From: goldstein@carafe.tay2.dec.com (Fred R. Goldstein)
Subject: Re: What is Category 5 (re: ISDN)
Organization: Digital Equipment Corp., Littleton MA USA
Date: Sat, 7 Aug 1993 03:49:00 GMT

In article <telecom13.539.8@eecs.nwu.edu> hhallika@tuba.calpoly.edu
(Harold Hallikainen) writes:

> Can anyone give a BRIEF description of the electrical coding
> scheme on all these high speed twisted pair systems (such as ISDN)?  I
> haven't taken the time to research it at all.  How many bits per baud
> are being run?  Is it a multi-level coding system or just two or three
> level (such as positive/negative or positive/zero/negative).  I've
> always thought that on an analog medium, such as a pair of wires, we
> should run as many levels as we can until noise starts making level
> determination at the far end difficult, making the error correction
> overhead exceed the data throughput benefit of running more levels.
> Do such systems treat the twisted pair as a transmission line,
> matching the characteristic impedance to prevent reflections?  Do they
> work full duplex using some sort of hybrid, or just go real fast half
> duplex?  Can we use these techniques over leased lines ordered from
> the local phone company (like 3002 lines)?

One of the little sound bites I like to throw into my Transmission
class is that "at the bottom, all transmission systems are analog"!
So it's not a simple answer to a very good question.

I'll give a summary of ISDN.  The BRI S/T interface (inside wire) uses
Modified Duo-Binary (MDB) coding, which is a three-level code.  A 1 is no
voltage and a 0 is a positive or negative voltage pulse, each pulse
the opposite of the last (with some exceptions).  The coding used for
the Primary Rate, that is to say T1 and E1, is Alternate Mark
Inversion.  That's sort of like MDB except that 0's are no voltage and
1's an alternating pulse, but note that the pulse shapes are specified
quite differently for the BRI (Basic Rate 2B+D) and the PRI (T1/E1),
among other details.

At the U interface of the BRI (outside wire, local loop), the coding
is now usually "2B1Q".  This takes two bits from the scrambled data
stream and encodes them as +2.5, +.7, -.7, or -2.5 (about) volts.  So
one pulse is two bits (thus 2 Bits one Quat).  Don't quote me on the
exact voltages; I don't have the spec handy.  Again there's pulse
shaping.

To make all this work, impedance is specified, with very picky
transformers needed for S/T to meet spec.  S/T uses separate transmit
and receive wires, thus a 4-wire interface.  The BRI U uses one pair,
so there's some fancy echo cancellation done inside the transceiver
("UBAT", in AT&T terms) chip.  Real fancy.

3002 lines are utterly obsolete.  Voice-grade is now used for dial-up,
and the "state of the art" is being presented in the developing V.fast
world.  In lieu of 3002 most phone companies (in America at least)
will sell you 56k service.  THis could be provisioned using ISDN
technology, but in practice there are cheaper purpose-built line
drivers.


Fred R. Goldstein   k1io   goldstein@carafe.tay2.dec.com
Opinions are mine alone; sharing requires permission

From: goldstein@carafe.tay2.dec.com (Fred R. Goldstein)
Subject: Re: What is Category 5 (re: ISDN)
Organization: Digital Equipment Corp., Littleton MA USA
Date: Thu, 12 Aug 1993 05:19:03 GMT

In article <telecom13.558.6@eecs.nwu.edu> hhallika@tuba.calpoly.edu
(Harold Hallikainen) writes:

Sheesh!  I post a glib answer and somebody comes back with a
substantive reply!  :-)

> I've gotten several mile local loops (for FM radio stations)
> with 70 dB or more of dynamic range.  So, why do they only use 4
> levels in the 2B1Q?  Seems like a lot more would be available
> (stuffing more bps into a baud) before noise starts making it
> difficult to determine what the actual transmit analog level is.  Or,
> is the problem intersymbol interference where the level of one quat
> has an effect on those surrounding it, making the level of this
> particular quat difficult to determine?

The four-level signal was selected over a couple of three-level
proposals.  Remember they have to reliably distinguish an attenuated
signal in the presence of crosstalk.  The bit rate is 160 kbps, the
baud rate (symbol rate) is 80 kbps, and the energy peak is 40 kbps.
That's a lot higher than FM music, thus the shorter range.

