PDA

View Full Version : IBM Parity Memory Card 128Kb Not working



lutiana
March 9th, 2017, 06:14 PM
It just stopped working. So I am wondering how best to go about fixing it? I am thinking some or all of the chips maybe bad, they look to be Oki M3764-15RS (8Kbyte each) so does anyone have a modern equivalent that could work?

Also, I have a whole bunch of mixed 32kByte (M5M4256P, W41256PP-15 and TMM41256AP-10) chips, would they work on this card (making it 512Kb total), or am I stuck to only using the 8Kbyte chips?

Stone
March 9th, 2017, 06:30 PM
Are they really 8K each? Or are they actually 64K x 1 bit?

It's unlikely that all the chips are bad.

lutiana
March 9th, 2017, 06:35 PM
Are they really 8K each? Or are they actually 64K x 1 bit?

It's unlikely that all the chips are bad.

Does it matter? I have no idea as I have zero experience IDing chips like this, which is why I posted the number on the chip. And yes, I realize it's probably only one or two in either bank that't the probkem, but whichever one is bad it causing the whole card to not be counted in the RAM count.

Stone
March 9th, 2017, 06:45 PM
They're 64K x 1 bit.

It's a tedious process but replacing each chip, one by one, with a known good chip will ferret out the bad one(s).

It's better if you have a bank of good chips to swap but it can be done with just one or two (or three) good chips, depending on how many chips are actually bad.

lutiana
March 9th, 2017, 06:54 PM
They're 64K x 1 bit.

It's a tedious process but replacing each chip, one by one, with a known good chip will ferret out the bad one(s).

It's better if you have a bank of good chips to swap but it can be done with just one or two (or three) good chips, depending on how many chips are actually bad.

Yes, I am aware of that process, I did it with the 5170 board I have. But what I need to know is where can I get some known good chips? And if I can put bigger chips in the thing and get some memory beyond 640kb.

Chuck(G)
March 9th, 2017, 07:09 PM
It's also worth trying the piggyback trick to ID the bad DRAM chip.

lutiana
March 9th, 2017, 07:12 PM
It's also worth trying the piggyback trick to ID the bad DRAM chip.

That's not one I've heard of. How does one do the piggyback trick?

Stone
March 10th, 2017, 03:05 AM
But what I need to know is where can I get some known good chips?How many are you looking for?

They're available on eBay.

Depending on how many you want I may have some.


And if I can put bigger chips in the thing and get some memory beyond 640kb.No

I have a 256K card that will give you 768K.

Flamin Joe
March 10th, 2017, 03:48 AM
That's not one I've heard of. How does one do the piggyback trick?

http://www.minuszerodegrees.net/5150/ram/5150_ram_piggybacking.htm

Stone
March 10th, 2017, 04:43 AM
That's not one I've heard of. How does one do the piggyback trick?While it is worth trying it cannot give you a conclusive answer since there are certain chip malfunctions that will defy this test.

IOW, if piggybacking reveals a bad chip you're home free and know which one is bad.

But, if it doesn't indicate a bad chip that's not proof that the chip actually is functional so you will still need to perform an absolute test.

lutiana
March 10th, 2017, 06:43 AM
I have a 256K card that will give you 768K.


How much?


One thing I did notice is that the 512k card and the 128k card (that I have) are nearly identical, the silk screens are different, but the traces and board layout are the same. I am wondering if it is possible to make mine into a 512k card by adding in the missing ICs and ram chips. I probably won't do this, it looks to be a pretty massive soldering job (though it's all through hole) and I don't want to risk f-ing up this RAM card.

Stone
March 10th, 2017, 08:40 AM
It's a NOS HIcard AMS from RYBS Electronics.

$35 shipped

modem7
March 10th, 2017, 01:53 PM
One thing I did notice is that the 512k card and the 128k card (that I have) are nearly identical, the silk screens are different, but the traces and board layout are the same. I am wondering if it is possible to make mine into a 512k card by adding in the missing ICs and ram chips. I probably won't do this, it looks to be a pretty massive soldering job (though it's all through hole) and I don't want to risk f-ing up this RAM card.
Both cards are described and pictured at [here (http://minuszerodegrees.net/5170/cards/5170_cards.htm#128_meo)]. Per that web page, they use different classes of RAM chips and address the RAM differently.


It just stopped working. ,,,, I am thinking some or all of the chips maybe bad.

And yes, I realize it's probably only one or two in either bank that't the probkem, but whichever one is bad it causing the whole card to not be counted in the RAM count.
If one day, one (or all for that matter) of the RAM chips on that card (card addressed from 512K to 640K) went bad, then the IBM 5170 would display "201-Memory Error". I just now proved it on my IBM 5170.

