PDA

View Full Version : IBM 5155 gone to "dollar sign hell" - help!



TheMaritimeMan
July 15th, 2016, 06:27 PM
I have an IBM 5155 Portable PC that did work perfect. Today I fired it up for the first time in about 3 months, and it worked fine as usual, and then after a few minutes dollar signs ($) started popping up all over the display. Now when you turn it on, it gives the POST beep code for a video card error. It still boots and shows video, albeit littered with dollar signs.

I tried reseating all the expansion cards, and I reseated the ROM chip on the video card. No help. I would try reseating the RAM, but I haven't yet because it takes a lot of time and energy to get to the motherboard in this thing.

Anyone ever experience this before and/or might have an idea of a solution? Someone mentioned to me that it might be failing tantalum caps - I've never heard of them doing anything other than exploding without so much as crashing the system, but does anyone else concur?

It kinda stinks that this worked perfect after sitting for years before I got it, and it worked perfect when I stowed it away 3 months ago, and now on its first run since then it's totally crapped out.

SomeGuy
July 15th, 2016, 06:44 PM
If the system boots, then it would not be the motherboard RAM.

It is almost certainly a bad RAM chip on the CGA card. I would guess it is probably not any of the intermediate logic chips since you are seeing video at all.

I assume the "$" is where spaces should be. A space is hex "20" and a "$" is hex "24", so one bit is getting stuck on. Somebody else here could probably tell you exactly which chip that is, but they are soldered on to the board.

TheMaritimeMan
July 15th, 2016, 07:22 PM
Bad RAM on the video card? Aw, rats!

You're right - looking at a picture I took (linked below), the dollar signs do seem to appear only in blank space. Specific letters are also being replaced by other letters - in the picture, many E's are replaced with A's - again a change in one bit, and actually the same bit as the space/$ replacement. I also see '-' getting replaced with '>' - again a change in the same bit!

Crap, you're totally right! I wish the system RAM was still a possibility because at least that stuff is socketed.

http://i.imgur.com/1caAZwm.jpg

Still odd that this happens after just three months in storage.

mlaferriere
July 16th, 2016, 07:35 AM
There is a possibility it may not be a bad chip since the solder is likely old as well. One thing you could try is warming and relflowing the solder to the chip. At a minimum its an easy thing to do and you dont have to wait for a replacement in the mail or you could do this while waiting for a replacement to arrive.

Stone
July 16th, 2016, 08:09 AM
There is a possibility it may not be a bad chip since the solder is likely old as well. One thing you could try is warming and relflowing the solder to the chip. At a minimum its an easy thing to do and you dont have to wait for a replacement in the mail or you could do this while waiting for a replacement to arrive.You say 'chip' like there's only one. There's at least eight RAM chips on that card. :-)

Chuck(G)
July 16th, 2016, 08:23 AM
I suspect that it's the chip--over time, the plastic encapsulation can fail and there's basically no other cure than replacement.

SomeGuy
July 16th, 2016, 09:03 AM
You say 'chip' like there's only one. There's at least eight RAM chips on that card. :-)
There are eight. One for each bit. In this case bit 2 is stuck "on", so it is only the chip that stores that bit. Someone who could be arsed to read the schematics could tell you exactly which one. Probably the third from one end or the other if they are in order.

If it were me, I would test the machine with another video card before breaking out the soldering iron. It wouldn't even have to be IBM CGA if you can hook up an external monitor.

TheMaritimeMan
July 16th, 2016, 09:04 AM
Yeah, I'm willing to take a crack at reflowing the solder to the ailing chip, if I can identify which chip it is, which is where I would need help. I suppose I can visually inspect all of them and see. This thing did undergo a 90 minute ride in the trunk of the car before this point, so I wouldn't rule out a cold solder joint.

While I don't have another CGA card I can test, I do have another XT machine I can test this card in, but unfortunately that machine is inaccessible right now.

