This is Richard, I'm the one who donated this to you. So glad you got this working! I didn't think this would turn into a multi-part series. I'd say I'm sorry for giving you such a problematic device but this is what you love and we the audience!
Adrian, I have refurbished a fair number of Atari ST's and have a spare MMU C025912-38. I'll just double check that it works (I have a 520STFM on my bench right now I'm fixing up to test it on). Happy to send it on over. Just tested...works fine.
This is one of the things i love about the retro community: almost every time one retro geek runs into a problem, another has the solution, usually within a few hours. You, Sir, deserve a medal! 🥇
Wait a second, how about trying to bake that broken one in the oven first? As Adrian himself experimented with: th-cam.com/video/QxeUJbIubHA/w-d-xo.html
Some background Information: The RAM bus and the system bus are separated. The Shifter is on the RAM bus. The separation is needed for the framebuffer (Shifter) and RAM refresh. The MMU is controlling the Shifter and it is fetching 16Bit every 256 clock ticks. Without the separation the system bus would be blocked every time a RAM refesh is running or part of the frame buffer is fetched into the Shifter. A Crazy little tidbit. To load the Shifter register only the lower 8 bit of the RAM bus is connected to the system bus to present the register to the CPU. The ROMs are on the system bus not the RAM Bus. The 68K can't run without RAM. Important Interrupt and Error-Jump variables are stored In the first 512 byte of RAM. The only reason why MMUs are diying is an overload on the output pins. (and over voltage) The chip can't really power a lot. With 4MB RAM the output limit is reached. This is the reason why you have to completely disconnect the RAM (lifting all the resitors). It is recommended to remove the old RAM chips. Not only to protect the MMU but also the power supply. The diconnect of the CAS0/1 lines is a botch job that can work but will stress the MMU. SOME versions of the TOS did crash when the keyboard is not connected. The reason is that the inteligent keyboard controller (ikbd) is also the system clock. And when the system clock is not initialized the TOS would crash. But the ikbd is a very bad clock with an high drift. But good enough for a day or two. So later Models like the MEGA ST, the MEGA STE, Falcon and TT had all clock chips.
Lots of people pointed out that the Atari does have a full 16 bit bus and that I "got it wrong." That may be, but consider the fact this was the first time I had ever worked on any 16 bit Atari, I had not yet examined the schematics for the 520ST and that the other 68000 machines I've worked on (the Macintosh 68000 machines and the Amiga 68000 machines) do not have two CAS lines for the high and low bytes of each memory location. All 16 RAM chips on these systems get the same CAS signal so the RAM is either selected or it's not selected. If one were to hook up 8 bits of RAM to a 16 bit system, two latches and two banks of 8-bits of RAM with two select signals would be one way to do it. This would easily allow Atari to install half the RAM in the system to allow for 256k of RAM. (For the 260ST)
While I'm pretty sure the ST (at least normally) uses the full 16 bit bus, you are correct that the existence of the latches between the MMU and the RAM suggest it is a possibility. The board I made in school used latches in the way to interface a single 8 bit ROM to an m68k, although for the RAM card we did have a 16 bit data bus with separate select lines for each byte connected to UDS/LDS in large part because our address decoding logic was simpler that way.
Using latches and two select signals *COULD* work since the 68000 doesn't have an A0 signal on the address bus. It instead has an upper data strobe (/UDS) and a lower data strobe (/LDS) which you could use as the selects. Those signals are there because Motorola intended RAM to be organised in parallel high byte/low byte "banks" and is precisely why Atari chose to organise the ST's RAM in two or four parallel high/low banks the way they did. Besides, you'd need to get the timings just right for it to work which would likely mean having to use much faster (and therefore *MUCH* more expensive) RAM chips more than wiping out any savings you'd make by using half the chips - well, back then anyway. Otherwise you'd have to find a way to insert wait states while loading those latches, I've not done any 68k hardware design in a couple of decades and I'm not sure that's even possible (maybe using the /DTACK signal?). All of which would slow the machine down. Also for 16/32 bit memory operations /LDS and /UDS both go low together, so even more complications... No, the only sensible way to create a 256K ST would be to have two banks of 128K RAM chips - e.g. HM48128P, MK4128N, etc.
@@nathanperry1997 RMC recently showed how to upgrade an Atari 520STFM from 512k to 4M but from what I can tell that probably wouldn't work in this 520ST.
Yeah here is just one memory bank 16bit width connected, you can solder a second batch of memory ontop of the first one and connect ras/cas to the mmu to get 1mb of memory. Atari did it on the 520St+
That MMU may still be alive, clean the pins and check if none of them are pushed in from the side, those PLCC mod boards wreck sockets and chips. TOS 1.00 is just bad, replace it with 1.04 ASAP, 1.00 is kinda buggy and don't even bother with harddisks on that version, the hardisk routines are seriously borked in 1.00 I have no idea what is in your SF314 or if it even works, but the signals that come from the atari itself are pure Shugart, any active conversion within SF354 or SF314 is to accomodate a non standard drive. An easy way to check if an atari 16/32 bit machine lives is to check the floppy led or the DS0 line on the YAMAHA, a faint blink indicates some lif in it. One of the biggest issues with these old Atari's is in my experience. bad contacts on sockets, especially the square plcc sockets, clean them or carefully re-bend them. If needed replace those pesky staggered sockets with SMD sockets, with some bending of the outer pins you can solder these sockets with smd and through hole at the same time. The easiest retro way of upgrading the 520st is by piggybacking another set of ramchips on the existing and connect their RAS and CAS lines to the other set at the MMU for a home made 1MB total 520ST+, or you can solder a 72pin socket to the board, it's a tedious job, but a fine DIY way to max it out, ST's need FP-RAM sticks. Boot speeds relating to floppy drives. 1: diskette with game or bootsie bootsector 2: regular diskette in drive 3: no disk drive attached, takes a second to detect a drive foor bootsector and gives up. Also no A or B drive on desktop. 4: empty drive, takes a long time to give the user time to insert a bootable diskette
I still have TOS 1.01 on my machine. You can get TOS 1.04 and boot it from the Floppy if you can't get the ROM. Also you can get the Clock speed up to 16 MHz. but if you want some old games or software to run you should install a switch so you can boot the machine stock without having to remove the mod.
You gave me quite a start there! I thought, "Oh no! Adrian broke my 1040STf!" :) Getting chips seated back in those PLCC sockets is such a hassle. I've had that kind of problems so many times with the Agnus chips in a Amigas.
The only thing more gut wrenching than bricking a piece of kit is bricking something that was entrusted to you! I felt that wrench even though I was only watching!
I had a 520 STFM, with a 4MB memory expansion, which used 30 pin SIMMS. I also had a 52MB quantum scsi hard disk, with a DMA host adapter, which cost me two weeks wages back in the day. But it allowed me to play The Secret of Monkey Island, without swapping floppies :-)). Keep up the great work.
Sounds pretty much like my Amiga 500 back at the time. Only I had a 40MB Micropolis drive, and with the SCSI controller I think it was almost three weeks of wages that went into the HDD. And I got that drive at about half the RRP... HDD's were horribly expensive back then.
This is a rare talent these days. My electronics instructor said to trace backwards and make your own schematics to the system your troubleshooting. I have not seen many people do this. They buy the original schematics for trouble shooting.
This is one of the most common failures on the ST. Turn the chip upside down and carefully push/bend the pins outward the end of the pin has a stop point on the package. Also bend the socket pins slightly so they put more pressure on the chip. If that doesn't work the MMU is definitely fried.
Glad you got this sorted, and very cool that another local TH-camr was able to loan you a test machine. I hope you can source a replacement MMU without too much hassle or cost.
Although the 68000 has a 16 bit data bus the memory is byte-addressable which requires separate chip selects for the high and low bytes - hence the two chip select bodge wires. Check the CPU pin diagram and you won't find an address line A0. It is replaced by a pair of upper and lower byte select lines. The 68008 (as used in the Sinclair QL) is basically a cut down 68000 with an 8 bit data bus, fewer address lines and a few missing control lines. I had a 520STF back in the day with an SM124 monochrome monitor and wrote my university dissertation on it (BSc Computing Science). I moved up to an Atari Falcon but after years gathering dust in the loft after I moved to PC's they both ended up in the local refuse dump! 😲 This was long before online auction sites made such machines valuable. I still get cold shivers when I remember dropping the Falcon into the dumpster 😭
You simply can't just wire up the memory with only 8-bits and call it a day on this iteration of the 68000, all 16 bits of the data bus are acquired whenever instructions are fetched, whenever data is accessed that is a word or longword, whenever reset or exception vectors are accessed, whenever any kind of stack operations are performed, etc. The ST schematic shows that the 68000 data bus isn't wired directly to the RAMs, data from the RAM is latched by 74LS373's, and writes to RAM are buffered by 74LS244's. I'd say this is probably used because without them, the MMU would have to stall the 68000 even for non-RAM accesses while the shifter is retrieving the display memory. The buffers split the data bus so that the 68000 can access ROM or non-Shifter I/O concurrently while the Shifter/MMU is fetching display data. The ST schematic suggests that the MMU can control two different banks of RAM, but on the 520ST only one bank (of 16-bit-wide data) exists.
