21 Apr
2015
21 Apr
'15
7:41 p.m.
I picked up a (known-faulty) 5160 a few weeks back, just because it had a
keyboard with an intact decal, and the one on my otherwise-perfect system
is missing.
The machine came with a CGA card, which I've tested as working in my other
5160.
Symptoms of the fault are one long beep and two short beeps at power-on,
with 1024 characters of garbage displayed on-screen (typically either a
solid block or a space, but with a handful of other random chars) - i.e. 12
rows of 80 and then a row of 64.
Following this on screen, at location 13,64, the memory count occurs, and
then it'll drop to ROM BASIC (no hard disk, and for the purposes of testing
I've not bothered with a boot floppy), with the BASIC startup text starting
at row 14 and all offset by 64 characters from the left margin. However,
the function key reference on the bottom line of the display is OK,
starting at position 25,1.
I can issue BASIC commands, and each new line is offset by 64 characters,
until I hit the bottom of the screen, at which point text appears at
position 1 (but still offset from the top of the screen by 13 lines).
Issuing a 'cls' doesn't remove the garbage from the display, and typing
then resumes from location 13,64. Throughout all of this the cursor seems
to appear where I'd expect it on the screen, however - I'm assuming it's
done via hardware entirely within the CGA card.
I've tried pulling all other cards, leaving just the CGA board, and the
problem persists. One long and two short beeps seems to be 'video failure',
but as mentioned I've verified that the card works in my other 5160.
Does anyone have any ideas what might be going on? It doesn't quite seem
like a memory fault - possibly some sort of address decoding error? It's
almost like the video board is pulling display data from the wrong part of
memory, but I'm not sure that makes sense given that the board has its own
local RAM rather than relying on RAM on the system board.
cheers
Jules
21 Apr
21 Apr
8:09 p.m.
On 04/21/2015 06:41 PM, Jules Richardson wrote:
Symptoms of the fault are one long beep and two short
beeps at power-on,
with 1024 characters of garbage displayed on-screen (typically either a
solid block or a space, but with a handful of other random chars) - i.e. 12
rows of 80 and then a row of 64.
A picture being worth a thousand words and all that...
http://www.classiccmp.org/acornia/tmp/5160.jpg
cheers
J.
23 Apr
23 Apr
12:20 p.m.
Does anyone have any ideas what might be going on? It doesn't quite seem
like a memory fault - possibly some sort of address decoding error? It's
almost like the video board is pulling display data from the wrong part of
memory, but I'm not sure that makes sense given that the board has its own
local RAM rather than relying on RAM on the system board.
To me this sounds like an address line fault on the expansion bus. The CPU thinks
it's writing to a particular location in the CGA card's memory, but due to the
fault it
actually accesses another location.
Yes, the cursor is generated by the 6845 on the CGA card.
Have you tried the CGA card in different slots? Just in case it's a bad contact
on the edge connector.
Other than that, stick a logic probe on the address lines on the expansion slot and
see if any are stuck high/low or always floating.
-tony
12:48 p.m.
On 4/23/2015 9:20 AM, tony duell wrote:
Yes, the cursor is generated by the 6845 on the CGA
card.
Since the messages start out skewed to the right, I'd consider that
there is a stuck bit in the data written to whatever 6845 location
controls that. If you can probe the bus and the data is okay most
places, it may be noisy or corrupted to the slots. Obviously a lot of
the memory has to work for the bios to pass and run, so that is
suggesting some problems getting data to the interface slots, though
most of that is working or the display card would have failed as well.
Fault may be on the card data path as well, since so much other I/O is
working fine.
thanks
jim
1:30 p.m.
Yes, the
cursor is generated by the 6845 on the CGA card.
Since the messages start out
skewed to the right, I'd consider that
there is a stuck bit in the data written to whatever 6845 location
controls that. If you can probe the bus and the data is okay most
places, it may be noisy or corrupted to the slots. Obviously a lot of
the memory has to work for the bios to pass and run, so that is
suggesting some problems getting data to the interface slots, though
most of that is working or the display card would have failed as well.
Fault may be on the card data path as well, since so much other I/O is
working fine.
I believe this CGA card works correctly in another machine.
-tony
1:58 p.m.
On 4/23/2015 10:30 AM, tony duell wrote:
The cursor actually has nothing to do with where the
data is written or
appears. And I understood the cursor appeard correctly, that is in the
location you would expect characters to be written, even though they
are not. This suggests the 6845 is being set up correctly but that the
processor is not writing to video memory where it thinks it is.
