Help with a 5170 type 3 motherboard

[Dave Farquhar]

So I got a so-called "untested" IBM 5170 type 3 motherboard. We all know what that means. I tested it for shorts and didn't find any, so I figured it would be safe to plug in a video card in and fire it up. The video card initialized and gave me a display right away, which was awesome, so I decided to attach a disk controller and try my luck.

Nothing bad happened, but the system just sat there with a blinking cursor. So I grabbed my POST card and plugged it in, and it's getting hung up on either code 3E or 36. Interestingly, my keyboard does light on power up, but as the system is sitting on 3E or 36, it doesn't respond to caps lock.

I've let the system sit for several minutes, and it just seems to hang on one of those two codes. Am I just being too impatient? Do I need to let it sit longer and see if it decides to try to boot or drop to Basic?

The chips are getting warm, so I know the system is doing something.

Thanks for any insights you may have!

 

[Dave Farquhar]

An update... I tried waiting longer, and what I found was that if I watched the keyboard instead of the POST card, the lights would flash from time to time. Sometimes just one, but usually all three. But the keyboard never came to life. Since POST code 36 is related to the keyboard controller, does it sound like a bad keyboard controller?

 

[modem7]

So I got a so-called "untested" IBM 5170 type 3 motherboard. We all know what that means. I tested it for shorts and didn't find any, so I figured it would be safe to plug in a video card in and fire it up. The video card initialized and gave me a display right away, which was awesome, so I decided to attach a disk controller and try my luck.

Nothing bad happened, but the system just sat there with a blinking cursor.

So you power on the motherboard, and all that you ever see is a blinking cursor, nothing else? Please confirm.

So I grabbed my POST card and plugged it in, and it's getting hung up on either code 3E or 36.
Since POST code 36 is related to the keyboard controller, does it sound like a bad keyboard controller?

The POST gets the motherboard's keyboard controller to do a self test at POST code 0Bh, per [here]. That passed (the controller indicated a pass). But such a self-test is rarely a complete test.

Looking at the 11/15/85 BIOS source listing:
- POST code 36h is one of the codes in a later (large) block of code described as 'RESET THE KEYBOARD AND CHECK THAT SCAN CODE "AA" IS RETURNED TO THE PROCESSOR. CHECK FOR STUCK KEYS.'
- POST code 3Eh is where the POST reads the RAM in the 146818 CMOS/RTC chip to see if a hard drive 0 is defined. Then interrupts are enabled.

So, when you see 3E, the POST must have seen AA from the keyboard, and at that time, the POST is happy about the keyboard controller and keyboard.

Out of interest, what POST code do you see if the keyboard is disconnected from the motherboard.

Interestingly, my keyboard does light on power up, but as the system is sitting on 3E or 36, it doesn't respond to caps lock.

When the keyboard is doing a self test (at getting +5V, and later when commanded by the POST), the keyboard normally flashes its LED's as part of that self test. It depends on the keyboard.

Later, once the POST has finished, and user keyboard operations enabled:
- In at AT, the keyboard turns off/on its CAPS LOCK light in response to a command from the motherboard BIOS.
- User presses CAPS LOCK ---> motherboard BIOS becomes aware of that ---> motherboard BIOS sends command to keyboard to turn CAPS LOCK light off/on as appropriate

I doubt that such user keyboard operations are enabled until near the end of the POST.

... does it sound like a bad keyboard controller?

Odd to me is that (subject to your confirmation) that nothing is seen on screen apart from a blinking cursor.
The only thing that I am aware of that comes comes close to that, is described in the 'No progress past POST's RAM count-up on screen' section of [here].

What do you see in a minimized configuration of only: motherboard + video card ?

 

[Dave Farquhar]

So you power on the motherboard, and all that you ever see is a blinking cursor, nothing else? Please confirm.

Almost. I do see the initialization screen from my VGA card, then the screen clears and all I see from then on is a flashing cursor. It starts in the upper left, and advances to about line 12 or so, always staying on the left side. It never gets to the memory test.

It's not much like any of the symptoms on your site, which I found odd.

Out of interest, what POST code do you see if the keyboard is disconnected from the motherboard.

With no keyboard it eventually gets through 36h and hangs at 06h. That's the test of the DMA controller at U111, right? Maybe a bad 8237A?

What do you see in a minimized configuration of only: motherboard + video card ?

With no keyboard it hangs at 06h. With a keyboard, at 36h.

 

[modem7]

It's not much like any of the symptoms on your site, which I found odd.

I don't find it odd. Ever since I joined these forums in 2006, it has been the case of continual discovery.

For example, the 'No progress ...' issue that I pointed to was something that I accidentally discovered myself a few years ago.

With no keyboard it eventually gets through 36h and hangs at 06h.

06h is executed before 36h, so 36h is not getting executed.

That's the test of the DMA controller at U111, right? Maybe a bad 8237A?

One has to be careful in interpreting these things. An analogy I sometimes make is: On the dash of my car, a red 'battery' light is flashing. Yes, the battery may be faulty, but perhaps the problem is that the alternator is what is faulty, not charging the battery.

When I look at the symptoms presented so far, I do not see a singular cause (yet). And that brings up another possibility - multiple problems. In either case, one picks a symptom, fixes that, and if more symptoms remain, deal with the remaining symptoms one by one. Some symptoms are easier to diagnose and fix than others.

With no keyboard it hangs at 06h. With a keyboard, at 36h.

Interesting that with no keyboard, the POST stops at 06h. I will take a peek at the source listing for the 11/15/85 motherboard BIOS, because I have seen IBM insert non-related code into their blocks of POST tests.

Are you able to get yourself a set (U27/U47) of Supersoft/Landmark Diagnostic ROM's - see [here]? It would be interesting to see what they make of the situation. Note the special video card requirements.