>> To make all this work, impedance is specified, with very picky
>> transformers needed for S/T to meet spec.  S/T uses separate transmit
>> and receive wires, thus a 4-wire interface.  The BRI U uses one pair,
>> so there's some fancy echo cancellation done inside the transceiver
>> ("UBAT", in AT&T terms) chip.  Real fancy.

> It seems that if the characteristic impedance of the line is
> matched, there should be no "far end echo", or , at least, it should
> be substantially attenuated.  The "near end echo" (side tone on POTS)
> seems like it could be cancelled pretty well if we precisely know the
> impedance the line presents to the interface.  If the far end has
> indeed terminated the line with its characteristic impedance, then it
> seems the near end should have the same impedance.  I haven't messed
> with long twisted pairs to play with the transmission line effects,
> but it would sure be fun.  

The NEXT (near-end cross talk) is easy.  The FEXT isn't terribly
tough.  What kills things is the junk in the middle.  While ISDN lines
are not supposed to have bridge taps (little stubs off the middle),
the reality is that bridge taps happen.  And a little bit of bridge
tap creates whopping echo somewhere, which the chip tries to cancel.
We actually verified (unwanted!) bridge taps on lines which were still
sort of working, but not working well enough to be useful.

> We sell transmitter control and telemetry equipment to radio
> and television stations.  They are generally using a 3002 type circuit
> to send 1200 to 2400 bps full duplex data.  I'm wondering what other
> sort of leased line circuits they could use, and how the data could be
> coded to go down that line.

If all you need is remote control and telemetry, low-speed analog is
fine.  Most datacomm can use 56k though, especially at the same price.
Digital lines are not like 3002s; they don't have the same terminators
and may have different repeaters.  And no bridge taps :-).


Fred R. Goldstein   goldstein@carafe.tay2.dec.com
Opinions are mine alone; sharing requires permission

Date: Thu, 12 Aug 93 17:49:34 CDT
From: varney@ihlpe.att.com
Subject: Re: What is Category 5 (re: ISDN)
Organization: AT&T

In article <telecom13.558.6@eecs.nwu.edu> hhallika@tuba.calpoly.edu
(Harold Hallikainen) writes:

> In article <telecom13.551.5@eecs.nwu.edu> goldstein@carafe.tay2.dec.
> com (Fred R. Goldstein) writes:

> I've gotten several mile local loops (for FM radio stations)
> with 70 dB or more of dynamic range.  So, why do they only use 4
> levels in the 2B1Q?  Seems like a lot more would be available
> (stuffing more bps into a baud) before noise starts making it
> difficult to determine what the actual transmit analog level is.  Or,
> is the problem intersymbol interference where the level of one quat
> has an effect on those surrounding it, making the level of this
> particular quat difficult to determine?

   I've written about loop plant characteristics in the
"comp.dcom.isdn" a few times, and won't repeat all the stuff here.
Much of my meager knowledge is based on the book "Digital Transmission
Systems and Networks", Vol. II, M. J. Miller & S.V. Ahamed, Computer
Science Press, Rockville, MD, 1988.  A sub-chapter is headed "The
major limitations for loop data transmission" and details the
following imperfections of subscriber loops and their effects on ISDN
data rates:

   Physical:

- 1 or more guage changes between CO and subscriber
(each of these junctions can be a reflection point, and obviously
alter the overall loop characteristics)

- Bridged taps (open circuit cable sections tapped off the "main" loop)
(these provide reflections of transmitted signals, and attenuation
of signals from the far end, as well as major shifts in impedance.
Imagine a quarter-wave unterminated circuit bridged onto your loop
at one or more arbitrary points -- a reflected pulse will return
just in time to occupy the next bit position.)

- Loading coils, used on loops longer than about 18000 ft (5486 km)
(these suppress high frequencies to the point that everyone just says
these won't support the "U" interface) In the US in 1973, such loops
accounted for 24% of subscriber loops.  This number has certainly gone
down due to DLC, etc. deployment.

