Root cause found regarding BAD DMA phenomena

[zChris]

Hello!

Havent seen this been posted here so i thought i would toss it out. Seems that someone has found the rootcause of the bad dma phenomenon that has plagued atari ste's.

https://www.chzsoft.de/site/hardware/ne ... stigation/

/zChris

[alexh]

I imagine everyone has been watching it over at Atari-Forum?

Interesting bit of detective work.

A timing fault like this explains why some bad DMA work and some don't. Process variation.

[exxos]

Great that someone did further research into the long known problematic RDY signal. I think it would have been little courteous to mention I documented the RDY signal problems a decade ago in my original research. Indeed I've been saying tolerances play a part in all this for years now. This is why you can swap a DMA chip from a STFM into a STE and it works, but also that the DMA which was in the STE, also then works in the STFM etc.

The GST side of the investigation is interesting. The "timeline" for the GST would make more sense, if an early version of the GST was present without those inverters. It would make more sense that Atari would try and do a "quick fix" by later adding those double inverters into the GST, , Assuming that is true,it obviously did not work as they had to re-spin the DMA anyway. But without knowing more about the chips and the history etc, I think it is difficult to come to a solid conclusion there. We don't know how many revisions of the schematics or GST chip there actually was ? There are multiple inverters driving ICLK8 elsewhere as well. Though is also important to not forget "DMA problems" are also present on the STFM. I documented problems with the RDY signal on the STFM as well. Also issues with the ACK and DRQ signals with gigafile. The -001 "GOOD DMA" does not solve all problems as claimed either.

They claim the HC CPU does not solve the problem for them which is odd, to be clear I never claimed it was a "fix all solution", though to have no effect is rather strange. To my knowledge it has always fixed people's problems and its been verified many many times.So they could possibly have some weird "corner case" machine. As we know all these machines seem to have their own unique quirkes.

Another problem is that people have also added a diode in series with the power, like they do with the IMP DMA , which would slow down the timings but this was reported to have fixed the problems. This is contradictory to what is stated about the longer delay helping matters when that particular test a shorter to delay would have worked. I also found noise on the signal which I squatted out with a capacitor which would also slow down the signal fractionally but I found that also solved the problem in some cases. But this could have technically been still a faster signal between the DMA and GST if there is a DC offset which got squatted by the capacitance. The problem is you really go down the rabbit hole with all these problems and it never ends which is ultimately why I gave up looking into all these issues.

They do not state what motherboard revision they are testing either. If they have one of the earlier ones, they have different pullups on various signals compared to later ones for example, this could may well be a factor they have completely missed. This is down to Atari's own fixes. So we do not know if these were even done or not.

We also don't know if they have verified they have a genuine HC CPU. It could be a restamped NMOS if they got it from China. There are of course other factors I found such as bad power supplies, if your power supply is not stable then the DMA circuit is 1st to start malfunctioning. So I'm not sure why they are trying to insinuate otherwise. I think a lot more clarification is needed in what they are exactly trying to say, as it sounds very misleading how it is worded to me.

Their RDY signal looks strange to start with. It looks more exponential than linear as it is supposed to be. Though it looks like they have ground bounce and voltage differential going on at the same time. I have documented up to about 1VDC offset signal which could also be playing a part. It could easily add 20ns onto their timing screw it up. If the HC CPU adds more delay, and that signal is that bad to start with, then it would explain why it didn't solve their problem. But again this could be down to the particular motherboard revision they are using. I can't ever remember seeing such a bad signal. But knowing how screwy these machines are I am really not surprised.

I just tried the timings between a original CPU and HC CPU an there is a 12ns extra delay there, which is claimed to be the root cause of the problems. So if this is the root cause of the problems, how does the HC CPU not fix them ? clearly there is some other factor going on here. I also tried the IMP DMA which gives a negative -8ns delay. So probably explains why the IMP never worked in the STE. The diode which gets added would slow it down more, so the STFM doesn't seem to care about the delay. But the RDY signal is still very sensitive in other respects as I documented.

