As it starts to happen after a bit of running, something could be heating up and going bad only when it's warm, of course determining what usually involves blowing hot air on components or using freezer spray to cool parts down, but the age of the components, it's not unheard of for a part to go down that wonky heat-death path... :\
I like this clue. Jan can try to add heatsinks one by one on chips an run tests in between. But it can be also issue with pcb itself. So next test can be to add small fan blowing air into different areas of the board, and testing, testing, testing.
I bet you did it already but I'd suggest inspecting the RAM chips, especially how they are seated. I once did a bit different - yet similar modification (different logic, the same board the same amount of extra RAM). And I had similar problems. And the diagnosis was... precision sockets. Yes. I know, they are considered to be better. But they not necessarily are. Since then I'm mounting only good standard sockets.
My understanding is that the turned pin socket is supposed to be the best option but that's for brand new chips, but for chips that has shorter legs (IE has been soldered and cut) the dual wipe sockets seems to often work better for a lot of people. Not very scientific but yeah, dual wipe seems to be safer even if they're technically more vulnerable than turned pin socket to various things like "chip creep" and a few other issues. This means you're more likely to need to push down the chip after X years with a dual wipe socket which I suspect is why the turned pin socket is required for Mil-Spec and most high-reliability applications - reusing chips with shorter legs just isn't something they consider at all.
@@Torbjorn.Lindgren i've heard previously soldered chips , even with ful length legs, can damage turned pin sockets, stretching the contacts too far open, unless you remove all traces of the solder from the pins...
Great video and no, you don't disappoint :-) its 'nice' and sometimes reassuring to see that guys like you can also have trouble fixing these machines. My advice? Get another machine and use this one for spare parts. Hope you have a nice Xmas :-)
I've done this mod a couple of times. I always lift PB5 and PB6 and solder the jump wires at the PIA leads bent up, and also adding some pullup resistors to this lines.
Great vid Jan, does seen heat related, maybe solder reflow ram sockets and surrounding logic. I assume the power supply voltage and current hold solid after a warm up.
I would add some 100nf ceramic bypass caps to the back of that PIA GND to VCC, TTL signals looked to have a bit of noise, unless that was just from the way you were probing the circuit. Bypass , bypass some more. Clean up the power rails. Cheap thing to try.
I wonder if there's a RAM test that prints out the stored and incorrectly retrieved values when a test fails (or if one could be written)? It seems like that might be useful in determining what exactly is happening when it fails.
Hey Jan, a really stupid left field thought - its not something to do with the switch you've added to switch between the ram mod and the standard config???? Anyway - its sounds like you've tried everything else, and while not likely..... ;) And I agree. These bloody 8bit Atari's are so temperamental. You only need to look at them and they start having a problem! Thanks for your awesome videos and merry Christmas!
Have you checked the switch you are using to switch between normal and expanded ram it might be good for switching powered on and off reliably but for carrying very little to no voltage it might be subpar.
I had a PK232 Z80 based modem which I upgraded with Z80 and Z85C36 CMOS version. I had to slow down some lines of the CMOS Z80 CPU, like the A14, A15, IORQ and the WE and OE lines on the SRAM adding 68pF capacitors.
with the Ram tester isnt there the stepping test or something that tests the older Ram better. I watch Adriran Black he just did a video on testing Amiga Ram he discovered the right way to test that Ram
One thing I would defo try is to swap the CPU, yours says Mexico and those are infamous. Had recent troubles with an SRAM upgrade myself and, at the end of the day, it was down to the CPU.
I suspect the additional multiplexed address line A8 on the ram chip. for normal 64kbit ramchips freddie is multiplexing address lines. the additinal 2 address bits we "steal" from the pia registers PB6 and PB7 may me not be multiplexed in sync with the other multiplexed data lines as the new A8 ram signal is generated on the gal. I would definitely scan with oscilloscope both A8 and A7 multiplexed ramchip pins at same time and compare whether the actual mutliplexing timing aligns. the other thing I would try is 3k pullup or pulldown resistor on the multiplexed ram A8 line. .... If I watched carefully the ram test error is always in the middle two line/banks when two PIA PORTB outputs (PB6 and PB7) are different and banks works when they are same - so if you multiplex between 0 and 0 its always 0 and timing does not matter - all works. simillar 1 and 1 always works. Maybe instead of using PIA pins 6 and 7 try putting switches on the input of gal to emulate all 4 PB6/PB7 combinations and see if the 64kb works, but everytime as different banks. Inserting one or more 74ls gate on the O2 signal input to the GAL or the A8 output on the GAL would put all ram multiplexed signals in sync. or use separate 74ls157 or 158 to multiplex PIA PB7 and PB6 for A8 ramchips as on some other upgrades (scott petersons mem upgrades) just to test what is wrong ... I appreciate Your attitude to isolate the problem instead of using PCB for spare parts. Sometimes it's easier to just do it on the other motherboard or put ultimate1mb inside, but that's like Agatha Christies book without last chapter!!! Don't give up.