The 1040 ST that appeared part way through is so nostalgic for me. It was the first machine I used that that felt like a "real computer", with its incredible 1MB of RAM and graphical interface.
The MMU and GLUE chips can go bad on ST’s. I’ve fixed a few with bad MMU or GLUE chips. I’ve fixed a 1040STf with a bad GLUE chip, it totally booted, but the screen was all glitching and no green desktop. But it gave the keyboard beeps. Replaced the GLUE chip and was working perfectly again.
I have absolutely no idea what your doing but by the end of your videos I’m understanding more than I did at the beginning.so thank you for your videos I appreciate your knowledge
I used to work at Chiang Computers in Ottawa, selling Atari and PC computers. Some of the common repairs my boss Gary would make to Atari computers was cleaning and bending the pins of the sockets and chips to improve connections, and replacing the power supplies. He fixed a lot of Atari computers. Apparently bad solder joints was common on some power supplies. Seeing this video brings back memories.
I was lucky enough to live near enough to an electrical engineering genius, we had originally piggy backed 520's to 1Megs, and then later he had a single chip with all three tos versions on it, selected by 2 switches set before you booted up, the last thing was a 4 MB ST had to swap out the original memory chips for that one though. And if I remember correctly, the bigger chips had 2 more pins on them so the sockets he used had 2 pins lifted up and hand wired / soldered to complete the modification. Never had a single problem with it.
When I saw 520ST in the video title I had to watch both parts. I enjoyed watching very much. I had both the 520 and the 1040 in my long journey of buying and building gaming computers. Actually 38 in total. The most recent I built from scratch and am enjoying. Just wish I could get a 3080 to replace my 2080 TI, but of course due to covid that's nearly impossible except for the rich folk which I am not. But anyway, thanks for the videos. They sure brought back some memories. Cheers.
I like the thumbnail picture :) It seems like a real smile for something that succeeded, after years of boring C64 repairs some old thing gave Adrian real brainwork :)
I have a lot of Atari, 1-800, 1-512, 2-1024, 1-mega STE, spare floppy drives, many DD, Hand scanner and even the SLM804 printer. all are working fine. 100's of software and most all the magazines and their disks. It takes a 4x8 table to set them all up. Thank you for the post very interesting to me.
I fell in love with the 520ST the day it came out. Then the 1040STF came out. Was a poor college student and couldn't afford one. In second year of electrical engineering, I found a job as a computer technician, I mainly worked on IBM PC, Apple II and North Star computers. Now, at 63, I find myself looking at the schematic and admiring the design of the 1040. I could never go back to working with one. It would drive me mad. I'm spoiled having an 8-core 4GHz PC and even now, I wish it was faster!
I was at the Summer Comsumer Electronics Show 1985 in Chicago. I knew of the development of the Commodore Amiga and the Atari ST, so when I saw a flyer advertising the introduction of the ST line in a room upstairs off the main show floor, I went (they were too late to get a spot on the main floor). In the room, there were two ST models shown: the 260ST and the 520ST. along with some accessories such as an external CD-ROM drive, which they demonstrated. The desktop image looked oddly familiar; I'll get to that in a later post, if there's interest. I recognized Jack Trameil and at least one of his sons at the other end of the room. There weren't a lot of people there, but it was early in the Show. There was not much oganization and I just played around with a 520 for a bit, including a QuantumLink chat with an anonymous stranger, and a short chat with one of the reps. I got some literature and left; I was much more interested in the Amiga.
Back in the day we had a trick to bring an ST back to life. Lift the whole unit an inch off the desk, then drop it. I gather this process reseated the chips. A more brutal version involved twisting the whole unit. I can't remember how many times some combination of these techniques saved my ST.
super cool . i wanted one of these machines pretty badly and i only saw one a single time but i played around with it and gave it a good once over .i ended up with an ibm8088 that had no hd but 2 x 5.25 floppys . i got a roland interface and slaved it to a 1/2 inch 8 track reel to reel . i learned a lot . the pcs where quickly getting faster and you had to know when to not upgrade and wait a minute . eventually i switched over to an all digital system and i am so much happier . a lot of younger guys will adore all the older tech but not me . i like the effeciancy workflow of a good daw with good plugins over tape and $$$$$ outboard gear. very cool retro stuff here !
The two banks are probably still in parallel, but iirc the MC68000 enables the upper and lower bytes independently, so as to support 8-bit writes without clobbering the other byte. Just guessing though.
One bank is lower bytes, the other is upper bytes. The 68000 CPU has no A0 so only word addresses are possible. It uses UDS and LDS pins to address either lower, upper or both bytes as required.
When you wanted to fix those socketed chips back in the day without opening the machine, you would just lift the computer up about 6 inches and drop it down on the bottom. That would usually reset the chips. This is why they sold retainer clips for the sockets to hold them in place.
I'm sure someone mentioned this but the reason for the two chip selects on the RAM is because of how the 68k works. It can individually access the lower or upper bytes of a word and even has individual "strobe" pins to do that.
The 68000 can access the two 8 but halves of its 16 bit bus independently using two strobes, *UDS and *LDS, I guess that is why those chips are connected in two groups of 8 bits.
The buffer chips are actually there to isolate the 68000 and other chips from the data bus when the shifter wants to use the ram. The Amiga has a similar arrangement
@@jaycee1980 Yeap, I was explaining, with more or less success, why there are two different enable signals, you have to allow the 68000 independent access to the two 8 bit halves. It's not a surprise the Amiga has a similar arrangement, it uses the same CPU. Cheers.
There are two banks of 8 bits to support byte wide addressing. From the processor (68K?) the UDS and LDS signals would be used to determine which 8bits to access. I used to work for Frontier back in the day and fitted many of those Xtra-RAM units to Ataris. Great fun.
Indeed, the MMU is not proven to be dead (they never died on me in my years). The memory expansion board pushed (cold) in the socket always damaged the pins. They are pushed back. I used a needle to bring them back for a good tension again. The memory expansion board pins are pushing the socket pins back because they are between the socket pins and the MMU pins. (This is a terrible way to do a mod. Hmm plug and play the said at the time) It always ended up in wite screens eventually. Just an addressing problem. So, see if you can restore and polish the MMU socket pins and try again. I did this a lot! 35 years ago, for Atari the Netherlands. I hope it’s working, … I have a lot of original spare parts. I do sell them occasionally. I’ll see if I have the MMU.
The Atari ST did not have an 8 bit data bus anywhere and the latches are not there to somehow map 8 bits to 16 on the CPU. It was a full 16 bit bus. The latches were probably there for buffering timing and managing bus contention. The reason the RAM is organised in 2 8-bit banks is because the 68k is byte addressable. If you look at the pinout of the 68k there is no A0...just 23 address lines even though it has 24 bits address space. Instead of A0 it has LDS and UDS (lower/upper data select). You can select either data bank OR BOTH at the same time! In 68k assembler you can only access a 16 bit value from RAM from an EVEN address and both UDS and LDS are asserted. If you do a byte access you can specify even OR odd addresses and only LDS is asserted for even addresses and only UDS for odd addresses. Hence why each 8 bit bank has its own select. One is even bank and the other is odd bank.
Yes others have mentioned this as well. Curious how the Atari 260ST would work (which I assume is this motherboard with half the RAM removed?) When this machine is working again, I should try disabling half the memory to see what it does.
@@adriansdigitalbasement The 260ST had 512K of RAM -- it was a marketing ploy to sell off old 520STs with TOS loaded from disk (the machine have around 256K of free RAM left after TOS had been loaded in from floppy). All ST ram banks are 16-bit wide, so technically, the ST can support two banks of 128k to support 256K of RAM but no machines were ever made like that (and one bank for a 128K ST -- only seen at the original CES demo).
You can contrast that to the TMS-9900 in the TI-99 which also doesn't have a A15 (TI counts bit the other way round Motorola A0 is TI A15 and TI A0 is Motorola A15) but does not have the equivalent of LDS/UDS. It always reads 16 bits and masks internally. This is particularly bad as it has only 256 byte of 16 bit memory, the rest is 8 bit and each memory access will trigger two sequential reads from memory. This means loading a 16 bit register needs 4 bus accesses (each having 2 wait states to make things even worse).