There seems to be some false assumptions going on as well. For example a link to my old blog (why?) where I state you can run the machine for several hours without a failure with Jookies DMA 's program. This is of course true, but not entirely due to later research which was done. Timings can drift between the machine being hot and cold for example which can cause intermittent failures. Also some machines can be on for several hours and pass, then reset and it will fail a couple of random times and then pass for hours again perfectly well. So it seems word access can fail also though, albeit more intermittently. Floppy access can also exhibit the same type of corruption as the hard drive.

The comments like "people have blamed noise" etc for issues I assume is a insinuation towards myself. However Atari themselves have also blamed noise for DMA issues. So I somewhat resent such comments personally. Atari themselves concluded these problems and the article seemingly, and incorrectly, tries to dismiss such claims as made by Atari Themselves. They even link to the page where that information is even present. So I wonder if they have even read the page properly.

I do find it interesting about the word versus long word hypothesis. It is certainly something to consider. I have contacted Jookie about that and asked him to update his program. It certainly cannot hurt to have additional tests.

As for Atari-Forum, I had a quick look at that thread and just seem to be the usual exxos bashing thread, people trying desperately to find some way of disproving my work as usual. Exactly the reason why I quit being moderator and no longer go to that toxic dumping ground. People have been arguing for the past decade about such things and no doubt will still continued a decade later. I'm just keeping out of it all.

If someone has a STE which does not work with a HC CPU and a -38 DMA, then I would be interested in obtaining that motherboard along with the original CPU with it for further analysis. Though again, if it is one of the earlier boards then Atari's own fixes need doing as the HC CPU may well then solve the problem as normal anyway.

In any case,I feel I'm starting to repeat myself again on many points. There are indeed multiple reasons why the DMA circuit can fail, on the STE and STFM. Timing problems between the DMA and GST are indeed interesting new findings though.

EDIT:

I was in two minds whether or not post this response, but I feel it's incredibly important point some things out yet again..

I’ve got to say, the “A New Atari STE Bad DMA Investigation” rubs me the wrong way. It’s not that his findings are baseless,he’s got a solid point about the RDY signal timing glitch causing those 12 missing bytes in his test case. Myself and putnik also documented similar years ago. But the way it’s presented feels like it brushes off years of my research into Atari ST/STE issues, and that’s hard to swallow.

My work spanning the Gigafile tests in 2013, the DMA deep dives from 2014-2017, and the “DMA Survival Guide” up to 2024,shows these problems are a tangled mess of factors, PSU instability, bus noise, motherboard quirks / revisions, CPU variations etc, you name it. I’ve proven with user reports and my own mods that these all play a role, yet his article zooms in on one glitch and calls it the “root cause,” sidelining everything else.

Take the PSU: my tests show a failing power supply can cripple DMA stability, causing voltage drops and corruption documented for over a decade. The article dismisses this as a “short-term fix,” barely engaging my evidence. Ignoring a 40+ year-old PSU risks misleading users into thinking it’s fine, when it could be the root cause. Troubleshooting without checking the PSU first is reckless.

Same with Atari’s own pull-up mods, upgrading resistors from 10k to 4.7k or P100 to 1.2k to solve issues,which I built on with 2.2k and 10K fixes that users have verified. He doesn’t mention those or even say which STE revision he even tested, these things are important and it's leaving a gap that makes his conclusions feel narrow. My HC CPU swap, shifting timings by up to 22ns with pullup changes, isn’t some random tweak, it’s a deliberate fix for issues, backed by real results. Dismissing that as a fluke ignores the data.

The article suggests that the HC CPU makes no significant difference, which feels dismissive of my extensive testing and countless users that have proven the fix to work and be reliable over the years. My results, however, demonstrate a measurable timing shift of up to 22ns with the HC CPU in similar tests. This discrepancy raises questions about the article’s findings,methodology and conclusions. While the article bases much of its argument on the HC CPU having no impact, my data clearly shows otherwise!