 

[Dave Farquhar]

For kicks I tried starting machine up again after it had plenty of time to sit and cool down and I watched the whole thing. Once again, no keyboard attached, and just a VGA card and the postcard. This time, it got to 36h, sat there a couple of minutes (I probably should time it), then reinitialized the video card and started over again, got to 36h, sat there a while, reinitialized the video card and started over again, and this time it hung at 36h. Does this sound like a boot loop? And maybe whatever marginal/bad chip overheats first determines where it sits until I power it down?

As far as the diagnostic ROMs, I do not have an eprom programmer, but have been thinking about getting one. This seems like a good reason to go ahead. Seems like the logical next step. I will check in again after I get some eproms burned and can give them a whirl on the board.

Thanks again for all your help so far.

 

[modem7]

For kicks I tried starting machine up again after it had plenty of time to sit and cool down and I watched the whole thing. Once again, no keyboard attached, and just a VGA card and the postcard. This time, it got to 36h, sat there a couple of minutes (I probably should time it), then reinitialized the video card and started over again, got to 36h, sat there a while, reinitialized the video card and started over again, and this time it hung at 36h.

Certainly an 'unstable' motherboard (unstable symptoms).

Using an AT-class power supply? Have you tried a different power supply, just in case?

Re-seated all socketed chips?

Tried a different CPU ?

Does this sound like a boot loop?

The 'boot loops due to a fault' that I have experienced were due to one or more 'bad' bits on either the address bus or data bus. The looping occurred very early in the POST, and was seen using a logic analyser.

And maybe whatever marginal/bad chip overheats first determines where it sits until I power it down?

One possibility of a few.

As far as the diagnostic ROMs, I do not have an eprom programmer, but have been thinking about getting one. This seems like a good reason to go ahead. Seems like the logical next step. I will check in again after I get some eproms burned and can give them a whirl on the board.

If the SuperSoft diagnostic runs in a stable manner, then a possibility is that one of your IBM BIOS ROM's is flakey.

 

[Dave Farquhar]

I'm back. I did try reseating chips over the weekend, and that didn't have any effect I could see. I got the programmer and the chips in the mail today, so I am hoping I can get the diagnostic ROM chips programmed and run the diagnostics. Fortunately I did order extra chips so if I need to program new BIOS chips, I will be able to do that.

 

[Dave Farquhar]

I got a chance to run the ROM diagnostics. Unfortunately I can't find my CGA card so I had to go just with what the speaker beeps tell me. I did get the 4-1 sequence that indicates system memory at address 00000. Is that informational, or is that indicative of a problem? I do have a clone 286 board I can borrow memory from if I need to. I *think* the other board also uses 41256 DRAM.

 

[modem7]

I got a chance to run the ROM diagnostics. Unfortunately I can't find my CGA card so I had to go just with what the speaker beeps tell me. I did get the 4-1 sequence that indicates system memory at address 00000.

So, the 16K CRITICAL MEMORY REGION test (first 16 KB of RAM)(AT beep code of 2-1) is passing. Refer to [here]. 'System memory at address 00000' will, I am sure, equate to the on-screen test of MEMORY TO 080000, which will be RAM from 16 KB to 512 KB. A video card will tell you more, e.g. address of error, e.g. maybe for you, the ROM is testing up to 640 KB.

If that test always fails when you start the computer, then that suggests that you now have stability, something you did not have before.

Have you also tried burning another set of IBM BIOS ROM's ?

 

[Dave Farquhar]

Thanks for that observation, it didn't occur to me that yes, I had achieved stability. As I search for a CGA card and fail, I burned a Quadtel BIOS just to see if it might tell me something different. And it did. It counts to 128K and dies with code 60 (base memory test from what I'm able to see). It's even nice enough to print the code on the screen, but it matches what's on the POST card. It's consistent, and it fails very quickly. Is this starting to sound like memory chips (and bad BIOS ROM chips) to you? Do I need to replace it 256K at a time or is there a way to replace 128K at a time on the Type 3 board? I'm certainly willing to order a quantity of 41256s for this.

I have more EPROMs on the way so I will put an IBM BIOS back when the system is healthy.

 

[modem7]

Thanks for that observation, it didn't occur to me that yes, I had achieved stability. As I search for a CGA card and fail, I burned a Quadtel BIOS just to see if it might tell me something different. And it did. It counts to 128K and dies with code 60 (base memory test from what I'm able to see). It's even nice enough to print the code on the screen, but it matches what's on the POST card. It's consistent, and it fails very quickly. Is this starting to sound like memory chips (and bad BIOS ROM chips) to you?

The IBM POST's test of RAM is pretty crude. You have the SuperSoft diagnostic suggesting (i.e. audio only) a RAM problem between 16 KB and 512 KB. And now you have a Quadtel BIOS suggesting a problem at address 128 KB. So yes, the latter two are suggesting a RAM related problem. Of course, it may not be a RAM chip, but something else in the RAM system.

Do I need to replace it 256K at a time or is there a way to replace 128K at a time on the Type 3 board? I'm certainly willing to order a quantity of 41256s for this.

Replacing the RAM chips is an easy way of ruling out the chips (assuming the replacements are good).
Considering the pointed-to address of 128K, you will need to replace all 18 chips.

 

[Dave Farquhar]

IT'S ALIIIIIIIIIIIIIIIIVE! And the problem was soo stupid too. I found one 41256 with pin 1 broken and the stub bent inward just right to hide the problem from me when I reseated the memory chips. I have Checkit's memory test running now. Of course I don't trust much of anything about the system yet but getting a DOS prompt is huge progress.

Thank you so much for your help. I am incredibly out of practice working on PCs of this age.