@@DrSteveBagley The 130ST prototype was likely equipped with 16x 4164 RAM chips, and yeah it looks like the 260ST was just a 512ST without ROM chips for TOS, and IIRC was only badged that way im Europe and for a short time.
@@adriansdigitalbasement I would be curious as well to see if the MMU does some sort of trickery, but I k ow the CPU requires the full 16 bit data bus to be present to function properly. It's like how 386 PCs used 8 bit wise SIMs but needed banks of 4 in the same size to work. In other 68k systems I've seen with 256k they used 32 4164 chips arranged in 2 banks of 16, and each bank having 2 groups of 8 like here, and in another case there were 8 4464 chips in banks of 4, each bank being divided into pairs.
So far you've made me get my 1040ST and C128 up and running again, now to find the power supply for my CD32 and Vic20 and I'll be back in 8/16 bit heaven...
On the original Atari 520 ST you needed the drive connected to boot up because the OS (GEM & TOS) was loaded from disk. But even without the drive connected you would get an error message. When the OS was moved to ROM, you no longer needed a drive connected to boot the ST, but it was pretty useless without one since almost all the software available for it was disk based.
The reason there are two banks of 8 RAM chips is so the CPU can read/write one byte at a time. The 68000 has no A0 pin so can only address words. It has UDS and LDS (upper and lower data strobe) pins for addressing the even and odd byte at a given address and uses the lower or upper 8 bits of the data bus to do so. If it wants to transfer 16 bits, it strobes both pins and uses the full 16 bits of the bus. It would be interesting to see if (with a bit of hacking) the onboard RAM could be relocated in the address space to have 4.5MB total rather than disabling it.
The PLCC sockets Atari used are notorious for having bad contact. I have heard it was common back then to try the 5-inch-fix first. That is, you lift your ST flat 5 inches above the table and then drop it. This was supposed to push the chips back into the sockets.
I really enjoyed this look back at the Atari ST. I had a 1040 ST with a hard drive (10 whole megabytes) that I used for years. It brought back memories of when I was one of the Compuserve Sysops for the Atari Forum. On that cartridge port, I used to keep an Action cartridge plugged in there (Action was a popular programming language on the ST).
I used to have one of these machines. One of the things these were notorious for was the chips in the square sockets would, due to thermal cycling, begin to work their way out of the sockets causing random issues. One of the "official" service bulletins was to hold the computer about 7 inches above the bench and drop it. That would reseat the chips! LOL.
I still am getting sick that I ever sold my ST 1040 (F) for a few guilders (about 100, should be 50 euro's now). That machine learned me to program with GFA Basic, finding out what MIDI is, how to cheat a brick game etcetera. And tham I bought my first PC, a 8086 with a 50 Mb HD. Man, I did not know what to do with it, until I got a copy of WP 5.1. That worked. But I still miss that ST 1040. Today I own a laptop, my 4th, and before that some PC's (tower and desktops) about 5 to 7, I don't really know anymore. But I still miss the ST 1040.... (and to be honest, I still have an Sinclair Spectrum, with a unit giving it a diskdrive and parallel printerport, and a video out (coax).) z)So, Adrian, in case you are interessted to find out how a Sinclair works? I can send it to you. (Oh, 220 V, 50 Hz)
Just wanted to point out that the Motorola 68000 is a 32-bit processor, with 32-bit instructions, but with a 16-bit data bus, and 24-bit adress bus. Motorola called it a 16/32-bit processor.
The Mega ST line was the follow up to the 520/1040 There was the Mega ST1 and the Mega ST2 -- I owned a Mega ST2 back in the 90-94 timeline. The number indicated the megabytes in the system. They were pizza boxes with separate keyboards. I think they came with an integrated HD but it's so long ago I'm not entirely sure. (Edit: Nope. No internal HDs. Did some googling...) Main reason for buying these were that (a) Apples cost so much more and (b) desktop publishing. Yeah, everyone associates the mac with desktop pub, but there were a few decent programs for the Atari ST line that worked just fine. We had a few 1040's at my work connected to an apple Laser -- whatever their first laserjet was called, and used it a ton.
Mega ST2 and ST4 came first then a little bit later the ST1. The Mega ST's were sold as "professional" machines and therefore would need more memory. They realized only later that there was demand for the only 1 MiB version. They hadn't an integrated harddisk. They had another pizza-box with harddisk called Megafile 20, 30 and 60. The Mega ST's had a better keyboard (real keys using cherry MX black ), blitter integrated and a battery backed realtime clock. They also had an extension connector with the CPU bus for easy extension. I still have mine with integrated PC-Speed emulator and Volksfarben ET4000 VGA card.
I wish I could get my hands on a decent ST. Maybe one of the later ones with more memory, built in floppy and PS. I had an original 520ST that I expanded to 1024KB by the piggy-back method. Did some awesome quality hours on that machine. Mostly working with N-dimensional intersections with 3D space .. for instance, how would a 4D object appear if it was inserted into a 3D space. Think about sticking a 3D sphere into a 2D space. It would appear as either a point or a circle, depending on how far it was inserted into the 2D space. Now think of what a random 3D object, maybe an odd shaped rock, would look if ti was inserted into the same 2D space. Now, think about 4D objects in 3D space.
The STe machines are good. Used to have a couple of those, wish I still did. They had a later TOS versions, more colours available & also used 30 pin simms so ram upgrades were a lot easier than previous versions.
@@CJ-rf9jm yeah we got an 520STE but a few of our friends games we borrowed that ran fine on STFM wouldn't work on the STE - so we got the store to give us a STFM instead. This was a really dumb thing to do when we tried to upgrade the RAM an saw we had to solder wires and mess about with jumpers etc... And we learned that the STE could pretty easily be upgraded to an insane 4MB just with socket SIMMs. There was a small amount of software incompatibility on STE early on - but most developers starting making sure their new releases worked on both models. And we never did fit our RAM upgrade.
@@adz693 I had a 1040STe from 1990 onwards that I'd later upgraded. I never had any incompatilibility problems. That store absolutely ripped you off badly. The problem were those badly cracked games. I'm not against cracked software when it's no longer available except those who do it so badly they end up screwing up the program in the process of cracking it.
It's 2 8bit banks because the 68000 can also access only one half of its 16bit data bus. For 16bit access, both banks are used at once (you actually can't do 16bit accesses on odd memory adresses, this started only to be possible with the 68020, iirc).
The reason for needing the keyboard is that its got an embedded z80 on board and provides the system clock. If you add a battery backup to the clock on the keyboard then it'll survive power off Seconded and thirded comments about the sockets. They were a notorious failure point on all 1980s machines. Pull and clean everything
Very good troubleshooting. You nearly reverse engineered how the ST works. :-) Bad MMU is not that common. I never had an ST with a bad MMU. But the PLCC sockets of the MMU and the GLUE are known to have contact problems from time to time. Especially if a RAM expansion was squeezed between chips and socket. I would suggest to take a closer look at the pins of the bad MMU with a microscope. Maybe some of them are not making good contact and the MMU from the 1040 is working because it has them still in order.
For a time I worked doing warranty repairs for both Atari and Commodore. I saw pretty few Atari ST machines coming in but a lot more Amiga. I remember that for a time Commodore had big problems with those PLCC sockets and chips not getting good contact. A lot of this was traced to the factory where completed and boxed machines got bumped around when the pallets were loaded onto the trucks. To get to the loading docks the forklifts drove over some very uneven parts of the floor. As I remember it they described it as all the cartons on the pallet going airborne only to slam down after free falling more than an inch, but that could be my memory playing tricks. Apparently that was enough to shift the chips just a little in the sockets and if you were unlucky it resulted in the machine failing. Often just pressing all the chips firmly into the sockets could get the up and running again. Even after addressing that problem with the forklifts Commodore choose to install a retention spring across the the sockets and it cut down on the number of DOA and early warranty returns.
Lol I stand corrected! I guess the original ST can boot, sort of, without the keyboard. Is the IKBD chip on the motherboard on these? The stuck key thing happens when there is an error with the keyboard controller, it also happens when a joystick is plugged in with the autofire on after a reboot. Or some times just randomly. Hitting a key will stop the beeping in the latter case.
Cool video thanks for making it. I still have a Atari 1040ste modded to have more memory. IT still works fine except the floppy as it needs cleaning from smokers.
The Atari 520 ST had a real problem with chips becoming unsocketed. At least for a while, the official fix was to drop the computer 4-6 inches onto the table. I agree that it sounds sketchy, it was. My point being you might want to reseat every chip that you can. I have an original 520 ST, and also a 1040 that I bought years later.
I added 520kb to my 520STm with an upgrade that piggy backed the MMU and something else, probably like this upgrade was done. Can't remember where I bought it but it was through a Swedish computer magazine. It was a lot of money for me but that extra sound and graphic effects in Falcon was worth it!
