Another (new) 130XE boots to red screen

[Beeblebrox]

@woj

 

I've done the simple XE vid mod with my XEs.  Good improvement everytime. 

 

Check it out:

 

https://forums.atariage.com/topic/155245-vertical-lines-on-many-xe-computers/?do=findComment&comment=4907992

 

Edited September 6, 2022 by Beeblebrox

[woj]

49 minutes ago, woj said:

the capacitors are 22pF (if understand the markings correctly) instead of 10pF and 18pF

Decoding the color stripes better says that would be 2.2uF in fact, ?

[woj]

Nevermind the last thing about the capacitor, from the description of the new 65XE board it is clear that these capacitors for the oscillator need to be 22pF. Also the resistors are 1M and 240 there, so I guess this is how it's supposed to be.

 

Today I cleaned up everything around Freddie, inspected visually and checked the continuity of all Freddie and RAM pins (I now think I might have forgotten one pin, will recheck tomorrow). I also properly desoldered and resoldered these two capacitors and resistors. Especially the capacitors are connected to a huge ground plate, so my previous reflowing was not necessarily the best, now its done properly. 

 

The first boot went fine, but then it crashed once half way through the memory test. Then rebooted fine again. I attribute this to the board being still warmer than usual from soldering and cooling down (if so, this is not good of course, but I am lost in how to find out what it is). After this I also decided I will screw the board to the case with the bottom shield on and fastened. Since it seemed to be very sensitive to ground mass changes, why not try. Now running the demo loop of PoP, so far so good (30 minutes). I will have to chance to stress test it tomorrow again, we will see. My next decision is that if I experience a single crash again, I solder the 220nF bypass capacitors for the memory back in, as with those oscillator components, there had to be some reason for those. 

 

And, it may as well still be RAM for all I know, SysCheck and new chips will tell the truth, I hope.

 

EDIT: PoP seems to nicely stress all of CPU, Antic, and GTIA, all kind of really warm

 

EDIT2: And this is how Albert looks like, I will defo need to try that Svideo mod...

Edited September 6, 2022 by woj

[woj]

So shield does not seem to have anything to do with it, but keeping it on is a good idea as I read in some other topics, especially with U1MB.

 

There are new developments, I now have considerable evidence this is RAM. This morning booted cold, survived the RAM test until 16K or so, crashed. Turned it off, waited for ca. 15 minutes. Similar thing. Did the same yet again (wait to cool, turn on), crash again. Then had to take an hour break due to logistics, when came back, I first heated up the RAM chips a bit with the heat gun (at the lowest temp, which for mine is 100C), for no more than a minute. Gave it some short time for the heat to sink in, booted up, it worked fine, memory test ran in cycles for 30 minutes or so. 

 

Then, just to make sure and I was supposed to do it yesterday anyhow, at this point I took the RAMs out, gave them one more clean (there were still some traces of solder here and there), cleaned the sockets and the pins with contact spray (which I realized I have only yesterday evening), made sure they sit the best they can. All this procedure took close to an hour. 

 

Booting after this - the memory test lasted until the last block of the extended memory (this ST goes all the way to 128K, I know some don't...), it did not crash, but the block was red. Took the opportunity, switched off, sprayed the memory with air duster, let the computer cool down for 10-15 minutes again, turned it on, crashed within the first row of the RAM test. Turned off quickly again, cooled with air duster, let it sit off for 10 minutes. Then warmed the RAM with the heat gun as before, turned it on, and it works without a glitch since then (now ca. 3 hours), running different things in both 64K and 128K mode.

 

Apart from taking the RAMs out for cleaning, I tried to disturb the computer as little as I could, not even sneezing in front of it. I should have a chance to repeat this test tomorrow too, but to me it seems memory now is the prime suspect. Yet, I still have no equipment or parts to verify this 😕

[Mrshoujo]

If the RAM is continually causing problems, try replacing it. What's the slowest speed RAM a 130XE needs to work reliably? If the RAM is too fast, that would cause problems. If you haven't already, check all circuit lines involving the RAM and addressing. Something's still really flaky if it runs one moment but could crash at any nanosecond.

[woj]

Once they arrive I will do that, two different brands are on their way to me. I have seen on pictures  120ns and 100ns (this one dominantly) mounted. 

Edited September 7, 2022 by woj

[Beeblebrox]

27 minutes ago, woj said:

Once they arrive I will do that, two different brands are on their way to me. I have seen on pictures  120ns and 100ns (this one dominantly) mounted. 

AFAIK stick to one speed of RAM chip and don't mix em up. I tend to go with 12n or 15n.

[woj]

Sure I won't mix them up. 

 

I think I found a benchmark program for diagnosing this, Last Squadron does not load, I mean it loads, but goes to black screen after loading. I have read around here about folks having problems with this one on unstable machines.

[Beeblebrox]

I can't recall but the free version of The Last Squadron - does it check and adjust for PAL and NTSC or just run on PAL?  Just working out if the crash you have is down to this. 

Edited September 7, 2022 by Beeblebrox

[woj]

The loading screen says Antic: PAL, GTIA: PAL, so I guess that's the one that adjusts...

