I admire your patience. If i was doing this, i already would have given up. At the end of the video you looked so defeated and fed up. I hope a replacement antec/graphics chip fixes it. Hope they're not too expensive.
If you haven't already know, I need to say that -XI-, the author of the XRAM test you were using, has recently passed away. A good test for RAM expansion seems to be a game titled LAURA, and the famous NUMEN demo. The most RAM related chips are FREDDIE, MMU, and EMMU (your GAL), and also SALLY, and ANTIC. These "precision" type IC sockets (especially DIP40 ones), brought me a lot of troubles during my build of Atari ITX.
Yeah, guys who work on arcade machines tend to **HATE** machined pin sockets. They apparently cause issues a lot. Talk to someone who works on arcade machines, and they'll tell you that high quality double wipe sockets are the most reliable.
@@Calphool222 I wonder if it's more of an issue with pins getting knocked out from jostling. Kids/young adults are not above hitting or shoving an arcade unit.
I have never heard anyone having issues with precision sockets. On the contrary, every time a board of mine had an issue, switching to precision sockets did the trick.
@@nickolasgaspar9660 Go read the Arcade Museum forums. There's a debate that breaks out there from time to time. The *majority* don't like machined pin sockets, though some prefer them. There are various reasons given.
I'm not actually sure why I like watching these videos. When it's me methodically troubleshooting mysterious and inexplicable faults I mostly just find it stressful but watching you do it is fun somehow. Hardware is also just more interesting to watch than software I guess.
I feel the same. But I guess I'll just do my own frustatring stuff in small doses and in the meantime enjoy Jan, Adrian and the other electronics necromancers and have a good time. If all fails I can still donate the wreck i gave up on to them. :P
I think we all have been in his position trying to make a retro machine work (I am in such a position for a moth with an xe board as we speak). The additional stress for him is that he HAS to produce content, while I can call it a day when ever I feel like it.
With all the tests now, and the fact that standard usage (w/o memory expansion) works fine let me point out the PIA(6520). The drives the memory expansio and bank select. Portbits PB2, PB3, PB6 and PB7 selects one of the 16 banks with 16K each. PB4 low = $4000-$7FFF is mirrored with extra bank, PB4 high = standard RAM active in the area. PB5 is used for dedicated ANTIC access, PB5 low = ANTIC "sees" extended memory in selected bank, PB5 high = standard RAM. It´s possible the set different (PB4 & PB5), to let use the ANTIC the extended memory and the CPU runs the code in standard memory. This trick is invented with the 130XE and was used by a very few demos which pre-calculate the screens and let them run from extended memory. Use the scope and run XRAM to check the signals an PB2...PB7 (Pin 12...17 of PIA). Because the test is quite slow, there must be a stable and sometimes a transistion visible - PB2 more often, because it´s the "LSB" of the 4 bit bank number. When there are flaky signals visible... the PIA is the culprit.
@3:36 The "N" doesn't mean anything important at all. It's part of the order code for the chip, indicating the plastic DIP package. Generally, the presence or absence of a letter like the "B" on the other hand can point to chips with slightly different behaviour, like an extra buffer stage, schmitt-trigger inputs or a different speed grade. The meaning of the letter suffixes are not necessarily standardized and might vary between different manufacturers. Still, what Jürgen most likely wanted to point out is not "you need a chip that has an 'N'", but "you need a chip that *doesn't* have a 'B'". In case of Texas Instruments, it seems the A or B suffix on the 74xx95 chips indicate power consumption, with "B" being lower power than "A". They only offer the the 7495A (standard TTL) and 74LS95B (low-power Schottky) which both have approcimately the same speed, but the LS chip has lower power consumption due to the more efficient implementation using Schottky logic. There seems to be no TI SN74LS95N.
The thing about the Atari 8 bit computers, is the clock design. It literally is daisy chained through the main LSI chips, and if one of them damages or deteriorates that clock for the next chip, issues like this can result. Use a scope and look at the clock going in the chip, versus the clock coming out of it. The CPU is at the end of this chain and can sometimes get a signal that is barely 5% duty cycle. Usually increasing the speed type of the logic chips also in the clock path can help this issue. A 74LS to a 74F for example. Also XEs with a quartz crystal instead of a clock generator "can" are less stable than the crystal oscillator designs. Most XEs have solder points for both. A quartz crystal with a couple transistors generates a sine wave clock. The dedicated clock generator cans generate a square wave. The square wave has less chance of deteriorating than the cheap crystal clock circuit.
Hi Jan, maybe a silly thing to mention but have you checked the capacitors (value) for the new ram? Also have you checked the voltages on the RAM’s while a demo is running? Keep up the good work, love these videos!
I got a Atari 65xe, had it for a couple of years but not used it yet! I'm 99% sure it works and the board is in lovely condition. Just need a quality SCART lead RGB/AV. I got the computer for free.
Hi Jan, it is great to see that you have saved Lenovo notebook for retro uses (I’m quite sure you can use it for more stuff as well). Great job as always 👍✌️
Here is an idea. Jurgen has a really great 512k upgrade! Restore the 65xe to its initial 64k memory and install the 512k board. (I know...not quite the idea you expected....lol)
Unsure but would certainly put your wire runs on the oscilloscope, then check voltages, try replacing PIA as Jurgen suggested in these comments. Then try a replacement ANTIC as merely a shot in the dark. Other than that maybe it's a failing passive component issue? Sorry to see you having difficulties with it. Also you might consider getting an XL/XE SysCheck 2.2 from Jurgen since it offers 512K.