Adrian, you could place a DPDT switch in place of the mod jumpers and chip select traces on the memory to effectively enable / disable the memory with the flip of a switch.
Mega STs are desktop case STs w/ Blitter chip and real time clock (in 2MB and 4MB models) added. Some of the later STf/STfm models have metal clips over the PLCC chips because it was a known issue they could get a little loose during shipping. An official 'fix' from Atari at the time for new units not booting was to drop the machines (as flatly as possible) from 6-8" onto a hard surface to reseat those chips w/o having to open up the case. :-) EXXOS Atari store in UK has MMUs. Not sure about US sources (Best Electronics or B&C (myatari on eBay)?).
The PLCC chips could also get loose due to thermal creep over a period of months. I had to reseat my MMU and GLUE regularly in my early 520STM, but it never gave me any long-term problems.
I believe if you examine the memory upgrade, and the little board that has pins sticking out of it that fits on top of the MMU, you will see that one or more of the pins has one side of the pin as bare metal, the the other side covered in epoxy or some type of insulating material. If that insulating material rubbed off onto the pin of the MMU, it might be preventing that pin from making contact with the pin in the socket. Also, I believe the MMU was a custom Atari chip.
Interestingly I repaired two Rev B ST's recently and it was the latches you were checking out that was the issue. Those PLCC sockets go bad by the way. Because that memory upgrade was pushed onto the original MMU it could have forced the pins in too far... you could try pulling them out a little.
Yeah I've heard that about the PLCC sockets from a few people -- I inspected it on my microscope (not seen on the video) and it was actually OK -- along with all the pins on the original MMU. Seems the Marpen pins are thin enough to not cause an issue, at least in this case. Had both latches failed on the machine you repaired?
My 1040ST got me through college with zero issues. Replaced it with a PowerMac 6100 66 DOS that I used until 2005. Two computers in sixteen years, until internet and software demands forced me to join the 21st Century. Simpler times.
RAM is actually organized as you thought it was. The chips correspond to bits 0-15. Just notice, that RAM is addressed with bytes, not words, so it's 16 bits per word, which occupies two consecutive bytes. By the way, it's exactly the same on an Amiga 500.
And, if you plan to connect your STF to OSSC you will probably to modify the mobo a bit filling the elements in the modulator area to obtain a proper synchronization signal for SCART connection.
I use a full spectrum grow light to retrobrite indoors and it works AMAZING. So much faster than outdoors. Just suspend it over your container of water/hydrogen peroxide and you are good to go.
I would be cautious with that memory expansion, give it a once over. Sounds like something might have shorted on it and sent 5v down the data line to kill the MMU. That hot glue that was on there (I think I remember the simms being hot snotted on that expansion) is a huge red flag to me
I agree with yours and others who pointed out that the RAM expansion board is suspect. It wasn't made for the 520ST anyway, so why is he trying to make it work with it if it wasn't meant to be used with it in the first place? He should really look at the Exxos or thff RAM expansion boards which are much smaller (only uses 2 or 4 chips) which produces less heat too. The designers also say they work with the original 520ST models.
That's very interesting. I wonder if the keyboard not being plugged in, is only true on some versions of the board, because I know I couldn't get mine to boot up without it plugged in. It also could be one of those strange coincidences as well, because my system has other faults as well where it won't boot up at all unless I push down all the chips. Probably have a cold solder joint or a bad socket somewhere I need to replace.
I know some versions of TOS are crashing because they want to set the system time in the keyboard controller and fail. It's not the board that is the cause but the TOS version.
@@boelwerkr Good call. I have rather 1.03 or 1.3 (can't remember which off the top of my head) in mine, which seems to be newer than the one that Adrian is playing with.
Yes Something like that. The RAM uses a Column/Row address select (CAS/RAS) schema and the MMU in the ST supports 4 CAS and 2 RAS lines. The person installing the 4Mb simply disconnected the RAS lines but not the CAS and address lines. This caused an overload on these lines an one line driver in the MMU burned out.
Macs, PCs and Amigas all use the pins in rows. In fact, you can't even buy the zig zag type other than pulls from Atari motherboards. The normal type is readily available from every source.
The software that was used to layout the ST boards worked better with the zig-zag socket, so they was used. This type of sockets is not often used otherwise.
Adrian - very interested in how you go about retrobrighting the ST. I've tried on a 1040ste case using 40 Cream + CA sun and it came out streaky and blotchy all over. Air bubbles under the plastic wrap seemed to get "ghosted" onto the plastic as well.
Yeah, from what I’ve seen from the 8Bit Guy, plastic wrap always comes with streaks, he always did better with a tub of solution, and weighting the pieces down to keep them submerged (often a zip lock full of the liquid on top is enough)
@@nichderjeniche Yeah - I've seen all his vids as well as Adrian's, Jan Beta's and others. I've used the 40 Cream method on other (smaller) pieces successfully.
@@Underestimated37 That is what I am going to try next (liquid in a tub). Not sure if this case is salvageable but at this point, will keep experimenting to see what happens. Worst case, I may take the path that Noel's Retro Lab did with a C64 and paint it custom. I've been pondering if I could reproduce the 8-bit Atari XL color scheme on it... LOL.
There has gotta be a better way, maybe there needs to be some research into creating a ‘gel’ using something like polyacrylamide (the stuff in diapers and water beads) that can retain the wetting properties and transparency but lose the texturization that the plastic wrap causes.
This is Richard, I'm the one who donated this to you. So glad you got this working! I didn't think this would turn into a multi-part series. I'd say I'm sorry for giving you such a problematic device but this is what you love and we the audience!
I'm pretty sure Adrian likes fixing old computers more than he likes using them. :)
Thank you for your sacrifice. 😜
Thanks dude, same my first pc XD
Have an ST520FM somewhere here in a box somewhere and it just needs a new floppy drive to function. At least it worked in 1992 when it was last used.
Thanks for giving us so much content in such a small package, Richard. I love these multi-part repair series
Adrian, I have refurbished a fair number of Atari ST's and have a spare MMU C025912-38. I'll just double check that it works (I have a 520STFM on my bench right now I'm fixing up to test it on). Happy to send it on over. Just tested...works fine.
This is one of the things i love about the retro community: almost every time one retro geek runs into a problem, another has the solution, usually within a few hours.
You, Sir, deserve a medal! 🥇
Wait a second, how about trying to bake that broken one in the oven first?
As Adrian himself experimented with:
th-cam.com/video/QxeUJbIubHA/w-d-xo.html
Some background Information:
The RAM bus and the system bus are separated. The Shifter is on the RAM bus. The separation is needed for the framebuffer (Shifter) and RAM refresh. The MMU is controlling the Shifter and it is fetching 16Bit every 256 clock ticks. Without the separation the system bus would be blocked every time a RAM refesh is running or part of the frame buffer is fetched into the Shifter. A Crazy little tidbit. To load the Shifter register only the lower 8 bit of the RAM bus is connected to the system bus to present the register to the CPU.
The ROMs are on the system bus not the RAM Bus. The 68K can't run without RAM. Important Interrupt and Error-Jump variables are stored In the first 512 byte of RAM.
The only reason why MMUs are diying is an overload on the output pins. (and over voltage) The chip can't really power a lot. With 4MB RAM the output limit is reached. This is the reason why you have to completely disconnect the RAM (lifting all the resitors). It is recommended to remove the old RAM chips. Not only to protect the MMU but also the power supply. The diconnect of the CAS0/1 lines is a botch job that can work but will stress the MMU.
SOME versions of the TOS did crash when the keyboard is not connected. The reason is that the inteligent keyboard controller (ikbd) is also the system clock. And when the system clock is not initialized the TOS would crash. But the ikbd is a very bad clock with an high drift. But good enough for a day or two. So later Models like the MEGA ST, the MEGA STE, Falcon and TT had all clock chips.
Lots of people pointed out that the Atari does have a full 16 bit bus and that I "got it wrong." That may be, but consider the fact this was the first time I had ever worked on any 16 bit Atari, I had not yet examined the schematics for the 520ST and that the other 68000 machines I've worked on (the Macintosh 68000 machines and the Amiga 68000 machines) do not have two CAS lines for the high and low bytes of each memory location. All 16 RAM chips on these systems get the same CAS signal so the RAM is either selected or it's not selected.
If one were to hook up 8 bits of RAM to a 16 bit system, two latches and two banks of 8-bits of RAM with two select signals would be one way to do it. This would easily allow Atari to install half the RAM in the system to allow for 256k of RAM. (For the 260ST)
Didn't BYTE mag publish a way to upgrade the memory for the Atari 520ST?