 

EDIT: I opened another can of worms, luckily not in the 130XE. LS does not start when loaded with RespeQT, but it does when loaded with AspeQT. Both running on my Ubuntu (spare hints to try Windows, won't happen). It also seemed to struggle to load with AspeQT, took very long, but the game started. I have a suspicion, luckily  easy to check one. Cliff hanger

Edited September 7, 2022 by woj

[woj]

OK, so I thought this was a USB driver problem on Ubuntu, which in another case of a USB device playing up was solved by connecting it through an external hub.

 

I declare RespeQT unusuable for me, LS is one thing, the other thing it cannot load is Blinky's Scary School with the High-Team crack / demo in front. AspeQT does both. Perhaps one day i can be inclined enough to check the sources of both to see what can possibly be going on...

[xrbrevin]

going back to the oscillator circuit, is it worth checking the crystal?

[woj]

7 minutes ago, xrbrevin said:

going back to the oscillator circuit, is it worth checking the crystal?

Perhaps, just that now it works OK most of the time (in fact, I was not able to crash it since the last time I described above, even after an hour of rest), how ever am I going to catch that it is failing apart from replacing it with another one, which I do not have...

Edited September 7, 2022 by woj

[woj]

So it is good and bad still, the computer did not crash cold this morning, ran totally fine. That part is good, the bad thing is that I might not ever be able to have it in a condition to nail the problem down (if it is still there, perhaps reseating and cleaning up RAM helped).

 

But, the SIO2PC AspeQt/RespeQt saga, I played a bit more with them yesterday, and even though I said it's not related to the machine itself, it actually might be, the first crash ever I had was during SIO2PC loading... This is not to say that there is a problem in the SIO circuitry, but that it perhaps gets affected by a problem elsewhere.

 

I have three different versions: older AspecQt manually compiled, RespecQt manually compiled, and AspeQt 2K20 from an installation package. Even though now I set the speed at 1x on all of them, let's only take LS (but I have seen different behaviours with PoP and other things too):

 

(a) loads LS successfully, in the sense that the game starts, the loading, however, stops several times in the process and behaves like there are transmission errors

(b) and (c) perform a flawlessly sounding data transmission when loading LS, no stops, but the game does not start.

 

Surely I will try disconnecting the caps on the SIO for data in/out and handshake to see what it does, but for 1x speeds that should not matter, hence I wonder if something to do with timing / frequencies / power noise in the computer somewhere else has anything to do with it... I do have oscilloscope, but I have little clue how to use it, or what to look for...

 

EDIT: Looking at the schematics and pictures, my computer has both the capacitors on the in/out lines (C77/C88) that I see mentioned everywhere, and on the SIO connector as such (C303/C304) that I have not yet seen in any of posts...

 

Edited September 8, 2022 by woj

[xrbrevin]

do you get failures with other SIO devices?

the POKEY controls SIO data and clocks - do you have a spare for testing?

[woj]

I don't have other SIO devices or spare Pokey unfortunately.

[Beeblebrox]

As mentioned the other day I had a partially failing Pokey chip causing SIO/Polling issues on an XE. Audio/music side of the chip had no issues. Pokey replaced and all good in that respect.

[woj]

11 minutes ago, Beeblebrox said:

As mentioned the other day I had a partially failing Pokey chip causing SIO/Polling issues on an XE. Audio/music side of the chip had no issues. Pokey replaced and all good in that respect.

Right, so I guess my next bet / plan is to remove the small caps next to the SIO port (I have seen them not done by factory, even on this specific kind of board), see how my SIO2PC behaves, and if I still get issues at speed 1x, my next suspect is POKEY. 

Edited September 8, 2022 by woj

[woj]

It was a good day for my Atari today, so I decided not to take the soldering iron out to do the s-video mod, even though the parts arrived. 

 

Caution, persistence, and patience are what is required for these little things. It hasn't crashed on me since yesterday, warm or cold, statistically speaking the problem is fixing itself as we go .

 

That gave me the time to investigate the SIO(2PC) issues. Instead of going in it with the iron and undoing the SIO port capacitors, I thought I would first check a Windows computer after all, just to rule out some unlikely weird behavior of my Linux. And what do you know, it all worked just fine, LS loaded quick and started, Blinky loaded nicely with the intermediate demo and the loading scroll (which I forgot should be there, saw it now for the first time). So not only it was all working, it never ever worked 100% before on my Linux, even when it technically did.

 

Switched back to my Linux computer, experimented a bit with settings. It turned out that it is not working reliably on with the DTR handshake. The one that works is NONE! Now same as on Windows, LS loads and works fine, Blinky with all the bells and whistles, and also (check my first post here) Lasermania xex now loaded with the black background! There is a down side to this - AspeQT uses 100% of the CPU for this, this is way too much and it gets warm.

 

It probably has something to do with some Linux kernel module incorrect default setting somewhere that the software takes for granted rather than checking / setting it. I will obviously have a look into this, however, all my previous adventures with serial comms on Linux only gave be gray hair. 

