11 Nov
2007
11 Nov
'07
11:37 p.m.
Subject: Re: Tarbell is making me insane
From: Grant Stockly <grant at stockly.com>
Date: Sun, 11 Nov 2007 13:21:32 -0900
To: "General Discussion: On-Topic Posts Only" <cctech at
classiccmp.org>
Tarbell Update
Ok you have 28k ram. Is the image sized and booting in that?
If the image is sized for say 32 or 48K it will crash, likely
as a bounce back to rom monitor.
>> > Also sounds like oneshot problems.
Check cpu timing. Even small timing
>> > errors tend to magnify bus noise issues and incompability problems.
>>(Snip)
>> > Also HEAT. That thing despite a very heafty noisy fan and cover mods
>> > didn't like heat. FYI: the timing of the oneshots drifts with heating!
I can't speak for the exact combination of the Tarbell, but the
system has been running all day long with:
-16k static card
-8k static card
-4k static card -2SIO
Is the boot image set up for MITS 2SIO and does it set it up?
-1k ROM
-8k Byte saver
I assume thse are in high memory and not below the address the image will
try to boot to. CP/M wants from 0000h to system size as configured.
When working on FDCs I load the system with a single known good RAM16,
Ram17 or similar 64k static card that can kill the last 4 or 8k. I use
a CPU with resident monitor and local ram (compupro CPU Z) and an
CPro Interfacer for IO. At that point I know I have a known IO,
known CPU and a monitor at F000h with POJ to F000h and ram at F800h.
Any time there is doubt about these they get a quick retest in a
known backplane (NS*Horizon). This means the system will be 4
boards max with the 4th being the FDC in question. I keep a CCS
terminator card handy as it seemed to work the best of all of them.
Things I've found in old backplanes.
Staples under the S100 connectors.
Metal under the backplane shorting to chassis.
Terminators with burned off tracks.
Metal bits in the connectors
Bits of nonconductor in the connectors.
Green cruft in the connectors (corrosion)
Black card edge connectors (gold over copper
without nickle flash, a bad thing)
Power supplies with excessive ripple. (tired caps)
Power traces fried on the backplane (altair 4slots)
Signal traces missing... from a former short?
That doesn't even touch on cards that just didn't play
well in some machines or with certain other cards.
Heres an example of a machine.
I have two explorer 8085s. One I built back when
and another aquired. They are identical as best as
I can see or test. Mine runs perfect, always did
with any card I've tried. the second is flaky,
won't run any S100 card without a bus terminator
and even then it's fussy. I've gone as far as to
completely swap every chip (they are fully socketed)
between the one that is cranky and my old one is
still happy and the newer one not. I've even swapped
PSs ( uses s100 style unregulated) and no dice. All
on board (there are a few) regulators replaced on the
bad one. No help. After a LONG time of testing and
all the problem was a high resistance trace on the
faulty board. I can see the signal at the source IC
solder side but between there and it's terminus it's
a 1000ohm resistor! One of the S100 status signals
was not quite there and floating. I bridged the
trace with blue wire between the two solder pads
and it's solid now. I cannot even with a 10x scope
see where the apparent crack is. That's chasing a
phantom.
But I do find the oddest of the odd.
-Tarbell FD
It has been running solid a 4K ROM monitor from Dave
Dunfield. Floppy interface still doesn't work on either card, and
these cards were tested by someone else on an IMSAI with an Altair CPU card.
If it's the monitor I think, it doesnt stress ram just sitting there.
You need a memory exerciser running at a minimum to be sure.
Some Z80 cards
complicate the issue. They have different timing and at 4mhz
some ram MAY NOT be fast enough. Either slow the z80 to 2mhz or more wait
states from ram.
The card I have has a switch for 2MHz. The tarbell with the reduced
clock speed will not work at anything higher like you suggested. I
have confirmed (from the person who shipped me the tarbell cards)
that the tarbell card with a reduced clock speed works normally with
a Z80 at 2MHz.
I attached a logic analyzer to the drive and watched all of the
signals. Every thing looks normal to me. I hooked it to a 5.25"
drive and then to a 3.5" drive. The drives are sending everything
that they should.
Then the FDC is potentialy not decoding the data.
Shows how rough the bus timing and noise can be.
Is that a one piece
backplane of the two sided variety or the one of the earlier Altair
4 slot chains or single sided backplane? The reason is the earlier
2x4 slot motherboards, but all of the wires are the same length
(there was a tip on that to reduce noise) I'm going to let the thing heat soak and then check
the bus with an
oscilloscope and the CPU clock.
if the oneshots are good ones once set they should say put. Then again
there was lore and fact about what was and wasn't good.
You know, I'm worried that the reliability issues
aren't related to
the Altair, but to a heavily modified tarbell card trying to write a
Tandon SSSD 5.25" onto a 3.5" disk???
Likely if the data rates are not set right for the FDC and it's data
seperator (usually also oneshots) is not set right it will fail.
Could even be a simple bad IDC crimp on the cable.
Would any of you like to see the data coming out of the disk
drive? I can upload the logic analyzer data to my web site...
Not I. If I did at the very least I'd also need to see the schematic and
jumper info for the tarbel. Other wise it's just a pretty picture.
I will never be happy until I find the problem.
I'd like to think
that with all of the information we have that I can make the Altair
as reliable as any other S-100 system. I have collected a lot of
errata and have applied it all to my system... That's why I won't
give up with this setup. : )
Most of the erata is lore not hard tried fixes. Most fall in the
"it worked for me" realm.
The problem is you have a lot of still possibles lurking.
Oh, SA400 drives... had a problem with them and all that used
that frame (sa400, 400l and 450). Seems if the spindle bearings
go bad the amount of jitter on the data can exceed that of simple
data seperators used with 1771 (especially RS early TRS80 with
early EI). I'm sure it's possible to have the same problem with
other drives. Showed up best with the SA alignment disk and
a dual trace scope set up for data read EYE pattern.
Allison
Thanks,
Grant