While I'm pretty sure the ST (at least normally) uses the full 16 bit bus, you are correct that the existence of the latches between the MMU and the RAM suggest it is a possibility. The board I made in school used latches in the way to interface a single 8 bit ROM to an m68k, although for the RAM card we did have a 16 bit data bus with separate select lines for each byte connected to UDS/LDS in large part because our address decoding logic was simpler that way.
Using latches and two select signals *COULD* work since the 68000 doesn't have an A0 signal on the address bus. It instead has an upper data strobe (/UDS) and a lower data strobe (/LDS) which you could use as the selects. Those signals are there because Motorola intended RAM to be organised in parallel high byte/low byte "banks" and is precisely why Atari chose to organise the ST's RAM in two or four parallel high/low banks the way they did.
Besides, you'd need to get the timings just right for it to work which would likely mean having to use much faster (and therefore *MUCH* more expensive) RAM chips more than wiping out any savings you'd make by using half the chips - well, back then anyway. Otherwise you'd have to find a way to insert wait states while loading those latches, I've not done any 68k hardware design in a couple of decades and I'm not sure that's even possible (maybe using the /DTACK signal?). All of which would slow the machine down.
Also for 16/32 bit memory operations /LDS and /UDS both go low together, so even more complications...
No, the only sensible way to create a 256K ST would be to have two banks of 128K RAM chips - e.g. HM48128P, MK4128N, etc.
@@nathanperry1997 RMC recently showed how to upgrade an Atari 520STFM from 512k to 4M but from what I can tell that probably wouldn't work in this 520ST.
Yeah here is just one memory bank 16bit width connected, you can solder a second batch of memory ontop of the first one and connect ras/cas to the mmu to get 1mb of memory. Atari did it on the 520St+
That MMU may still be alive, clean the pins and check if none of them are pushed in from the side, those PLCC mod boards wreck sockets and chips.
TOS 1.00 is just bad, replace it with 1.04 ASAP, 1.00 is kinda buggy and don't even bother with harddisks on that version, the hardisk routines are seriously borked in 1.00
I have no idea what is in your SF314 or if it even works, but the signals that come from the atari itself are pure Shugart, any active conversion within SF354 or SF314 is to accomodate a non standard drive.
An easy way to check if an atari 16/32 bit machine lives is to check the floppy led or the DS0 line on the YAMAHA, a faint blink indicates some lif in it.
One of the biggest issues with these old Atari's is in my experience. bad contacts on sockets, especially the square plcc sockets, clean them or carefully re-bend them.
If needed replace those pesky staggered sockets with SMD sockets, with some bending of the outer pins you can solder these sockets with smd and through hole at the same time.
The easiest retro way of upgrading the 520st is by piggybacking another set of ramchips on the existing and connect their RAS and CAS lines to the other set at the MMU for a home made 1MB total 520ST+, or you can solder a 72pin socket to the board, it's a tedious job, but a fine DIY way to max it out, ST's need FP-RAM sticks.
Boot speeds relating to floppy drives.
1: diskette with game or bootsie bootsector
2: regular diskette in drive
3: no disk drive attached, takes a second to detect a drive foor bootsector and gives up. Also no A or B drive on desktop.
4: empty drive, takes a long time to give the user time to insert a bootable diskette
Adrian's 520 ST has 1.4 already
@@Motolav I think S B is referring to the 1040 STF, which has TOS 1.0 in this case.
@@TPau65 which isn't Adrian's lol
I still have TOS 1.01 on my machine. You can get TOS 1.04 and boot it from the Floppy if you can't get the ROM. Also you can get the Clock speed up to 16 MHz. but if you want some old games or software to run you should install a switch so you can boot the machine stock without having to remove the mod.
@@kevin34ct You mean TOS 1.0, there wasn't a 1.01. In my 520 STM I have switchable TOS 1.0 and KAOS-TOS. 😁
You gave me quite a start there! I thought, "Oh no! Adrian broke my 1040STf!" :) Getting chips seated back in those PLCC sockets is such a hassle. I've had that kind of problems so many times with the Agnus chips in a Amigas.
The only thing more gut wrenching than bricking a piece of kit is bricking something that was entrusted to you! I felt that wrench even though I was only watching!
Nothing worse then an Atari person touching an Amiga. (The Amiga / Atari wars continue). LOL!
I had a 520 STFM, with a 4MB memory expansion, which used 30 pin SIMMS. I also had a 52MB quantum scsi hard disk, with a DMA host adapter, which cost me two weeks wages back in the day. But it allowed me to play The Secret of Monkey Island, without swapping floppies :-)). Keep up the great work.
Sounds pretty much like my Amiga 500 back at the time. Only I had a 40MB Micropolis drive, and with the SCSI controller I think it was almost three weeks of wages that went into the HDD. And I got that drive at about half the RRP...
HDD's were horribly expensive back then.
"That pin is just high all of the time!"
Welcome to the club
I love the fact that they used the fancy font every time the word deluxe is referenced in the manual
I like that section 24 tells you to use "some small snipers..."
@@merseyviking Makes sense... You do want your snipers to be as inconspicuous as possible.
Glad you did this video. Helped me out with my 520!
Insane how you’re able to troubleshoot without schematics! This is a new level! Good job on the repair!👍
This is a rare talent these days. My electronics instructor said to trace backwards and make your own schematics to the system your troubleshooting. I have not seen many people do this. They buy the original schematics for trouble shooting.
This is one of the most common failures on the ST. Turn the chip upside down and carefully push/bend the pins outward the end of the pin has a stop point on the package. Also bend the socket pins slightly so they put more pressure on the chip.
If that doesn't work the MMU is definitely fried.
Glad you got this sorted, and very cool that another local TH-camr was able to loan you a test machine. I hope you can source a replacement MMU without too much hassle or cost.
Another viewer has already offered to send a working replacement! :D
Really enjoyed seeing the debug session live! Reminds me the old days watching my dad fixing TVs...
Although the 68000 has a 16 bit data bus the memory is byte-addressable which requires separate chip selects for the high and low bytes - hence the two chip select bodge wires. Check the CPU pin diagram and you won't find an address line A0. It is replaced by a pair of upper and lower byte select lines. The 68008 (as used in the Sinclair QL) is basically a cut down 68000 with an 8 bit data bus, fewer address lines and a few missing control lines. I had a 520STF back in the day with an SM124 monochrome monitor and wrote my university dissertation on it (BSc Computing Science). I moved up to an Atari Falcon but after years gathering dust in the loft after I moved to PC's they both ended up in the local refuse dump! 😲 This was long before online auction sites made such machines valuable. I still get cold shivers when I remember dropping the Falcon into the dumpster 😭
You simply can't just wire up the memory with only 8-bits and call it a day on this iteration of the 68000, all 16 bits of the data bus are acquired whenever instructions are fetched, whenever data is accessed that is a word or longword, whenever reset or exception vectors are accessed, whenever any kind of stack operations are performed, etc.
The ST schematic shows that the 68000 data bus isn't wired directly to the RAMs, data from the RAM is latched by 74LS373's, and writes to RAM are buffered by 74LS244's. I'd say this is probably used because without them, the MMU would have to stall the 68000 even for non-RAM accesses while the shifter is retrieving the display memory. The buffers split the data bus so that the 68000 can access ROM or non-Shifter I/O concurrently while the Shifter/MMU is fetching display data.
The ST schematic suggests that the MMU can control two different banks of RAM, but on the 520ST only one bank (of 16-bit-wide data) exists.
My Saturday evening entertainment. 😀
Adrian,anche se non sono un fan della atari,ti dico che sei in gamba,ci sei saltato fuori,bravo
The 1040 ST that appeared part way through is so nostalgic for me. It was the first machine I used that that felt like a "real computer", with its incredible 1MB of RAM and graphical interface.
The MMU and GLUE chips can go bad on ST’s. I’ve fixed a few with bad MMU or GLUE chips. I’ve fixed a 1040STf with a bad GLUE chip, it totally booted, but the screen was all glitching and no green desktop. But it gave the keyboard beeps. Replaced the GLUE chip and was working perfectly again.
The most articulate and informative channel for EE and vintage electronics.
I have absolutely no idea what your doing but by the end of your videos I’m understanding more than I did at the beginning.so thank you for your videos I appreciate your knowledge
I used to work at Chiang Computers in Ottawa, selling Atari and PC computers.
Some of the common repairs my boss Gary would make to Atari computers was cleaning and bending the pins of the sockets and chips to improve connections, and replacing the power supplies. He fixed a lot of Atari computers. Apparently bad solder joints was common on some power supplies.
Seeing this video brings back memories.
