apratt@isrnix.UUCP (Allan Pratt) (02/02/86)
SUMMARY: My upgrade didn't work at first, then did, and I'm not sure what I fixed. THE INSTURUCTIONS I FOLLOWED NUMBERED THE PINS OF THE MMU BACKWARDS! The correct numbering is silkscreened in white around the chip. My ST didn't have any metal shield around the RAMs, and the MMU and Glue chips were not soldered in, but there was a blue MOD stamp on the box, and there were some fixes inside. For even moderately-experienced hardware hackers, this upgrade is fairly easy, and worth it. I recently upgraded my 520ST into a 520ST+, and I would like to share my experiences with the net. When I received my 520ST by UPS, the first thing I noticed was the word MOD stamped in blue on the outside of the CPU box. I don't know what that means, but I would guess that it means my ST has been somehow modified. Just below the word MOD is the sticker with the serial number of the ST, and on that sticker there's the notation "7 1985" in the same color of blue as the MOD stamp. If anybody has any clues about this, please let me know. I ordered 20 Hitachi 150ns 256Kx1 RAMs from Microprocessors Unlimited, and was, of course, impressed with their performance: all the good things you've heard about them are true. They took my VISA card number when I ordered, so they could ship my RAMs right away, and also sent a stamped return envelope that I could use to send them payment by check. More companies should do this -- it speeds the process tremendously. I ordered the chips on Monday, and they arrived Thursday. By Friday morning, I was in the lab on campus with my foil-covered desk, solder-vacuum, and high hopes. Nothing like having the right tools for the job: the vacuum-operated de-solderer made removing the capacitors a breeze. When I got my mother board out of the casing, I looked it over for any MODs I might find. I saw some obvious hand-work around the MMU chip - on the pins to the left of Pin 1 as I looked at the solder side. There was also some blue solder mask in that area, but I could not see any actual modifications; maybe they just had to resolder that part of the socket. There was also a resistor which had obviously been installed after manufacture: it was in the upper-left quadrant on the component side, near the CPU, and came from one side of a capacitor to ground. Both sides of the capacitor in question were still soldered in place. Finally, I noticed a capacitor which had apparently been replaced, in the middle of the row of capacitors which parallels the mouse and joystick ports. Also, my board was Rev B, and my serial number is A1-102-6982. If anybody knows the meaning of all this, please clue the rest of us in. Thanks. Piggybacking the RAMs was not rough, but be sure to use a narrow soldering tip: there's not much room between chips, and you have to solder in that channel. Here's a tip: bend the pins of the new RAMs so they come out not-quite-straight down from the package, then seat them on the old ones: you get a decent electrical connection just by the friction, and the solder just has to hold everything together. (Somebody please post something if sliding the pins of the new RAMs over the old ones builds up dangerous static charge.) The first four chips I soldered on had the wrong pins bent up: I am used to seeing the solder side of ICs, and I counted backwards. I had to de-solder the correct pins and bend and solder the wrong ones... It turned out later that one of the chips involved was bad, so maybe this is when I blew it. The rest I did correctly, and it was really no trouble. Once the pins are pressed against each other, just add enough solder to flow, and it will work fine. Wiring the busses across the tops of the chips was easy, once I bent the pins almost vertical. If your bus wire is thick enough, you can weave it between the pins to get some tension while you solder. Weave the wire between the pins before soldering any of them, and press down on it so it rests on the chips. Solder the whole Pin 15 (or is it Pin 4?) bus at once, and then cut it in the middle after you're finished: you get fewer loose ends that way. The instructions I had called for six resistors, and it took me a couple of minutes to find all the places to connect them. What the instructions don't tell you is that the three lower-bank bus control signals come out of the MMU chip together, three traces on the component side of the board. I found that I could slip some spaghetti (insulation from a stripped wire) over the leads of the resistors to get insulated leads, and these worked well for me. In fact, most resistors have long-enough leads to make those three connections: cut the traces on the top of the board, but solder the resistors on the solder side. You will need low-profile reistors for this: the cylindrical ones are too fat. Use the barbell-shaped ones. As instructed, I connected the upper-bank bus control pins of the MMU to the busses in the middle, not at one end, for a T shape. This is just as easy, and seems to be a good noise-reducer. I soldered the resistors to the pins from the MMU socket, then soldered wires to the ends of the resistors, and finally brought the wires up through the holes in the PCB to connect to the busses on the component side. Leave a little slack in these wires, because they have to get out of the way of the plastic case which comes up through those holes. Finally, I was ready to test it. I put some paper down over my foil, connected all the cables (Atari returns us to the Age of Rat's Nest Computers!) and powered up. Glory be, she booted! But my free-ram desk accessory showed the same number as before, not 1/2 meg more! So the top bank hadn't passed the memory test. I got lucky: I used the memory editor I got from bammi@case, and it demonstrated a one-bit error in the upper bank. Bit Twelve of each word was always zero. Fortunately, the instructions I had told which chip was which bit, so I could pinpoint the problem: it was one of the chips I'd made the counting mistake with. I removed that chip from the two busses, and de-soldered all its pins. The best way to do this is to get a sturdy hook of some kind (the lead from a resistor bent into a hook is good), catch the pin of the upper pin in the hook, and pull while applying the soldering iron. The pin should come free, and when you've de-soldered them all, the chip will rattle in its position. It was a little tricky to get it out without disturbing the bus wires above it, but I managed it. Then in with the new chip, and power up, and it booted, and ... GEM still recognized only 1/2 meg. This time, when I got into the memory editor, each reference to the upper 1/2 meg gave a bus fault of some kind (lots of mushroom clouds; more than my list describes). I'm glad this didn't happen the first time, because I would never have been able to pinpoint the problem. This time, I knew which chip was suspect. The trouble is, I don't know how I fixed it. I re-soldered the pins of that suspect chip, and it didn't help. I poked at the bus wires and resistors, and it didn't help. But then, magically, it worked! I have no idea what I fixed, but I can only imagine that my futzing with it broke some short I had introduced. It hasn't given a lick of trouble since then (two weeks ago or more), so my faith is growing. In conclusion, I suggest that rank amateurs not do this, but that middleweights should give it a try -- especially if they have an expert or two around whom they can ask if they get in trouble. For experts, this should be a piece of cake. In any event, you don't have to pay $150 for somebody else to do it; just $50 for the chips and a few hours of your time. Naturally, if you have the bucks, and don't want to take your ST into your own hands, find a pro, or a friend who has done it. -- Allan Pratt ...ihnp4!inuxc!iubugs!isrnix!apratt