It's not that 74 series being fake is it - ie. a HC re-labelled as an LS? I think the N just means its plastic btw! I also wonder if the wiring (proximity of 2 cables next to each other - or length of them might be a factor?
@@GadgetUK164 I've not seen that one.. seems bizarre though, not something i've ever come across but then I dont buy components from eBay or Aliexpress any more. They could have at least relabelled HCT as LS, which would largely pass
I would re-route the wires, I see a few wires going under the resistors perhaps it picks up some disturbance and also the switch I would try shorter wires ...
This is also what I thought. I had a lot of annoying problems with random wires like this. But I guess the GTIA was at least partially to blame for the issues, so maybe the Atari is slow enough that long wires don't matter too much
You look like you have another signal overlaid on your square waves around the 17 minute mark. Its amplitude in your samples mostly isn't big enough to trip a false hi or lo when it's overlaid, but I bet sometimes it is. You either have something actually crossed or it's just so physically close that you're getting some kind of RF bleed or something. All I know is that's not just the usual ringing you expect on a square wave. Maybe try taking the pin you were testing where you kept seeing a sudden fluctuation in the middle of the top plateau of the square wave (edit: 17:12), and see if you can tie its source hi or lo. That'll basically let you isolate the overlaid signal, if there is one. Then you might recognize the pattern in the overlaid signal and then you can look at whatever generates _that_ signal and how it might be bleeding over into the memory bus. Edit: Sorry if I'm not describing the issue well, I come from the software side of retro computing and my knowledge of hardware mostly comes from channels like yours and some mentoring I once got from a guy who worked in aerospace. It's not my field, but I get the basics and I've seen enough oscilloscope diagnostics to know that your sample looked wrong.
I would write my own assembler program to do my own ram test. We need to get to the point where we have *precision* on how *exactly* it is failing. Are *reads* occasionally failing, or are *writes* occasionally failing. Does it happen only when specific address bits are high or low? Does it fail when only specific data bits are high or low? We don't have enough info right now, and we're shotgunning chips into it with our fingers crossed. That's not going to work when we get to this level of troubleshooting.
You might want to check out the Retro Chip Tester Pro. It's an Arduino board that can test many old 74xx devices but also DRAM chips, dump ROMs and, with adapters, check SIMM modules and other things. It doesn't reliably detect marginal chips (it's not fast enough to check actual chip timings) but it should be able to tell you if a chip is completely dead
Were you testing the bare board just sitting on your table? Make sure you clean it of solder dust, etc. Could have been an intermittent short from tiny solder crud on the table. Or maybe place the board on an insulator and not the bare table that you used for soldering...
It's very curious to me that the GTIA chip works perfectly fine with the RAM expansion disabled, but breaks when enabled. Is it possible there's nothing really broken with the original GTIA chip, but it just works differently, and is incompatible with the mod? It might be curious to send your original GTIA chip to the guy who designed this mod and see if he has the same problem. Perhaps he can provide assistant,since he's the one most familiar with the mod
I dont know if you checked the power supply before all this, but if it was me, I would be checking it. Under full load for several hours. I would also make sure power getting to all the chips is at spec. Ur using different chips now so power demand could be less on the power supply as well, alleviating some of the load and symptoms. But, either way, you have an issue that is system wide that seems to be power/bus issues... did you check any bus transceivers?
Jan, would you like to try my CO25953 EMMU replacement? It'll get you only 128KB RAM but might be useful to check if you see similar behaviour as with the 320KB RAM mod..
Jan... I love your videos... :) but... in this case I think the problem is with interference between elements and that "spider mounted" cables over motherboard. I had a Atari XE motherboard that had custom 1MB RAM upgrade inserted on 3 sockets in 4-bit RAM motherboard version. So... Seller climed that is ok, but when you checked it with Jurgen's SysCheck it was possible to see some kind of glitches / artifacts between couting digits during memory tests. I removed this upgrade and now it is ok. If you want to have for sure that this Atari works, it is very easy, but you should go enter to "modern retro computring" world... So 1) remove all DRAM chips (on XE motherboard it is always problem - a) not working b) vertical bends on the screen in the background c) others) 2) put Lotharek's SRAM 64K V3 under 6502C. DONE. You will have stable 64KB machine. If you want to have a more RAM, use solid Ultimate 1MB upgrade, that provide RAM up to 1MB with different configurations, support for HSIO (so important when you have more RAM), support for SIDE3 and a lot of more.. if you don't want to have "modern" elektronics inside Atari, use Jurgen's SysCheck that provides HSIO + extra 512KB. That's all! Greetings from Warsaw and Polish Atari users! BTW. I invested in Keysight DSOX1204G 200Mhz and U1272A - great devices! Thanx for you videos!
I once had a logic board with 64h DRAM and Z80 CPU. It was a proprietary control board for some machine. However, in time it started developing random glitches. After few days, I soldered on the bottom of the board DIRECTLY ON THE RAM POWER PINS a 1uF capacitor, on each ram chip. It solved the problem. Maybe try it to see what happens. My guess was that the decoupling capacitors went bad so I tried his solution that worked. Edit: the glitches started to appear after the Z80 PIO and the Z80 CPU were replaced; they died from a power surge so, the new chips were less tolerant to the unclean signals on the bus. The added capacitors maybe helped to keep the power on the ram chips clean, thus making the signals better. While you have a oscilloscope, watch the power on the ram chips, see if there are something interesting when the ram test fails.