TheMaritimeMan
July 16th, 2016, 09:40 AM
I just noticed something really interesting. When I type commands in IBM PC BASIC, and the computer obviously injects erroneous characters in them, the commands won't execute - it throws a syntax error. So it seems like the characters are not just passively appearing - the computer actually sees them and tries to execute them.

For example, if I type "run", it changes to "ruj" and it won't execute. If I type "cls", it changes to "cls $", and the cls command actually executes, and then it throws a syntax error, as if it's due to the dollar sign being there afterward.

Does this still seem like something that just a video card problem would cause? It's not the keyboard because if I boot with the keyboard disconnected it still outputs bad characters during the memory count.

retrogear
July 16th, 2016, 09:59 AM
From looking at the schematic the ram paths are labeled D0-D7 and D2 is U55. Looking at my card, I see there is a tantalum cap right below it. Look for crud, scratches etc

Edit: just saw your last post. Doesn't seem like video ram if it's affecting keyboard syntax too

Chuck(G)
July 16th, 2016, 10:02 AM
I just noticed something really interesting. When I type commands in IBM PC BASIC, and the computer obviously injects erroneous characters in them, the commands won't execute - it throws a syntax error. So it seems like the characters are not just passively appearing - the computer actually sees them and tries to execute them.

For example, if I type "run", it changes to "ruj" and it won't execute. If I type "cls", it changes to "cls $", and the cls command actually executes, and then it throws a syntax error, as if it's due to the dollar sign being there afterward.

Does this still seem like something that just a video card problem would cause? It's not the keyboard because if I boot with the keyboard disconnected it still outputs bad characters during the memory count.

Entirely possible that BASIC reads back the screen contents. Wouldn't surprise me one bit.

TheMaritimeMan
July 16th, 2016, 10:34 AM
From looking at the schematic the ram paths are labeled D0-D7 and D2 is U55. Looking at my card, I see there is a tantalum cap right below it. Look for crud, scratches etc

Thanks for the info! I looked at all the RAM chips, and the solder joints looked good as far as my vision could see. But I suppose there's no harm in reflowing U55 for the heck of it.


Entirely possible that BASIC reads back the screen contents. Wouldn't surprise me one bit.

Looks like you're right - I booted DOS, and everything I typed and executed worked as it should.

I keep forgetting that when I first powered this on last night, it worked perfect for about 5 minutes before it got sick, so it really does seem like the chip just gave up and died. Looks like I can get replacements cheap enough on eBay.

VileR
July 17th, 2016, 06:56 AM
Entirely possible that BASIC reads back the screen contents. Wouldn't surprise me one bit.
IIRC they went through the trouble of doing that in graphics mode too, as bonkers as that sounds. To read the input, BASIC checks those 8x8-pixel blocks from video RAM, compares them with the 8x8 characters in the PC BIOS, and tries to find a match for each so that it can tell what you just typed. (I believe that this faux-'OCR' routine is part of the PC BIOS too.)
This also led to the baffling situation of using a non-IBM EGA/VGA card, where the 8x8 graphics mode font comes from the card's BIOS ROM, and may be somewhat different from the one in the PC BIOS... in which case, INPUT statements would promptly stop working, etc.

bobba84
July 17th, 2016, 04:15 PM
Why on earth would they bother doing that? Was it to save a few K of RAM on a 16-64k 5150? It seems like a huge length to go to, for not much result.

Bobby.

Chuck(G)
July 17th, 2016, 04:31 PM
There's a function in the BIOS to read the screen, whether or not it's in graphics mode, so not so much in BASIC per se. You can see how this works in BASIC by repositioning the cursor over text lines that have been displayed and hitting "enter".

TheMaritimeMan
August 17th, 2016, 05:36 PM
Thought I should post an update - I re-flowed U55; no dice.

Here's something really interesting - my friend with the card he may be able to send me recently tested the card, and his too spontaneously started displaying incorrect characters in text mode (Link (http://imgur.com/4oVOBNr)). Shortly afterwards, one of the tantalum caps exploded! He replaced the cap, and now his card is working fine again!

