jbm@uncle.UUCP (John B. Milton) (11/12/88)
In article <169@rtjpc.UUCP> jacob@rtjpc.UUCP (tom jacob) writes: >My university has recently acquired a Telebit Trailblazer for its Vax >11/780 (BSD 4.3). It has no trouble talking to all of its connections >EXCEPT my 3B1 (internal modem, 1200 baud). When calling my machine, the >Vax connects OK and logs in OK, but when it tries to "sendfile" it >gets "pkcget: alarm #####" signals. > >I have heard that Unix PCs have trouble talking to Trailblazers. Has >anyone figured out how to solve this problem? I have a UNIXpc with a Trailblazer (TB), and am connecting to other UNIXpcs. For a while I had this problem too. After a lot of munging it just went away. I have thought about this problem on other occasions, and never have been able to connect up the problem. Someone else may have already figured this one out. Here is what I think is going on: The serial chip the UNIXpc is using is the NEC 7201 (Intel 8274). This is really a hell of a fancy chip. I think AT&T had more ideas for the chip than what the UNIXpc currently does with it. (SDLC, etc.) I heard from someone else on the net that this chip has a very nasty problem: If it does not get at least 1.000 stop bits, it drops the character. The TB has a corresponding nasty habit: when it's really trucking along fast, it shortens up the stop bits a little. This is why the connection to the UNIXpc starts out fine, but when the TB BLASTS out a full size UUCP packet, it bunches the characters up a little too much and the UNIXpc's 7201 chokes. The TB only bunches up characters when it has the opportunity to do so: when it is getting them from the host computer faster than it can send them. One would think this would be a generic problem with Zilog 8530s (the DUART in the TB) sending to NEC 7201s, but it doesn't seem to be. This brings up an interesting question. The whole idea behind the TB is to get it there faster, right. Consider the case where you know you're going to get an interrupt to a 68000 (the uP in the TB) everytime the DUART needs a character. What would happen if you reduced the stop bit just a tad so that the added time of the interrupt routine would bring the total stop bit time down to just one stop bit? What about if you take it even further down, knowing that there is supposed to be some slop in the serial chip on the other side? Hmm. Maybe stuff like this would break an old ASR-33 :) ****** Anyway, the solution is use a serial speed of 1200 baud from the computer to the TB when there is a UNIXpc on the other end of the phone line. If your host also shaves stop bits, you might have to increase the host to TB stop bits to something >1.0 ****** Let me know my mail if this fixes the problem. If you have found similar problems, let me know about those too. John -- John Bly Milton IV, jbm@uncle.UUCP, n8emr!uncle!jbm@osu-cis.cis.ohio-state.edu (614) h:294-4823, w:764-4272; MS-DOS is a beautiful flower that smells bad!