I was lucky enough to live near enough to an electrical engineering genius, we had originally piggy backed 520's to 1Megs, and then later he had a single chip with all three tos versions on it, selected by 2 switches set before you booted up, the last thing was a 4 MB ST had to swap out the original memory chips for that one though. And if I remember correctly, the bigger chips had 2 more pins on them so the sockets he used had 2 pins lifted up and hand wired / soldered to complete the modification. Never had a single problem with it.
When I saw 520ST in the video title I had to watch both parts. I enjoyed watching very much. I had both the 520 and the 1040 in my long journey of buying and building gaming computers. Actually 38 in total. The most recent I built from scratch and am enjoying. Just wish I could get a 3080 to replace my 2080 TI, but of course due to covid that's nearly impossible except for the rich folk which I am not. But anyway, thanks for the videos. They sure brought back some memories. Cheers.
I like the thumbnail picture :) It seems like a real smile for something that succeeded, after years of boring C64 repairs some old thing gave Adrian real brainwork :)
I have a lot of Atari, 1-800, 1-512, 2-1024, 1-mega STE, spare floppy drives, many DD, Hand scanner and even the SLM804 printer. all are working fine. 100's of software and most all the magazines and their disks. It takes a 4x8 table to set them all up. Thank you for the post very interesting to me.
I fell in love with the 520ST the day it came out. Then the 1040STF came out.
Was a poor college student and couldn't afford one.
In second year of electrical engineering, I found a job as a computer technician, I mainly worked on IBM PC, Apple II and North Star computers.
Now, at 63, I find myself looking at the schematic and admiring the design of the 1040.
I could never go back to working with one. It would drive me mad.
I'm spoiled having an 8-core 4GHz PC and even now, I wish it was faster!
I was at the Summer Comsumer Electronics Show 1985 in Chicago. I knew of the development of the Commodore Amiga and the Atari ST, so when I saw a flyer advertising the introduction of the ST line in a room upstairs off the main show floor, I went (they were too late to get a spot on the main floor).
In the room, there were two ST models shown: the 260ST and the 520ST. along with some accessories such as an external CD-ROM drive, which they demonstrated. The desktop image looked oddly familiar; I'll get to that in a later post, if there's interest. I recognized Jack Trameil and at least one of his sons at the other end of the room. There weren't a lot of people there, but it was early in the Show. There was not much oganization and I just played around with a 520 for a bit, including a QuantumLink chat with an anonymous stranger, and a short chat with one of the reps. I got some literature and left; I was much more interested in the Amiga.
I've got a spare MMU here in the UK. Yours if you want it.
You can email him about this. You’ll be able to see the email if you access his page on a computer
@@sonicunleashedfan124 Got it. Thanks.
@Adrian's Digital Basement
this guy got one mmu
@@andrewclegg9501 you’re welcome. Hope he gets the MMU
Back in the day we had a trick to bring an ST back to life. Lift the whole unit an inch off the desk, then drop it. I gather this process reseated the chips. A more brutal version involved twisting the whole unit. I can't remember how many times some combination of these techniques saved my ST.
This is great content. Glad you got to the bottom of it. Brings me back to the days when I had a Tandy 1000.
The chillest of retro channels. Always fun.
the ST is such a fun computer... the little green screen and Gem is so iconic.... I run Gem/3 under DOS ... very similar.
Excellent video and work. When your friends Atari didn't work, I sat here feeling your pain. Glad it was a simple fix. See you on your 2nd channel
super cool . i wanted one of these machines pretty badly and i only saw one a single time but i played around with it and gave it a good once over .i ended up with an ibm8088 that had no hd but 2 x 5.25 floppys . i got a roland interface and slaved it to a 1/2 inch 8 track reel to reel . i learned a lot . the pcs where quickly getting faster and you had to know when to not upgrade and wait a minute . eventually i switched over to an all digital system and i am so much happier . a lot of younger guys will adore all the older tech but not me . i like the effeciancy workflow of a good daw with good plugins over tape and $$$$$ outboard gear. very cool retro stuff here !
Glad you nailed the issue with this machine. As others already mentioned, maybe the MMU is still ok and only needs a closer look at the contacts.
The two banks are probably still in parallel, but iirc the MC68000 enables the upper and lower bytes independently, so as to support 8-bit writes without clobbering the other byte. Just guessing though.
One bank is lower bytes, the other is upper bytes. The 68000 CPU has no A0 so only word addresses are possible. It uses UDS and LDS pins to address either lower, upper or both bytes as required.
@@thisnthat3530 Yeah, that.
When you wanted to fix those socketed chips back in the day without opening the machine, you would just lift the computer up about 6 inches and drop it down on the bottom. That would usually reset the chips. This is why they sold retainer clips for the sockets to hold them in place.
I have an early 520 and I can confirm that this was a problem, and this was the official fix.
@@jimgsewell I can remember when the 520's came out and this was the first thing to try if your new ST didn't boot!
I'm sure someone mentioned this but the reason for the two chip selects on the RAM is because of how the 68k works. It can individually access the lower or upper bytes of a word and even has individual "strobe" pins to do that.
The 68000 can access the two 8 but halves of its 16 bit bus independently using two strobes, *UDS and *LDS, I guess that is why those chips are connected in two groups of 8 bits.
The buffer chips are actually there to isolate the 68000 and other chips from the data bus when the shifter wants to use the ram. The Amiga has a similar arrangement
@@jaycee1980 Yeap, I was explaining, with more or less success, why there are two different enable signals, you have to allow the 68000 independent access to the two 8 bit halves. It's not a surprise the Amiga has a similar arrangement, it uses the same CPU. Cheers.
There are two banks of 8 bits to support byte wide addressing. From the processor (68K?) the UDS and LDS signals would be used to determine which 8bits to access.
I used to work for Frontier back in the day and fitted many of those Xtra-RAM units to Ataris. Great fun.
Indeed, the MMU is not proven to be dead (they never died on me in my years). The memory expansion board pushed (cold) in the socket always damaged the pins. They are pushed back. I used a needle to bring them back for a good tension again. The memory expansion board pins are pushing the socket pins back because they are between the socket pins and the MMU pins. (This is a terrible way to do a mod. Hmm plug and play the said at the time) It always ended up in wite screens eventually. Just an addressing problem. So, see if you can restore and polish the MMU socket pins and try again. I did this a lot! 35 years ago, for Atari the Netherlands. I hope it’s working, … I have a lot of original spare parts. I do sell them occasionally. I’ll see if I have the MMU.
Hi Adrian, glad you were able to find the culprit, hope you get a replacement part! Best wishes from Germany, Michael
The Atari ST did not have an 8 bit data bus anywhere and the latches are not there to somehow map 8 bits to 16 on the CPU. It was a full 16 bit bus.
The latches were probably there for buffering timing and managing bus contention.
The reason the RAM is organised in 2 8-bit banks is because the 68k is byte addressable. If you look at the pinout of the 68k there is no A0...just 23 address lines even though it has 24 bits address space. Instead of A0 it has LDS and UDS (lower/upper data select). You can select either data bank OR BOTH at the same time!
In 68k assembler you can only access a 16 bit value from RAM from an EVEN address and both UDS and LDS are asserted. If you do a byte access you can specify even OR odd addresses and only LDS is asserted for even addresses and only UDS for odd addresses. Hence why each 8 bit bank has its own select. One is even bank and the other is odd bank.
Yes others have mentioned this as well. Curious how the Atari 260ST would work (which I assume is this motherboard with half the RAM removed?) When this machine is working again, I should try disabling half the memory to see what it does.
@@adriansdigitalbasement The 260ST had 512K of RAM -- it was a marketing ploy to sell off old 520STs with TOS loaded from disk (the machine have around 256K of free RAM left after TOS had been loaded in from floppy).
All ST ram banks are 16-bit wide, so technically, the ST can support two banks of 128k to support 256K of RAM but no machines were ever made like that (and one bank for a 128K ST -- only seen at the original CES demo).
You can contrast that to the TMS-9900 in the TI-99 which also doesn't have a A15 (TI counts bit the other way round Motorola A0 is TI A15 and TI A0 is Motorola A15) but does not have the equivalent of LDS/UDS. It always reads 16 bits and masks internally. This is particularly bad as it has only 256 byte of 16 bit memory, the rest is 8 bit and each memory access will trigger two sequential reads from memory. This means loading a 16 bit register needs 4 bus accesses (each having 2 wait states to make things even worse).
@@DrSteveBagley The 130ST prototype was likely equipped with 16x 4164 RAM chips, and yeah it looks like the 260ST was just a 512ST without ROM chips for TOS, and IIRC was only badged that way im Europe and for a short time.
@@adriansdigitalbasement I would be curious as well to see if the MMU does some sort of trickery, but I k ow the CPU requires the full 16 bit data bus to be present to function properly. It's like how 386 PCs used 8 bit wise SIMs but needed banks of 4 in the same size to work.
