5150 Not Passs Mem Refresh Test

[KLund1]

I have a 5150 16K-64K board that dose not pass the SUperSoft Mem refresh test. All banks of ram were populated for the test.
I reflowed the 8253 and got the same results.
I then pulled all the socketed ram leaving just the solder 16K. Now SuperSoft reports critical ram area failure on all bits.
Does this suggest that the 8283 is bad? Or 8237?
Before this, I had to replace several caps on the -12 line to get the board to boot.
Suggestions?
Thanks

 

[mikey99]

There are some specific notes about SuperSoft and the 16-64KB motherboard on modem7's site:

http://www.minuszerodegrees.net/supersoft_landmark/5150/16KB-64KB.htm

What symptoms do you observe when the board boots with the original ROM chips.

 

[KLund1]

Thanks mikey99, I was already familiar with those issues.
On normal boot I get 1 long beep. This, I think, means something with 8253, or the 8237.
With critical ram test passed with all ram installed, I'm not quite sure where I should look next.
OH, thanks for the quick reply, much appreciated.

 

[modem7]

What symptoms do you observe when the board boots with the original ROM chips.

On normal boot I get 1 long beep.

Which of the three possible IBM BIOS' do you have on your 16KB-64KB motherboard ?

What does Ruud's diagnostic ROM (found [here]) report ?

 

[KLund1]

The system had the 1983 BIOS upgrade. There was a HD in the computer so they need that upgrade.
I do not know what size eprom to use with Ruud's ROM. I have wanted to try it. I have a lot of eproms to use in many sizes.
Thanks for the reply!

 

[KLund1]

I just took a guess on a Eprom for Ruuds diagnostics and it programmed and boots on the 5150 MB !!
I get 8253 timer channel pass 0,1,2, 8237A DMA Controller pass. 8255 Parity check pass
Then " Check First 2K of Ram" Critical error stop....
I did the piggyback on all 8 chips with probably good 16K chips I pulled from the other ram sockets on this board and get no change.
Which IC is the 1st 2K chip?

 

[Stone]

I did the piggyback on all 8 chips with probably good 16K chips I pulled from the other ram sockets on this board and get no change.

What about the parity chip, #9?

'Probably good' means you 'could' be wasting your time.

Which IC is the 1st 2K chip?

Unless you specify which bit is in question you can't tell which chip it is.

 

[KLund1]

Well Stone, I have about 100 16K, 100 64k, 100 256K chips but I do not know if they are all working. I do not know of a quick cheap easy way to test them all. The same when you buy 'new'. No way to know for sure if they are good. I assume all are good until proven otherwise. With Atari 8-bit's, it is repetitively easy. If the ram chip gets hot quick on power up, it is bad. Not so on 5150-70's.
Ruud's diagnostics does not show which chip is bad. Supersoft also is unreliable for the 1st 16K.
I was hoping the piggyback method might show which is bad. But no joy so far.
See pics for both diagnostics output.
Both programs seem to contradict each other.
Some thoughts on what might be wrong here?
Thanks

 

[Ruud]

I get 8253 timer channel pass 0,1,2, 8237A DMA Controller pass. 8255 Parity check pass
Then " Check First 2K of Ram" Critical error stop....

My BIOS tests counters and registers that can be written and then be read again. That the individual test went OK is not a guarantee that the refresh as a whole works fine. A damaged gate between two of the ICs is all what is needed to create havoc.

The 2 KB test test the first 2 KB of RAM. In this case it isn't a single chip but the first 2 Kbit of every chip in the first row. I test it by writing various values to it and reading them again. But the reading is done after waiting some time. If the reading goes wrong it simply means that the wrong value has been read. The wrong value can be caused by one or more bad DRAM ICs or a bad refresh (no contradiction in this case). Maybe I should extend the test with showing what the written and read value were.

About piggybacking: just don't do it. It is indeed quick but also dirty, very dirty. If a gate is broken, it can be so broken that a good one on top of it won't help. And it con be so good broken that it destroys the good one as well. Just take some time, remove the original IC and replace it with a scocket.

 

[KLund1]

Ruud, Thanks for the reply. Your diagnostics are a great effort.
This my 3rd 5150 with similar issues with the first 16K. There is still know way to ID which individual chip(s) are causing the problem.
That is a a lot of chips to pull and put sockets in for. Also the ground and + volt pins are really hard to get out without some damage to the hole and/or the board.
Is there are procedures to use a logic probe to find the bad chip(s), perhaps?

 

[Ruud]

There is still know way to ID which individual chip(s) are causing the problem.

I just have to find some time to alter my diagnostic ROM as proposed above. The difference in the written and read value should point to the offensive bits.

Also the ground and + volt pins are really hard to get out without some damage to the hole and/or the board.

I recognize the problem. In the worst case AND having enough spares of the IC, I just cut it out the board (cutting it at the top of each pin). I then only have to worry about one pin each time.

Is there are procedures to use a logic probe to find the bad chip(s), perhaps?

Assume the 8253 timer signals the 8237 that it is time to refresh the DRAMs but the signal never arrives for what ever reason. I doubt very much if you can see the little blip wit a probe. A logic probe? No, or you have to be very lucky.
A oscilloscope would do better. Much better would be a logic analyzer. But the prizes are "better" as well.
Unless you have a lot of time at hand, wait for my new ROM that will tell you why the 2 KB RAM test failed.

