PDA

View Full Version : Testing RAM by piggybacking - How reliable?



tezza
November 18th, 2008, 03:05 PM
Hi,

I'm about to start desoldering and replacing the 4116 RAM on my PET 3032.

One question. Assume a DRAM chip is faulty. Will piggybacking a good DRAM chip on top ALWAYS fix the fault, or does it depend what's wrong with the chip?

I know I asked a similar question months ago, but that was in relation to any IC rather than DRAM specifically.

Thing is, I've piggybacked a good chip over the bad ones and there is no change to the garbage screen. None of the chips get hot. However, this does not mean they are all good, right?. One or more may STILL be faulty, yes? Is this a correct assumption, or has the piggyback test actually verified these chips are OK?

Tez

Chuck(G)
November 18th, 2008, 04:05 PM
Hi,

I'm about to start desoldering and replacing the 4116 RAM on my PET 3032.

One question. Assume a DRAM chip is faulty. Will piggybacking a good DRAM chip on top ALWAYS fix the fault, or does it depend what's wrong with the chip?

The latter. Suppose you have a chip with a bit that's stuck one way or the other. How will a good chip on top of the bad one fix things? Suppose a chip has been damaged by ESD and an input or output is shorted to the substrate. Piggybacking won't help you either.

BG101
November 18th, 2008, 06:01 PM
I'm afraid it's a case of "socket and see" ;)




BG

Druid6900
November 18th, 2008, 07:03 PM
Shorted RAM chips, especially 16K ones will be HOT, hot enough to burn the chip info into your fingertip in less than a second.

On a chip with an open circuit, in most cases, piggy-packing will complete the circuit through the known good chip.

If you have MORE than one open chip (and I have seen it happen), you will probably notice a change in the "garbage" on the screen. You should leave a chip piggy-backed on that chip and use another to continue.

Ideally, if you have enough chips, each one should have a chip PB'd on it and, if the garbage clears up, remove one at a time until you get garbage, then replace it and continue, the chips that are still PB'd at the end is/are the bad one(s).

It doesn't work 100% of the time, but, damn close.

If having a PB'd chip on all the RAMs doesn't clear up the garbage, you should be looking somewhere else.

tezza
November 18th, 2008, 07:21 PM
Shorted RAM chips, especially 16K ones will be HOT, hot enough to burn the chip info into your fingertip in less than a second.

On a chip with an open circuit, in most cases, piggy-packing will complete the circuit through the known good chip.

If you have MORE than one open chip (and I have seen it happen), you will probably notice a change in the "garbage" on the screen. You should leave a chip piggy-backed on that chip and use another to continue.

Ideally, if you have enough chips, each one should have a chip PB'd on it and, if the garbage clears up, remove one at a time until you get garbage, then replace it and continue, the chips that are still PB'd at the end is/are the bad one(s).

It doesn't work 100% of the time, but, damn close.

If having a PB'd chip on all the RAMs doesn't clear up the garbage, you should be looking somewhere else.

Thanks Druid, I do have enough chips to piggyback them all so I might try this. However, one question. Is there a risk of damaging the good chips (either the piggy-backers or the piggy-backees) or any other part of the circuit by having all DRAM chips piggybacked?

Tez

Druid6900
November 18th, 2008, 07:28 PM
Thanks Druid, I do have enough chips to piggyback them all so I might try this. However, one question. Is there a risk of damaging the good chips (either the piggy-backers or the piggy-backees) or any other part of the circuit by having all DRAM chips piggybacked?

Tez

Very, very slim and probably only in the case of trying to piggy-back a shorted chip. The "hot finger" test is the first thing you try.

I know I wouldn't unsolder a whole bank of chips without trying it first and I'm good at it.

Chuckster_in_Jax
November 18th, 2008, 08:08 PM
I know I wouldn't unsolder a whole bank of chips without trying it first and I'm good at it.

