  LD Provider  CO  bbs

or:
home  CO  tandem  LD Provider  LD Provider  tandem 
CO  bbs

with all paths from local CO to remote CO being digital (and likely
fiber). Paths across town may not take an all digital route (though most
now seem to) or the route may consist of multiple links that are not
properly set up. They should be improving over the next couple of years,
though.

DH>It's real hard to get any sympathy for the telcos when your complaint
DH>goes:

DH> I can't seem to get any better than 24000 bps connections on my
DH>voice line.

TD|Once V.34 is "officially" approved I'm going to try the "I'm using a
  |protocol the was approved internationally for use on voice grade
  |lines, why aren't *your* lines up to snuff?" line. Be interesting to
  |see what they say.

That's the proper approach, in my opinion. A simple remark that these
are designed to operate within the specs for a standard voice line
should be sufficient.

DH>  SLMR 2.1a #40  Why does 14400 bps seem so slow now?

TD|ROTFL!  I still remember my first 14400 HST connect, I just couldn't
  |believe it. Now, if I call a board that's still "stuck" at 14400 it
  |seems like it takes an enternity for anything to happen. How soon we
  |get spoiled... <g>

See the next generation tagline ... Ŀ

---
  SLMR 2.1a #40  Drat! Only 24kbps!
 * Telephone Exchange 407-791-2474 V.32bis ZyXel 19200!
 * PostLink(tm) v1.20 TELEPHNE (#222) : RelayNet(tm)
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 06-22-94 (04:57)             Number: 211
From: TOM DEVLIN                   Refer#: 10851
  To: BILL GARFIELD                 Recvd: NO  
Subj: New 28.8 Modem                 Conf: (2) TechInf-LO
---------------------------------------------------------------------------
BG>There are 24 time slots (channels) on a standard T1 line. Each time slot
BG>(channel) is divided into 8-bits and sampled at an 8-kilobit rate....

BG>....This is of no consequence for voice transmission but means data is
BG>effectively restricted to 7/8ths of the maximum 64k bps....

BG>In theory, if I understand all this smoke & mirror magic, the max
BG>speed one can expect running data over a Mode-1 or Mode-2 voice channel
BG>on an SLC-96, under the most favorable of conditions, is 26,400 bps (due
BG>to the theoretical restrictions of Shannon's Law as applied to the
BG>reduced bandwidth of a 56 kilobit voice channel.

I've saved your message for further study but this is *finally*
starting to make sense!  Shannon says that you need to sample at a
frequency of (at least) twice the data rate. An 8K samples/second rate
should still allow the 3600 V.FC frequency to pass but that 7/8 figure
(which I've *never* seen before) gives us a top frequency limit
(assuming a "brick wall" low pass filter) of 3500 cps. Presto!  A
26.4K connect!

Does anyone know what frequencies V.34 will use?

---
 * Channel 1(R) * 617-354-7077 * Cambridge MA * 100 lines
 * PostLink(tm) v1.11  CHANNEL1 (#15) : RelayNet(tm)
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 06-23-94 (19:33)             Number: 212
From: BILL GARFIELD                Refer#: 10842
  To: TOM DEVLIN                    Recvd: NO  
Subj: New 28.8 Modem                 Conf: (2) TechInf-LO
---------------------------------------------------------------------------
TD>Then you see the SLC's as a temporary expediant, right?  There was (as
  >I'm sure you recall) a lot of talk in the Field Trial conference about
  >the effect SLC's were having on connect speeds, you seem convinced
  >that they *should* allow a 28.8K connect but other folks noticed a
  >direct correlation between connect rates and the presence (or absence)
  >of a SLC in the path.

Tom, my contacts inside Southwestern Bell say that everything from the
technology standpoint is "temporary".  The SLCs will likely remain for
the forseeable future.  The new SLC's (SLC-5 and SLC-2000) represent
only the latest technology is subscriber loop concentrators, or "pair
gain devices" as most telcos refer to them as. The concept is still
the same - a device intended to bring hundreds of telephone circuits to
an area which is either:
  (A)  Too far from the switching office for conventional copper, or
  (B)  Rapidly outgrowing the capacity of existing cabling, or
  (C)  A newly platted subdivision
A single SLC-96 can bring 96 phone lines into an area and use only 4
pairs of conventional copper wires to make it happen.  It can do this out
to 50 miles from the switching office, though a T1 repeater is required
every 6,000 feet.