[woj]

It was not so good day, with some good stuff, and some at least conclusive I hope.

 

1. For SIO2PC on Linux, the best thing is the AtariSIO cl / ncurses stuff. Works with the DTR, no heavy CPU load, just brilliant.

2. Did the s-video mod. At first I got black and white picture, twisted the 500K pot back and forth, nothing, after a quick peek at the schematics I concluded the Q2 resistor probably went bad in the process (soldering nearby), because composite was fine (had color). That did not help. It turned out that I cannot use 820ohm for R205 for whatever reason (my TV?). Going lower to 300ohm gave better color, still very gray. Went back to 150ohm and that solved the problem. In fact, the juiciest color was when I was bridging the R205 connection, but I did not dare to put it together like this.

3. Generally it got better in terms of color, otherwise nothing much changed. In fact, the black and white picture with the 820ohm was really good and crisp, but, well, it was black and white.

4. Because I was playing with the pot in point 2. I played quite a bit with it to get back the ideal color. On one approach, when I touched the pot with the screwdriver the computer crashed (screen went blank green). 

5. Well, I thought I have hit the nerve. Next boot went fine, but the picture was a bit "nervous" (like noise from touching the cable pins or similar). The pot seemed sensitive (in terms of picture noise, color distortion, and loosing picture all together) to what I was doing with it, so I had the idea that I now regret, to re-flow its solder points.

6. That actually made things worse, after resting a lot the computer boots fine with good picture, but only for few seconds, then colors start to change, there is also audio noise. Letting it stay on makes things progressively worse. 

7. I know one should not measure a pot that is connected, nevertheless, when the board is "rested", I get sensible readings on the pot, when I measure it right away after turning off the computer, I get all kinds of weird things. Also voltage readings when it is on on the different pins of the pot are not what I would expect.

 

In any case, the pot definitely needs replacing, and of course I don't have one. I am really hoping this is it though. 

 

[woj]

Yesterday I pre-adjusted the pot to the position I knew it should work and let everything rest for the night. This morning it booted (almost) fine, the color was good and did not slide away. But, the "almost" part is that the there is still slight audio noise, and now when I hold a key down in Basic, they keyboard auto repeat and according t-t-t-t-t click sound causes a corresponding noise on the screen. I actually had this before (I mean last week or so) once, but it went away. I even suspected the cable, but re-checking it did not reveal anything suspicious.

 

Now, for the 500K pot, I read somewhere it should be 0.5W. But there is absolutely no way to get this in my local shops, and I also noticed that for the 65XE remake board a much smaller (150mW) part is listed. Will I be able to go with 0.25W? 

[woj]

The fact that I see 9V across the outer legs of the pot can't be right, can it?

[woj]

The edit time out is defo too short... Looking at the circuit, I guess this 9V is actually right, that's what all these diodes and capacitors nearby are for I guess. 

 

But, then, if the pot leaks that voltage to ground somehow, then all kinds of weird things can be happening...

[woj]

I managed to get my hands on a big panel type 470k pot, soldered it in for test. My original pot is defo flaky, after the test it's now back in and clearly having contact issues, while the panel one did not. It got a proper clean / bath in the contact fluid, perhaps it will get better (I experienced before the fluid needs its time to do its magic). New parts in the mail, but they are only 0.25W (but looking identical to what I have).

 

But, the pot was not the core of the problems with the video / audio noise and clear interference from the keyboard click sound. It was actually the s-video mod. I put back the original R204 and R116 (R205 was already "stock" after yesterday when the modded one gave me B&W picture) and the symptoms are fully gone. Or I cured something totally different by soldering on the board again...

 

If it is the mod then it's totally beyond me why a mod that so many of you managed to do successfully with great results and that was properly developed by trial and error by someone that knows this stuff does not work on my machine 🤕 The only possible answer I have is that this newer 4xRAM board has also some parts of the video circuitry different (though the components that were obvious for me to check turned out to be the same as in older XEs...)

 

The saga continues...

[woj]

Took a little step back and looked into schematics of what could have possibly gone wrong with the s-video mod and the problem I got of video / audio interference / noise. As far as I can see it had no right to cause the symptoms it did. What I did notice is that the 500K pot is right next to the audio circuitry on the board, so the audio noise and the corresponding video interference could have been easily caused by soldering nearby (perhaps poor flux cleaning...?). Since it's good now until the new pot arrives, I let it rest...

 

I also looked at some pictures I took before and after the s-video mod, there really wasn't any difference to speak of to be honest. So I decided to see what else I can push out of my TV (I still think its upscaler simply sucks). Turns out I can get the best color results when using the composite output instead of s-video and playing with the sliders on the TV. So the computer mixes up chroma and luma much better than my TV. The only way to kill the picture noise that the built in upscaler actually amplifies a lot when using composite was to turn down the sharpness slides all the way down. The picture is a bit blurry, but this blurriness also hides the noise artifacts. In any case, I finally got really nice color out of it, Albert screenshot attached as an example.

 

My target mod is to actually have Sophia to replace all this mess, just need to wait some really long time before it is available again I guess... So I will probably just give this particular thing a rest