The article’s tone “new investigation,” “ending speculation” sets up this vibe like my findings are just noise to be cleared away. That stings, especially since I flagged RDY signal issues back in 2013, and Wolfgang (Suska’s creator) beat me to it even earlier. "The lightning team" acts like they are the first to crack this, but they are standing on groundwork I laid.

Worse, they push the C398739-001 DMA as Atari’s grand fix when I’ve shown it’s no magic bullet multiple times, my video proves it can fail or even worsen things in some STEs as others have also proven. Claiming Atari “knew the root cause” feels optimistic when my tests say swapping DMAs is a lottery, not a solution, and skips basic troubleshooting like checking PSUs or hard drives etc. The best fix Atari did is change the pullups and swap to the SGS brand CPU. That resolves more DMA issues than using the C398739-001 DMA chip in my research. I have never, not even once, needed to fit the C398739-001 DMA chip to resolve hard drive problems.

I get why some might like his clean, lab style take, it’s tidy, with a possible software workaround to boot. But it’s misleading to boil DMA woes down to only one issue when I’ve documented a dozen flavors, including floppy errors which are completely ignored, HDD corruption, read vs. write failures. It’s not “exxos bashing” in some grand conspiracy sense, but it fits a pattern I’ve seen,folks in the Atari scene quick to challenge my fixes like the HC CPU or pull-ups, calling them ineffective, despite them saving countless machines. This constant dismissal isn’t just frustrating , it’s exhausting.

Why the resistance? Maybe it’s purists clinging to “bad DMA” myths, or sellers hyping “good DMAs” over cheap mods. Beats me. All I know is I’ve sunk thousands of hours into this because no one else was, only to see it downplayed by a crowd that’d rather nitpick than say thanks. By ignoring Atari’s own mods and overstating the unreliable C398739-001 DMA as a solid fix, the article’s “end speculation” tone downplays my user-verified work, obscuring the problem’s true complexity.

I’m not here to offend, just vent. The article adds a footnote, fine, but it doesn’t rewrite the book. I’m tired of rehashing it and I regret diving into it all now. Future investigations? Not my circus anymore. The community’s got my notes, they can run with it, or not. I just want folks to see the full picture, not this one-note narrative.

[alexh]

As for Atari-Forum, I had a quick look at that thread and just seem to be the usual exxos bashing thread, people trying desperately to find some way of disproving my work as usual. Exactly the reason why I quit being moderator and no longer go to that toxic dumping ground. People have been arguing for the past decade about such things and no doubt will still continued a decade later. I'm just keeping out of it all.

A lot of the posts in that thread are by dlfrsilver, don't take them too seriously, I love him dearly, he's a super nice guy who is very enthusiastic about what he does. But English is not his first language, he is not technical and he does "make stuff up" if he doesn't know the answer and insist it must be the truth because he says so. (Something probably lost in translation)

[exxos]

A lot of the posts in that thread are by dlfrsilver, don't take them too seriously, I love him dearly, he's a super nice guy who is very enthusiastic about what he does. But English is not his first language, he is not technical and he does "make stuff up" if he doesn't know the answer and insist they must be the truth because he says so. (Something probably lost in translation)

He is very enthusiastic yes, but I don't generally have a problem with him to be honest. I know he has got branded a lot of times for bashing Atari, but I don't think that is really his intention. I have found loads of problems wrong with these machines over the years and I state them a lot of times, but it does not mean I'm necessarily hating on Atari, It could have be seen that way though. Of course I don't really know what goes on on other forums or how people behave, or the millions of past arguments which may or may not have happened. I just try to keep away from all the drama these days.

[Dlfrsilver]

As for Atari-Forum, I had a quick look at that thread and just seem to be the usual exxos bashing thread, people trying desperately to find some way of disproving my work as usual. Exactly the reason why I quit being moderator and no longer go to that toxic dumping ground. People have been arguing for the past decade about such things and no doubt will still continued a decade later. I'm just keeping out of it all.