As far as 28.8k modem performance is concerned, a SLC-96 configured for
Alternate Mark Inversion has to "steal" 1/8 of each channel's timeslot
to conform with the T1/D4 AMI format. Therefore we could logically say
by so doing that it also robs 1/8 of the available bandwidth. This has
no effect on voice transmission, but when the modem is looking for
bandwidth, a 1/8 reduction in the timeslot equates to a 12 percent
reduction in bandwidth and easily 1 rung off of the speed ladder.

During the 28.8 Courier beta trials we experienced this loss of
bandwidth and reduction in speed capability every time we encountered
T1/D4 analog-to-digital and then back digital-to-analog conversion
(which happens at each end of a T1/D4 interface).  As the SLC-96 was
designed around D4 channel bank technology, it behaves similarly.

I can route my call over 2,800 miles of 100% digital (B8ZS) circuits and
back to another modem across the room and GUARANTEE that we'll have a
solid 28,800 bps connection every call.  I can send the same call out to
Southwestern Bell through a D4 channel bank and a 3 mile loop and
guarantee that you will never, ever see a connection faster than 26,400
(that will hold in).

---
  OLX 3.0  CONNECT 21600/ARQ/V32/LAPM/V42BIS - 2500 cps!! YEAH!!
 * IBM Net Connection <> Indpls., IN <> (317)882-5575 21.6 Duals All Lines
 * PostLink(tm) v1.20 IBMNET (#5) : RelayNet(tm)
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 06-30-94 (10:14)             Number: 216
From: TOM TCIMPIDIS                Refer#: 11055
  To: ALL                           Recvd: NO  
Subj: A quick review                 Conf: (2) TechInf-LO
---------------------------------------------------------------------------
This message was from BILL GARFIELD to ALL
originally in conference USRobot-RI on MOGUR (The MOG-UR'S EMS)
and was forwarded to you by TOM TCIMPIDIS
                    ----------------------------------------
A quick review....

Although we've spoken of these SLC-96 remote terminals as a device which
combines (multiplexes) 96 telephone circuits onto two or four broadband
high speed synchronous lines (T-1 lines), perhaps another word or two
of explanation is in order.

These "slicks" or "remote terminals" or "pair gain" devices as they are
called, (names used interchangeably) are not simply a single stand-alone
piece of gear.  Rather several of them will be installed in a "hut" or
small building.  This hut is normally a non-obtrusive building,
about 10' X 20' that blends into its surroundings. Chances are most
folks have seen them but not known what's inside. The phone company
tries to locate these near the center of a new (normally) residential
subdivision. These then become the "local hub" for all neighborhood
phone service out to about 3 miles or so in all directions from the hut.

No "switching" is done in these remote terminal huts. Their sole purpose
is to gather or combine the neighborhood cluster of phone circuits
together (often as many as 2,000 or more lines) which are then routed to
the switching office (Central Office) via just a few high speed digital
lines.  In this way, the phone company is able to bring phone service
to 2,400 subscribers and use only 200 copper pairs to do it.  In this
specific case, a 2400 line hut would contain 25 individual SLC-96
concentrators inside. If the main line between the remote terminal and
the distant Central Office happens to be a fiber optic cable, a mere two
strands of fiber is all that would be needed to provide service to those
2,400 subscribers. (In actual practice more than just 2 strands of fiber
are used, but I'm sure you get the idea).

These remote terminals provide the subscriber (you) with loop current
and ringing voltages, but the actual dial tone you hear is coming from
the real Central Office (switching center) several miles away.

Obviously, the SLC-96 can potentially save the phone company millions of
dollars with its capability to provide service to so many while
requiring so few lines all the way back to the Central Office, but they
can be a horrible vexation to our new technology modems.  The issue is
bandwidth. Our new 28.8k bps modems are hoping to see a phone line with
at least 3200 Hz of bandwidth, centered on 1920 Hz and extending from
320 Hz on the low end to 3520 Hz on the high end. Any less than this and
our new ultra warp speed modems may operate at something less than
optimum speed.

Typical bandwidth for phone service coming through a SLC-96 is about
3100 Hz in the best case and sometimes as little as 3000 Hz. Therein
lies the focal point of the issue.  The possible reasons for that
reduced bandwidth and the effect it has on modem performance is what
we've been discussing.