Removing the card from the 5170 (i.e. simulating complete card failure) also results in a 201 error.

If I go into CMOS SETUP then down the base RAM amount from 640 KB to 512KB, then restart the 5170, I see "164-Memory Size Error" displayed.

Are you not seeing an error message ?

[ Note: The above assumes that your 5170 motherboard has IBM's BIOS for the 5170. ]

lutiana
March 10th, 2017, 09:38 PM
With the card in I get:


080002 0001 201-Memory error

When I take the card out I get:



080002 0001 201-Memory error
164 - Memory Size Error (Run SETUP)


So, yeah, it's the RAM on the card, which I already suspected due to having done this troubleshooting before I posted my original post (this is not my first time at the rodeo).

As to the comparison between the two cards, yes they obviously have different RAM and do things differently with the addressing, but I can tell you that the board layout and traces look damn near identical, and it makes perfect sense that they would design one circuit board that could accommodate both types then simply populate them as needed, would make manufacturing much cheaper than having to produce two different circuit boards (which in the 80s was very expensive to begin with IIRC).

I'd be willing to bet that with some time, some solder, extra ICs, dip switch blocks and some appropriate RAM chips I could turn this thing into a 512kb card. The RAM chips are shifted to the right, so I'd need to remove all the sockets in Bank0 and then add all the sockets for Bank1 and then fill in the missing ICs and dip switch blocks. But I have neither the time nor the desire to test this theory out.

modem7
March 10th, 2017, 11:51 PM
With the card in I get: 080002 0001 201-Memory error
080002 ---> Error in word at address 512K
0001 ---> error in bit 0

A chip reseat rather than replace-with-new may be the fix.

http://minuszerodegrees.net/5170/cards/5170_memory_board_128K_ram_layout.jpg

lutiana
March 11th, 2017, 01:48 PM
I pulled and reseated all the chips, same error.

I then swapped 0 and 15, thinking if I could get the location of the error (bit number) to change I could ID the bad chip. Needless to say this did not change the error still 080002 0001. So then swapped the two parity chips, no difference there either.

Is my understanding flawed here? Each chip is not considered a "bit" as per the error message decipher?

Stone
March 11th, 2017, 03:01 PM
Maybe the flaw is in the board and not the RAM chips. That would be consistent with the address being fixed. You could probably ascertain that by putting all the board's ram into another card that utilizes similar ram chips. If it functions normally the chips are likely good and the board is bad.

modem7
March 11th, 2017, 04:12 PM
Is my understanding flawed here? Each chip is not considered a "bit" as per the error message decipher?
Yes.

In the IBM '128K Memory EXpansion Board' within my function IBM 5170, I removed the bit 0 chip.
As expected, I then saw "080002 0001 201-Memory Error" message on power-up.

I then restored the bit 0 chip, and instead removed the bit 3 chip, then as expected, I saw the "0001" portion change to "0008".

Doing some experiments with various combination of two chips removed, quite often, only one bit was reported as at fault. That would be due to the way the IBM is testing. But I would 'repair' the reported bit, and then the other bit would be reported as at fault.

In your scenario, bit 0 chip proven good, then the other causes that I can think of are:

* Bad IC socket
* Bad solder joint on card
* Bad connection between card and motherboard socket - try cleaning the card's edge connector
* Chip U8 on card

lutiana
March 16th, 2017, 05:28 PM
Bought new chips from Ebay (8 of them, same make and model). Put one in Bank 0, same error, put back the chip and moved it to the Low Parity, same error. Tried the High Parity (because, why not?), same error.



* Bad IC socket


How does one test for this, visually, the all look ok to me.




* Bad solder joint on card


Again, not sure how to test for this. All the solder joints look good to me.




* Bad connection between card and motherboard socket - try cleaning the card's edge connector


Tried it in different ISA slots, cleaned the contact on the card with rubbing alcohol no change.




* Chip U8 on card


I am thinking this is the most likely culprit. But why U8 and not say U6 (both are SN74ALS245AN)? And is there an easy way to test these with something like a multimeter?

modem7
March 16th, 2017, 09:09 PM
I am thinking this is the most likely culprit. But why U8 and not say U6 (both are SN74ALS245AN)?
Refer to the diagram at [here (http://minuszerodegrees.net/temp/3/temp_89567812642949275784.png)].
Bits 0 to 7 go through U8, and bits 8 to 15 go through U6.


And is there an easy way to test these with something like a multimeter?
No.

But in relation to possible connectivity issues, yes.
The following will give you good confidence in the U50 socket (pins 2 and 14), the solder joints, and traces.

