#Hardframe-A2052-DMA
21 messages in this thread
Don. You seem to be the Hard Drive expert here so maybe you can help me
with something. I have an A2000, A2052 2Meg RAm card, HardFrame and Quantum
105S Harddrive. I appear to have a DMA conflict between the A2052 Ram Card
and the HardFrame. When I boot from power on I get strange results;
sometime lines in the Startup-sequence are skipped, Echo cmds give errors
or Shells come up with the history buffer not working. Also running
FixDisk1.2 on my Old File System partition gives random CHECKSUM ERRORS (no
errors on my FFS partitions) Removing or disabling the Ram Card seems to
cure the boot up problem (Warm boots work with card in or out) and the
random CHECKSUM errors. Memory tests on the Ram card show nothing and at no
time do I ever get any READ/WRITE errors. MaxTransfer is set correctly and
I have LowLevel formated the drive twice. Reading an OFS sector with DEDIT
(and Ram card in) shows off/on CHECKSUM errs but removing the Ram card and
reading same sector yields no chksum errs. How many times will DOS try and
reread a sector (I assume it checks the Checksum) before it gives an error?
I assume it must be reloading sectors because I have no problem running
pgms out of the OFS partition. Any thoughts on the problem would be greatly
appreciated. Thanks Michael Maryka 72321,3504
Michael,
Some others had problems similar to yours, and the cure was to make
sure the drive was powered directly from the A2000's power supply (use one
of the free drive power plugs) rather than from the power plug on the
board.
I'm not sure what the exact cause of this is, but I do know that
the A2052 draws quite a bit of power…and thus the bus may not be able to
properly handle providing power for both the RAM card and the HardFrame
with drive included.
Don
Typical A2000 problem. ;^D
Chris,
Why would you say that? The only place I've heard of it occuring
is with the HardFrame. I've not experienced it, nor have I heard of anyone
else having problems drawing power from the 2000's slot bus.
And, in reference to the HardFrame…this message is only the 2nd
that I'm aware of describing the problem.
Don
Excuse me Don, but I was just being opurtuninstic in pointing out A2000
expansion pitfalls. Oh, powering boards in particular is probably the
rarest of problems (although it helps not to have certain A2000
motherboard-revisions from a power perspective). I suppose I should have
picked on a device-conflict situation instead.
Chris,
Ok, point out an A2000 expansion conflict between products that are
made 100% in accordance with Commodore's published specifications.
Don
Commodore's own 2MB RAM board for the A2000.
What does it conflict with? Alot of things, especially '030 cards.
Commodore's 2090A can conflict with software as well as hardware.
While I know these are not current products, alot of people bought them
and don't expect to sacrifice them for future expansion.
Also, I think whether or not many products conform completely to
Commodore's specs is irrelevant in this case; My observations are on the
general risks of expansion hardware, and if a good number of A2000 products
cause conflicts out of not meeting Commodore specs, then that is still a
black mark on the A2000.
Chris,
I disagree. If 3rd party hardware has problems, but Commodore's
similar hardware works ok…then the fault isn't the 2000…it's the 3rd
party vendors who didn't follow the specs.
If you bought "Joe's off-brand Tires" for your car, and they kept
blowing out…yet if you bought a recognized brand, and they didn't blow
out…would you be upset with the car maker, and decide it was a bad
product…or would you be reasonable enough to figure "Joe's off-brand
Tires" might not be all they seemed at first?
I've got a 2090…I haven't have any problems with it for either
ST506 or SCSI drives or combination of both. No idea on the 2052 and and
the 2630, but yes…both are old, out of production products.
Don
Don, I cited Commodore products, primarily. Frankly, if I had a 2630, 2052
and 2090A in the same machine, I'd be using the machine as warily as when
my first Synergy SCSI interface started going nuts (walking on eggs and
probably pulling my har out). I'd probably wanna get most or all of it
replaced.
Don,
Thanks for the reply but from day 1 I have been using the Power
Supply drive plugs not the buss power. I just received V1.9 Rom and
software from Microbotics but no change. I still think it is a DMA speed
problem because I've measured the Voltages on the Ram card with a digital
meter and got 4.9Volts DC right on the Ram chips (should be close enough to
5.0V) All I have in the expansion slots is the HardFrame and the A2052.
Perhaps some of the memory or buffer chips can't handle the DMA speed ??
Have you heard of a "U605 Patch" for Rev 4.5 and less mother boards ( I
have Rev 4.2) supposed to be expansion slot fix for Hard disk controllers.
Any way looks like I'm going to replace the A2052 I've been looking at
the Microbotics 8-UP Simm board do you have any suggestions ( I would like
to get a board that uses Simms not DIPS to slow up obsillecense (any body
want to by 64 256k*1 chips cheap ;^) . Maybe I should trade in the
HardFrame and get a GVP Series II with ram; who knows !!
Thanks for the help
Michael Maryka 72321,3504
Michael,
The "U605 patch" involves soldering a small capacitor across two pins of
one of the DMA controller chips, and has been recommended by Microbotics.
I don't recall the exact details (like which pins or what value cap), but I
did have it done to my system. Your Amiga service tech. should have the
details, as it was in one of the tech notes CBM supplied to their service
techs.