---
  OLX 3.0  OS/2 VirusScan - "Windows found: Remove it? (Y/y)"
 * USRobotics Customer Support, U.S.A., 708-982-5092 (28.8k)
 * PostLink(tm) v1.11  USRUSA (#174) : RelayNet(tm)
---
  QMPro 1.52  MOG-UR'S EMS  Internet: sysop@mogur.com 
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 07-07-94 (17:32)             Number: 217
From: BILL GARFIELD                Refer#: 11200
  To: FRANK HABER                   Recvd: NO  
Subj: New 28.8 Modem  1/2          Conf: (2) TechInf-LO
---------------------------------------------------------------------------
FH> to support customers down there).  I'm here in Manhattan, where they
FH> gave up gutta-percha cable insulation about five minutes ago.

ROTFLMAO!!

So others can share the humor in Frank's comment, Gutta-Percha is an old
latex substance first discovered back in 1847. It first found use about
1851 as the insulation in the first international -telegraph- cable
which ran between England and France.  Gutta-Percha was also the first
insulator to survive in undersea applications and was still the
insulator of preference as late as 1947 when polyethylene finally began
to gain acceptance. (Just how old are you Frank?)

FH> This is a mostly 5ESS, big-gauge copper, lots o' crosstalk, <3mi to the
FH> CO -that sort of environment. Casually calling around New York, I get
FH> 28.8 about 30% of the time.  The rest of the connections are split
FH> evenly between 26 and 24.

Pretty much what I'd expect to see in an all-copper, short loop length
environment.  Envious, but not enough to justify moving to the big apple.
We sleep with 12-ga slide action Mossberg now.

FH> Yes, the rosy glow does seem to disappear when the two modems are of
FH> different makes.

Yes, but that will become less of an issue with the passage of time
and the maturing of the protocols. Remember the problems of the early
V.32 and V.32bis modems?

FH>  Years ago, telcos were experimenting with adaptive PCM and
FH>  delta-PCM.  Is that devilishness dead now?

Unfortunately, No.  Fortunately it also hasn't gained much popularity.
To my knowledge all ADPCM stuff is limited to dedicated lease circuit
applications and 'el-cheapo' tielines into hole-in-the-wall businesses
with complete morons and penny pinchers as their communications
consultants. Most of that crap also relies on 3rd party hardware.  For
those reading along, ADPCM (Adaptive Delta Pulse Code Modulation) for
lack of a better description, is basically a digital circuit with
compression. Most of it runs at 32k bps and has very narrow bandwidth.
ADPCM works *swell* for strictly voice applications, and might have made
some inroads were it not for the advent of the office fax machine about
the same time (whew!). Modems and fax machines don't get along well with
"compressed" phone lines, hence ADPCM has been judged unsuitable for
widespread applications, thank goodness.

FH>  2) You guys are debating a lot over differences of 5% in bandwidth.

Actually the tussle is over a 12% discrepancy in bandwidth. For a 28.8k
bps modem which -hungers- for usable bandwidth beyond 3500 Hz, a 12%
bandwidth impairment spells the difference in achieving 28.8 versus 26.4
or 24.0 & less.  My goal in conducting the discussion is not so much to
cast aspersions upon the phone company as it is to -explain- to other
modem users *why* some can achieve better results than others.  I'm
trying to help users understand that it's not the modem's fault.

FH> What's the actual spec for a the frequency response of a subscriber

The numbers I've seen are 300 to 3000 Hz within +2 and -8 dB over the
whole range, referenced against 'milliwatt' (1004 Hz) injected at +/-0
dBMO at the serving office.  This is a very generous spec, and it's
unlikely that anyone's subscriber loop will actually be found to be out
of tolerance.  LD circuits are spec'd at 300 to 3400 Hz (CCITT G.132).
This might also explain why we hear occasional reports of users getting
higher connect speeds on an LD call than on a call down the lane.

The shape of the bandwidth response curve will generally be 'humped',
depending on loop length and whether or not it's loaded cable (loaded
meaning with load coils, used to improve frequency response on long
cables).  At the high end, the subscriber loops going through the SLC-96
and other pair-gain devices will start stonewalling around 3300 Hz.

Though the subscriber loops are spec'd at 300~3000, we all know that in
actual practice most of them perform better, but not many perform as well
as we'd like them to. As the North American telephone infrastructure
gets rebuilt and upgraded to fiber over the next few years, the
remaining bandwidth issues should gradually begin to disappear.