I have an 8032 that appears to be dead. Power supply voltages are fine,so I unsoldered all the RAM and ROM chips and installed sockets and new chips. That was real tedious! The ground trace on the PCB is wide and acts like a large heatsink. It takes a lot of heat to get the solder to melt! I am lucky to have a nice desoldering gun(with vacuum pump) and a temperature controlled soldering station.
After all that it still doesn't work. I'll get back to that project some other time. It's one more on "the repair list".

MikeS
November 19th, 2008, 09:18 AM
Surely there are a few things that you can do before replacing all the soldered-in RAM; seems to me you're more likely to *create* a problem than solve one...

Unlike the static RAM in the older PETs, the dynamic chips are actually pretty reliable.

If you have a scope, or even just a logic probe, you can test the address and data lines to some extent by wiring up an adapter for the CPU to force NOPs on the data bus; another approach would be to temporarily cut a trace or two to disable the on-board RAM completely and wire up a 32K SRAM chip (e.g. 486 cache chip) to connect to the expansion bus.

Assuming the ROMs are socketed, can you read them in another system and compare to the original?

Good luck in any case!

Chuck(G)
November 19th, 2008, 10:21 AM
I still can't rationalize why piggybacking is particularly useful in the (as stated) vast majority of cases.

Suppose a bad bit is stuck low. Piggybacking a good DRAM won't change that, will it? The bad chip will still continue to drive the output low.

As Mike said, once DRAM has been checked out, it tends to stay good unless damaged by ESD or maltreatment.

Druid6900
November 19th, 2008, 11:14 AM
I still can't rationalize why piggybacking is particularly useful in the (as stated) vast majority of cases.

Suppose a bad bit is stuck low. Piggybacking a good DRAM won't change that, will it? The bad chip will still continue to drive the output low.

As Mike said, once DRAM has been checked out, it tends to stay good unless damaged by ESD or maltreatment.

The bit would be stuck on the output of the of the chip (internally), not the input (the problem, in that case, would be somewhere else) and the piggy-back RAM would take the good input and produce a good output back to the bus superimposed on the stuck-bit output.

Chuck(G)
November 19th, 2008, 11:52 AM
The bit would be stuck on the output of the of the chip (internally), not the input (the problem, in that case, would be somewhere else) and the piggy-back RAM would take the good input and produce a good output back to the bus superimposed on the stuck-bit output.

Which would give you what? ISTR that the V(OH) on the early DRAMs was a pretty far cry from 5v--more like 2.5V.

I could see piggybacking DRAMs and disconnecting CS\ on the one in the lower bunk, however.

BG101
November 19th, 2008, 04:03 PM
My thoughts on this are as follows. If a bit is stuck low (effectively the output signal for that bit is at zero potential) then bridging it with another chip is only going to pull the potput of the bridging chip down as well - a bit like bridging a short-circuit transistor and hoping to get an output?



BG

Druid6900
November 19th, 2008, 07:34 PM
Ya know something, fine, don't do it. Unsolder RAM chips one at a time until you either fix it or ruin the board.

tezza
November 21st, 2008, 07:24 PM
The piggyback test seems a good "quick-and-dirty" method to diagnose some cases of bad RAM. While obviously not definitive, it did work with my Osborne.

However, it didn't reveal anything in the PET. I suspect, there are likely to be a number of things wrong with it, and I think seeking a MB replacement might be the way to go. See pics at http://www.vintage-computer.com/vcforum/showthread.php?p=77599#post77599

Tez

rittwage
November 28th, 2015, 12:35 PM
It is a good, time-tested troubleshooting step. It doesn't hurt anything if you're careful. As stated above, if they don't get hot, they are OK to piggyback to test.

It doesn't always pinpoint the exact problem depending how RAM is used and wired in the machine, but much more often that not.

The C64 makes a great 4164 RAM tester, as the machine doesn't work at all with any of them bad since each chip is 1 bit of 8 total.
The Apple II makes a good 4116 tester- just use the first or second chip, they are zero-page and stack (or maybe each bank is wired 1/8 like the C64, I forget). Anyway, machine doesn't work with those bad.