So I'm wondering if it's actually one of the caps on my card causing the problem? Other evidence of this is that if I let the computer sit for a long time (ex. I powered it up today for the first time in a month), the video card will actually function normally for a few minutes, before the problem surfaces again. To me this sounds like a cap issue, as I wouldn't think semiconductor failures would fade in and out like that.

Does anyone concur? All my caps visually look fine, but I wouldn't mind pulling some to see if that changes anything. Problem is there's 30 of the things and I have no idea where to start (well, I guess U55...)

giobbi
August 17th, 2016, 05:44 PM
Does anyone concur? All my caps visually look fine, but I wouldn't mind pulling some to see if that changes anything. Problem is there's 30 of the things and I have no idea where to start (well, I guess U55...)

Are you sure all of them are tantalum? Probably most of them are ceramic...

And, yes, tantalum caps use to explode...

SomeGuy
August 17th, 2016, 05:46 PM
the video card will actually function normally for a few minutes, before the problem surfaces again. To me this sounds like a cap issue, as I wouldn't think semiconductor failures would fade in and out like that.
They can. Especially RAM chips. They can operate normally under some circumstances, such as lower temperature, and then fail under others. Which can be annoying when trying to trace a specific issue.

Tantalum capacitors are usually just for filtering/stabilizing power, so it is very strange that that other card started to work.

TheMaritimeMan
August 17th, 2016, 06:13 PM
Are you sure all of them are tantalum? Probably most of them are ceramic...

Here's a pic showing all the different types of caps on the board. (http://imgur.com/a/HdKhe)


They can. Especially RAM chips.

Well, it was worth a thought anyhow.

The RAM chips are Motorola MCM4517P12 - does anyone know if equivalent chips exist? The cheapest I can get any in Canada is $18 for five of them on eBay, which is a bit crazy.

jmetal88
August 17th, 2016, 07:06 PM
Here's a pic showing all the different types of caps on the board. (http://imgur.com/a/HdKhe)



Well, it was worth a thought anyhow.

The RAM chips are Motorola MCM4517P12 - does anyone know if equivalent chips exist? The cheapest I can get any in Canada is $18 for five of them on eBay, which is a bit crazy.

Hmm, looks like a single-supply version of the 4116. You could replace those with 4164s by connecting pin 9 on each chip to VCC or Ground.

KC9UDX
August 18th, 2016, 11:53 AM
Why on earth would they bother doing that? Was it to save a few K of RAM on a 16-64k 5150? It seems like a huge length to go to, for not much result.

Whomever was responsible for that bit of code was used to using Commodores, if I had to hazard a guess.

Retrom
August 18th, 2016, 01:56 PM
I'm the guy who's card TheMaritimeMan was talking about, the cap that was dodgy was C19 (it's top had cracked open), I swapped it with C8 which was the same type and looked ok visually (So now C8 was in C19's position and C19 in C8's), the moment I powered on the computer, the suspected bad cap that I'd put in C8's place exploded instantly (I've never turned a computer off faster than when that happened. lol). So I knew for sure now that that cap was bad, so I replaced it with a junkbox spare and hoped for the best. When I powered up the computer again, the card was working pefectly, and I did leave the system running for quite awhile, to see if the intermittent character fault would return, it did not, and what's more, is that CGA graphics mode was now working perfect as well (Before replacing the cap, if I put the card into 320x200 graphics mode, the screen would flicker and go blank at random). So all in all, I'd say capacitors can definitely cause this problem.

Retrom
August 18th, 2016, 02:14 PM
BTW, of the 30 caps, I think only 7 of them are tantalum capacitors (Unless I counted wrong) they are the following:
C8
C6
C4
C15
C24
C29
C19

They are a weird type I haven't seen before (they have three legs), however, from what I can tell, the center pin is positive, and the pins on either side are both negative, so you can replace them with regular tantalum caps.

TheMaritimeMan
September 27th, 2016, 03:07 PM
Just to kind of update/close this thread, I received the replacement card from Retrom, and the computer is working perfectly once again. :)