In other 68k systems I've seen with 256k they used 32 4164 chips arranged in 2 banks of 16, and each bank having 2 groups of 8 like here, and in another case there were 8 4464 chips in banks of 4, each bank being divided into pairs.
I love the thumbnail. It looks like you are playing the Atari like a guitar :)
This was thee one I had before the PC... Looking forward to seeing this one!
So far you've made me get my 1040ST and C128 up and running again, now to find the power supply for my CD32 and Vic20 and I'll be back in 8/16 bit heaven...
On the original Atari 520 ST you needed the drive connected to boot up because the OS (GEM & TOS) was loaded from disk. But even without the drive connected you would get an error message. When the OS was moved to ROM, you no longer needed a drive connected to boot the ST, but it was pretty useless without one since almost all the software available for it was disk based.
27:06 Deoxit Fader F5 spray on the knobs can free up the seized knobs. That is how I fixed mine. Same exact monitor
The reason there are two banks of 8 RAM chips is so the CPU can read/write one byte at a time. The 68000 has no A0 pin so can only address words. It has UDS and LDS (upper and lower data strobe) pins for addressing the even and odd byte at a given address and uses the lower or upper 8 bits of the data bus to do so. If it wants to transfer 16 bits, it strobes both pins and uses the full 16 bits of the bus.
It would be interesting to see if (with a bit of hacking) the onboard RAM could be relocated in the address space to have 4.5MB total rather than disabling it.
The PLCC sockets Atari used are notorious for having bad contact.
I have heard it was common back then to try the 5-inch-fix first. That is, you lift your ST flat 5 inches above the table and then drop it. This was supposed to push the chips back into the sockets.
I really enjoyed this look back at the Atari ST. I had a 1040 ST with a hard drive (10 whole megabytes) that I used for years. It brought back memories of when I was one of the Compuserve Sysops for the Atari Forum. On that cartridge port, I used to keep an Action cartridge plugged in there (Action was a popular programming language on the ST).
I wonder if some problem with the memory expansion fried the MMU.
There's a chance that probably that has happened
was wondering myself if the simple wiring chip enable hack could have damaged the mmu over time
I used to have one of these machines. One of the things these were notorious for was the chips in the square sockets would, due to thermal cycling, begin to work their way out of the sockets causing random issues. One of the "official" service bulletins was to hold the computer about 7 inches above the bench and drop it. That would reseat the chips! LOL.
There are also metal clips available that hook on to opposite corners of the socket and apply pressure in the middle of the chip.
@@dnwheeler My Atari ST had those.
The Apple III had the same issue, and solution.
I still am getting sick that I ever sold my ST 1040 (F) for a few guilders (about 100, should be 50 euro's now). That machine learned me to program with GFA Basic, finding out what MIDI is, how to cheat a brick game etcetera. And tham I bought my first PC, a 8086 with a 50 Mb HD. Man, I did not know what to do with it, until I got a copy of WP 5.1. That worked.
But I still miss that ST 1040. Today I own a laptop, my 4th, and before that some PC's (tower and desktops) about 5 to 7, I don't really know anymore.
But I still miss the ST 1040....
(and to be honest, I still have an Sinclair Spectrum, with a unit giving it a diskdrive and parallel printerport, and a video out (coax).)
z)So, Adrian, in case you are interessted to find out how a Sinclair works? I can send it to you. (Oh, 220 V, 50 Hz)
Just wanted to point out that the Motorola 68000 is a 32-bit processor, with 32-bit instructions, but with a 16-bit data bus, and 24-bit adress bus. Motorola called it a 16/32-bit processor.
16/32-bit processor, hence the name of the ST.
24:58 not true, the ST works without the keyboard. Just makes annoying key click sounds!
Your reaction at 7:16 kind of caught me off guard. To me, that is the normal style of PLCC socket, seeing anything else would be strange.
Congrats on getting the 520 ST working. The only thing I liked about about my 1040 ST was the 1MB of ram and the 68000.
The Mega ST line was the follow up to the 520/1040 There was the Mega ST1 and the Mega ST2 -- I owned a Mega ST2 back in the 90-94 timeline. The number indicated the megabytes in the system. They were pizza boxes with separate keyboards. I think they came with an integrated HD but it's so long ago I'm not entirely sure. (Edit: Nope. No internal HDs. Did some googling...)
Main reason for buying these were that (a) Apples cost so much more and (b) desktop publishing. Yeah, everyone associates the mac with desktop pub, but there were a few decent programs for the Atari ST line that worked just fine. We had a few 1040's at my work connected to an apple Laser -- whatever their first laserjet was called, and used it a ton.
Mega ST2 and ST4 came first then a little bit later the ST1. The Mega ST's were sold as "professional" machines and therefore would need more memory. They realized only later that there was demand for the only 1 MiB version.
They hadn't an integrated harddisk. They had another pizza-box with harddisk called Megafile 20, 30 and 60.
The Mega ST's had a better keyboard (real keys using cherry MX black ), blitter integrated and a battery backed realtime clock. They also had an extension connector with the CPU bus for easy extension.
I still have mine with integrated PC-Speed emulator and Volksfarben ET4000 VGA card.
@@galier2 They where also the first STs where the Bitter was installed from the factory.
Nice to see at some are so deep in those:) I have Atari ST 520, ST 1040 and Mega ST4 at early days.
Wow! In 1984 I bought an Atari 800XL, instead of a hard drive it had a cassette tape machine for storage.
I wish I could get my hands on a decent ST. Maybe one of the later ones with more memory, built in floppy and PS. I had an original 520ST that I expanded to 1024KB by the piggy-back method. Did some awesome quality hours on that machine. Mostly working with N-dimensional intersections with 3D space .. for instance, how would a 4D object appear if it was inserted into a 3D space. Think about sticking a 3D sphere into a 2D space. It would appear as either a point or a circle, depending on how far it was inserted into the 2D space. Now think of what a random 3D object, maybe an odd shaped rock, would look if ti was inserted into the same 2D space. Now, think about 4D objects in 3D space.
The STe machines are good. Used to have a couple of those, wish I still did. They had a later TOS versions, more colours available & also used 30 pin simms so ram upgrades were a lot easier than previous versions.
I'd love either an ST or an Amiga.
@@CJ-rf9jm yeah we got an 520STE but a few of our friends games we borrowed that ran fine on STFM wouldn't work on the STE - so we got the store to give us a STFM instead. This was a really dumb thing to do when we tried to upgrade the RAM an saw we had to solder wires and mess about with jumpers etc... And we learned that the STE could pretty easily be upgraded to an insane 4MB just with socket SIMMs. There was a small amount of software incompatibility on STE early on - but most developers starting making sure their new releases worked on both models. And we never did fit our RAM upgrade.
@@adz693 I had a 1040STe from 1990 onwards that I'd later upgraded. I never had any incompatilibility problems. That store absolutely ripped you off badly. The problem were those badly cracked games. I'm not against cracked software when it's no longer available except those who do it so badly they end up screwing up the program in the process of cracking it.
You can never find a proper 4D object when you want one... not even on eBay. Apparently they are incompatible with 3D packaging.
It's 2 8bit banks because the 68000 can also access only one half of its 16bit data bus. For 16bit access, both banks are used at once (you actually can't do 16bit accesses on odd memory adresses, this started only to be possible with the 68020, iirc).
The reason for needing the keyboard is that its got an embedded z80 on board and provides the system clock. If you add a battery backup to the clock on the keyboard then it'll survive power off
Seconded and thirded comments about the sockets. They were a notorious failure point on all 1980s machines. Pull and clean everything
Nice job Adrian, was a great moment when the machine booted. 😆🖒
Waiting on part 3 - Episode: A new MMU
I just wanna see retrobriting on that keyboard. I can't bear it 😂
Very good troubleshooting. You nearly reverse engineered how the ST works. :-) Bad MMU is not that common. I never had an ST with a bad MMU. But the PLCC sockets of the MMU and the GLUE are known to have contact problems from time to time. Especially if a RAM expansion was squeezed between chips and socket. I would suggest to take a closer look at the pins of the bad MMU with a microscope. Maybe some of them are not making good contact and the MMU from the 1040 is working because it has them still in order.
I was thinking this same thing. As you say, check all the pins on the MMU.
For a time I worked doing warranty repairs for both Atari and Commodore. I saw pretty few Atari ST machines coming in but a lot more Amiga. I remember that for a time Commodore had big problems with those PLCC sockets and chips not getting good contact. A lot of this was traced to the factory where completed and boxed machines got bumped around when the pallets were loaded onto the trucks. To get to the loading docks the forklifts drove over some very uneven parts of the floor. As I remember it they described it as all the cartons on the pallet going airborne only to slam down after free falling more than an inch, but that could be my memory playing tricks. Apparently that was enough to shift the chips just a little in the sockets and if you were unlucky it resulted in the machine failing. Often just pressing all the chips firmly into the sockets could get the up and running again. Even after addressing that problem with the forklifts Commodore choose to install a retention spring across the the sockets and it cut down on the number of DOA and early warranty returns.
