Atari 800 RAM test

[JoSch]

I got Atari 800 some weeks ago.

I'm trying to get it working, but I think that some RAM is bad.

The 800 passes all the tests, it should pass.

So, how do I test the RAM chips? Is there some test program?

[ACML]

Try this Atari diagnostic program SALT. Press "?" for the help menu, then "R" for RAM. If its a 48K 800, enter "5" for the number of blocks of 8K to test. It can only test 40K as 8K is taken up by the program.

 

Atari SALT diagnostic CART v2.05 (1982).xex

[JoSch]

Thank you very much.

[JoSch]

Seems, I have a problem with the SALT program.

ACID800 runs and passes all tests, but SALT doesn't even want to start.

I put the file on my SIO2SD and booted. It crashes.

Then I tried starting via DOS. The only DOS, which boots at the moment is MyDOS. I started SALT from it. It crashes.

I tried some cartridges like Asteroid and Star Raider. They work, so it seems safe to say that the problem are most probably are RAM modules.

 

What suggestions, links or hints can someone give me to troubleshoot my problem?

[HiassofT]

Do you have a flash cart or some other possibility to create a cart (eg programming an 8k EPROM)?

 

If yes, download the SALT cartridge dump from here http://atariage.com/forums/topic/161828-hooked-on-8-bit-carts/page-8?do=findComment&comment=2476148

 

SALT 2.05 is a standard 8k cart, SuperSALT a standard 16k one.

 

Keep in mind that the carts occupy an 8k/16k memory area that can't be tested ($A000-$BFFF/$8000-$BFFF) so you'll need to swap around the memory modules in your 800 to ensure you tested all RAM locations.

 

so long,

 

Hias

[JoSch]

SALT 2 did run and could confirm my suspicion that one of the 16K modules is bad.

SuperSALT should run on a 800, but I could only get it running on my 130XE.

Is this a sign, that something else is bad?

[JoSch]

I looked at the bad 16K module and I noticed that half of the RAM ICs are not 4116, but HM4716AP-4 (see pictures.)

Are these legitimate replacement chips? OR should I always use 4116?

Edited April 26, 2016 by JoSch

[mr-atari]

-removed post- 4116 use 12 and -5 volt......

Edited April 26, 2016 by mr-atari

[1050]

They would not be legitimate based on speed rating of 410ns, you want something in the range of 100 to 150 nanoseconds for ram speeds. So, going by the downloaded datasheet, HM4716AP-1 might work fine here. 4116 are also available in the too slow speeds so it's not a cut and dry type of thing when you leave the speed rating unspecified. Can't zoom in on your pictures as posted due to hosing site (or?) to see any numbers well enough here, but you can see them first hand, and that's the important part. Speed rating usually follows the hyphen. Could very well be that when you replace these slow chips with good faster ones, it will become a good module?

 

Both these chips and all 800 memory chips that I know of use an odd 12 / 5 volt negative, positive voltage system. Thank goodness this only went on for a very short time before they simplified things with one 5 volt power pin.

[JoSch]

So I take it, that even the 4116 on this board are too slow because there are TMM416P-4, not TMM416P-1(5). Correct?

 

@1050: I hosted the pictures on Atariage, so if you click on them, you should see the number quite good

The data sheet, you mention, would be interesting, I only could find data sheet behind pay walls.

Edited April 26, 2016 by JoSch

[1050]

Absolutely correct sir and to the best of my understanding of the issue, so now I'm wondering how you got a full complement of stinkers on there? I'm pretty sure they weren't there from the factory, but I've been wrong before.

 

I just googled HM4716AP

http://www.datasheetlib.com/datasheet/732494/hm4716ap-4_hitachi-semiconductor.html

 

Looking close you can find a download link

http://www.datasheetlib.com/datasheet/732494/hm4716ap-4_hitachi-semiconductor/download.html

 

Where you have to enter some Captcha type nonsense and Bob's your uncle.

 

If you fancy a 310 page book, here is one for you

http://bitsavers.trailing-edge.com/pdf/toshiba/_dataBooks/1983_Toshiba_MOS_Memory.pdf

 

Used to be Datasheetarchive was pretty good but I stay away from them anymore. If you can find a Jameco site they often will have several free datasheets for each chip they are selling, DigiKey, Mouser, Futurlec too. And of course don't turn down a direct link from National, Texas Instruments, Hitachi, etc., ...

http://www.jameco.com/1/1/2199-4116-15-dynamic-ram-16kx1-150ns-dram.html

 

And then my firefox browser allows me to download any PDF I'm viewing online so I'll often take advantage of that toy as well. It does seem to be an evolving art form to get free datasheets, it always has been too.

 

Your pictures don't act like most pictures uploaded to this site though, they are about the same size as posted when clicked on instead of supersized right away and there is no readily available 'save' option either. Only thing I got going was yet another page to open with the same size photo on it. That's when I gave up since I knew YOU could see the full number and it didn't really matter if I could or not. Maybe I just need to get tricky with firefox?

[JoSch]

It gets funnier and funnier: I looked at the other RAM modules.