I had a weird problem with an apple IIe with Ramworks III and the VGA adaptor that you plug in the side. it would not show color some times and ram errors but not in any ram test. turn out to be the a 74LS125 and GAL IC next to the AUX connector. the 74LS125 would only act up on the Ramworks giving ram errors only on some programs.. the GAL would mess with the color funny they where right next to each other. I don't have any logic test equipment just had to look at the schematic and guess where would waldo be
Maybe a bit of testing with a logic analyzer is the way forward with diagnosing it? Might give some more clues as checking only 1 pin at a time only gets you so far. Must admit i am not at all familiar with atari.
You need to write a RAM test that reports way more detail about the failure ... knowing how it fails can be insanely helpful to make hypothesis about what goes wrong.
I hate desoldering ICs. I'm always afraid of lifting a trace on the topside (under the chip). Then to add insult to injury having a new socket become the problem, ie not being tight enough on the IC legs, creating intermittent issues. The XE series specifically have really weak PCBs.
Here is a suggestion. Change the logic chips 74LS08 and 74LS138 with a pair of 74F08 and 74F128 for better timing. Jurgen told me this trick and it always seems to straight things up with misbehaving upgraded machines. If the above tip doesn't work I would suggest to remove everything and buy a compo solution (Cartridge) that can be used on all your Ataris. A. Acquire an AVGcart with an SIO and an ECI cable. It will enable you to run ALL known types of files(floppy images ,executable, cassette,cartridges) with amazing compatibility plus you will enjoy 512kb RAM expansion and a video/music player! B. or a SUBcart with an SIO, an ECI cable and a DAC audio cable. It will enable you the same file compatibility but with 1mb ram expansion and a Pokey stereo upgrade. If you buy the PBI cable you will be able to enjoy the same upgrades on your XL machines with any of those cartridges.
@@PG-gs5vb indeed it does, amazing piece of hardware ! All these extra features can really be appreciated by power users like you! Huge fan of your work sir!
Finding genuine fast "F" chips is a problem. With so many fake chips around many are likely to be remarked standard/LS devices (if they are real logic chips at all!)
Hello, it seems like a timing problem. I had a similar problem with a spectrum +3 and 128k games. First I thought it was a ram problem, and in the end the problem was one of the multiplexers that controlled the "lower ram", a crazy thing...and yes all test passed ok 😂
The thing is, the xe line has so much potential. Especially with all the mods out. The vbxe, dual pokey, u1mb. The various rom upgrade kits. You can really turn these into modern usable computers, let's not forget fujinet. Now if we can get more home brews out that take advantage of all these cool new mods.
As it starts to happen after a bit of running, something could be heating up and going bad only when it's warm, of course determining what usually involves blowing hot air on components or using freezer spray to cool parts down, but the age of the components, it's not unheard of for a part to go down that wonky heat-death path... :\
I like this clue. Jan can try to add heatsinks one by one on chips an run tests in between. But it can be also issue with pcb itself. So next test can be to add small fan blowing air into different areas of the board, and testing, testing, testing.
I bet you did it already but I'd suggest inspecting the RAM chips, especially how they are seated. I once did a bit different - yet similar modification (different logic, the same board the same amount of extra RAM). And I had similar problems. And the diagnosis was... precision sockets. Yes. I know, they are considered to be better. But they not necessarily are. Since then I'm mounting only good standard sockets.
Might definitely be the issue
My understanding is that the turned pin socket is supposed to be the best option but that's for brand new chips, but for chips that has shorter legs (IE has been soldered and cut) the dual wipe sockets seems to often work better for a lot of people. Not very scientific but yeah, dual wipe seems to be safer even if they're technically more vulnerable than turned pin socket to various things like "chip creep" and a few other issues. This means you're more likely to need to push down the chip after X years with a dual wipe socket which I suspect is why the turned pin socket is required for Mil-Spec and most high-reliability applications - reusing chips with shorter legs just isn't something they consider at all.