- Open wire (inductive 60 Hz coupling/crosstalk and impedance changes
with weather).

- Temperature (can change reactive component of impedance significantly).

   Electrical:

- Crosstalk (including crosstalk from adjacent T1 lines;)
- Lightning surges;
- EMF;

   In a 1973 loop survey of (pre-divest) AT&T loops, the average loop
was 7748.63 ft. long (2.262 km), had 4 sections and 1.64 bridged taps.
The average tap was 922.42 ft. (0.282 km).  Loops of less than 1000
ft.  had taps averaging 1333 ft. in length.  Characteristic impedance
toward the CO is markedly different than toward the subscriber on the
same loop.

>> To make all this work, impedance is specified, with very picky
>> transformers needed for S/T to meet spec.  S/T uses separate transmit
>> and receive wires, thus a 4-wire interface.  The BRI U uses one pair,
>> so there's some fancy echo cancellation done inside the transceiver
>> ("UBAT", in AT&T terms) chip.  Real fancy.

> It seems that if the characteristic impedance of the line is
> matched, there should be no "far end echo", or , at least, it should
> be substantially attenuated.  The "near end echo" (side tone on POTS)
> seems like it could be cancelled pretty well if we precisely know the
> impedance the line presents to the interface.  If the far end has
> indeed terminated the line with its characteristic impedance, then it
> seems the near end should have the same impedance.  I haven't messed
> with long twisted pairs to play with the transmission line effects,
> but it would sure be fun.  

   See above -- remember that ISDN deployment SHOULD be possible
without doing fancy impedance matching on loops, and SHOULD continue
to work as the loop plant changes with weather, time, repairs, etc.

> It seems like the throughput of a line is going to be limited by its
> analog dynamic range and the attenuation versus frequency.  Is the
> attenuation versus frequency (frequency response) reversible with an
> equalizer?  It seems that most equalizers that adjust the amplitude
> response also adjust the phase response (making it nonlinear).  

> It also seems that a long twisted pair would have a linear phase
> response (propogation delay relatively independent of frequency).  Can
> we feed a high speed multilevel pulse waveform in one end of a twisted
> pair and, with equalization, pull it back out the other end?  

   The trickest part for ISDN seems to be handling echo since the
2-wire line is both sending and receiving at the same time (in the
2B1Q version).  You seem to be looking at a uni-directional signal on
a two-wire pair.  The four-wire version of ISDN will work over dozens
of miles, since this is essentially a low-speed verison of a T1
(1.544MHz) line.


Al Varney - just my opinion

From: goldstein@carafe.tay2.dec.com (Fred R. Goldstein)
Subject: Re: What is Category 5 (re: ISDN)
Organization: Digital Equipment Corp., Littleton MA USA
Date: Mon, 16 Aug 1993 01:31:07 GMT

In article <telecom13.571.10@eecs.nwu.edu> varney@ihlpe.att.com writes:

> In article <telecom13.565.2@eecs.nwu.edu> goldstein@carafe.tay2.
> wrote:

>> ...  While ISDN lines are not supposed to have bridge taps (little
>> stubs off the middle), the reality is that bridge taps happen.  And a
>> little bit of bridge tap creates whopping echo somewhere, which the
>> chip tries to cancel.  We actually verified (unwanted!) bridge taps on
>> lines which were still sort of working, but not working well enough to
>> be useful. [...]

>   Hmmm!  Assuming your idea of an ISDN "line" is the same as mine
> (2B1Q U interface outside customer prem.), then howcome there's
> something like 16 test configurations that a "U" interface has to
> handle, and most have multiple bridge taps and multiple cable
> guages???  One has, if I recall correctly, a bridge tap off of a
> bridge tap.  Are you saying ISDN wasn't reliable over a 2B1Q two-wire
> interface with a bridge tap?  Or were you talking about the S/T
> interface (four-wire)?  Even the pre-ANSI AT&T U interface would
> handle 12000 ft. with a bridge tap or two.

Yes, our idea of an outside line is whatever was on the pole ahead of
time, at the "U" interface.  What we found out, all unofficially of
course because the actual trial details are not subject to being
publicized, lest the guilty be named :-), was that you can stretch a
line just so far.