One also uses the TMM4716AP-4 chips:

 

The other uses MB8116E chips:

 

 

Both are considered good by SALT.

Any comments?

[Joey Z]

They would not be legitimate based on speed rating of 410ns, you want something in the range of 100 to 150 nanoseconds for ram speeds.

No, Atari is 1.79 MHz, even 500ns should be plenty for an <<2MHz system, if the system is designed right. Besides that, atari was known for using RAMs which were 'too slow' for the system they were in. As supposedly quoted from an Atari engineer, "You have to understand, DRAMs are really analog devices."

 

I'll add that my 800XL's original RAMs are -2's, which I would guess are 200ns.

Edited April 28, 2016 by Joey Z

[JoSch]

No, Atari is 1.79 MHz, even 500ns should be plenty for an <<2MHz system, if the system is designed right. Besides that, atari was known for using RAMs which were 'too slow' for the system they were in. As supposedly quoted from an Atari engineer, "You have to understand, DRAMs are really analog devices."

 

I'll add that my 800XL's original RAMs are -2's, which I would guess are 200ns.

Very interesting. More options to choose from

[1050]

Sorry, been away for awhile. Wow, others are suggesting strongly that 100 to 150 is the sweet spot for XL/XE and I'm in agreement with that camp. So this is somewhat disturbing news to learn that 400 can even work. I did have to really wonder how they got into that module since if they didn't work, why didn't the guy replace them with ones that did? That part of it caused some nagging doubt to grow I will freely admit. And the truth is out now - they do work.

 

I'm stumped as to how. But I'm accepting the facts as presented at the same time and grateful you called me on the carpet. I've never noticed on the 800 actually, I may have some myself. I am aware of our 500ns clock and that even slow 400 should work if other timing is done right, it's just that I know some eproms need to be one speed and not another depending on their purpose. From this knowledge then I assumed the same type of sweet spot existed for ram without really searching for examples probably right under my nose. By rights there shouldn't be a sweet spot, in my experience with eproms I know there is just the same without knowing the exact why of it. I've assumed too much again and gotten a bite in the hindquarters for it. A learning experience.

 

So the 'other' timing on board these modules is the CAS signal derived from RAS signal via inductor/capacitor/resistor time delay combo - the problem should be there instead I would then think. L501, R501, C518 for both 16K and 8K modules. Two 74LS158 Z503, Z504 then take that signal and do address multiplexing with it so a bad 158 could do it and the 74LS10 Z501 has the RAS signal running thru it first so if it's bad we can have no joy too.`With a bit of luck, some chip swapping should be able to pinpoint the guilty party in a short amount of time if one has spare parts known to work in other modules.

 

Now I want to play with my 800s, but I don't recall that I've got any bad modules to start with. Very interesting is an understatement, two platforms to perform experiments with no less. Thanks guys.

[JoSch]

Well, SALT II says it's the third 8K block, which is bad. The result of SALT II implies that no other IC is bad.

But surely I have to check that too.

[1050]

Ok, but that strongly points the finger at Z504, a 158 address multiplexer instead of half of the ram chips you might be thinking it does. Each bit of all addresses covered by a module are held in ALL 8 ram chips, it's impossible to find a bad ram chip with a ram testing program by the very way Atari uses ALL the ram chips in unison like this then. BUT we have one 158 chip that only has lower address lines as inputs and that's Z504, the other 158 contain high address lines in it's input mix. So if there is situation where lower addresses are failing, odds would favor that it's the low address multiplexor chip rather than the one that is involved with higher addresses. Third 8K block would be the first half of the second 16K module, right?

 

But when you know that there is no 'half' to a module in the first place and we still have this kind of report as if it were an actual possible, then something else is afoot. This just gets better and better.

[ClausB]

No, each DRAM chip holds one bit of each byte in tge 16KB range. So a good RAM test program can determine which chip is at fault.

 

The first 3 generations of DRAM, the 1Kb, 4Kb, and 16Kb chips did require mulriple power supplies. They weren't odd, nor were they short-lived. Every major microcomputer of the late 70s used them: PET, Apple, TRS-80, Atari, and others. It was not until 1982 that the cost per bit of the 5V-only 64Kb DRAM approached that of the venerable 16Kb 4116 chip.

[JoSch]

Ok, but that strongly points the finger at Z504, a 158 address multiplexer instead of half of the ram chips you might be thinking it does. Each bit of all addresses covered by a module are held in ALL 8 ram chips, it's impossible to find a bad ram chip with a ram testing program by the very way Atari uses ALL the ram chips in unison like this then. BUT we have one 158 chip that only has lower address lines as inputs and that's Z504, the other 158 contain high address lines in it's input mix. So if there is situation where lower addresses are failing, odds would favor that it's the low address multiplexor chip rather than the one that is involved with higher addresses. Third 8K block would be the first half of the second 16K module, right?

 

But when you know that there is no 'half' to a module in the first place and we still have this kind of report as if it were an actual possible, then something else is afoot. This just gets better and better.

To test your thesis, I switched Z504 with Z503, but the error code is the same.

So I have to wait for the RAM chips I ordered.