FH>  I'm inferring from your messages that the "proper" line level is
FH> -20 or -22, that V.34's equalization makes that "zero," and that the

I'm not privvy to what the manufacturer expects to see.  Most can
tolerate moderate level swings down to about -43 at which point the
receiver becomes deaf.  These new V.FC and V.34 modems can do some
magic with the levels (preemphasis and transmitter offset) to help
compensate for "normal" high end rolloff characteristics, but are
helpless in trying to find a signal down in the mud below -43 or
trying to find bandwidth which just isn't there to find.

cont'd next message...

---
  OLX 3.0  Shopping tip: Shoes are only $.85 at bowling alleys.
 * USRobotics Customer Support, U.S.A., 708-982-5092 (28.8k)
 * PostLink(tm) v1.11  USRUSA (#174) : RelayNet(tm)
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 07-07-94 (17:32)             Number: 218
From: BILL GARFIELD                Refer#: NONE
  To: FRANK HABER                   Recvd: NO  
Subj: reply cont'd... 2/2          Conf: (2) TechInf-LO
---------------------------------------------------------------------------
Continued from previous message...

FH> Could you discuss group delay a little bit?  Is there a situation
FH> where the simple response curve in the frequency domain could be OK, and
FH> phase errors could bollix V.34?  Admittedly, analog loops are getting
FH> less and les important, but we still have loading coils and other old
FH> stuff in use here.

Hmmm... Definition time again. For those reading along, "Group delay" is
the distortion resulting from the non-uniform speed of transmission of
one frequency versus another frequency in the same cable. Normally it's
not an issue until we get involved with LD calls or excessively long
subscriber loops, usually beyond 18,000 feet. Load coils are a blessing
in flattening out the frequency response curve in those long loops but
extract another penalty at both the high and low frequencies, slowing
one range of frequencies down a few microseconds while not affecting
another range. To answer the question, yes, group delay can adversely
impact modem performance if it is sufficiently large. I don't know what
the modem can tolerate, however. Perhaps one of the USR guys can take
up the discussion for a moment here.  (Oh puhleeze?)

FH> Could someone define a "bridge tap" so someone with a general EE
FH> background can understand it?

You're the second one to ask that question in as many days.

They can be anywhere, but seem to proliferate in multi-family apartments
and older commercial buildings.  In telco-speak a circuit is said to be
"bridged" if that same circuit appears in two or more locations. The
term differs from "looped" in that a bridge connection (or tap) has an
actual physical connection to the main feeder cable in two or more
locations, often thousands of feet apart. In a looped cable, the cable
pair may "appear" in two or more locations but it's integrity has not
been violated by a physical connection having been made to it. Bridge
taps can be sources of imbalance (hum) and static.  It's a connection to
the line that is redundant and doesn't need to be there to get your
service to you. It's a potential problem waiting to happen.  Locating
bridge taps usually requires use of a device known as a TDR (Time Domain
Reflectometer) to "scope the line".  Getting them removed requires a
service order for a conditioned line ($$$).

FH> What are the "fast connect" options for V.34?  V.FC takes a
FH> looong time to negotiate, sometimes, and makes a really obnoxious set of
FH> noises

You haven't heard "obnoxious" until you hear a genuine V34 handshake!

FH> day, call setup time becomes a big factor.  USR already is one of the
FH> best "fast handshakers" in the V.32 world.  Can you do that with V.34?

Oh Mr. Frankiewicz, HELP!!!

FH> In V.34, are we still supposed to "train down" when we want to drop
FH> the connection, and twiddle our thumbs.  If we use &D3, will we still
FH> wind up in a Swiss jail (g)?

As part of the error correction protocol in both MNP as well as
V42/LAP-M there is a unique Link Disconnect Frame which is sent by the
modem. Another compliant modem, upon recognizing the link disconnect
request will drop the call virtually in an eyeblink.  USRobotics modems
have always issued and responded to the -MNP- link disconnect request
when in MNP or HST mode. Compliance with the V42 Link Disconnect Request
didn't come along in USR products until this last year (thanks in part
to yours truly flaming USR on the internet for omitting the feature).

&D3 is not supported in USRobotics modems. It's counterpart, S13=1,
works fine in accomplishing the same task (modem reset upon loss of
DTR).  In an earlier incarnation, the unsupported &D3 command would
emulate S13=1 but you couldn't "save" the &D3 in nvram whereas you could
save S13=1.  And yes, in that same approximate vintage there was a
recycle trap that you could fall into until you cycled power off and
back on.  That's an obscure bug that -you- weren't supposed to know about.
It got fixed anyway, about a year or two ago, also with no fanfare.