Jan....check all those wires for stray frequencies using your scope. You may be getting some spikes. Use some shielded wiring and check your power supply as well to make sure the extra chips are not dragging it down which could definitely cause an issue. If you have a bench supply with higher current capability, hook that up and try those tests and demos again.
Consider also Sys-Check 2.2, it's not just a diagnostic device, it provides 512K RAM expansion, simply plug it in the back. Then there's the Ultimate 1MB which is excellent too, but internal and a little more involved...
I've been following your channel for quite a while and I just got to this video. I usually don't write any comments on English-speaking channels, because I don't know English well enough to write a meaningful post myself, I'm from Slovakia. Now I'm making an exception and using Google Translate because I thought of one potential solution to the RAM expansion problem. I noticed that the plus (+5V) that you lead to the expansion switch is connected to the terminal of the resistor and that you said that you measured +5V there. It occurred to me that it might be a pull-up resistor and that it has a log. H most of the time, but there may be short pulses to log. L, which the multimeter does not pick up. It's just a shot into the unknown, but try to check with an ohmmeter whether the terminal of the resistor where you have the switch connected is really directly connected to the +5 V branch, or if it's not the other terminal. And while I'm writing a comment, I want to thank you for your videos.
Thanks for this! I tried several different 5V connections in the meantime and I am pretty sure the fault is somewhere else. Currently suspecting the PIA to be slightly wonky. I’m going to revisit the Atari soon and try to figure it out once more. :)
Have you ever considered getting a Retro Chip Tester Pro? It's a very powerful way to validate good/bad chips with a minimum of effort. I know Noel's Retro Lab and Adrian's Digital Basement both use them, and they have saved a bit of time compared to trying to swap out so many untested chips. I understand it's a rather involved project, but it seems to be worth while in the end.
@ 18:07, that falling edge is something that I would absolutely be worried about. And the rising edge that follows, I would be TERRIFIED about. And do the board over.
seems that the problem only occurs when the keyboard is laying on the mainboard. could you be getting some weird interferance from it? maybe put some plastic between the board and it? as a random guess
Can it be something like a bad/false contact pcb trace? Or a dryed out electrolytic not filtering out the power supply of the ram chips under certain loads?
logic analyzer mit Visualisierung des timings? ich hab den digiview dv3100. bedienung vielleicht etwas ungewöhnlich, aber trotzdem ein gutes teil. da könnte man das timing gerade mit unterschiedlichen anderen componenten gut vergleichen und auch mit dem cpu datenblatt.
I would try adding a small value resistor in series with the signal wires to reduce reflection. Say 22 ohms. Preferably at the driving end of the wire.
I've always wondered how much harder it would be to go the rest of the way - to 512K. You're already laying the address lines for the 256-bit chips over the 64-bit chips. Why not remove the other 64K and make the system bank in all 256K chips? Just curious - not at all my area of expertise - just asking.
It's a MOS 6502 variant. It can only ever see 64k. I'm not entirely sure how all of these extended memory ranges work on these computers, but I suspect they can't exceed a 64k base and everything else would be extended memory.
@@awilliams1701 I'm not quite sure "extended memory" would mean the same thing as it does in the PC universe. The norm with 8-bit hardware was bank switching, in which pages of 64k would be swapped in and out of the conventional memory space. The switching of memory pages should be handled by outboard logic, usually an MMU or similar type of chip.
@@awilliams1701 I didn't consider that since the processor requires the base 64K to be always present (for the stack and other zero page settings) that putting a swapping mod might create issues. Good thought!
@@horusfalcon I think that was the implication. An MMU with bank switching would be the only way an eight-bit chip could pull that off. Most chips use the base for stack/zero page operations in the base 64K. Even in the early days of the IBM PC where a 16-bit processor was used, the segments were still limited to 64K. Exceeding that in programming required setting registers in the processor to a stack segment (so the stack had 64K), a code segment, (where code in the segment had to be within 64K - but could jump to another segment), and a data segment 64K. However, 8-bit processors didn't have the built-in segment management registers so required an MMU.
Have you tried putting the original RAM chips back into their sockets? I noticed you still had the new ones in at the end, and while I believe you said they were tested and working, I'm wondering if they're faulty, or marginal at higher speeds? That might also account for the background color; I imagine the system at some point passes the color through it's RAM, and I also imagine the RAM it uses for that would be in the original bank, not the extended one, though maybe it does just go ROM -> CPU register -> graphics chip
If the machine works in 64K version and not with 256K expansion enabled, then there is a conflict which causes the machine to not work as expected. I would question the originator for more detailed info on how the mods actually work. I would suspect the 'N' chips are non-buffered outputs
Yeah this is especially true if the add-on chips have different timings, especially if those timings invoke special functions, as some newer RAM chips do. If you can't find chips that are made for the same timing, it may be necessary to design your own memory controller circuit to help the faster ones out.
I’m catching up on videos from months ago - did you ever solve this? GTIA *shouldn’t* have had much of anything specific to do with memory failures, except inasmuch as it’s part of the system clock circuit. The OSC inputs goes into GTIA (pins 16 and 28 on a PAL system), which generates the Fast Phi0 signal, which goes into ANTIC. ANTIC produced Phi0, which goes to SALLY, which then creates Phi1 and Phi2. It’s possible a tiny irregularity in the GTIA might affect a little of that timing but it’s surprising if that’s the only issue. It looks like your machine has a FREDDIE chip - I would definitely look into sourcing a replacement for that. XEs are much more common in Europe than the US and they shouldn’t be impossible to find, but might be getting pricey. I think Lotharek in Poland is either making or has announced a modern replacement for FREDDIE but I don’t know about availability in light of current worldwide chip shortages.
