Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!ames!ucbcad!ucbvax!kitty.UUCP!larry
From: larry@kitty.UUCP
Newsgroups: comp.dcom.telecom
Subject: Submission for mod.telecom (Telephone line quality)
Message-ID: <8704190535.AA24165@seismo.CSS.GOV>
Date: Sun, 19-Apr-87 00:35:32 EST
Article-I.D.: seismo.8704190535.AA24165
Posted: Sun Apr 19 00:35:32 1987
Date-Received: Sun, 19-Apr-87 23:39:48 EST
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 143
Approved: telecom@xx.lcs.mit.edu

In a recent article Greg Earle (earle@jplpub1.JPL.NASA.GOV) writes:
> I live in the 213 Area code in L.A.  I recently moved, and want to get two
> lines for my new abode, one of which I will use exclusively for a modem
> line.  When I talked to Pacific Bell I was told I could (for a nice high
> fee, of course) get a `Data Access Line' which would (presumably) run from
> the local switching office to my home; a higher grade line would replace the
> normal voice grade phone line.  I was told that this was recommended for
> anyone doing data transmissions of 2400 baud or higher.  I almost bit; but
> then I thought, what about the rest of the way?  I would be calling JPL in
> Pasadena 99% of the time, which is in Area code 818, prefix 354.  Since I'm
> not a TELECOM expert, I just surmised that the calls I would make would go
> from my home, over my `good' data line, to the local switching office; then
> to whatever the local switching office for Pasadena is, and then over a 
> (presumably) standard voice grade line to my other modem.
>  
> My question for you experts is (a) is this something like the real path that
> the call will take (3 hops; home <=> switching office <-> s.office#2 <-> work)
> and (b) if this is so, then is there any point in getting a higher grade line
> for one's home, when one has no control over the line quality for the other
> 2/3 of the connection ?!?

	Let's break up this discussion into two areas: (1) quality of central
office subscriber lines (i.e., between your home/office and the telephone
company central office); and (2) quality of lines between telephone company
central offices.

	I'll answer (2) first, because it is the easier answer.  In general,
the quality of an interoffice trunk (i.e., a line connecting two telephone
company central offices) is FAR superior to the quality of any subscriber
line.
	In keeping with the DDD network operating goals and an overall
transmission design plan called VNL (Via Net Loss), the transmission loss
on most interoffice trunks originating at End Offices (Class 5) trunks is
carefully kept below 4.0 dB.  Interoffice trunk transmission loss on
Toll Center (Class 4) and up to Regional Center (Class 1) switching offices
is carefully kept below 2.6 dB, or even below 1.4 dB, depending upon the path.
	Such interoffice trunk design is generally done so precisely that
loss is kept within +/- 0.1 dB of the design goal on any interoffice trunk
of a given path to assure a uniform transmission quality.  In addition, such
interoffice trunks are generally equalized to have a reasonably flat
transmission characteristic between 300 and 3,000 Hz.  Furthermore since most
interoffice trunks originating in End Offices (except in high-density urban
areas where adjacent central offices are close together) are four-wire (i.e.,
separate receive and transmission paths - one for each direction) and are
terminated in a precision hybrid-network, the transmission quality will be
far superior to anything which could ever exist on a two-wire subscriber loop.
All interoffice trunks of Toll Center and up origin are four-wire.
	Noise level, ERL (Echo Return Loss), and other parameters which affect
the quality of transmission are also kept within precise design goals on
interoffice trunks.
	The BOC's and larger independent operating telephone companies
check the transmission quality of interoffice trunks on a regular basis,
often using automatic test apparatus such as ATMS (Automatic Transmission
Measuring System), CAROT (Centralized Automatic Reporting of Trunks), TFMS
(Trunk and Facility Maintenance System), etc.  Trunks which fail to pass
these automatic tests are disabled until repair is effected.
	So the point is: under virtually all circumstances, you should
have little concern about the transmission quality of interoffice trunks,
as compared to your own subscriber loop.
	(IMPORTANT NOTE: The above applies to what is traditionally known
as the DDD network; some of this standardization has gone to hell with the
advent of Alternate Long Distance carrier. The above information should
still be safely applicable if your call is intra-LATA in length, is
inter-LATA but served by the same operating telephone company at both ends,
or is routed through AT&T.  This is NOT a "plug" for AT&T; it's just a
simple fact of life since AT&T still runs all the major toll switching
centers in the U.S.)

