Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!rutgers!ames!ucbcad!ucbvax!hplabs!well!spg
From: spg@well.UUCP
Newsgroups: comp.sys.amiga
Subject: Re: flakey memory boards/CHIP contention
Message-ID: <3077@well.UUCP>
Date: Sat, 16-May-87 12:55:04 EDT
Article-I.D.: well.3077
Posted: Sat May 16 12:55:04 1987
Date-Received: Sun, 17-May-87 02:59:47 EDT
References: <8705110100.AA27374@cogsci.berkeley.edu>
Reply-To: spg@well.UUCP (Stephen P. Grant)
Organization: Whole Earth 'Lectronic Link, Sausalito, CA
Lines: 40


[line eater food............]

In Article <8705110100.AA27374@cogsci.berkeley.edu> 
 bryce@COGSCI.BERKELEY.EDU writes:

<Several different memory expansion boards all cause the same problem:
<Streaks and missing bits start to show up on the screen. Sometimes entire
<objects like the resize and front/back gadgets become corrupted on all
<windows.  It would seem that these boards, all which have their own
<memory controllers, are affecting CHIP memory.
<The esiest way to crash these defective expansion products is to bring up
<a 'clock' window or two and let it sit for a few hours. The clock display
<will trash and eventually crash with an address error or illegal
<instruction trap.
<What in the design of these boards is defective?  Would any designer that
<has solved or avoided this problem like to comment on the technical
<aspects?
This problem could have a number of causes.  It is possible that the
memory board is malfunctioning in some way.  However, it is more likely
that the design of the memory is marginal.  Since the expansion connector 
is DIRECTLY connected to the MC68000, the Amiga is very sensitive 
to the effects of noise and the expansion spec reccommends that ALL of
the signal lines be buffered.  The spec further states that the trace
lengths should be kept as short a possible to reduce capacitive loading.

The prototype of the Alegra was built on a wire-wrap board and exhibited
the same types of crashes described above.  I spent about a week and a half
chasing this "ghost" and finally was convinced that the problem would
be solved by the 4 layer pcb that we designed.  We did a proto run of 5
boards and sure enough the flakiness was gone and the level of signal
noise was significantly reduced.  The moral of the story is that to design
reliable expansion devices for the Amiga 1000, one should follow the spec
and use good pcb design techniques.  Don't forget lots of decoupling
capacitors also.  We used a 4 layer board, both for the increased noise
immunity and to increase to odds of passing FCC.

Steve Grant
Engineering Manager
Access Associates