You said "Happy Ending" and I suddenly thought about cooling down the chips, not that they are overheated, but cooling down chips could worsen the beginning of malfunctioning. Maybe reduce the clock frequency, so that if it is a timing problem things might clear up.
Der Antic und der GTIA arbeiten in allen Ataris, in denen sie vorkommen, als Gespann. Der Antic entlastet dabei den Prozessor von Grafikberechnungen und liefert diese dann dem GTIA, der die weiteren Grafikberechnungen vornimmt. So, wie ich es gelesen habe, haben sich beide Chips über ihre Entwicklungsdauer nur wenig verändert, daher ist es unerheblich, aus welchem Baujahr sie stammen... Vor 1982 gab es den ctia,den Vorgänger des GTIA....
if theres an mmu, and its not addressing properly, that might be it, if the ram isnt banking correctly, i would suspect cpu/or something that is responsible for bank switching. if its not that, think about some of the glue logic again, perhaps its the way everything is tied together? can you loop pin 1 was it on the ram under side on both banks? i almost suggested a timer on the ram side from a faster 100ns to 150ns or 120ns or so, maybe a resistor pack? maybe a series? just taking a shot there. *smirk* what if its the joystick ports? its atari, expect strange happenings. graphics... its probably ram, switch out the ram again with the ram you had in the machine. are the grenades supposed to look black om the atari version of commando? i thought that was odd. check chip select? check clock lines? thats all i can think of. check all the glue i guess. good luck
Da der Tausch des GTIA ja soviel gebracht hat, könnte - soweit du sicher Verbindungsfehler ausschließen kannst - auch der ANTIC Chip eine Fehlerquelle sein. ANTIC und GTIA arbeiten immer als "Gespann" für die Grafikberechnungen / Kollisionsberechnungen und können dank DMA direkt auf das RAM zugreifen .Da "deine" Abstürze immer auftreten, wenn Grafikberechnungen / Darstellungen auftreten, könnte das noch eine Möglichkeit sein. Der ANTIC entlastet dabei die CPU von Grafikberechnungen recht ordentlich, ist also fast so etwas wie ein "Urahn" der heutigen GPUs. Gut, dass diese Chips vom Baujahr unabhängig zueinander kompatibel sind, solange du darauf achtest, dass du auch immer die PAL-Version einsetzt - es gibt auch GTIA für NTSC oder SECAM...Um Restzweifel am 74LS auszuschließen ,das "N" sagt nur was über das Package aus, das passt schon so (es gibt auch noch eine Variante mit "D", aber das ist dann ein SOP-Package, also quadratisch im Stile der Agnus im Amiga)...
The `BN` has nothing to do with the electronics - its a manufacturer specific code, for TI and Motorola (`SN` manufacturer) it indicates packaging... `N` and `BN` both indicate a DIP package and I'm not sure what the difference is
Which kind of power supply do you use? It certainly draws more power than without the RAM expansion. And is the timing of the original RAM ICs and the addon ICs the same? I guess they should all have the same timing
This is not "troubleshooting", you are just trying to replace parts at random. If the RAM test is failing, how exactly does it fail? What test exactly fails and how? Is the failure specific to particular chip(s), address lines, data lines? Does it affect just the extra RAM chips? If you're suspecting a timing problem, you should examine the signals on your scope, check them against the datasheet, look at timing diagrams in RAM chip datasheet, in processor datasheet, verify that they are compatible.
I admire your patience. If i was doing this, i already would have given up. At the end of the video you looked so defeated and fed up. I hope a replacement antec/graphics chip fixes it. Hope they're not too expensive.
If you haven't already know, I need to say that -XI-, the author of the XRAM test you were using, has recently passed away.
A good test for RAM expansion seems to be a game titled LAURA, and the famous NUMEN demo.
The most RAM related chips are FREDDIE, MMU, and EMMU (your GAL), and also SALLY, and ANTIC.
These "precision" type IC sockets (especially DIP40 ones), brought me a lot of troubles during my build of Atari ITX.
Yeah, guys who work on arcade machines tend to **HATE** machined pin sockets. They apparently cause issues a lot. Talk to someone who works on arcade machines, and they'll tell you that high quality double wipe sockets are the most reliable.
@@Calphool222 I wonder if it's more of an issue with pins getting knocked out from jostling. Kids/young adults are not above hitting or shoving an arcade unit.
I have never heard anyone having issues with precision sockets. On the contrary, every time a board of mine had an issue, switching to precision sockets did the trick.
@@nickolasgaspar9660
Go read the Arcade Museum forums. There's a debate that breaks out there from time to time. The *majority* don't like machined pin sockets, though some prefer them. There are various reasons given.
I'm not actually sure why I like watching these videos. When it's me methodically troubleshooting mysterious and inexplicable faults I mostly just find it stressful but watching you do it is fun somehow. Hardware is also just more interesting to watch than software I guess.
I feel the same. But I guess I'll just do my own frustatring stuff in small doses and in the meantime enjoy Jan, Adrian and the other electronics necromancers and have a good time. If all fails I can still donate the wreck i gave up on to them. :P
I think we all have been in his position trying to make a retro machine work (I am in such a position for a moth with an xe board as we speak).
The additional stress for him is that he HAS to produce content, while I can call it a day when ever I feel like it.
