jackson@ttidcb.UUCP (Dick Jackson) (02/07/85)
This is a sort of followup to the request for info re dial up modems at 4800 and 9600 bd. (I don't have the patience to figure out how to do a "followup"). We recently tested regular 4-wire 9600 bps modems from three vendors between Santa Monica into Huntington, Long Island. The modems were: General Data Comm 9600 QP (EP has same transmission system and is virtually equivalent), the UDS 9600 A/B (the regular 9600 has similar transmission circuitry) and the Codex 2660 (14.4 kbps with Trellis). Of course we had to make two dial calls using DAAs (sometimes bundled into various dual-dial auto answer units - every vendor has his own version). The one you'd think worked best - did! The worst showed about a 10 times worse block error rate. But! worst worst was good enough so that any self-respecting link protocol could take it in stride, i.e. less than five block errors per minute (511 bit blocks). About every five call attempts the connection was so bad (BLERT counter berserk) that an abort was necessary. Remember that we were dialling two lines, so its really about 1 in 10 switched circuits called. The signal quality light on the modems by the way provides a fairly good indication of go/no go. If it was anything but steady the block error rate was approaching what I considered uncomfortable to unnacceptable ~ 3-10% retries. (BLERT tests were done with the remote modem in loopback.) The path we used is probably not the worst you could pick but is fairly testing. Santa Monica is General Telephone country. The lines probably have to pass through Pactel before hitting ATT&T (anyone know?). Then they have to get around the NYC mess at the eastern end. In a nutshell, our experience says - Go for it, it will probably work.
lauren@vortex.UUCP (Lauren Weinstein) (02/10/85)
Actually, test calls from Santa Monica to Huntington, Long Island are not a very taxing test of a modem. Most of GTE Santa Monica is new EAX/ESS switching, and there is a major GTE long-haul toll switching center right there in Santa Monica as well. This makes the probability very high that the tests used nice, clear, digital trunks over most of the route. For real testing, you need to call (and be called from) areas that are more rural (where many Usenet sites are located, by the way) and farther from their local telco central offices and toll centers. Overall and specific path variability is usually much higher when one or both ends of the call are in "odd" locations. For example, I frequently call from my "well-located" (close to central office) Los Angeles location to Douglasville, Georgia (17 miles outside Atlanta). Only about 50% of the calls are usable for ordinary 1200bps 212-type data. --Lauren--
dsi@unccvax.UUCP (Dataspan Inc) (02/12/85)
One of the major modem manufacturers (UDS) have a facility whereby
you can dial a number in a distant city, dial a second number in same
distant city, and have the two connect. These were in place in several
major markets, and were used presumably by UDS to do real world testing
for the 9600 A/B.
Has anyone been able to confirm the effects of SLIC compressors and/or
noise gates on the 9600 A/B? We were having one heck of a time in the 662
exchange at Buffalo, NY. When the lines were requested with RJ-45S jacks,
the "effect" that was munging the training burst magically went away. Several
other circuits which are "normal" still fail to work correctly.
The effect was that the first part of the training burst was being cut
off by something like an expander with a very long attack time. In fact, if
you wrote software to take DTR high, low, then high again, the faulty
circuits would then work about 50 $% of the time. (high=assert, low = negate..
no flames please)
What was * REALLY * funky about this compression business was that
within Orchard Park, nothing would ever work at all. When dialing from
or into Orchard Park via long distance from about 10 major American cities,
the equipment worked perfectly. Where would such a noise gate or other
device be located ?
dya
/