The Museum of HP Calculators

HP Forum Archive 08

[ Return to Index | Top of Index ]

HP 9820A
Message #1 Posted by on 9 July 2002, 3:13 a.m.

I've just received a 9820A I bought in Germany.

The unit seems to be in good condition, lest the "5" key is missing.

When I plug it and I turn it on, only the fan works (quite silently) and nothing appears on the screen. I opened it and noticed that 2 small and dimmed lights burn close to the card reader.

I have to this stage these questions : - does anyone have an idea of what I could do or check to have my unit work ? - does anyone know where I could find som magnetic cards ? - can anyone service this unit (preferrably in Europe, for shipping is quite expensive (unit weights 18 kg)) ?


Re: HP 9820A
Message #2 Posted by Tony Duell (UK) on 9 July 2002, 2:25 p.m.,
in response to message #1 by

I don't know how similar this machine is to the HP9810, but I think some boards (at least) are the same. I have a 9810 in bits on the bench at the moment, you see, and anything I say below relates to that machine. Firstly, what test equipment do you have? The processor in the 98x0 series is a bit-serial design (mostly...) so it's hard to do any serious logic debugging without a logic analyser. The display on these machines is driven by the main CPU (which has to select each digit and turn on the right segments -- presumanbly in the 9820 it selects a column of dots and turns on the right dots). So unless the processor, RAM, ROM, and memory controllers are all working, you won't get a display. The lights in the card reader are normal -- there should be 4 of them, 2 on top and 2 hidden inside. These are part of optical sensors to detect the card, etc. The fact that they're appearing indicates that there's something on the main +5V line, but you still need to check the PSU voltages. And that would be my first test. On the 9810 there are clearly marked testpoints on the PSU boards for all the voltages apart from the +5V rail. Check the lot. The machine will not work if any of the supplies are missing (OK, the +24V one is only for the printer, I think). The +/-12V lines are needed by the firmware ROMs, the +16V and +20V lines are needed by the RAM. Then check the +5V rail at the power pins of the convenient chip -- on the display board or keyboard encoder, for example. If all the power lines are OK< next unplug the keyboard connector from the main backplane, remove the printer (if fitted -- does the 9820 have this option?) and the card reader interface. None of those are needed to get a display, and one could be locking up the I/O bus. Still noting? If you have a 'scope or logic analyser, look for activity on the pins of the test connector on the 09810-55613 board (CPU control/microcode). Those pins are bascially the microcode program counter, the microcode address change bits, and the condition flag. If you see changing logic levels there, then at least the CPU clock and microcode program counter are running. After that, we need to get into _serious_ troubleshooting...

Re: HP 9820A
Message #3 Posted by on 10 July 2002, 4:37 a.m.,
in response to message #2 by Tony Duell (UK)

Thanks !

I'll check with a voltmeter the different supplies and if unsuccessful will disconect the reader and the printer... That's all I'm able to do, since I've got _no_ electronics/electrics education or skills at all !

Re: HP 9820A
Message #4 Posted by Achim on 13 July 2002, 12:27 p.m.,
in response to message #3 by

I've repaired some 9810's which also where apparently dead. They had serious electronic defects on the logic boards and it took some weekends to repair. To do this you need test equipment (e.g. a logic comparator HP-5011T), the schematics, and deep understanding of TTL logic. And last but not least a "licence to solder". A second, working machine is also of great help. Greatings from germany, Achim

Re: HP 9820A
Message #5 Posted by Tony Duell (UK) on 13 July 2002, 4:50 p.m.,
in response to message #4 by Achim

I've repaired logic faults in the 9810 as well, and it's not totally trivial :-). The 9810 is (mostly) a bit-serial machine which means you _need_ a storage 'scope or logic analyser to grab one of the serial signals and look at the individual bits. If you don't have a way to do that, you're working 'blind'. The 3 most important tools in repairing a 9810 are a (simple) logic analyser (HP's LogicDart is excellent if you can find one), a good soldering iron, and even if I say so myself, a brain. You need a good understanding of how the circuit _should_ work, and what the signals are telling you about the fault. Given that, it's not impossible, though. The last fault I had was the keyboard LEDs malfunctioning, and some of the mode select keys doing the wrong things. I decided to attack the former fault first. After I verified that the latch strobe signal for the LED output latch was occuring correctly, I realised that this fault could be caused by the logic that selects between 8 and 16 bit I/O transfers. It's a couple of chips on the -66511 card, and a little bit of testing showed that there was a 7404 that was not a NOT gate :-). Replacing it cured _both_ faults :-) To monitor points on the PCBs in the cardcage, I normally solder short lengths of wire to the appropriate pins, then put the card in place, power the machine up, and probe the free ends of the wires. I should make a set of extender boards I guess. Feel free to ask me about these machines, I'll happily share the little knowledge I have.