The worst case of bridge-tap-itis occurred on an AMI line, using the
non-standard (but still more common here) AT&T 5E4 line card.  It is
rated for 12.5kf (given the loop in question) but actually worked fine
at 12.8, but NOT with a bridge tap!  But we also had a 2B1Q line or
two :-( fail with bridge taps on them.  All were beyond 12kf long.

The phone company folks say that ISDN is supposed to be installed on
unloaded lines without bridge taps, and the chip folks say that ISDN
is supposed to work over some degree of bridge taps, but a clumsy
technician working without line records can do amazing damage.
Remember these guys have no mobile radios or cellular with them; when
they want to talk to the office, they just climb a pole and tap in to
any old dial tone.  And if there's no dial tone on the buttinski, the
pair is free.  And ISDN doesn't put dial tone on a buttinski.  (Your
average lineman doesn't carry an ISDN buttinski and wouldn't know what
to do with one.)  


Fred R. Goldstein goldstein@carafe.tay2.dec.com
Opinions are mine alone; sharing requires permission.

From: varney@usgp4.ih.att.com (Al Varney)
Subject: Re: UDI vs RDI in ISDN
Organization: AT&T
Date: Fri, 7 Oct 1994 12:11:44 GMT

In article <telecom14.382.7@eecs.nwu.edu>, perpetual psycheness
<psyche@metronet.com> wrote:

> In the world of ISDN, what exactly does UDI and RDI mean?  And when
> would a person know which one to use?  (e.g. you try making a UDI call
> and then realize that it doesn't work, so then you try RDI?  What kind
> of indications does the network use to determine what to do?)

   UDI = unrestricted digital information -- also called 64U or "64 clear"
         or 64C
   RDI = restricted digital information -- also called "64 restricted"
         or 64R or sometimes 64I

   An ISDN user does not make a UDI or RDI call.  UDI/RDI is one part
of the Bearer Capability information passed to/from the network.
Specifically, the "information transfer capability" can be set to
either UDI or RDI.  The "information transfer rate" for a data call
would then be set to "64 kbps".  Together, the two parts are used to
request 64K IDSN data calls, commonly abbreviated as 64U or 64R calls.
Sometimes the 64U calls are calls "64 clear" or 64C; the "clear"
signifies that the channel is "clearly" or transparently carrying all
data values.  The 64R calls can be used to allow calls over facilities
(or trunks or circuits) that are not capable of transporting all data
values -- specifically the 8-bit value "0" (or "11111111" if you are
inverting data before transmission) cannot be carried.  The requester
of an RDI call is responsible for ensuring no "0" values are
transmitted -- if they occur, the network will alter at least one bit
to enforce the restriction.
   
   The "network" attempts to complete calls using the information in
the Bearer Capability you provide and information provided by the
operator of the network.  If you request UDI, the call will route (or
attempt to route) over facilities that have been designated by the
operator as supporting UDI.  If the facilities don't, in fact, support
UDI -- you get some of your bits mangled.  But this is no different TO
THE NETWORK than routing you over a bad facility -- the facility is
not working as the switches have been told it does.

> I think that UDI is supposed to be a 64kbps clear channel xmission
> (and I think you can have an RDI call over a trunk conditioned for UDI).

   If you request RDI, then the network (switches) will route over
facilities designated to support RDI, and will use UDI routes if no
RDI facilities are available.  Regardless of what that switch selects,
the switch at the other end of the selected facility will receive the
same Bearer Capability you originally requested, so that it can
attempt to select a trunk from all the original possibilities.

   The above "rules" apply to SS7-connected switches which are
carrying the selected UDI/RDI request via SS7 signaling.  I understand
RDI can be a problem when calling out of or into the USA, since many
countries use E1 (vs. T1) facilities that do not have a data value
restriction.

>   And I think that RDI means that the data is restricted in the
> sense that you can't have some number of contiguous 0's, which
> effectively reduces the maximum bit rate to 56kbps, right?  But then,
                                                      ^^^^^
                                                       NO!

> could you try sending data at 64kbps for an RDI call or is one of the
> bits for each channel used to keep sync (I'm assuming this is over T1
> type trunks for RDI; I guess E1 type trunks don't have this problem
> and are 64kbps clear channel trunks by nature?).