I guess it's also possible that the 6845 is not being set up correctly
to start the display at the first location of video memory. The problem
then is that the 6845_is_ being set up correctly to generate valid timing,
etc.
That was what i suggested. it looks like the video memory, and infact
most all of the memory operations are working or basic would not come
up, or the bios codes for memory errors would occur.
So the mechanism which sets up the 6845 is broken, with the skew of the
location of the first message being significant, if it is a multiple of
perhaps one or more bits stuck when the 6845 registers are written.
I suspect the setup of the addressing of where the display memory is not
relevant, so once you mess up where in the memory to fetch from, all the
memory accesses are skewed but still reference the data the processor is
writing to that memory (which appears to be correct).
since there is a lot of crap at the top which is not set up correctly,
perhaps there is a skew factor to be guessed at from that number of
lines or characters.
2:06 p.m.
That was what i suggested. it looks like the video memory, and infact
most all of the memory operations are working or basic would not come
up, or the bios codes for memory errors would occur.
Remember it does give the beeps for a video failure.
So the mechanism which sets up the 6845 is broken, with the skew of the
location of the first message being significant, if it is a multiple of
perhaps one or more bits stuck when the 6845 registers are written.
The 6845 and the logic on the CGA card must be good, it works in another
machine.
The 6845 must be initiallised essentially correctly, the thing gives a sensible
video output. So it must be possible to write to the 6845.
The next issue is that block of garbage on the start of the screen. AFAIK all the memory
on the CGA card is initialised by the ROM BIOS. So even if the 6845 was being mis-set, it
would have spaces, not garbage, there.
I suspect the setup of the addressing of where the
display memory is not
relevant, so once you mess up where in the memory to fetch from, all the
memory accesses are skewed but still reference the data the processor is
writing to that memory (which appears to be correct).
I think the reverse. The CGA card is set up correctly but the CPU can't write
to the correct location in the video RAM due to an addressing problem.
Possibly a faield buffer on the address bus somewhere.
since there is a lot of crap at the top which is not set up correctly,
perhaps there is a skew factor to be guessed at from that number of
lines or characters.
Isn't it exactly 1024 characters? This suggests a CPU address fault, not
a 6845 one to me.
-tony
5:55 p.m.
On 04/23/2015 11:20 AM, tony duell wrote:
Yes, could be. CGA base address is B8000h, I think, and I'm seeing display
text offset by 800h (2KB - 1024 characters plus 1024 attribute bytes).
Assuming that the BIOS is responsible for clearing video memory (rather
than it being a function of the CGA card itself), then my on-screen garbage
could be down to whatever happens to be in the first 2KB of the CGA card's
memory at power-on, and address bit A11 is jammed high on the expansion bus.
What's puzzling if that's the case is why the ROM BASIC 'function key'
strip appears normally at the bottom of the screen - I'd expect it to be
stuck in a 'non visible' portion of the CGA card's memory.
Does anyone have any ideas what might be going on? It doesn't quite seem
like a memory fault - possibly some sort of address decoding error? It's
almost like the video board is pulling display data from the wrong part of
memory, but I'm not sure that makes sense given that the board has its own
local RAM rather than relying on RAM on the system board.
To me this sounds like an address line fault on the expansion bus. The CPU thinks
it's writing to a particular location in the CGA card's memory, but due to the
fault it
actually accesses another location. Have you tried the CGA card in different slots? Just
in case it's a bad contact
on the edge connector.
Yes - the same thought had occurred to me.
I'll take a look at the address lines (albeit just with a DMM, but that
will hopefully be enough if it's a stuck bit)
cheers
Jules
9:56 p.m.
On Thu, 23 Apr 2015, Jules Richardson wrote:
Date: Thu, 23 Apr 2015 14:55:56
From: Jules Richardson <jules.richardson99 at gmail.com>
Reply-To: "General Discussion: On-Topic and Off-Topic Posts"
<cctalk at classiccmp.org>
To: General at classiccmp.org,
"Discussion at classiccmp.org": On-Topic and Off-Topic Posts
<cctalk at classiccmp.org> ;
Subject: Re: IBM 5160 oddball fault
On 04/23/2015 11:20 AM, tony duell wrote:
Yes, could be. CGA base address is B8000h, I think, and I'm seeing display
text offset by 800h (2KB - 1024 characters plus 1024 attribute bytes).