FH> That should do it for a first cut (g).

Whew!  You're not trying to set me up are you?  You sound like an OT
with some pre-divestiture whiskers longing for the days gone by of
the old panel and step office, strowger switches and of course a
P.K.Neuses burnishing tool in your shirt pocket. >very wide smile<

---
  OLX 3.0  GUN CONTROL means using BOTH HANDS.
 * USRobotics Customer Support, U.S.A., 708-982-5092 (28.8k)
 * PostLink(tm) v1.11  USRUSA (#174) : RelayNet(tm)
===========================================================================
 BBS: The MOG-UR'S EMS
Date: 07-12-94 (17:51)             Number: 220
From: BILL GARFIELD                Refer#: 11232
  To: ALL                           Recvd: NO  
Subj: Conditioned Lines              Conf: (2) TechInf-LO
---------------------------------------------------------------------------
I know this is a resend, but Joe Frankiewicz liked the first chart
so much that I was spurred on to gather even more information for it.

This is the kind of line specs the phone company is talking about with
"C" and "D" conditioning.  "D" conditioning is sometimes referred to as
High Performance Data Conditioning (HPDC).

(courtesy of the Bell System Technical Journals)
Ŀ
                    Without     With C-1     With C-2     With C-4  
                 Conditioning Conditioning Conditioning Conditioning
Ĵ
Frequency Range  300 - 3000    300 - 2700  300 - 3000    300 - 3200 
 in hertz (Hz)   +3 to -12 dB  +2 to -6 dB +2 to -6 dB   +2 to -6 dB
 and levels                                                         
 (ref. to 1004   500 - 2500   1000 - 2400  500 - 2800    500 - 3000 
 Hz tone level   +2 to -8 dB   +3 to -1 dB +1 to -3 dB   +2 to -3 dB
 between C.O.                                                       
 and subscriber                300 - 3000                           
 premesis)                    +3 to -12 dB                          
Ĵ
Group Delay in    < 1750s     < 1000s     < 500s      < 300s    
microseconds     800-2600 Hz  1000-2400 Hz 1000-2600 Hz 1000-2600 Hz
                                                                    
                               <1750s      < 1500s     < 500s    
                               800-2600 Hz 600-2600 Hz   800-2800 Hz
                                                                    
                                            < 3000s     < 1500s   
                                           500-2800 Hz   600-3000 Hz
                                                                    
                                                         < 3000s   
                                                         500-3000 Hz
Ĵ
Impulse Noise          15 counts in 15 minutes               Zero     
Ĵ
Bit error rate   1 bit error per 100,000 bits transmitted    Zero     
Ĵ
                               With the addition of "D" Conditioning  
Ĵ
Signal to Noise      24           28           28           28      
 ratio (in dB)                                                      
Ĵ
2nd harm. Dstn.     -25          -35          -35          -35      
Ĵ
3rd harm. Dstn.     -30          -40          -40          -40      


From the chart above, we see that a C-4 conditioned line provides us
with the best bandwidth.  "D-4" conditioning provides the enhanced
bandwidth of C-4 conditioning plus a 4 dB better signal-to-noise ratio
and 10 dB better 2nd and 3rd order harmonic suppression.

As we've seen from many of the Y11 frequency & level reports, some
people's phone lines may already meet or even exceed D-4 conditioned
status. Lucky them. For the rest of us, we can either order and pay for
C-4 or D-4 line conditioning or simply accept the fact that for the time
being, 24,000 bps may be the best we can do.

At this point I would like issue a disclaimer and in so doing make it
very clear that by no means am I or anyone else advocating that anyone
go out and spend the money for line conditioning! Unless you're going to
be modeming over a dedicated full-period "nailed up" circuit, line
conditioning will only help -your- leg of the call between you and your
own local phone company.  Most telephone companies will advise against
conditioning of dial-up circuits just for that reason. Also, remember
that conditioning merely guarantees a certain quality of line. It does
*not* guarantee that your modem performance will improve.

---
  OLX 3.0  Guns no more cause crime than flies garbage
 * USRobotics Customer Support, U.S.A., 708-982-5092 (28.8k)
 * PostLink(tm) v1.11  USRUSA (#174) : RelayNet(tm)