With all the tests now, and the fact that standard usage (w/o memory expansion) works fine let me point out the PIA(6520). The drives the memory expansio and bank select. Portbits PB2, PB3, PB6 and PB7 selects one of the 16 banks with 16K each. PB4 low = $4000-$7FFF is mirrored with extra bank, PB4 high = standard RAM active in the area. PB5 is used for dedicated ANTIC access, PB5 low = ANTIC "sees" extended memory in selected bank, PB5 high = standard RAM. It´s possible the set different (PB4 & PB5), to let use the ANTIC the extended memory and the CPU runs the code in standard memory. This trick is invented with the 130XE and was used by a very few demos which pre-calculate the screens and let them run from extended memory.
Use the scope and run XRAM to check the signals an PB2...PB7 (Pin 12...17 of PIA). Because the test is quite slow, there must be a stable and sometimes a transistion visible - PB2 more often, because it´s the "LSB" of the 4 bit bank number. When there are flaky signals visible... the PIA is the culprit.
Continua cosi Jan,nonostante i problemi
"Never give up, never surrender!"
By Grabthar's hammer!
@3:36 The "N" doesn't mean anything important at all. It's part of the order code for the chip, indicating the plastic DIP package. Generally, the presence or absence of a letter like the "B" on the other hand can point to chips with slightly different behaviour, like an extra buffer stage, schmitt-trigger inputs or a different speed grade. The meaning of the letter suffixes are not necessarily standardized and might vary between different manufacturers. Still, what Jürgen most likely wanted to point out is not "you need a chip that has an 'N'", but "you need a chip that *doesn't* have a 'B'".
In case of Texas Instruments, it seems the A or B suffix on the 74xx95 chips indicate power consumption, with "B" being lower power than "A". They only offer the the 7495A (standard TTL) and 74LS95B (low-power Schottky) which both have approcimately the same speed, but the LS chip has lower power consumption due to the more efficient implementation using Schottky logic. There seems to be no TI SN74LS95N.
The thing about the Atari 8 bit computers, is the clock design. It literally is daisy chained through the main LSI chips, and if one of them damages or deteriorates that clock for the next chip, issues like this can result. Use a scope and look at the clock going in the chip, versus the clock coming out of it. The CPU is at the end of this chain and can sometimes get a signal that is barely 5% duty cycle. Usually increasing the speed type of the logic chips also in the clock path can help this issue. A 74LS to a 74F for example. Also XEs with a quartz crystal instead of a clock generator "can" are less stable than the crystal oscillator designs. Most XEs have solder points for both. A quartz crystal with a couple transistors generates a sine wave clock. The dedicated clock generator cans generate a square wave. The square wave has less chance of deteriorating than the cheap crystal clock circuit.
This is an awesome video. It's always nice when something just works but you learn so much more when something doesn't work.
Hi Jan, maybe a silly thing to mention but have you checked the capacitors (value) for the new ram? Also have you checked the voltages on the RAM’s while a demo is running? Keep up the good work, love these videos!
Yes, a software RAM test is too high level, we need to check signal and power integrity.
I got a Atari 65xe, had it for a couple of years but not used it yet! I'm 99% sure it works and the board is in lovely condition. Just need a quality SCART lead RGB/AV. I got the computer for free.
Hi Jan, it is great to see that you have saved Lenovo notebook for retro uses (I’m quite sure you can use it for more stuff as well). Great job as always 👍✌️
+1 for perseverance, Jan! 😁👍
I truly enjoy your thorough method of investigating. I was riveted by both these Atari 65EX videos! Mahalo from your fans here in Hawai'i!
Here is an idea. Jurgen has a really great 512k upgrade! Restore the 65xe to its initial 64k memory and install the 512k board. (I know...not quite the idea you expected....lol)
Jan, I admire your perseverance. I hope that you eventually get a complete resolution to the issue.
amazing perseverance Jan! i'm sure i would've defenestrated the whole computer after the first few hours! great vid. next please :D
Chin up Jan. These old Ataris seem to be a mix of whatever they had on hand.
Unsure but would certainly put your wire runs on the oscilloscope, then check voltages, try replacing PIA as Jurgen suggested in these comments. Then try a replacement ANTIC as merely a shot in the dark. Other than that maybe it's a failing passive component issue? Sorry to see you having difficulties with it. Also you might consider getting an XL/XE SysCheck 2.2 from Jurgen since it offers 512K.
It's not that 74 series being fake is it - ie. a HC re-labelled as an LS? I think the N just means its plastic btw! I also wonder if the wiring (proximity of 2 cables next to each other - or length of them might be a factor?
N just means "plastic DIP" yes... I very much doubt anyone would bother remarking HC as LS.
@@jaycee1980 Clearly you didn't watch Adrians video - they ARE relabelling HC chips as LS (and vice versa).
@@GadgetUK164 I've not seen that one.. seems bizarre though, not something i've ever come across but then I dont buy components from eBay or Aliexpress any more.
They could have at least relabelled HCT as LS, which would largely pass
Are there good decoupling caps on each ram chip?
I would re-route the wires, I see a few wires going under the resistors perhaps it picks up some disturbance and also the switch I would try shorter wires ...
This is also what I thought. I had a lot of annoying problems with random wires like this. But I guess the GTIA was at least partially to blame for the issues, so maybe the Atari is slow enough that long wires don't matter too much
Good one yes, if they are not shielded it may very well do that and also mess up timings when they are too long
At very least he should throw a scope on those wires to see if they are picking up any stray frequencies.
You look like you have another signal overlaid on your square waves around the 17 minute mark. Its amplitude in your samples mostly isn't big enough to trip a false hi or lo when it's overlaid, but I bet sometimes it is. You either have something actually crossed or it's just so physically close that you're getting some kind of RF bleed or something. All I know is that's not just the usual ringing you expect on a square wave.