A lot of the posts in that thread are by dlfrsilver, don't take them too seriously, I love him dearly, he's a super nice guy who is very enthusiastic about what he does. But English is not his first language, he is not technical and he does "make stuff up" if he doesn't know the answer and insist it must be the truth because he says so. (Something probably lost in translation)

To make the answer short: what happened on AF just smelt flame war, and i perfectly understand what some were trying to do.
Exxos did lots for the Atari Community, and some people inside just choosed to dismiss him or spat at his face. And then, one guy comes and he barks "okkkk, i found the root cause". All this without respecting what has been done these last 10 years.

The reports i made were correct, but instead of having people noting down "ah another bug/problem spotted", they simply tried to go after me (easier).

One of them is the floppy drive write disable function override due to the DMA problem when you use the floppy drive in conjunction with an hard drive, when the CPU/DMA problem has not been fixed. What happened is that the original i was imaging with the pasti tool on my STE got trashed exactly the same as was my ultrastan content. The game 4 first tracks were simply destroyed in the process (hopefully, i had a second original with which i have restored my 1st original).

Much later, another guy on a french forum wrote that he met exactly the same behaviour with his STE. Ultrasatan content trashed and the floppy in the drive got its first tracks trashed.

The next one is about those on AF or facebook or any other forum stating that Atari Computers are made to last forever, with no need to maintain or service them, with capacitors that don't need to be replaced, which is a myth.

I'm not expert on the hardware field of the Atari ST (not at the chip level or even deeper), however, Exxos gave enough proofs after doing tests on enough Computers to be believed.

[exxos]

One of them is the floppy drive write disable function override due to the DMA problem when you use the floppy drive in conjunction with an hard drive, when the CPU/DMA problem has not been fixed. What happened is that the original i was imaging with the pasti tool on my STE got trashed exactly the same as was my ultrastan content. The game 4 first tracks were simply destroyed in the process (hopefully, i had a second original with which i have restored my 1st original).

I remember you mentioning that before. I've had floppies trashed while write protected. I think it's the same problem as when it triggers a write protected error when it's not. Think that was fixed with the DMA pull-ups.

It's a odd one in your case. I'd assume the floppy and harddrive select bit in, I presume ram (maybe glue?) Gets corrupted. That could be down to bus noise on the databus.

It would be a bizzare sequence of events to confuse the hard drive and floppy drive then also ignore the write protect bit. Though nothing much surprises me anymore.

If such a fault if replicatable reliabley with a normal floppy drive and ultra satan and all my current fixes are done as well, then I'd be interested in borrowing such a motherboard. Though can't promise anything.

I think there are just endless problems with these machines. It's why I gave up and did a new motherboard in the end as I had had enough trying to find solutions for all those problems.

. Exxos did lots for the Atari Community, and some people inside just choosed to dismiss him or spat at his face

Yes it generally seems the case and why I setup this forum as I had enough of it all. You either work with people or against them. Just a great shame some people's mentally just wants to destroy all the time.

I've had to deal with these types of people since I first came online in the 1990s. It's only got worse as the years go on. I have a very short fuse for such people these days. "I've heard it all before" and such people are just not worth my time.

.To make the answer short: what happened on AF just smelt flame war, and i perfectly understand what some were trying to do.

Always the same isn't it ? Some people don't care about the truth, just want anything to dismiss anything I say or do. They didn't even understand what was being said in that article. Just jump to conclusions as use it to argue against me.

It wasn't just myself who noticed this. Our founding members got tired of it all years ago as well. But anyway, I sometimes get emails from people who visit atari-forum and they can't believe how toxic it is. Anyway, all old news and all being said before many times.

I found many problems and many solutions. They helped a lot of people over the years. Not my problem if people think otherwise. At the end of the day, where's the harm to try the HC CPU when the "good DMA" is hard to find and expensive ?. People should be thanking me for finding such things, not making it their lifelong mission to disprove such findings by any nonsense they can think up. You would expect people would show a little respect as it was only myself who bothered to take the time into looking into various issues in over 30 years.