It seems a little backwards to put a square peg in a round hole. Call me old fashioned.
@@Torbjorn.Lindgren i've heard previously soldered chips , even with ful length legs, can damage turned pin sockets, stretching the contacts too far open, unless you remove all traces of the solder from the pins...
Great video and no, you don't disappoint :-) its 'nice' and sometimes reassuring to see that guys like you can also have trouble fixing these machines. My advice? Get another machine and use this one for spare parts. Hope you have a nice Xmas :-)
I've done this mod a couple of times. I always lift PB5 and PB6 and solder the jump wires at the PIA leads bent up, and also adding some pullup resistors to this lines.
Are you sure that the ram checker software isn't dodgy?
I think you said it. Some kind of timing issue. Looks like a bugger to try and figure out though. Good video Jan. Have a great Xmas.
I have had same problems, and it came out to be to much resistans on the ground - part of the main board.
Timing issue? Do you have the correct speed versions of all the 74 series chips?
Great vid Jan, does seen heat related, maybe solder reflow ram sockets and surrounding logic. I assume the power supply voltage and current hold solid after a warm up.
The ultimate 1mb mod is really the way to go with the Atari 8-bit line.
Is the switch that switches the expansion floating? Does its pull up/pull down have a bad reference to 5+v or ground?
Merry Christmas!
Looking forward to that mystery add-on you mentioned :D
It might look slightly familiar to you. :D
Given you said the timing looked wrong on that one sample, maybe your crystal is wonky when it gets warm? You could try replacing that too.
I would add some 100nf ceramic bypass caps to the back of that PIA GND to VCC, TTL signals looked to have a bit of noise, unless that was just from the way you were probing the circuit. Bypass , bypass some more. Clean up the power rails. Cheap thing to try.
I wonder if there's a RAM test that prints out the stored and incorrectly retrieved values when a test fails (or if one could be written)? It seems like that might be useful in determining what exactly is happening when it fails.
Why not leave the additional base ram in there so its a 130xe instead of a 65xe?
Hey Jan, a really stupid left field thought - its not something to do with the switch you've added to switch between the ram mod and the standard config???? Anyway - its sounds like you've tried everything else, and while not likely..... ;) And I agree. These bloody 8bit Atari's are so temperamental. You only need to look at them and they start having a problem!
Thanks for your awesome videos and merry Christmas!
Have you checked the switch you are using to switch between normal and expanded ram it might be good for switching powered on and off reliably but for carrying very little to no voltage it might be subpar.
something with the memory expansion are there different kinds of memory expansion boards that might work better?
I had a PK232 Z80 based modem which I upgraded with Z80 and Z85C36 CMOS version. I had to slow down some lines of the CMOS Z80 CPU, like the A14, A15, IORQ and the WE and OE lines on the SRAM adding 68pF capacitors.
A possible issue could be that there is a floating pin somewhere, especially look at the ram chips with more ram/address lines.
with the Ram tester isnt there the stepping test or something that tests the older Ram better. I watch Adriran Black he just did a video on testing Amiga Ram he discovered the right way to test that Ram
the chips on the 2nd ram bank don't look seated properly.........
One thing I would defo try is to swap the CPU, yours says Mexico and those are infamous. Had recent troubles with an SRAM upgrade myself and, at the end of the day, it was down to the CPU.
I suspect the additional multiplexed address line A8 on the ram chip. for normal 64kbit ramchips freddie is multiplexing address lines. the additinal 2 address bits we "steal" from the pia registers PB6 and PB7 may me not be multiplexed in sync with the other multiplexed data lines as the new A8 ram signal is generated on the gal. I would definitely scan with oscilloscope both A8 and A7 multiplexed ramchip pins at same time and compare whether the actual mutliplexing timing aligns. the other thing I would try is 3k pullup or pulldown resistor on the multiplexed ram A8 line. .... If I watched carefully the ram test error is always in the middle two line/banks when two PIA PORTB outputs (PB6 and PB7) are different and banks works when they are same - so if you multiplex between 0 and 0 its always 0 and timing does not matter - all works. simillar 1 and 1 always works. Maybe instead of using PIA pins 6 and 7 try putting switches on the input of gal to emulate all 4 PB6/PB7 combinations and see if the 64kb works, but everytime as different banks. Inserting one or more 74ls gate on the O2 signal input to the GAL or the A8 output on the GAL would put all ram multiplexed signals in sync. or use separate 74ls157 or 158 to multiplex PIA PB7 and PB6 for A8 ramchips as on some other upgrades (scott petersons mem upgrades) just to test what is wrong ... I appreciate Your attitude to isolate the problem instead of using PCB for spare parts. Sometimes it's easier to just do it on the other motherboard or put ultimate1mb inside, but that's like Agatha Christies book without last chapter!!! Don't give up.