Maybe try taking the pin you were testing where you kept seeing a sudden fluctuation in the middle of the top plateau of the square wave (edit: 17:12), and see if you can tie its source hi or lo. That'll basically let you isolate the overlaid signal, if there is one. Then you might recognize the pattern in the overlaid signal and then you can look at whatever generates _that_ signal and how it might be bleeding over into the memory bus.
Edit: Sorry if I'm not describing the issue well, I come from the software side of retro computing and my knowledge of hardware mostly comes from channels like yours and some mentoring I once got from a guy who worked in aerospace. It's not my field, but I get the basics and I've seen enough oscilloscope diagnostics to know that your sample looked wrong.
I would write my own assembler program to do my own ram test. We need to get to the point where we have *precision* on how *exactly* it is failing. Are *reads* occasionally failing, or are *writes* occasionally failing. Does it happen only when specific address bits are high or low? Does it fail when only specific data bits are high or low? We don't have enough info right now, and we're shotgunning chips into it with our fingers crossed. That's not going to work when we get to this level of troubleshooting.
You might want to check out the Retro Chip Tester Pro. It's an Arduino board that can test many old 74xx devices but also DRAM chips, dump ROMs and, with adapters, check SIMM modules and other things. It doesn't reliably detect marginal chips (it's not fast enough to check actual chip timings) but it should be able to tell you if a chip is completely dead
You have far more patience than I do Jan!
Were you testing the bare board just sitting on your table? Make sure you clean it of solder dust, etc. Could have been an intermittent short from tiny solder crud on the table. Or maybe place the board on an insulator and not the bare table that you used for soldering...
For the Thinkpad (5:25): Sometimes it needs a retro computer to do retro computer stuff 😀
This has to win the head scratcher of the year award ( if that was a thing ) i fink this was pain for you jan
Love the poke command for the black border.
hi Jan, I know you upgraded the PSU, but this all looks like voltage instability to me. I would check the voltage under load as first thing...
It's very curious to me that the GTIA chip works perfectly fine with the RAM expansion disabled, but breaks when enabled.
Is it possible there's nothing really broken with the original GTIA chip, but it just works differently, and is incompatible with the mod?
It might be curious to send your original GTIA chip to the guy who designed this mod and see if he has the same problem. Perhaps he can provide assistant,since he's the one most familiar with the mod
How stable are the supply voltages? Is there interference? And how stable is the xtal frequency?
I dont know if you checked the power supply before all this, but if it was me, I would be checking it. Under full load for several hours.
I would also make sure power getting to all the chips is at spec.
Ur using different chips now so power demand could be less on the power supply as well, alleviating some of the load and symptoms.
But, either way, you have an issue that is system wide that seems to be power/bus issues... did you check any bus transceivers?
I second this..
I love your videos they help me sleep for some reason xD
A+ for effort!
Jan, would you like to try my CO25953 EMMU replacement? It'll get you only 128KB RAM but might be useful to check if you see similar behaviour as with the 320KB RAM mod..
Jan... I love your videos... :) but... in this case I think the problem is with interference between elements and that "spider mounted" cables over motherboard. I had a Atari XE motherboard that had custom 1MB RAM upgrade inserted on 3 sockets in 4-bit RAM motherboard version. So... Seller climed that is ok, but when you checked it with Jurgen's SysCheck it was possible to see some kind of glitches / artifacts between couting digits during memory tests. I removed this upgrade and now it is ok. If you want to have for sure that this Atari works, it is very easy, but you should go enter to "modern retro computring" world... So 1) remove all DRAM chips (on XE motherboard it is always problem - a) not working b) vertical bends on the screen in the background c) others) 2) put Lotharek's SRAM 64K V3 under 6502C. DONE. You will have stable 64KB machine. If you want to have a more RAM, use solid Ultimate 1MB upgrade, that provide RAM up to 1MB with different configurations, support for HSIO (so important when you have more RAM), support for SIDE3 and a lot of more.. if you don't want to have "modern" elektronics inside Atari, use Jurgen's SysCheck that provides HSIO + extra 512KB. That's all! Greetings from Warsaw and Polish Atari users! BTW. I invested in Keysight DSOX1204G 200Mhz and U1272A - great devices! Thanx for you videos!
It's ALIVE!!!!
I once had a logic board with 64h DRAM and Z80 CPU. It was a proprietary control board for some machine. However, in time it started developing random glitches. After few days, I soldered on the bottom of the board DIRECTLY ON THE RAM POWER PINS a 1uF capacitor, on each ram chip. It solved the problem. Maybe try it to see what happens. My guess was that the decoupling capacitors went bad so I tried his solution that worked.
Edit: the glitches started to appear after the Z80 PIO and the Z80 CPU were replaced; they died from a power surge so, the new chips were less tolerant to the unclean signals on the bus. The added capacitors maybe helped to keep the power on the ram chips clean, thus making the signals better.
While you have a oscilloscope, watch the power on the ram chips, see if there are something interesting when the ram test fails.
Jan....check all those wires for stray frequencies using your scope. You may be getting some spikes. Use some shielded wiring and check your power supply as well to make sure the extra chips are not dragging it down which could definitely cause an issue. If you have a bench supply with higher current capability, hook that up and try those tests and demos again.
Perhaps a twisted pair with one line tied to ground?
@@JanEringa8k indeed!