   The 56 kbps bit rate vs. RDI is confusing.  I can't assure you that
I can un-confuse the issues.  RDI is a restriction on 8-bit data
values; the value "00000000" is prohibited.  But there is no rate
adaption or change to the data rate implied by RDI.  You send at 64
kbps, period.  The reason for RDI is due to T1 facilities that don't
use B8ZS or other schemes to remove the original T1 restriction that
no more than 15 "0"s could be sent over the facility -- more "0"s
meant loss of sync.  (Some pre-B8ZS schemes used non-adjacent channels
for data or borrowed bits from a control channel to remove the RDI
restriction.)

   The use of 56 kbps in the USA/Canada and a few other T1 countries
came about because one bit of the channel was sometimes used as a
signaling bit (on-hook,off-hook indicator).  In standard T1, this
happens in every sixth frame.  Because the sixth frame on one circuit
isn't the sixth frame on another, the 1-out-of-6 position cannot be
predicted end-to-end.  This effectively removes the ability to use 8
bits -- 7 bits at the 8000 frames/second T1 rate yields 56 kbps.

   When SS7 signaling is used, switches don't need the signaling bit
in the T1 channel.  For some switches, this permits 64 kbps over RDI
as soon as SS7 is implemented between 2 switches.  Others require
hardware upgrades to remove the "hard-coded" signaling bit.  But even
when the signaling bit is removed, the T1 transport facility still
can't handle unrestricted data -- that requires a change to the
facility (sometimes a hardware upgrade, sometimes just changing data
in the facility).

    From an ISDN end-point, 56 kbps is a V.120 rate adaption scheme.
The network knows which bit to ignore (and force to a "1").  The far
end end-point will get 64 kbps data, and will ignore the eighth bit.

> But, 64kbps or 56kbps doesn't necessarily mean UDI and RDI, respectively, 
> does it?

   Correct.  You can, in theory, use UDI or RDI with 64 kbps calls.
Or you can use UDI or RDI with 64 kbps calls rate adapted to 56 kbps.
To confuse the issue, Bellcore requirements for ISDN don't recognize
RDI, so they only talk about 64 kbps via UDI and 56 kbps rate adaption
over 64 kbps via UDI.  Unfortunately, there are some vendors who
adopted the rule that 56 kbps calls would use an "information transfer
capability" of RDI.  When connecting to networks/equipment that only
recognizes an "information transfer capability" of UDI, the call will
be refused.

   Also, some carriers (LEC and IXC) may still have some 56Kbps
facilities (trunks and "switched 56" customers).  If such trunks are
in the path of the call (which could start out 56 kbps with UDI or
with RDI, the switches/customer beyond the 56Kbps facility receive
only the called number -- and must assume a value of UDI or RDI for
your 56Kbps call.  The value assumed depends on the network, the
switch and lots of other semi-random information.

> And how does rate adaption come into the picture?  Is it possible to
> have 9600bps data stream rate adapted up to 56kbps for RDI?  or up to
> 64kbps for UDI?  Is V.120 or V.110 common in the U.S.?  What is
> commonly used overseas?  What are the advantages of V.120 over V.110
> or are there not any?

   V.120 common in USA.  V.110 common in Europe.  V.120 is more
flexible, but has some added overhead.  V.110 has less overhead if you
can keep its fixed sub-channels full of data.  9600 bps is supported
by both.  Since 56 kbps is itself a rate adaption, it's hard to say
9600 bps is supported by 56 kbps -- I believe you could indicate 56K
in the Bearer Capability rate adaption and would use LLC to indicate
the actual partitioning of one or more 9600 bps over the 64 kbps
B-channel.  (Keep in mind that the ISDN end-points never actually
transmit at anything other than 64 kbps.)  If I remember correctly,
V.120 rate adaption for 9600 bps forces 1 out of the 8 bits to a "1",
so the same bit pattern would work over both 56 kbps or 64 kbps calls,
whether UDI or RDI.

   Asking a more specific question on comp.dcom.isdn might get you a
more correct (and shorter) answer ...


Al Varney - just my opinion