JOE@oregon.uoregon.edu (09/27/90)
We've been experiencing a severe problem with noise on our dialins which began, coincidentally, immediately after we got a new in-house phone system here at the University of Oregon. :-) In talking with an E.E. friend and describing the symptoms (phantom curly right braces and other characters typed "player-piano" style with no keyboard assistance from the user required!), he suggested that we are experiencing "slippage." Because we run full-duplex, we've been able to determine that sometimes the noise is introduced "inbound" (i.e., the phantom characters appearing on the user's screen are also seen by the system they've dialed into), while other times the noise is introduced "outbound" only (i.e., the phantom characters appear on the user's screen, but are never received by the remote minicomputer -- the garbage disappears when the user forces a screen refresh). The phenomena is stochastic, and apparently uncorrelated with anything else we've been able to monitor (weather, system load, time of day, type of modem user has, modem speed, particular modem dialed-in-to, etc.). Since this is driving our users crazy, we'd really like to resolve this problem. Can anyone provide me with a citation to some technical references on slippage? Has anyone come up with a fix for this sort of problem? (I'd love to hear, "Well, if you just put an xxpF filter capacitor on each of the modem lines...") Thank you, Joe St Sauver (JOE@OREGON.UOREGON.EDU or JOE@OREGON) Statistical Programmer and Consultant University of Oregon Computing Center
brian%cyberpunk@ucsd.edu (Brian Kantor) (09/29/90)
When you get "twinklies", consisting of characters having a lot of bits on (especially high order bits) like }, you are probably seeing your modem attempting to resynchronize. (1200 bps 212 modems use synchronous transmission between each other even though you are using async to talk to them.) My experience is that the A-#1 cause of this is a defective or misconfigured interface card on one or both ends of one or more of the circuits that connect your university's phone switch to the local telco's digital switch. What happens is that the a/d and d/a conversions at opposite ends of the trunk occasionally drop a little data. In other words, one or more of the 8k/sec samples was damaged and was discarded at the receiving end. This has NO measureable effect on voice - completely inaudible - but it makes the modems lose sync and they blow 1's bits at each other until they resync, so you see lots of twinklies. Sometimes they switches are misclocked so that they drop one sample out of every N, so you see a periodic burst of twinklies every M seconds. This is always repairable, but it will probably take a transmission specialist to bring his special test equipment and check for it as the normal telco voice quality measurement stuff won't show the problem. We had this problem big-time here at UCSD when the main campus was on one machine and the student housing on another in the same telco office - the two switches in the same building couldn't talk to each other without sync slips. The DMS-100 switch was famous for this - I heard they had a production run of line cards that came from the factory misconfigured slightly so that they worked ok for voice but got lots of slips. I understand they had to pull every single card out of the switch to check the jumpers or some equally boring task. Now that PacBell has fixed that problem with our local switch, we see sync slip storms only once or so a year - typically when they've just upgraded one of the central office switches in some other part of town. A quick call to their technical people handling the campus gets it fixed right fast. I get the impression we find out about it before they do, sometimes. (We've got over 200 dialup lines and about 8,000 students and faculty using them 24 hours a day, so we have a large window of opportunity.) My experience parallels others in this regard - once you get high enough in the telco to find someone who can understand what you're saying, they'll get it fixed. If you're not in a position to bang on them from an official campus position, try to talk to whoever runs the switchroom in your campus phone facility and explain to them what's going on. They can get to the right people in the telco, eventually. Brian
decoste@iro.umontreal.ca (Ronald Decoste) (10/02/90)
We are also experiencing transmission errors with are dialup lines since our new phone switch went into operation. Things were so bad that we had to revert to direct lines from the CO until a solution is found. The switch (multiple Meridian SL1's) is connected to the DMS-100 by a fiber DS-3 link for aapprox. 350 trunks. Brian Kantor's article suggests that some jumpers must be set properly on some interface cards. Does anyone know more about this ? Which jumpers on what cards for the DMS and for the SL1 ? Ronald
mdv@uunet.uu.net (Mike Verstegen) (10/04/90)
decoste@iro.umontreal.ca (Ronald Decoste) writes: >We are also experiencing transmission errors with are dialup lines >since our new phone switch went into operation. Things were so bad >that we had to revert to direct lines from the CO until a solution is >found. >The switch (multiple Meridian SL1's) is connected to the DMS-100 by a >fiber DS-3 link for aapprox. 350 trunks. >Brian Kantor's article suggests that some jumpers must be set properly >on some interface cards. Does anyone know more about this ? Which >jumpers on what cards for the DMS and for the SL1 ? I have experieced this problem when a T-1 has been loop timed at each end of the circuit. In timing distribution, one end (the "better" timing) must always be master and the other end loop (or "slave") timed. By definition the telco will be the master timing source. If your DS-3 is terminating into an external M13 multiplexer, check the timing connection between the M13 and the T-1 terminations on the SL-1. The telco fiber DS-3 should time the M13 mux and the M13 will in turn provide master timing via the T-1 to the SL-1 which should be loop timed. If the the SL-1 is master timed (or the T-1 side of the M13 is loop timed) you will see slippages that will cause data transmissions problems. If the DS-3 terminates directly into the SL-1, the integrated interface probably has the equivalent options settings, either in hardware or software. Mike Verstegen Domain Systems, Inc Voice +1 407 686-7911 ..!uunet!comtst!mdv 5840 Corporate Way #100 Fax +1 407 478-2542 mdv@domain.com West Palm Beach, FL 33407