Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!brl-adm!adm!CONLEY%UKANVAX.BITNET@wiscvm.wisc.EDU
From: CONLEY%UKANVAX.BITNET@wiscvm.wisc.EDU
Newsgroups: comp.unix.wizards
Subject: Re: uda50 delay settings
Message-ID: <7105@brl-adm.ARPA>
Date: Sun, 26-Apr-87 10:12:42 EDT
Article-I.D.: brl-adm.7105
Posted: Sun Apr 26 10:12:42 1987
Date-Received: Sun, 26-Apr-87 23:30:04 EDT
Sender: news@brl-adm.ARPA
Lines: 26

The jumper is there for those folks who have problems with Unibus data late
errors due to heavy DMA usage ( e.g. having a UDA50 and RK07 on the same
Unibus ). The jumper selects the delay ( microseconds ) that the UDA50 will
wait before issuing an NPR request to the Unibus; i.e. it allows one to change
the DMA 'load' imposed by the UDA50.

The considerations for which setting to use depend on the number and type of
devices you have on that Unibus. If you have several fast DMA devices, you'll
likely need some delay. In general, if you decrease the delay, and encounter
no Unibus data late or timeout errors, be happy. It requires some time and
tinkering. Whether a smaller delay buys you anything is another issue, I've
always intended to explore that and never had the time.

After several months of experimentation, we're using the 0 delay. I should
mention that we migrated most devices off of that Unibus, leaving only the
UDA50, a DMF32 and a TU80 ( used infrequently ). I should also mention that
DEC recommends *not* using the 0 delay.

        Cheers,

        Dennis R. Conley
        Computer Science Dept
        University of Kansas

        CSNET:  conley@csvax.cs.ukans.edu
        BITNET: conley@ukanvax