Consider also Sys-Check 2.2, it's not just a diagnostic device, it provides 512K RAM expansion, simply plug it in the back. Then there's the Ultimate 1MB which is excellent too, but internal and a little more involved...
I've been following your channel for quite a while and I just got to this video. I usually don't write any comments on English-speaking channels, because I don't know English well enough to write a meaningful post myself, I'm from Slovakia. Now I'm making an exception and using Google Translate because I thought of one potential solution to the RAM expansion problem. I noticed that the plus (+5V) that you lead to the expansion switch is connected to the terminal of the resistor and that you said that you measured +5V there. It occurred to me that it might be a pull-up resistor and that it has a log. H most of the time, but there may be short pulses to log. L, which the multimeter does not pick up. It's just a shot into the unknown, but try to check with an ohmmeter whether the terminal of the resistor where you have the switch connected is really directly connected to the +5 V branch, or if it's not the other terminal. And while I'm writing a comment, I want to thank you for your videos.
Thanks for this! I tried several different 5V connections in the meantime and I am pretty sure the fault is somewhere else. Currently suspecting the PIA to be slightly wonky. I’m going to revisit the Atari soon and try to figure it out once more. :)
@@JanBeta Thanks for the answer, I hope you manage to solve it
Have you ever considered getting a Retro Chip Tester Pro? It's a very powerful way to validate good/bad chips with a minimum of effort. I know Noel's Retro Lab and Adrian's Digital Basement both use them, and they have saved a bit of time compared to trying to swap out so many untested chips. I understand it's a rather involved project, but it seems to be worth while in the end.
Don’t you need 5v to the Gal to properly enable the extra ram ?
@ 18:07, that falling edge is something that I would absolutely be worried about. And the rising edge that follows, I would be TERRIFIED about. And do the board over.
A system with DMA chips seems to make troubleshooting more difficult. Can you reproduce the crash of each program?
Nice video. some proper rabbit hole deep dive action. :P
Would it not be worthwhile looking at the Freddie Chip?
looking at the video you bend pin 6 out on the MMU but after the cut it looks as if it was placed in the socket is this just visual?
seems that the problem only occurs when the keyboard is laying on the mainboard. could you be getting some weird interferance from it? maybe put some plastic between the board and it? as a random guess
Check all the buffer capacitors and the power supply!
Can it be something like a bad/false contact pcb trace?
Or a dryed out electrolytic not filtering out the power supply of the ram chips under certain loads?
logic analyzer mit Visualisierung des timings?
ich hab den digiview dv3100. bedienung vielleicht etwas ungewöhnlich, aber trotzdem ein gutes teil.
da könnte man das timing gerade mit unterschiedlichen anderen componenten gut vergleichen und auch mit dem cpu datenblatt.
Great Video, thanks.
I would try adding a small value resistor in series with the signal wires to reduce reflection. Say 22 ohms. Preferably at the driving end of the wire.
Curious, what would happen with the original CPU and the new GTIA.
I've always wondered how much harder it would be to go the rest of the way - to 512K. You're already laying the address lines for the 256-bit chips over the 64-bit chips. Why not remove the other 64K and make the system bank in all 256K chips? Just curious - not at all my area of expertise - just asking.
It's a MOS 6502 variant. It can only ever see 64k. I'm not entirely sure how all of these extended memory ranges work on these computers, but I suspect they can't exceed a 64k base and everything else would be extended memory.
@@awilliams1701 I'm not quite sure "extended memory" would mean the same thing as it does in the PC universe. The norm with 8-bit hardware was bank switching, in which pages of 64k would be swapped in and out of the conventional memory space. The switching of memory pages should be handled by outboard logic, usually an MMU or similar type of chip.
@@awilliams1701 I didn't consider that since the processor requires the base 64K to be always present (for the stack and other zero page settings) that putting a swapping mod might create issues. Good thought!
@@horusfalcon I think that was the implication. An MMU with bank switching would be the only way an eight-bit chip could pull that off. Most chips use the base for stack/zero page operations in the base 64K. Even in the early days of the IBM PC where a 16-bit processor was used, the segments were still limited to 64K. Exceeding that in programming required setting registers in the processor to a stack segment (so the stack had 64K), a code segment, (where code in the segment had to be within 64K - but could jump to another segment), and a data segment 64K. However, 8-bit processors didn't have the built-in segment management registers so required an MMU.
@@jeffreyphipps1507 It doesn't have to be the whole 64k but it does need a constant memory portion for it to work I believe.
Did you do a RAM test before doing the upgrade to see if the standard RAM had any issues?
Have you tried putting the original RAM chips back into their sockets? I noticed you still had the new ones in at the end, and while I believe you said they were tested and working, I'm wondering if they're faulty, or marginal at higher speeds? That might also account for the background color; I imagine the system at some point passes the color through it's RAM, and I also imagine the RAM it uses for that would be in the original bank, not the extended one, though maybe it does just go ROM -> CPU register -> graphics chip
If the machine works in 64K version and not with 256K expansion enabled, then there is a conflict which causes the machine to not work as expected. I would question the originator for more detailed info on how the mods actually work.
I would suspect the 'N' chips are non-buffered outputs
Isn't the RAM test software itself unstable?
Check the rams capacitors you installed, maybe one is bad?
I suspect marginal RAS/CAS timings causing the chips to corrupt
Yeah this is especially true if the add-on chips have different timings, especially if those timings invoke special functions, as some newer RAM chips do. If you can't find chips that are made for the same timing, it may be necessary to design your own memory controller circuit to help the faster ones out.
