From: floyd@polarnet.com (Floyd Davidson)
Newsgroups: comp.dcom.telecom.tech
Subject: Re: How can this T1 be half duplex???
Date: 12 May 1996 03:03:57 GMT

wamacher@rahul.net wrote:
>Dave Gellerman (dave_gellerman@newbridge.com) wrote:
>: In article <4moc53$8ms@unet.net.com>, razel@tac00.net.com (Ray Hazel) wrote:
>
>: >Which leads me to one other question.  Is there a way to run data over an 
>: >"unframed" DS1/T1 where you could get 1.544Mbs data?  Not that losing 8k of
>: >bandwidth is that critical, but I was curious if there was something out
>: >there that operated data that way.
>: >Ray Hazel
>: >razel@net.com
>
>: Actually, back in the dark ages of T-1 (early 80s), a lot of customers ran
>: an "unframed" T-1 signal (using stuff like Timeplex muxes that had

Heh heh, calling the early 80's the dark ages of T1 is kinda
funny!  TDM carrier had been in common use for over a decade by
then, and I'd even venture to say that about 1980 marks the
beginning of "modern" T1 systems, though most of them installed
about that time are now being replaced by fiber systems.  But they
still function perfectly well, it's just that more bandwidth is
now required and the wire facilities are being replaced by fiber
provide it.

>: proprietary framing patterns).  The phone company really twisted the
>: customers arms hard to move all traffice to standard framed T-1 in order
>: to allow the phone company to monitor the quality of the traffic (by
>: watching for frame bit errors in SF, and CRC-6 errors in ESF.  You might
>: still be able to RUN unframed, but I suspect most phone companies would
>: PREFER you run framed.  From a transmission point of view, there's really
>: nothing limiting a T-1 to framed service (at least if we are really
>: talking HICAP service here, and not fraction T-1, or DID trunks, or ISDN
>: PRI).

That is very true!  You can't imagine the lost feeling (and
annoyance) that technicians have when a customer reports a data
corruption problem on an unframed T1.  The telco simply has no way
to test it short of having someone, at the customer location,
completely break the circuit and apply a known test signal.
Invasive tests are hard on the customer, and "windshield wiper"
time to put a technician on site is expensive.

>: The first poster is just completely wrong on the full/half duplex issue. 
>: T-1 is inherently full duplex.  And actually, the source of the clock for
>: the T-1 is sometime negotiable.  The T-1 transmission infrastructure is
>: asynchronous...as long as the clock is stable, it shouldn't matter where
>: the timing comes from (if it's not stable, frame slips will occur).

The infrastructure isn't really asynchronous, but isochronous.
That means async data is send over a synchronous bit stream.  The
timing is plesiochronous, which means that the clocks can be
separate, (as stated above) but must be accurate.  On a stand
alone basis only Stratum 1 clocks are that accurate, and all
others must be synced to a Stratum 1 source.  Stratum 2, 3, and 4
clocks are progressively less stable (rate of accuracy change)
after synchronization is lost.

>The clock for T1 and E1 is normally derived from the incoming T1/E1 
>signal, so that it is exactly the same, pulse for pulse. One of the CCITT 
>G.xxx specs specifies the maximum jitter and wander at the input 
>(+/-50ppm).  Normally a crystal controlled phase locked loop is used 
>to reduce jitter.  There is supposed to be a master clock in Omaha or 
>somewhere to which the whole system is synchronized.  

The single master clock concept was the original plan, but of
course now that The Bell System isn't the only system there are
many Stratum 1 clock sources (GPS and Loran C being two that can
be made available just about anywhere, as are relatively cheap
cesium beam clocks now.)

>Clock derivation from the incomming signal goes on and on at each 
>successive channel bank and CO, so that the whole system is exactly 
>synchronized. 

Well...  no and yes (just not that way).  Each CO will have a
master distribution system, the synchronization of which can be
traced directly to a Stratum 1 clock.  Every DS-1 interface in the
CO will be timed with clocking from that distribution system and
will be totally independent of clock timing at its distant end.
Generally a major location (Toll Center, etc.) will have a Stratum
2 clock, a CO will have a Stratum 3 clock (e.g., in the switch).
Individual equipment, like channel banks, have Stratum 4 clocks
which are never used to provide sync to anything other than the
specific channel bank at the distant end if it is using loop
timing.

>The clock must be more than stable, it must be absolutely 
>phase-locked.  Otherwise the slightest difference in frequency will 
>result in slips fairly often.  

Phase is (relatively) insignificant, and only absolute accuracy is
important to network synchronization.  In general, buffered receive
circuits are used to allow for phase variations (wander and
jitter).  As long as the accuracy of the clocking is maintained
(i.e.  the frequency) there will be no buffer over or under runs
(which result in slips).

>Slip may not affect voice much, but is death to data.  

At least for controlled slips where no loss of frame occurs, but
"uncontrolled slips", where the input buffering is not large
enough to repeat or drop and entire frame, cause a loss of frame
condition and a re-sync, and the result is dramatic even for voice
usage.  Most channel banks cannot do controlled slips.

Floyd

-- 
Floyd L. Davidson          Salcha, Alaska         floyd@tanana.polarnet.com