	Now we'll get back to the first topic, which is the local subscriber
loop.  Subscriber loops are generally designed based upon only two parameters:
(1) DC resistance and (2) transmission loss at 1.0 KHz.
	Since most central office apparatus has subscriber loop resistance
limits between 1,200 and 1,500 ohms, resistance of a subscriber loop is
controlled to be within this range by selecting cable layout with sections
that have large-enough wire gauges (the SMALLER gauge sections are generally
CLOSEST to the central office).  If the resistance limit still cannot be
economically met with wire gauge selection alone, then a signaling range
extender (loop extender) will be connected to the line; this device is always
located in the central office.  Under this condition, the subscriber loop
resistance may be >> 1,000 ohms, but the loop extender has the sensitivity
to support such a higher resistance.
	In simple terms, the transmission loss of a subscriber loop is
directly proportional to its DC resistance - so a long loop will also have
a large transmission loss.  Invariably, subscriber loops greater than 10 kft
(kilofeet) in length will have loading coils installed every X-kft (there are
different loading schemes which use different spacing between loading coils);
these loading coils add inductance which compensates for the attenuation of
the loop that results from distributed capacitance.
	The worse case loss that any reasonable telephone company would impose
on a subscriber loop is about 9 dB.  New York Telephone, as an example,
tries to keep loop loss to no more than 6 dB - but not every loop is that
lucky. :-)
	If a maximum subscriber loop transmission loss goal of 6 to 9 dB
cannot be met through loading and cable routing, then a voice-frequency
repeater is installed on the subscriber line; this repeater is almost
always installed in the central office on the required lines.  Sometimes a
combination loop extender-repeater is used, but in many cases there will
be two discrete devices in the central office.
	Under most circumstances, subscriber loop loss is measured at only
1.0 kHz.  However, a frequency-vs-attenuation plot of a subscriber loop
can look like a roller coaster!  Since the human ear is rather forgiving,
for voice applications most telephone companies care little about the
frequency-vs-attenuation curve on a POTS (Plain 'Ole Telephone Service)
subscriber loop. 
	However, MODEMS can care about this curve!  Subscriber loops
which run through mixed gauges of loaded cable, and/or run through
voice-frequency repeaters (especially of the older E6 variety) can have
some pretty ugly frequency-vs-attenuation curves.  The only way to flatten
the curve (and thereby make the line more attractive to data) is by means
of an equalizer, a better quality repeater (hybrid rather than a negative
impedance type like the E6), along with more careful design engineering of
the particular loop.  
	This equalizer, repeater change, and additional engineering is not
necessary for most voice applications - so it isn't done.  However, an
equalizer and additional engineering DOES result in a superior subscriber
loop for data purposes.
	So, telephone companies generally charge more money for a better
subscriber loop design for data applications.  If you are making serious
use of data transmission at > 2,400 bps, the comparatively small additional
monthly and installation charge is well worth it to get a better subscriber
loop.
	I don't make a habit of defending telephone companies, but I must
say that I feel such an additional charge is reasonable.  They do have to
install additional equipment (under most circumstances), and certainly do
have to perform specific engineering on the design of the subscriber loop.
	As far as noise level on subscriber loops is concerned, this is
generally caused by wet terminals and splices and is really a repair problem.
There is little that can be done to reduce noise level on a subscriber loop
other than to track down wet or poor splices.  However, for the price of
a better quality loop, one generally gets a quantitative noise measurement
with some attempt at repair if the noise level is beyond normal limits.

	Now, to sum up and answer the $ 64 question: In my opinion, for data
transmission > 2,400 bps on LONGER SUBSCRIBER LOOPS (say > 2 miles from the
central office), an additional charge for a better quality loop (i.e.,
flatter frequency response and lower attenuation) IS a worthwhile expense.
At least, with a known good loop of known characteristics, one can look
elsewhere should data errors become a problem.

<>  Larry Lippman @ Recognition Research Corp., Clarence, New York
<>  UUCP:  {allegra|ames|boulder|decvax|rocksanne|watmath}!sunybcs!kitty!larry
<>  VOICE: 716/688-1231        {hplabs|ihnp4|mtune|seismo|utzoo}!/
<>  FAX:   716/741-9635 {G1,G2,G3 modes}    "Have you hugged your cat today?"