If you get enough lights and a proper tub, retrobrighting inside usually works better than trying to do it outdoors. :)
FLYNN LIVES!
Odd Tinkering, Tysy Tube, The 8-Bit Guy, among others I've forgotten have proved that.
Lol I stand corrected! I guess the original ST can boot, sort of, without the keyboard. Is the IKBD chip on the motherboard on these? The stuck key thing happens when there is an error with the keyboard controller, it also happens when a joystick is plugged in with the autofire on after a reboot. Or some times just randomly. Hitting a key will stop the beeping in the latter case.
Cool video thanks for making it. I still have a Atari 1040ste modded to have more memory.
IT still works fine except the floppy as it needs cleaning from smokers.
This is our new Bob Ross ☺️
The Atari 520 ST had a real problem with chips becoming unsocketed. At least for a while, the official fix was to drop the computer 4-6 inches onto the table. I agree that it sounds sketchy, it was. My point being you might want to reseat every chip that you can. I have an original 520 ST, and also a 1040 that I bought years later.
When the 1040 ST failed to boot after you replaced the MMU, a wave of dread came over me. LOL.
I added 520kb to my 520STm with an upgrade that piggy backed the MMU and something else, probably like this upgrade was done. Can't remember where I bought it but it was through a Swedish computer magazine. It was a lot of money for me but that extra sound and graphic effects in Falcon was worth it!
Never heard the term "bodge wire" before that engineer in Australia used it. Here in the states, I though we always just called them "jumper wires"?
Adrian, you could place a DPDT switch in place of the mod jumpers and chip select traces on the memory to effectively enable / disable the memory with the flip of a switch.
Mega STs are desktop case STs w/ Blitter chip and real time clock (in 2MB and 4MB models) added. Some of the later STf/STfm models have metal clips over the PLCC chips because it was a known issue they could get a little loose during shipping. An official 'fix' from Atari at the time for new units not booting was to drop the machines (as flatly as possible) from 6-8" onto a hard surface to reseat those chips w/o having to open up the case. :-) EXXOS Atari store in UK has MMUs. Not sure about US sources (Best Electronics or B&C (myatari on eBay)?).
The PLCC chips could also get loose due to thermal creep over a period of months. I had to reseat my MMU and GLUE regularly in my early 520STM, but it never gave me any long-term problems.
Well done Sir, good work!
I believe if you examine the memory upgrade, and the little board that has pins sticking out of it that fits on top of the MMU, you will see that one or more of the pins has one side of the pin as bare metal, the the other side covered in epoxy or some type of insulating material. If that insulating material rubbed off onto the pin of the MMU, it might be preventing that pin from making contact with the pin in the socket. Also, I believe the MMU was a custom Atari chip.
an interconnection line is cut off on the board between the chip SN74LS373N ( socket U22 ) and 5E1 R HD46850P ( socket U21 ) in the middle
"That pin is high all the time"
Hey, leave that pin alone. Times are tough.
We all have our own coping mechanism. :)
Interestingly I repaired two Rev B ST's recently and it was the latches you were checking out that was the issue. Those PLCC sockets go bad by the way. Because that memory upgrade was pushed onto the original MMU it could have forced the pins in too far... you could try pulling them out a little.
Yeah I've heard that about the PLCC sockets from a few people -- I inspected it on my microscope (not seen on the video) and it was actually OK -- along with all the pins on the original MMU. Seems the Marpen pins are thin enough to not cause an issue, at least in this case.
Had both latches failed on the machine you repaired?
My 1040ST got me through college with zero issues. Replaced it with a PowerMac 6100 66 DOS that I used until 2005. Two computers in sixteen years, until internet and software demands forced me to join the 21st Century. Simpler times.
RAM is actually organized as you thought it was. The chips correspond to bits 0-15. Just notice, that RAM is addressed with bytes, not words, so it's 16 bits per word, which occupies two consecutive bytes. By the way, it's exactly the same on an Amiga 500.
And, if you plan to connect your STF to OSSC you will probably to modify the mobo a bit filling the elements in the modulator area to obtain a proper synchronization signal for SCART connection.
Has anyone noticed that SysInfo (30:26) only updates the time every 2 seconds?
Tgettime (GEMDOS 44) returns the system time with a granularity of 2s.
I use a full spectrum grow light to retrobrite indoors and it works AMAZING. So much faster than outdoors. Just suspend it over your container of water/hydrogen peroxide and you are good to go.
5:30 Shouldn't it be: pulling the chip select to 5V?
I would be cautious with that memory expansion, give it a once over. Sounds like something might have shorted on it and sent 5v down the data line to kill the MMU. That hot glue that was on there (I think I remember the simms being hot snotted on that expansion) is a huge red flag to me
I agree with yours and others who pointed out that the RAM expansion board is suspect. It wasn't made for the 520ST anyway, so why is he trying to make it work with it if it wasn't meant to be used with it in the first place?
He should really look at the Exxos or thff RAM expansion boards which are much smaller (only uses 2 or 4 chips) which produces less heat too. The designers also say they work with the original 520ST models.
That's very interesting. I wonder if the keyboard not being plugged in, is only true on some versions of the board, because I know I couldn't get mine to boot up without it plugged in. It also could be one of those strange coincidences as well, because my system has other faults as well where it won't boot up at all unless I push down all the chips. Probably have a cold solder joint or a bad socket somewhere I need to replace.
I know some versions of TOS are crashing because they want to set the system time in the keyboard controller and fail. It's not the board that is the cause but the TOS version.
@@boelwerkr Good call. I have rather 1.03 or 1.3 (can't remember which off the top of my head) in mine, which seems to be newer than the one that Adrian is playing with.
"That Pin is High all the Time"..., Was I the only one thinking "So am I" lol?
I Think the 2 Chip Enable Lines are for accessing the high and low byte independently.
Yes Something like that. The RAM uses a Column/Row address select (CAS/RAS) schema and the MMU in the ST supports 4 CAS and 2 RAS lines.
The person installing the 4Mb simply disconnected the RAS lines but not the CAS and address lines. This caused an overload on these lines an one line driver in the MMU burned out.
MMU Chip: "...and I would've gotten away with it too if it hadn't been for you meddling hackers!"
This is a fun channel. Great stuff.
Adrian, i love this stuff. i am a collector also. if you have anything you dont want let me know! will take it!
I might have my old Atari ST Mega 4 around somewhere. That was the ultimate ST that ever made proper production.
Logic probe! Didnt know they existed. Getting one for sure, thanks for a great video!
interesting! I remember when I was a kid (in the ~90s) I had one of these, I found it on a dumpster. I brought it back and it booted perfectly fine.
7:10 I don't think I have ever seen a PLCC socket that did NOT have that zig-zag pin layout.. no idea why you consider that so unusual.
Macs, PCs and Amigas all use the pins in rows. In fact, you can't even buy the zig zag type other than pulls from Atari motherboards. The normal type is readily available from every source.
The software that was used to layout the ST boards worked better with the zig-zag socket, so they was used. This type of sockets is not often used otherwise.
@@adriansdigitalbasement Which limits the options when it comes to replacing the socket.
hey Boss what sorta mic r u using there - is that a bluetooth one? or wifi....great vid btw
Adrian - very interested in how you go about retrobrighting the ST. I've tried on a 1040ste case using 40 Cream + CA sun and it came out streaky and blotchy all over. Air bubbles under the plastic wrap seemed to get "ghosted" onto the plastic as well.
Did you watch how the 8bit guy is retrobrighting? He does it quite often and has usually good results.
Yeah, from what I’ve seen from the 8Bit Guy, plastic wrap always comes with streaks, he always did better with a tub of solution, and weighting the pieces down to keep them submerged (often a zip lock full of the liquid on top is enough)
@@nichderjeniche Yeah - I've seen all his vids as well as Adrian's, Jan Beta's and others. I've used the 40 Cream method on other (smaller) pieces successfully.
@@Underestimated37 That is what I am going to try next (liquid in a tub). Not sure if this case is salvageable but at this point, will keep experimenting to see what happens. Worst case, I may take the path that Noel's Retro Lab did with a C64 and paint it custom. I've been pondering if I could reproduce the 8-bit Atari XL color scheme on it... LOL.
There has gotta be a better way, maybe there needs to be some research into creating a ‘gel’ using something like polyacrylamide (the stuff in diapers and water beads) that can retain the wetting properties and transparency but lose the texturization that the plastic wrap causes.