 

[Ruud]

Big bummer: I found a bug in the memory test. I already found the test quite fast and I thought that were my programming skills. Checking to find a way to output the reason why the 2 KB test halted the diagnostics, I saw that I used register DX twice: once to save the contents of CX and once as I/O register. It meant that, instead of restoring CX with its original value, it was written with this I/O value, 61h, instead. 97 or something like 2048 is a big difference. And why did I use DX in the first place? 61h can be used directly as well in an IN or OUT command.

To make things sure: this bug could never cause the error KLund1 ran into.

 

[modem7]

The system had the 1983 BIOS upgrade. There was a HD in the computer so they need that upgrade.

Per [here], that will be a 1982 dated upgrade (10/27/82).

What symptoms do you observe when the board boots with the original ROM chips.

On normal boot I get 1 long beep.

Ruud's soon-to-be-modified diagnostic may reveal a bad RAM chip (or chips) in bank 0, however I am curious about the long beep you are hearing from the 10/27/82 BIOS.

The breakdown of the POST in the 10/27/82 BIOS is at [here]. The only single beep is at step 27, and that is a short beep, not a long one.

And, if there is indeed bad RAM in bank 0 (16KB sized on the 16KB-64KB motherboard), the POST would halt the CPU at step 13, which is before any of the steps that issue any beeps.

So, puzzling.
How long, in seconds, is "long" ?

 

[KLund1]

Ruud,
Thanks for the input. I am glad to be of help. Let me know when the new ROM code is ready and I will burn an eprom, and test on this board. For me, finding the issue with the 1st 16K has been difficult on many boards. Some I have yet to solve. Those boards are gathering dust, awaiting a fix so a new user can appropriate them.

modem7. I as gong from memory on the beep. I will reconfigure to original chip set and report back with accurate data.

As always, I very much appreciate all your help in my learning process with these vintage machines!!

 

[KLund1]

Hi, all I am back to the board again.
I put the original ROM back in and tested. I get no beep at all on power up. I must have been thinking of another machine, sorry.
I tried another speaker to verify, and get nothing. I do get a small click when I power it down.
I do get the same screens from RUUD's and Supersoft ROM's.
I forgot I have an ISA/PCI PC Analyzer card like this:
https://www.amazon.com/4-Digit-Display-Analyzer-Diagnostic-Motherboard/dp/B00RX56KWC
I only get red LED's from 3.3+v and 12+V. I do not think that is right. AS mentioned above I did have to replace several caps near the Mem chips to get it to boot. Maybe there are others I should replace. if so, which?
Suggestions.
Thanks

 

[modem7]

I forgot I have an ISA/PCI PC Analyzer card like this:
https://www.amazon.com/4-Digit-Display-Analyzer-Diagnostic-Motherboard/dp/B00RX56KWC
I only get red LED's from 3.3+v and 12+V. I do not think that is right.

That card is detailed at [here]. The +3.3V LED is monitoring the -5V contact in the ISA slot. Therefore, expected to be lit, are all four voltage LED's.

You must have +5V present, otherwise the Supersoft diagnostics and Ruud's diagnostics would not be executing.

BTW. Do not expect any diagnostic codes to display when that particular model card is plugged into an IBM 5150, even if you know for fact that a diagnostic ROM outputs codes. Vintage POST cards are a different story.

 

[modem7]

I do get the same screens from RUUD's and Supersoft ROM's.

The Supersoft ROM screen shot at post #8 shows failure of the MEMORY REFRESH test (specifically bits 6 and 3).
At [here] is an example of the MEMORY REFRESH test failing because of a bad RAM chip in bank 0.
Bad RAM chips in your bank 0 ????

 

[KLund1]

Many thanks for the posts.
So I replaced bits 3 and 6, chips U41 & U44. I did my best to not damage the MB. I had to clip a few pins that did not want to let loose. I have a vacuum solder remover. But that could not help with the suborn pins. Is there some trick to getting these to just fall out of the board like the many videos I have watched to try and learn how to remove them without damaging the MB?
I them put a bit of flux on the holes from the back,and put in a socket. I them soldered the pins in. I did a bit of clean up with ISO. I picked 2 random chips from those I pulled from the MB sockets a couple weeks ago.
I powered it up with SuperSoft and now bits 1-7 , not zero are showing as bad. I know this test is not 100% on the first 16K, but did I damage the board with all the heat I had to use to get those stubborn pins out? Or do I need to just proceed the removing chips and testing until I can pas the test?
Also I do not know 100% if the pulled ram chips are good. I looked online for a cheap 4116/4164 ram tester and there dose seem to be such a unit available. How do I test chips to see if they are good?
Thanks

 

[modem7]

I powered it up with SuperSoft and now bits 1-7 , not zero are showing as bad. I know this test is not 100% on the first 16K, but did I damage the board ...

With the banks of RAM, a lot of signals jump from one RAM chip to the next, to the next, ...

An open trace would explain the change in symptom - see [here].

The circuit diagram will show which pins across the bank 0 chips are meant to be joined. Use a multimeter to see if those joined pins are no longer joined.

How do I test chips to see if they are good?

I see if they work in a functional motherboard.

 

[KLund1]

Thanks Model7 !!
your link helps.
I checked the continuity from U3 to U45 in bank 0. All looks good. Not every pin shows 0.0 on my meter, (.LO = no continuity on my meter). Some have numbers that change. But I think that might be the circuit? Pins 2 and 15 show .LO, but the same shows on banks 1,2,& 3.
What should SUperSoft display if 1 or more ram chips in bank 0 are missing?