Hope this helps…
…BobR
Bob,
Thanks for the reply on the "U605" patch. I called the nearest CBM
Dealer and not to my surprise he was unable to locate that particular mod.
He was able to find one that related to Accelerator cards but not the U605
one. I got about as much help from CMB Canada east.It's enough to make you
pull out your hair. I was how ever able to Order a Technical Service manual
from CBM. The parts lady said it should have all the schematics and update
bulletins in it. So hopefully I'll find it in there. (All I wanted is to
talk to someone who runs a Soldering iron for a living NOT A TELEPHONE )
I'll probably call MicroBotics since they mentioned it in their Docs
they probably can tell me exactly what to do. I've spend more time trying
to get this HardDrive to work properly than actually using my machine for
something usefull. I guess that's the way it goes sometimes.
Thanks again for the info
Michael Maryka 72321,3504
The one problem that I've observed building my own A2000 accelerator and
getting the A2052 to work reliably was that The Address strobe (AS*) had to
exactly emulate the 68000 signal. Actually, if 68000s drifted to the
extremes listed as valid by the 68000 Technical Reference Manual, the
Expansion bus wouldn't work at all. According to it, AS* can fall as late
as 35 nsecs (or so, I'm going from memory) from the start of state s2. For
the Expansion bus to work, AS* can't fall any later than 15 nsecs to meet
setup requirements of the bus control logic. Fortunately, all 68000s are
faster than 15 nsecs at room temperature. Beyond this issue, the A2052
needs the shape of AS* from the hardframe to look exactly like the 68000s
in order to be reliable. In DMA, the hardframe is generating the AS*
signal. If it is close, but not perfect, then memeroy failure is possible.
The U605 fix was for accelerators that had a lazy AS*. The resistor makes
the signal inactivate faster, compensating for poor performance of the
Signetics chip used in U605 on some revisions of the motherboard.
The real problem is who to blame the problem on. The A2052 shouldn't be so
sensitive to the precise timing within the spec of the 68000, the hardframe
must be slightly off as well. That's the wonder of the ZorroII bus. Two
people can have properly designed cards that have trouble together.
-Greg
Greg.
Building your own Accelerator WOW!! Wizards like you BobR and Don
Curtis here on CIS more than make up for Dealer and CBM's short falls. I
don't think AS* is my problem I have a Texas Instr. U605 what do you think?
1) On cold boot (pwr up) I get problems reading the startup file (never any
R/W requestors) lines are skipped or misread. Warm boots usually have no
problems. 2) Running FixDisk or Dedit gives semi-random checksum err's but
only on old File system partitions not FFS partitions. (partitions have
been low/ high level formatted several times). Larry Philips Scan pgm never
gives any err's on OFS or FFS partitions (probably does't do chksum check).
Say sector 1,000 gives a chksum err. Sometimes it will sometimes it won't.
IE not all sectors will give errs; it seems related maybe to the data in
that sector ??
Checksum read = $F3BFC1DB Checksum Calculated = $F3BFC1A5 The 2
chksums are usually only a couple of hex bits out never totally different.
Could the timing of the AS* not be allowing the A2052 ram chips enough time
to fully load the data properly? Wouldn't a faulty AS* cause a more
consistant err IE on both file systems and on more sectors not just certain
sectors? Any ideas Greg:
– Michael
Besides AS* controlling the data transfer over the expansion bus, it may
affect the A2052 by way of a parameter of DRAM called RAS precharge time. A
150 nsec dram is OK for the 7.16 Amiga bus, but care must be taken in
timing to assure that it has at least a 100 nsec time that RAS is inactive
between memory cycles. Some designs rely on the time AS* is inactive to
accomplish this. This causes problems when something other than a perfect
68000 bus cycle occurs. Many accelerators sync the forward edge of the AS,
but don't have as good control over the trailing edge. In my accelerator
design, I had to reset AS high exactly at the start of 68000 bus state 7 to
get the A2052 to behave reliabily. A slightly lazier design, with AS* going
inactive in the middle of state7 produced a memory board that goes erratic.
However that lazier AS* worked perfectly with all other expansion bus cards
I had at the time. In that case, the A2052 is a good test for accelerator
compatibility.
For now, I'd pull the A2052 and see whether the other problems continue. If
not, you know where the problem is. If they do continue, I'd see if the
hardframe could be exchanged or repaired.
-Greg
Greg– Your message seems to imply that you are working on a cpu
accelerator. If so, will it be released as a product? I haven't seen much
need for the rejuvenator yet for my A1000, but more cpu speed is definitely
on my wish list.
Tom
The U605 patch is a 1K resistor from pin 20 to pin 11 of U605, if and only
if it is a Signetics IC. Not a cap!!!
-Greg
Thanks for the correction, Greg. I should know better than giving out
hardware answers. <grin> At least I _was_ vague.
…BobR
Hey! Looky whats stuck on your name there, fella! Now I can flame you for
real! <grin>
Ben
Amateur Radio Callsign is A A 7 A S
And now I have recourse for real! <back atcha>
…BobR
THis is a very good point <he said meekly>
Ben
Amateur Radio Callsign is A A 7 A S