How to check RAM on a PCB?

[ken_cinder]

I have a Taito F3 with a Puzzle Bobble 2 cart that has developed a problem.

A problem I was able to fairly accurately reproduce by compiling my own MAME in order to leave suspect ROMs out etc.
I narrowed my problem down to 2 SPRITE ROMs on the cart, but I don't know how to tell if it's the actual ROMs or RAM that is causing the problem.

This is the problem pretty accurately reproduced with MAME. ROMs e10-01 and e10-02 are responsible for the overlaid sprites you are seeing messed up. Disabling either one or both leads to this problem pretty much looking the same.



What I don't know, is if the ROMs themselves are actually the problem. A respectable source diagnosed another board I have with similar problems in background tiles as being a RAM problem, which led me to believe I better damn well make sure this one is or isn't before ordering new ROMs.

I just don't know how to diagnose bad RAM chips.

[MonMotha]

Many commercial "device programmers" (things used to program EPROMs, EEPROMs, Flash, microcontrollers, etc.) are also capable of testing RAMs and ROMs of various standard types.  If you happen to know somebody with such a unit, they should be able to test these devices for you.

[ken_cinder]

Unfortunately I don't, I had enlisted the services of a person I was referred to, to burn the SPRITE ROMs for me.
After realizing it could be a RAM problem, I put him on hold for now.

I was hoping there was a way I could at least check them to be physically functioning, and go from there.

[MonMotha]

Physical function is what you can easily check in a device programmer.  You're not using it as a programmer in this case (to permanently put data on the device), but rather a tester.  For RAMs, the device programmer will write patterns into the device (which will of course be lost when power is removed) and then read them back to see if the readback matches what was written.  It then repeats with several other patterns specifically designed to expose problems.  If everything compares "right", then the device is good.

ROMs can be tested in a similar manner by reading them and comparing with what is known to supposed to be on the device.  A good tester will also test over varying Vcc.

There's no way to verify functionality via physical (visual, mechanical measurement, etc.) inspection.  What you could do if there's more than one identical chip (not familiar with the F3 hardware) is swap the RAMs around.  If the problem manifests itself differently, then a RAM is bad.  If the problem stays 100% the same, then it's probably not the chips you swapped (though part groups that make up a wider data bus than a single part can sometimes produce similar results when swapped around - pay attention to all details).  Of course, having a known-good spare is even better: swap that in for each candidate "bad" part, and see what happens.  If the problem goes away, you've found the bad part (and fixed the problem).  If the problem remains, look elsewhere (or suspect multiple bad devices).

There are generally several things that can cause bad sprites ranging from bad ROMs to bad RAMs to improper address decoding to simply corroded traces (resulting in poor/broken connectivity).  ROMs and RAMs are usually the easiest to test without a logic analyzer and/or detailed information on how the system is otherwise built.

[ken_cinder]

I don't have another board to borrow any parts from, so I'm rather limited here. F3 game boards aren't cheap either.

Both boards are in excellent physical shape, and I've gone over them countless times looking for traces or shorts or anything like that.

The RAM for these ROMs is located right next to them, so I'll give the suggestion of swapping them around a try.

[MonMotha]

You can't swap the RAMs with ROMs or vice versa.  That won't work (and could easily fry the chips, depending on pinout).  You can, however, swap identical RAMs or ROMs with other identical chips should there be more than one.  Note that swapping sprite ROMs around will likely lead to terrible graphics corruption, and swapping program ROMs around will often get you something that doesn't work at all, so swapping ROMs for non-identical data (i.e. from a known good board) is not of great utility.

For things in good physical shape, visual inspection rarely turns up anything.  Best bet if you can't do swaps to find anything wrong (including from another board) is to yank the chips off and put them on a tester.  Just need to find somebody that has one, since those programmer/testers aren't overly cheap.

[ken_cinder]

I'm not THAT dumb lol

There are 4 Sanyo LC321664AM-80 on the motherboard right next to the connector that leads to the game board area where the sprite roms are.
I'm going to swap those around and see what happens, if I can clean up the garbled mess in 1 area, I have bad RAM I take it.

Wonder if I can even find replacements if thats the case.

[SirPeale]

Are the RAM socketed?  That would be an easy test if it is.

[ken_cinder]

No SOP40 mounted right on the board, and I've done some swapping and determined they're the problem.

Now I need to replace them, but here's the real problem! I can't find anywhere I can buy them, without mimimum orders of $100 or hundreds of pieces at a time.

[MonMotha]

That's a 64Kx16 (1Mbit) Fast-Page Async DRAM with byte selection lines.  Pinout looks like what I remember to be the industry standard, so that's probably good.  You can usually substitute something that's faster.  You may be able to substitute a larger part of the same width (e.g. a 128Kx16) due to the address multiplexing and pinout compatiblity features.  With larger parts, you may have some lines that were N/C on your original device that become address on the new one, but usually it's OK to tie those high or low, and the extra space will go unused.

Parts of this nature have been long EOL, and the 16-bit wide parts were kinda hard to come by to begin with, so this probably won't be easy to find NOS, especially in low quantity.  You'll likely end up ripping it off something else, probably another arcade game.

I looked at the boards I have laying around, and I didn't see anything suitable as a replacement.  Closest I found was a 256Kx16 EDO with an incompatible pinout.  Sorry.

[ken_cinder]

I decided to go the route of a supplier search site, and much to my surprise found many suppliers willing to fill a 4pc order. My inbox is still being flooded with offers after making a request.

HKinventory.com is where I went.

[MonMotha]

Ah, those places.  I've dealt with those sorts of places before.  Sometimes you get great service, good product, and good prices.  Sometimes you end up with (I kid you not) lead frames with no die in them and a company that mysteriously disappears with the money (which you sent via wire transfer, so no recourse).  Just make sure the company seems reputable.  I doubt many would go to the lenghts required to actually cheat you out of your money on a 4 piece order, but who knows.  You'll likely be getting pulls, not NOS parts, and they aren't usually well tested, so order extras.  If you are expecting NOS, make sure you get that confirmed in writing so that if you get pulls with damaged leads or something, you can attempt to get things straightened out.

[ken_cinder]

Yeah I've decided I'm not dealing with anyone who only accepts TT/Wire Transfer as a method of payment. As long as they accept a payment method that covers my butt, I will deal with them.