Re: HP 9820A
Message #6 Posted by Achim on 14 July 2002, 6:21 p.m.,
in response to message #5 by Tony Duell (UK)

That sounds very interesting because I still have a small defect on one of my 66511-boards which I couldnt fix yet: the STOP-key doesnt work under certain circumstances, e.g. when FIXED-mode or one of the function-block LEDs is lighted. Can you tell me which of the 7404's has to be replaced?

HGP98x0 I/O interface repairs
Message #7 Posted by Tony Duell (UK) on 14 July 2002, 7:00 p.m.,
in response to message #6 by Achim

In my case it was the 7404 in the front bottom corner of the card, but the fault was not the same. Still, it could be worth checking (see below). I don't generally like 'lucky dip' servicing, though, I prefer to think about the fault and work out what could be causing it. I am going to number the ICs starting with U1 in the rear top corner, going along the top row of the card, then starting again at the rear of the second row, and so on. The 7404 I just mentioned is then U20. OK, are you getting an STP/ signal from the keyboard under all conditions? Check at the input of U3a. This should go low whenever Stop is pressed. If not, you have a fault in the keyboard. Now look at the output of U3a (it's a7404 again). Does it go high when Stop is pressed. No, U3 is bad. The output of U3 goes into the AOI gate U9 (7453), which seems to generate some kind of I/O flag. The 'enable' input for the stop signal comes from (wait for it), the I/O size (8/16 bit) logic. Check U19d (3001/74H08), U20 (7404) and U12b (7420). Also check U9 itself, of course. STP/ also goes into a circuit that produces a condition code signal to the microcode sequencer when a key is pressed. It consists of U5d (7400), U1b (3001/74H08), U4a (7400). I don't think this is the problem, but it might be worth checking. You are sure the problem is on the -66511 card, I trust.

Re: HGP98x0 I/O interface repairs
Message #8 Posted by Achim on 15 July 2002, 3:52 a.m.,
in response to message #7 by Tony Duell (UK)

THX for your reply. Yes I'm shure that the problem is in the 66511 card because changing this causes the fault to disappear. I forgot to mention one more fact: often it helps to press STOP twice! It seems to be a kind of timing problem or weak output signal of one IC which is very hard to find. I also observed that the fault vanished when the machine was warm.

Re: HGP98x0 I/O interface repairs
Message #9 Posted by glynn on 15 July 2002, 5:11 a.m.,
in response to message #8 by Achim

...which also brings to the discussion other methods of diagnosis, especially in the case of those faults which are both repeatable and seem to be thermally exacerbated.

Have you tried the stuff in a can which "freezes" components? Selectively isolating chips by subjecting the candidates to this treatment, and looking to see when the fault reappears?

This is admittedly a crude technique for finding the fault, but it has been known to work-- and it does have the advantage of not requiring a logic analyzer and the associated training.

Re: HGP98x0 I/O interface repairs
Message #10 Posted by Achim on 15 July 2002, 3:57 p.m.,
in response to message #9 by glynn

I used this cooling agent years ago and indeed forgot about it. This really could be a way to find out weak parts. Maybe I'll try this. Another challenge is to build extender cards to raise the electronic boards for better access. Has anyone build something like this yet?

Re: HGP98x0 I/O interface repairs
Message #11 Posted by Tony Duell (UK) on 15 July 2002, 6:17 p.m.,
in response to message #10 by Achim

I've not tried extender boards yet (I use the poor-man's solution of soldering a few inches of wire onto the pins I I want to look at, then putting the board back into the machine, and probing the free ends of the wires). But the 98x0 is slow enough that you probably could use extenders with no problems. Be warned that although all 4 processor boards have a 18 pin (per side) and a 15 pin 0.156" edge connector, the connectors are in different positions on the 4 cards. I guess this is to prevent you putting the cards in the wrong slots. Anyway, you'd either have to make up 4 extender boards, or more sensibly, separate 15 and 18 pin ones. For the memory box, you could get away with an extended for the rear (22 pin per side) connector only -- the front connector only carries the ROM address, data and select signals which are taken to the backplane behind the display PCB. You wouldn't need that for testing. You would need to make up a lot of extenders to work on the board in the memory box, though. Incidentally, has anyone else noticed that (with one exception), if you put the board component side up, edge fingers towards you, then the colours of the handles give the last 2 digits of the part number (using the resistor colour code). That -66511 card has 2 brown handles, for example. The exception is the display. It's -66542, but has yellow (4) and brown (1) handles. But on my display board, you can see where the word 'red' was etched into the copper and then scratched off.

