NCR Tower 32/650 Restoration
Tower650 has been giving problems since about 2005: disk errors and a weird power-on problem (the machine turned on only sometimes, and some other times it suddenly shut off). Then we moved from the Freaknet Medialab base in Catania to the Poetry Hacklab in Palazzolo Acreide. It never power on anymore since then...
Some months ago we started some testing to the machine; the power supply suffered of a strange problem, so we decided to try fixing it.
In April we made some tests, we noticed a "tick tick" on a relay inside this power supply, and so we decided to check it.
Unfortunately, while dismounting the relay we broke it, so we ordered a spare part. While waiting for the shipping, we decided to simulate the relay using a manual switch.
So, we turned the power supply on... nothing. Then we switched, closing the contact on the relay, trying another power on ... *BOOOOM!*
An entire bank of power transistor exploded loudly!
Checking the circuit we noticed that the relay does a very weird trick: it shorts the main 220V phases directly to the transistors making what we think is a ... massive short circuit! Why this? Maybe this relay was only used to "short circuit" the bank of electrolytic condensers.... BOH! The only thing that's sure for now is that everything blow out with smoke and flames.
New transistors was then ordered and after about 2 weeks they was changed, the relay was installed, and the power supply was restored. But ... no way to make it work: it seemed totally dead. A great problem, because now I can't check it to copy the output voltage, to simply adapt a new ATX power supply to it!
The power supply box talks about +5, +12, -5 and -12 Volt, plus some test points. So ...
My decision was taken: I have to do a massive reverse-engineering of the Power Supply to adapt a normal ATX power supply to Tower650. I suddenly dismounted the power supply and started analysing the circuit, looking for IC on google, checking connections with the tester, and so on...
The power supply is a very strange one: it's composed by 2 big and complicated boards.
The first is the power supply itself:
that generates +5V and +12V, there are very BIG diodes and other power components on it.
The second is a charging circuit for a 6V battery, used as a sort of internal UPS for the RAM boards:
So when the main power fails, the entire machine will shut down BUT the RAM cards will be powered by the battery. Also, this board generates some power lines (-5V, -12V), and a bunch of signals that go to the BUS or come from it (i had to discover this later ) These 2 boards are interconnected by a flat cable, complicating the search of signals going (or coming!) to/from all boards to the tower650 backplane and CPU board.
A first check on the first board and to his flat connector was useful to find +5V and +12V, aka the mail power line for the hard drives and magtape connectors.
The power supply box talks about +5, +12, -5 and -12 Volt, plus RST and DOWN signals. So I started taking out connectors and checking connections between the two boards and to the main backplane.
After 2 hours, here was an early version of schematic:
Now it comes the hard game: We must be *SURE* about the -12V, the -5V, and the nature of the other signals: where do they come from? And where do they must go? What kind of signals are they? Data signals or power supply? And, do they come FROM the backplane TO the power supply logic board, or do they come FROM the P.S. to the backplane? We *NEED* to be sure about that, because a mistake can FRY all the electronics.
Checking the second PS board, I happily discovered the -5V source:
Here we have a LM337 negative voltage regulator connected directly to the -5V test point on the board. This was pretty simple to find. After that, I checked the -12V from the SN75188 serial line driver on one of the serial board, thru the backplane, reaching the correspondent pin on the power supply.
The hard work was to reach every other signal from/to the P.S. to/from the system backplane. This required an entire day, leading to this kind of "simplified" schematics for interconnection between the Power Supply and the connectors on the Backplane:
This scheme was an early version, showing just how the lines are connected, and it was debugged in the next day, discovering the nature of every signal, the connection of every pin to the backplane and also how the CPU main board "sense" the power supply, with a bunch of components around them, just to be sure
The "C" signal required some additional reverse engineering work to fully understand if I must consider it as an INPUT signal coming from the PS to the backplane, or an Output one, from the CPU/Backplane to the Power Supply:
That pin of LM339 (voltage comparator) is an output one, used to trig the base of 2N3904 transistor, in which the C signal enter from the collector. So I'm almost sure that this is a signal coming from the Backplane to the Power Supply. The 555 supply a kind of clock, but here I become a bit confused and i DECIDED that this is an OUTPUT signal from the backplane. Maybe it was used to shutdown the computer?
After all this, I decided to start the BIG JOB, connecting a good ATX PC Power Supply to the electronics of Tower650. Got a 50pin scsi connector from trash, scissors, solder and so on ... after a while I was ready for the first test, that was positive
Checking voltages on the backplane was OK, the main CPU will turn ON! NO SMOKE, NO FIRE, NO BANG! FIRST PART OF THE MISSION WAS ACCOMPLISHED, GOOD JOB!!
After this I made all the other test. I made a console cable and connected Tower650 at a PC serial port. Connecting and checking SCSI hard disk was a REAL MESS! I had EVERY KIND OF PROBLEM but after all I finally obtained a working system, and the taste of victory was so sweet.
The final work was to adapt the ATX circuit inside the old original power supply case, the one that you can see in the previous image, the gold monolyth on the table
As you can see above, this was also time to do some wiring optimizations, to connect the ATX Power On to the original ON/OFF button of the main computer, and to finally have a closed and full working system. Now it's online, full working, and it "sounds" exactly as the old power supply because I connected all the old fans to the main 12V line.
- asbesto, June 6, 2008, Poetry Hacklab