I had a weird problem with an apple IIe with Ramworks III and the VGA adaptor that you plug in the side.
it would not show color some times and ram errors but not in any ram test. turn out to be the a 74LS125 and GAL IC next to the AUX connector.
the 74LS125 would only act up on the Ramworks giving ram errors only on some programs..
the GAL would mess with the color funny they where right next to each other.
I don't have any logic test equipment just had to look at the schematic and guess where would waldo be
Could voltage be fluctuating?
Maybe a bit of testing with a logic analyzer is the way forward with diagnosing it?
Might give some more clues as checking only 1 pin at a time only gets you so far.
Must admit i am not at all familiar with atari.
nice to see you again
those transitions on youre scope are those square waves I assume
You need to write a RAM test that reports way more detail about the failure ... knowing how it fails can be insanely helpful to make hypothesis about what goes wrong.
I hate desoldering ICs. I'm always afraid of lifting a trace on the topside (under the chip). Then to add insult to injury having a new socket become the problem, ie not being tight enough on the IC legs, creating intermittent issues. The XE series specifically have really weak PCBs.
Voltages?
Dropping as it heats up?
/guess.
Timing issue?
Look over schematics of 130XE mother board to see if any factory mods are made of at least a working 130CD mother board for comparison.
Sorry for typo. 130XE.
You might get it under a microscope and do a close-order inspection for traces that are scratched or cracked?
I agree with the others. It has to be a socket at this point, or a bad trace.
Here is a suggestion.
Change the logic chips 74LS08 and 74LS138 with a pair of 74F08 and 74F128 for better timing. Jurgen told me this trick and it always seems to straight things up with misbehaving upgraded machines.
If the above tip doesn't work I would suggest to remove everything and buy a compo solution (Cartridge) that can be used on all your Ataris.
A. Acquire an AVGcart with an SIO and an ECI cable. It will enable you to run ALL known types of files(floppy images ,executable, cassette,cartridges) with amazing compatibility plus you will enjoy 512kb RAM expansion and a video/music player!
B. or a SUBcart with an SIO, an ECI cable and a DAC audio cable. It will enable you the same file compatibility but with 1mb ram expansion and a Pokey stereo upgrade. If you buy the PBI cable you will be able to enjoy the same upgrades on your XL machines with any of those cartridges.
SUBAVG also adds COVOX emulation, realtime clock, OS switching and 4MB AXLON expansion.
@@PG-gs5vb indeed it does, amazing piece of hardware ! All these extra features can really be appreciated by power users like you! Huge fan of your work sir!
Finding genuine fast "F" chips is a problem. With so many fake chips around many are likely to be remarked standard/LS devices (if they are real logic chips at all!)
@@spacedock873 I never had any issues with my ebay sellers (Europe) tbh.
@@spacedock873 what about 'HCT' types?
Instead of modifying the socket, use the trackside of the board, it would be much neater.
Maybe firmware related?
You might contact Adrian Black in the USA he repairs older computers and I think he has worked with Atari's Im not sure about memory expansion.
Maybe a wonky PSU?
Haven't seen the wrongest possible solution here - Deglitching caps! Bwahahahaaa!
😁👍
Hello, it seems like a timing problem. I had a similar problem with a spectrum +3 and 128k games. First I thought it was a ram problem, and in the end the problem was one of the multiplexers that controlled the "lower ram", a crazy thing...and yes all test passed ok 😂
PCBWAY is the way... to get it done
Talk to flashjazzcat he is a very smart dude
The thing is, the xe line has so much potential. Especially with all the mods out. The vbxe, dual pokey, u1mb. The various rom upgrade kits. You can really turn these into modern usable computers, let's not forget fujinet. Now if we can get more home brews out that take advantage of all these cool new mods.
FIND A REAL JOB BRO 😊
If it is intermittent, it could easily be a single trace or component contact somewhere...
Yes, that might be the case. I gave up on it in the meantime. Spent many hours both on and off camera trying to figure it out to no avail.