[1050]

Well, that's not a really valid test because we are only assuming that the test is accurate, it may not be and the bad chip is still in the circuit. This is two separate possible defects with that testing solution. I haven't disassembled the test to know if it is a lousy test program or not, has anyone?

 

Change the original suspected bad chip for one from a working module is the proper way to do this theory test. You may have the same results anyway. Which tells us my proposed theory may be all wet for reasons we just don't know about yet. It's probably dripping wet to start with. It could just as easily be that the test program defaults with this output during the first portion of the test rather than the latter portions, so the results are misleading as to where the defect is. The defect was found early in other words rather than later so the test program reports first half. NOT of the module though - of the TEST. I can't know this either until it's disassembled. All we have is a bag of possibles and taken as a whole, it's not a very valuable bag so far.

 

No, each DRAM chip holds one bit of each byte in tge 16KB range. So a good RAM test program can determine which chip is at fault.

Thanks Claus, but didn't you mean to say instead that a good ram test can determine exactly which module is at fault? Which is what I thought I had said too.

 

To put a finer point on it, a GOOD test can determine exactly which ram chip is at fault. BUT one of those GOOD ram tests doesn't seem to exist coming from Atari at least as far as I've seen yet. BUT I've only had a look at the XL/XE self test which can not determine which bit fails at all, they do it only by bytes there which can not pinpoint the bad ram chip. I only assume the SALT tests are written the same way, but I really don't know even that much.

 

So, can the SALT test really determine which half of the module is at fault? Which was my point. I was proposing that it really can't do half a module unless the address multiplexing is taken into account. I see now that instead of just one higher address line there is actually two used on the 16K Z504 (A8 AND A9) which cuts into our odds a good deal more than I was originally thinking. I'm thinking now that my theory might be a little bit wet. Odds before the recent re-appraisal were 1 to 6, now they are 1 to 3 which is enough of a change to give a bookie a heart attack depending on how much he had invested in the bad advice. I'm not so sure half a module can be determined by the test in the first place.

 

A really good test might be able to expose a fault in address multiplexing but that would be quite a mess to write code for too.

 

I didn't mean to denigrate or cast aspersions upon the venerable 4116, sorry if it was taken that way. You are absolutely right that it was the go to workhorse of the entire industry. I'm so very grateful I didn't have to deal with those power supplies for it was my entire point there. Not so bad when somebody like Atari has already done it for me.

[JoSch]

If you read the 400/800 service manual, there it says, that the SALT II test, even say whether some of the other ICs on the RAM boards are bad.

Sorry, I go with the test program, because it was the ATARI standard test system for their field techs. I don't believe they can't write diagnose programs for their own hardware.

BTW, if both Z503 and Z504 were bad, shouldn't I get more test errors? The SALT test doesn't stop on the first error. It scans the whole memory and then gives you matrix, which tells what is wrong.

[ClausB]

Thanks Claus, but didn't you mean to say instead that a good ram test can determine exactly which module is at fault? Which is what I thought I had said too.

 

No, I meant that a good test program can tell you which bit is bad in each byte, and that can lead you to the bad chip, because each DRAM chip holds one bit from each byte in the full range of the module. I don't know whether SALT reports bad bits.

[Joey Z]

BTW, if both Z503 and Z504 were bad, shouldn't I get more test errors? The SALT test doesn't stop on the first error. It scans the whole memory and then gives you matrix, which tells what is wrong.

Yes, you would get more errors than a single location if it were 503 or 504. Also, I've seen bad RAM *way* more often than a bad TTL chip (although it does happen).

[1050]

Ok Claus, we are going at it cross purposes mainly because I left out the word Atari as to memory tests in post #17 pointing at the defect in Atari memory tests as I know them - they don't test bits and can't show which memory chip is bad. This same defective test then also can't test half a module, yet SALT is reporting the 3rd 8K block is bad where only 16K modules are being tested - that's not possible either, I'm thinking. Intermittent operation of ram might be one explanation for that and I'll just have to be satisfied with that until a solution is found to see what it really was. Someday I might even be able to run the test myself just to see a matrix result.

 

I don't know about showing 'more' errors as I've never even run the test. You can place your faith where you want to and I don't mind a bit, but if you knew what a toy the XL/XE self test really was you might not be so faithful. I firmly believe they didn't write a memory test that could isolate a bad memory chip. It would still be in wide use and most would know about it. SALT, if it's anything like the code that was to come later, will be found lacking in many aspects. This part will have to wait for disassembly.

[ClausB]

I suppose it makes some sense that SALT would report RAM errors in 8KB blocks, since that is the unit of RAM selection in the 800 and 400 (see the thread below). So a repair tech would look at the sizes of the RAM boards (8, 16, or 32KB) and figure out which board contains the bad block, and then replace the whole board. So SALT is a board level diagnostic, not a chip level one, right?

 

As you said, the 3rd 8KB block is in the 2nd 16KB board. (Although it's actually the upper half, if you look at the details of RAM selection presented in the thread.)

 

http://atariage.com/forums/topic/177885-atari-800-ram-selection/