1. Leave RAM chip U50 in.
2. Using multimeter, ensure zero ohms (or very close to that) between pin A09 on card's ISA edge connector and pin 9 of U8 (the actual pin, not the solder side).
3. Using multimeter, ensure zero ohms (or very close to that) between pin 11 of U8 and pin 2 of RAM chip U50 (the actual pins, not the solder side).
4. Using multimeter, ensure zero ohms (or very close to that) between pins 2 and 14 of RAM chip U50 (the actual pins, not the solder side).

lutiana
March 19th, 2017, 03:02 PM
1. Leave RAM chip U50 in.


Done



2. Using multimeter, ensure zero ohms (or very close to that) between pin A09 on card's ISA edge connector and pin 9 of U8 (the actual pin, not the solder side).


0 Ohms measured



3. Using multimeter, ensure zero ohms (or very close to that) between pin 11 of U8 and pin 2 of RAM chip U50 (the actual pins, not the solder side).


0 Ohms measured



4. Using multimeter, ensure zero ohms (or very close to that) between pins 2 and 14 of RAM chip U50 (the actual pins, not the solder side).

If you mean pin 2 on U8, then I am reading 10Meg Ohms. If you mean pin 2 on RAM chip and 14 on the RAM chip, then yes I am getting 0 ohms.

So I am thinking the conclusion here is that U8 is the most likely culprit here?

modem7
March 20th, 2017, 12:33 AM
So I am thinking the conclusion here is that U8 is the most likely culprit here?
My money is on U8 too.

lutiana
March 20th, 2017, 05:25 PM
So I picked up a 512Kb RAM card (the IBM one (http://minuszerodegrees.net/5170/cards/5170_cards.htm#512_meo)) and plugged it in, same error message. But am I correct in saying that this card cannot be used to fill the RAM to 640KB? Or are there jumper settings that would allow me to do this?


The manual shows how to configure the card as the 1st, 2nd, 3rd etc, and the jumpers are set to the 1st card.

On another note, I managed to find what I think is an Enhanced Memory Expansion Adapter, which should arrive this weekend. Not sure how much RAM on that card, but time will tell.

bear
March 20th, 2017, 09:07 PM
Correct.

modem7
March 21st, 2017, 10:38 PM
So I picked up a 512Kb RAM card (the IBM one (http://minuszerodegrees.net/5170/cards/5170_cards.htm#512_meo)) and plugged it in, same error message. But am I correct in saying that this card cannot be used to fill the RAM to 640KB? Or are there jumper settings that would allow me to do this?

Correct.
"What bear said."

Extended memory only.
'1st' setting = 1M to 1.5M
'2nd' setting = 1.5M to 2M
'3rd' setting = 2M to 2.5M
'4th' setting = 2.5M to 3M
'5th' setting = 3M to 3.5M

lutiana
March 24th, 2017, 06:00 PM
Okay, replaced U8, painstakingly checked that each solder joint was continuous with some part of the trace coming off of it, so I can say that all there are no dry solder joints and there is continuity from the pin all the way to the board and down the traces.

That said, I am still getting the same error.

Any suggestions on where to go from here? Could it be some sort of clash with the video card (~512Kb Cirrus Logic GD5420 VGA card)? Bios issue?

modem7
March 25th, 2017, 08:16 PM
Okay, replaced U8, painstakingly checked that each solder joint was continuous with some part of the trace coming off of it, so I can say that all there are no dry solder joints and there is continuity from the pin all the way to the board and down the traces.
Back to socket U50. Apart from the DATA IN (pin 2) and DATA OUT (pin 14) pins, all the other pins are tied to the same pin on the other RAM chips (pin 1 to pin 1, pin 3 to pin 3, etc.). Do those other pins have continuity to the same pin on say, U51 ?


Could it be some sort of clash with the video card (~512Kb Cirrus Logic GD5420 VGA card)?
I think it very unlikely, but I have been surprised so many times.
Pity that you do not have another make/model of card to try.


Bios issue?
Again, very unlikely. If your 5170 motherboard has its original IBM BIOS, then I know that the card works in the 6 MHz 5170 with 01/10/84 revision BIOS, and 8 MHz 5170 with 11/15/85 revision BIOS.


Any suggestions on where to go from here?
Back at post #16, you wrote that you swapped the bit 0 chip with the bit 15 chip. Are you certain they got swapped? Didn't get distracted mid course?

Earlier, I observed that multi-bit failure was sometimes reported as single-bit failure. So try swapping other chips (besides bit 15) into the U50 socket. I am pretty sure that you have already tried that.

Do you have a logic state analyser? What about a logic probe ?