Assuming that the BIOS is responsible for clearing video memory (rather than
it being a function of the CGA card itself), then my on-screen garbage could
be down to whatever happens to be in the first 2KB of the CGA card's memory at
power-on, and address bit A11 is jammed high on the expansion bus.
What's puzzling if that's the case is why the ROM BASIC 'function key'
strip
appears normally at the bottom of the screen - I'd expect it to be stuck in a
'non visible' portion of the CGA card's memory.
Since it comes up in BASIC, have you tried BASIC's POKE command to
stuff some values into the first 1,024 characters (2048 bytes)?
I dug up the old IBM PC BASIC manuals and refreshed my memory, and
took a few minutes to write a BASIC program that will POKE values
into the first 1,024 characters. You can use it to test if the
CGA memory is OK when accessed directly. If the memory access is
OK it should place 1,024 capital "A" letters that fill in the
first 12 lines and 64 characters of the 13th line.
100 ' A test of the first 1,024 characters of CGA character memeory.
110
120 ' Set to 80x25 text mode, with color, page 0 active, page 0 viewed.
130
140 SCREEN 0, 1, 0, 0
150 WIDTH 80
160
170 ' Set segment to CGA memory.
180
190 DEF SEG = &HB800
200
210 ' Poke bright white on black into CGA page 0.
220
230 FOR I = 1 TO 2047 STEP 2
240 POKE I, 15 ' Poke 7 instead of 15 for normal white.
250 NEXT I
260
270 ' Poke capital A into CGA page 0.
280
290 FOR I = 0 TO 2046 STEP 2
300 POKE I, 65
310 NEXT I
320
330 ' Reset segment to BASIC's data segment.
340
350 DEF SEG
/s/ Ernest
Does anyone have any ideas what might be going on? It doesn't quite seem
like a memory fault - possibly some sort of address decoding error? It's
almost like the video board is pulling display data from the wrong part of
memory, but I'm not sure that makes sense given that the board has its own
local RAM rather than relying on RAM on the system board.
To me this sounds like an address line fault on the expansion bus. The CPU
thinks
it's writing to a particular location in the CGA card's memory, but due to
the fault it
actually accesses another location. Have you tried the CGA card in different slots?
Just in case it's a bad
contact
on the edge connector.
Yes - the same thought had occurred to me.
I'll take a look at the address lines (albeit just with a DMM, but that will
hopefully be enough if it's a stuck bit)
cheers
Jules
24 Apr
24 Apr
7:33 p.m.
On 04/23/2015 08:56 PM, Ernest G. Allen wrote:
Since it comes up in BASIC, have you tried BASIC's
POKE command to
stuff some values into the first 1,024 characters (2048 bytes)?
I'd thought of it, but then of course POKE wouldn't accept a linear address
of sufficient size to 'reach' the memory. As per my other message, I've
fixed the machine now (bad PCB trace for A11 between slots 3 and 4), but
your code was still handy because I didn't know that 'DEF SEG=$hxxxx' was a
thing. I've filed it away in my brain as I'm sure it will be useful again
one day! :)
cheers
Jules
7:26 p.m.
On 04/23/2015 04:55 PM, Jules Richardson wrote:
Aha! I'd tried it in slot six (originally it was in seven) with the same
result, and then I gave it a go in slot five too, just to confirm that it
wasn't 'luck' that two slots were bad - the problem remained.
Now while I had the system board out of the machine (as it makes it easier
to poke around the bus signals without a riser card / extender), I figured
I'd try the other slots just to verify that they were all bad.
Well, it worked in slots 2 and 3 (I didn't try 1 as I've got a hazy memory
telling me there's something different about that one), but was failing in
slots 4 and up.
There was no visible damage to the PCB, but checking with a meter showed
there to be a break in continuity on A11 between slots 3 and 4, so I've now
bridged them with a wire, and all is now working OK :-)
cheers,
Jules
Checking with a meter, it seems that I've got a break in the PCB track for
A11 between slots three and four. There's no visible damage
Have you tried
the CGA card in different slots? Just in case it's a bad
contact
on the edge connector.
Yes - the same thought had occurred to me. I'll take a look at the address lines (albeit just
with a DMM, but that
will hopefully be enough if it's a stuck bit)
cheers
Jules
3377
days inactive
3380
days old
10 comments
4 participants
participants (4)
-
ard＠p850ug1.demon.co.uk -
eallen＠owt.com -
jules.richardson99＠gmail.com -
jws＠jwsss.com