fredc@petsd.UUCP (Fred Cassirer) (01/24/89)
I am involved in building a memory board (well I wrote some memory tests and other S/W stuff ...) for an Amiga 1000, and we are having a problem with what we think is bus noise. The board checks out ok, we can read/write and read-modify-write all fine by reading/writing before doing an ADDMEM (we have disabled the autoconfig stuff). We can successfully "poke" an assembly program into the memory and the execute it via an indirect function call in C. All seems well. We can ADDMEM, but soon afterwards (like after run emacs) we will guru. Usually during a disk I/O (but I can't prove that). Most of the time it looks like something has written all thru chip memory because the screen goes nuts. On the analyzer we are seeing noise glitches on the control signals (AS, R/W). We have heard that some kind of bus termination may clean up the problem. Does anyone know what buses or signals should be terminated and what to use for termination? We have new pals (10ns) and have done the grounding trick. I'm told the board is very "clean" as far as signals go, the memory chips are very fast (in the 25ns range) Following is a rough sketch of what we see happening: AS -------. /\._______ | /\______________________/\/\____________| \/ Data .---------. Bus ________________________________| |__________ |_________| RW ________________________________/\ __________________ | \/ | `- - - - - - - - - - - - - -. /\.- - - - -' \/ GRND _______________________________ /\___________________ \/ Sorry if my graphics are a bit poor. Any help would be greatly appreciated. Thanx. -- /\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\ < Fred Cassirer ... rutgers!petsd!fredc or > < Concurrent Computer Corp (preferred)... rutgers!petsd!pecnos!fredc > \/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/
anakin@gpu.utcs.toronto.edu (Anakin Research) (01/25/89)
I have found that the best termination is to terminate the 16 data lines and the four control signals (*AS, *LDS, *UDS, & R/*W) with a 4.7K resistor to 5V+ and a 1K resistor in series with a 1000 pf. ceramic cap to ground. I was told this idea by the kind people at Microbotics. It has greatly increased the reliability and the number of peripherals on the bus. Brad Fowles
randy@bcsaic.UUCP (Randy Groves) (01/26/89)
In article <1430@petsd.UUCP> fredc@petsd.UUCP (Fred Cassirer) writes: > > I am involved in building a memory board (well I wrote some memory tests > and other S/W stuff ...) for an Amiga 1000, and we are having a problem > with what we think is bus noise. > A side note: Anybody built or know of a commercial board for the 1000 that can use either 256kbit or 1Mbit SIMMS that are used in the MAC?? They seem to be readily available. Thanks much!! -- -randy groves - Boeing Advanced Technology Center UUCP: ..!uw-beaver!uw-june!bcsaic!randy USNail: Boeing Computer Services CSNET: randy@atc.boeing.com PO Box 24346 M/S 7L-68 VOICE: (206)865-3424 Seattle, WA 98124
wn0e+@andrew.cmu.edu (William Nichols) (01/31/89)
Brad Fowles writes
>I have found that the best termination is to terminate teh 16 data lines and
the four control >signals(*AS,*LDS,*UDS,&R/**W) with a 4.7K resistor to 5V+ and
a 1K resistor in series with a >1000 pf ceramic cap to ground.
I have also had some problems with bus noise, and would like to try the
terminator, but I do not know which lines on the bus those 20 lines correspond
to. Can anybody help me out?
Thanks in advance.
WRN
Dept. of Physics
Carnegie Mellon University
anakin@gpu.utcs.toronto.edu (Anakin Research) (01/31/89)
I've also had a number of questions in my E-mail on the bus terminator to help with bus noise on the A1000, so here t'is pin 1,2,3,4 Ground pin 5,6 5V+ *Note: Grounds also at 13,25,37,49,61,73,85 Signals to terminate: (20) pin 63 D15 pin 65 D14 pin 67 D13 pin 68 R/*W pin 69 D12 pin 70 *LDS pin 71 D11 pin 72 *UDS pin 74 *AS pin 75 D0 pin 76 D10 pin 77 D1 pin 78 D9 pin 79 D2 pin 80 D8 pin 81 D3 pin 82 D7 pin 83 D4 pin 84 D6 pin 86 D5 The pins on the 86 pin expansion connector are numbered 1 on the top front pin, 2 on the bottom front pin, etc. ie. odd numbers on top, even on the bottom of the connector on the A1000 For the record, here is the network which is connected to each of the 20 signals listed above: 4.7K Signal<----*--------/\/\/\/\/\-------------->5V+ | | 1000 pf. | 1.0K | | | *--------/\/\/\/\/\----| |------->Ground | | *NOTE: must be ceramic Terminator should be put on the last peripheral in the chain. Brad anakin@utcs.gpu.toronto.edu
tomb@hplsla.HP.COM (Tom Bruhns) (02/04/89)
> > Terminator should be put on the last peripheral in the chain. > > Brad anakin@utcs.gpu.toronto.edu > ---------- As a satisfied user of the termination scheme mentioned, I'd like to point out that it's easy to build on an 86-pin edge connector and just plug it on the end of the chain. Easy to change the pile of peripherals around that way, if they all have pass-through. (I suspect this is what Brad had in mind, but I had a vision of someone soldering a bunch of parts directly into the last board/ expansion peripheral...) Using such a terminator, I have been able to put two Starboards on my A1000, where I couldn't before. Haven't had a chance to try three of 'em yet... Cheers, Tom Bruhns tomb%hplsla@hplabs.hp.com