Re: HGP98x0 I/O interface repairs
Message #12 Posted by Tony Duell (UK) on 15 July 2002, 6:07 p.m.,
in response to message #9 by glynn

Freezer spray is _very_ useful :-) (for finding thermal faults, and for a few other applications that I won't go into here). However, I do somewhat take exception to your last comment '...and the associated training'. FWIW, I've never had _any_ formal training or education in electronics, fault finding, or anything like that. I'm entirely self-taught. These days some people seem to believe that the only way to know how to do somethign is to be 'trained' to do it. Meaning that the person doesn't really understand the subject, he can just follow a few instructions and will be totally lost if anything unusual comes up. And as you might have guessed by now, the 'unusual' has a habit of occuring :-) Sorry if this sounds like a flame, it's not directed at anyone here. It is directed, thought, at the idiots who believe that I can't possibly understand <foo> because I've not been on sone stupid 2 week course, even though I've been doing <foo> for the last 10 years!

Re: HGP98x0 I/O interface repairs
Message #13 Posted by Tony Duell (UK) on 15 July 2002, 6:00 p.m.,
in response to message #8 by Achim

Have you tried a can of freezer spray? You know, that stuff that cools components down to -20C or so? Squirt each chip on the -66511 card in turn and see if you can make the fault come back. Apart from that, you could try tracing the signals with a good 'scope to check they are all valid logic levels. Thermal faults are often (electrolytic-) capacitor related, but I can't think of a timing capacitor in this area. There are a few on the keyboard encoder PCB, but you know that's not the cause of the problem.

Re: HP 9820A
Message #14 Posted by on 10 July 2002, 4:34 p.m.,
in response to message #2 by Tony Duell (UK)

I've checked the voltage tester points and yes, all were amazingly pretty accurate : exactly the voltage on my (analog) voltmeter.

There are indeed 4 lights (I did not see at first sight the 2 hidden lights) on the card reader.

I took off the printer and the reader, but still no response...

The good news is that I met a former HP employee who gave me the name of the guy who took care of the customer service in the seventies... maybe he could give me a hint...

Anyway many thanks for your precious adivec.

Best regards, Thibaut

Re: HP 9820A
Message #15 Posted by Tony Duell (UK) on 10 July 2002, 6:22 p.m.,
in response to message #14 by

I have now seen the pictures of the 9820 on this site (my home PC has a text-only display, so it wasn't possible to see them last night). It looks _very_ similar to the 9810 inside -- I think that apart from the firmware ROMs, the keyboard and the display, it's pretty much identical. So my comments on the 9810 might be appropriate. OK, your PSU is working. In one way that's good, because it's not a simple PSU in some ways. For reference, the +5, -12, and +16V lines track the +12V line, the +24V line tracks the +16V line and the +20V line is a zener on top of the +16V line. So if one line is out it can bring others down with it. The bad news is that CPU problems are not going to be easy to diagnose without a logic analyser and the ability to use it. To get a display, much of the machine has to be working (all 4 CPU boards, the memory control boards, ROM and RAM). Figuring out which has failed without being able to check signals is going to be a nasty job. How I would attack this (given the equipment and knowledge I have ) would be : Check Clocks -- bit clock and micorcode clock. If mising, debug the clock PCB. Then check that the microcode PC is changing. Are the ROM outputs doing something sane (you can't really tell if they're right, but at least you can spot totally dead microcode PROMs). Is it trying to access the main firmware ROM (check the address shift register and address decoder). Is it trying to access RAM (ditto, and also the memory timing circuit). Is it getting data back to the ALU? Is it ever loading bits into the I/O register? DO they appear on the I/O bus? Is it generating a display strobe (the last problems could be caused by the I/O interface or the CPU control cards -- Q register, I/O control, etc). Is the display PCB itself working? Doing all that 'blind' would seem to be rather difficult!

[ Return to Index | Top of Index ]

Go back to the main exhibit hall