I’m catching up on videos from months ago - did you ever solve this? GTIA *shouldn’t* have had much of anything specific to do with memory failures, except inasmuch as it’s part of the system clock circuit. The OSC inputs goes into GTIA (pins 16 and 28 on a PAL system), which generates the Fast Phi0 signal, which goes into ANTIC. ANTIC produced Phi0, which goes to SALLY, which then creates Phi1 and Phi2. It’s possible a tiny irregularity in the GTIA might affect a little of that timing but it’s surprising if that’s the only issue. It looks like your machine has a FREDDIE chip - I would definitely look into sourcing a replacement for that. XEs are much more common in Europe than the US and they shouldn’t be impossible to find, but might be getting pricey. I think Lotharek in Poland is either making or has announced a modern replacement for FREDDIE but I don’t know about availability in light of current worldwide chip shortages.
You said "Happy Ending" and I suddenly thought about cooling down the chips, not that they are overheated, but cooling down chips could worsen the beginning of malfunctioning. Maybe reduce the clock frequency, so that if it is a timing problem things might clear up.
Der Antic und der GTIA arbeiten in allen Ataris, in denen sie vorkommen, als Gespann. Der Antic entlastet dabei den Prozessor von Grafikberechnungen und liefert diese dann dem GTIA, der die weiteren Grafikberechnungen vornimmt. So, wie ich es gelesen habe, haben sich beide Chips über ihre Entwicklungsdauer nur wenig verändert, daher ist es unerheblich, aus welchem Baujahr sie stammen... Vor 1982 gab es den ctia,den Vorgänger des GTIA....
Erst liken - dann gucken 😊🍻
As far as I remember to check the logic chip you skip LS part... So look for 7495 👍
No happy conclusion and worse still I ran out of popcorn! Flashjazzcat might be able to help!
Tweak the colour pot to adjust the background colour to the right shade of blue.
No idea what's causing the other issues.
Yay Speicher nachfüllen :)
Might be worth checking the logic levels when one of the problem games is playing up, now that the memory checker is working.
if theres an mmu, and its not addressing properly, that might be it, if the ram isnt banking correctly, i would suspect cpu/or something that is responsible for bank switching. if its not that, think about some of the glue logic again, perhaps its the way everything is tied together? can you loop pin 1 was it on the ram under side on both banks? i almost suggested a timer on the ram side from a faster 100ns to 150ns or 120ns or so, maybe a resistor pack? maybe a series? just taking a shot there. *smirk* what if its the joystick ports? its atari, expect strange happenings. graphics... its probably ram, switch out the ram again with the ram you had in the machine. are the grenades supposed to look black om the atari version of commando? i thought that was odd. check chip select? check clock lines? thats all i can think of. check all the glue i guess. good luck
Da der Tausch des GTIA ja soviel gebracht hat, könnte - soweit du sicher Verbindungsfehler ausschließen kannst - auch der ANTIC Chip eine Fehlerquelle sein. ANTIC und GTIA arbeiten immer als "Gespann" für die Grafikberechnungen / Kollisionsberechnungen und können dank DMA direkt auf das RAM zugreifen .Da "deine" Abstürze immer auftreten, wenn Grafikberechnungen / Darstellungen auftreten, könnte das noch eine Möglichkeit sein. Der ANTIC entlastet dabei die CPU von Grafikberechnungen recht ordentlich, ist also fast so etwas wie ein "Urahn" der heutigen GPUs. Gut, dass diese Chips vom Baujahr unabhängig zueinander kompatibel sind, solange du darauf achtest, dass du auch immer die PAL-Version einsetzt - es gibt auch GTIA für NTSC oder SECAM...Um Restzweifel am 74LS auszuschließen ,das "N" sagt nur was über das Package aus, das passt schon so (es gibt auch noch eine Variante mit "D", aber das ist dann ein SOP-Package, also quadratisch im Stile der Agnus im Amiga)...
The `BN` has nothing to do with the electronics - its a manufacturer specific code, for TI and Motorola (`SN` manufacturer) it indicates packaging... `N` and `BN` both indicate a DIP package and I'm not sure what the difference is
The MEXICO 6502s are infamously known for being faulty. Try getting a different one. Also update your XRAM test to the latest version.
Well, that's unpleasant! There are still a lot of possibilities: dry solder joints, bad sockets, marginal ROM...
Oh! A cliffhanger…!
lol 53281 days without accident........ thats 145.878544 years... :D pretty good jan .
Which kind of power supply do you use? It certainly draws more power than without the RAM expansion.
And is the timing of the original RAM ICs and the addon ICs the same? I guess they should all have the same timing
I also thought of a wonky power supply or marginal on-board converters with these kinds of symptoms.
Hard one there. You can figure it out, I am sure Jan.
Send the whole computer to Jurgen. He may need to update his mod design.
Most gals I know would take offense at any attempt to program them. They fancy themselves as being independent women.
This is not "troubleshooting", you are just trying to replace parts at random. If the RAM test is failing, how exactly does it fail? What test exactly fails and how? Is the failure specific to particular chip(s), address lines, data lines? Does it affect just the extra RAM chips? If you're suspecting a timing problem, you should examine the signals on your scope, check them against the datasheet, look at timing diagrams in RAM chip datasheet, in processor datasheet, verify that they are compatible.
Timings are a kind of mystery here.
I think you mean mystery
@@raypalmer7733 Yep. My bad. But still better than misery ;-)
Frieddie chip? RAM address multiplexer
Pull all the chips off and trash the motherboard it obviously has a problem that can't seem to be fixed.