Debugging the HP37201 HPIB Extender

The HP37201 is a unit that allows an HPIB (GPIB, IEEE-488) bus to be sent over a serial connection. It supports various forms of the latter, including asynchronous and synchronous modems and twisted-pair cable. Physically, it's a 2U rackmount unit. To get inside, remove the top and bottom panels with the captive screws at the back. The right side panel can then be removed in much the same way. To get the left side panel off, remove the 2 screws at the ends of the handle, the plastic covers under them, and the handle itself. The panel then slides off towards the back of the instrument. Inside there is one large PCB filling most of the case. It contains a 6800 processor, 8K of frimware EPROMs (either 8 off 2708, 4 off 2716 or 2 off 2732, sleected by soldered links on the board, mine has the middle configuration), 1K of RAM in 2 off 2114, a 6850 async serial chip, a 6852 sync serial chip 5 off 6821 PIAs (1 for the diagnostic connector and front panel LEDs, one for the GPIB data lines, one for the GPIB handshake/control lines, one for the internal configuration switches and GPIB control, and one for serial control and handshake lines), GPIB buffers, RS232 buffers, and a bit of TTL. The latter is mostly for the encoder/decoder for the twisted pair connector, but of course there's a little for the GPIB port, system address decoders, microprocessor address buffers, etc. The PSU is on the same board, but HP have been kind to people like me who like to test it on dummy (or no) load first. There is a 12 pin (6 pin double-sided) edge connector on the PCB. Some of the pins go to the PSU outputs, others carry the power lines to the rest of the machine. In normal operation, a little jumper PCB (no components, or even soldered connections) is plugged in there. It can be removed to test the PSU without the rest of the machine conencted. Incidentally, the marked PSU testpoints are on the PSU side of this connector, and are thus useful when the jumper is removed. The only other things in the case are the mains transfomer and associated input connector and power switch, and a 'DTR OFF' changeover switch that connects the DTR pin on the RS232 connector either to the DTR buffer output or to another buffer output that is permanently at -12V. There are 2 DIP switches and a rotary switch on the PCB. The latter selects the transmission mode/baud rate and self-tests. OK, so what did mine do. Well, in the process of fully dismantling and reassembling it (not to be recomended if you don't like fiddling tiny nuts into even smaller spaces -- the connectors are all soldered to the mainboard and have to be unbolted from the back panel to get the board out), I'd removed the 3 fuses from the PSU section of the PCB. There's a 3A fusr for the +5V line, a 0.5A fuse for the +12V line, and another 0.5A fuse for both the -12V and -5V lines. I put them back in suitable places and powered up with the jumper pulled. The +5V and -ve lines were fine, the +12V line was missing. I quickly traced this to a blown fuse. Since I'd not seen it blow when I flipped the power switch, I replaced it, the +12V line now came up correctly. I fitted the jumper and powered up again. The -ve supply fuse flashed and failed. I suspect.actually, the fuse I had put in the +12V holder had originally come from the -ve one, it had failed long before I started fiddling with things. A bit of checking showed that there was a short from the -5V line to ground, and that a 75110 buffer (line driver for the twisted pair cable) was getting rather hot. I desoldered it, replaced the fuse, and tried again. This time the fuse held, and the instrument should work for other-than-TP links without this buffer. Alas it didn't seem to be working. None of the LEDs came on other than the power light (which simply runs from the 5V line). Fiddling with the CTS and DSR lines had no effect on the associated LEDs (note : The 'link' is in software, the RS232 lines are read via pins on one PIA, the LEds are driven by another). More worrying still, at least one of the PIAs was not being initialised, a pin that drives a select input on a mux to set the async baud rate [1] was left configued as an input. [1] The async baud rates are quite clever. The unit can officially do 150, 300, 600, 1200 baud. The 6850 can do either /16 or /64 between the clock input and the internal bit clock, so one clock input frequency will do for 300 and 1200, half of it would do for 150 and 600. There's a mux that selects the appropriate output of a counter depending on the state of a PIA pin. There's also a soldered link that would appear to double all the baud rates, I've not tried it yet. I twiddled the rotary switch and powered up again. Now the first interesting thing was that in 2 of the self-test positions I did get some front panel LEDs on. So the CPU was clearly running. It could do enough to intialise the PIAs, read the configuration switches, and drive the LEDs. I'd already pulled the socketed EPROMs and dumped their contents. I couldn't be sure it was correct, obviously, but they looked sane. One of the test positions was clearly a memory test (I forget the labelling). Now, for very good reasons, namely that they were 2114s, I suspected the RAMs. Getting to them is not easy, they're at the very front of the board, partly hidden by the front casting. So I had to get that out of the way first. Disconnect the DTR OFF wires at the back of the PCB (note the order, they are not marked on the PCB). Pull the cable through the grommet, undo the 2 screws and nyloc nuts holding the cable clips to the right side casting. Prise off the trim strip at the top of the front casting, remove the 6 screws holding the front panel. Pull that forwards, remove the 2 screws holding the mains swtich to the back of it, then remove the front panel. Don't worry about the lack of the DTR OFF switch, you don't need the DTR pin to sort out the memory. With the front panel out of the way, the front casting comes off with the 4 corner screws.. OK, I desoldered the RAM, and fitted 18 pin DIL sockets. With no RAM chips at all, I got a different set of LEDs on the memory test (interesting, at least it was doing somethign with the RAM). Putting the RAMs back gave the old pattern, putting them back the other way round (swapping which chip stored each nybble) gave a different pattern. Had the RAM been working correctly, swapping them would have made no difference of course. So it was likely I had a RAM fault. I found an old Apple 80 column card with some socketed 2114's on it. Pulled a couple and put them into the 37201. Powered up. Got yet another patten from the memory trst. The self-test now gave a patten of flashing LEDs (test-complete alternating with all the others). The 'operating' modes now gave a single LED on, which indicated 'remote signal lost' -- not suprising as I had nothing connected to it. More usefully, the RS232 handshake pins now did affect the frontpanel LEDs, the baud-select pin from the PIA was now driven correctly. Trying the origianl RAMs back one at a time proved that only one was dead. Still to do : Get a new 75110 driver chip for the TP interface. And investigate the lowest hardware test, which doesn't even use the firmware ROMs, with a logic analyser