|Re: 9810 Power Problem?|
Message #6 Posted by Tony Duell (UK) on 28 Aug 2003, 4:39 p.m.,
in response to message #4 by Katie
Firstly a couple of points from your first message. The fan doesn't tell you a lot -- it's an AC fan running off the mains side of the transformer (in you case (120V), straight off the mains, in my case (240V), the mains transformer primary acts as an autotransformer). Secondly, you say you tested the 5 labelled PSU rails, but presumably you didn't test the 5V line (which is, of course, the supply for all the logic). There's a convenient capacitor to use for this on the display PCB (I think the ends are even marked!), or you
can check it on any TTL IC in the unit. Of course noise on this line causes problems!
I don't have a set of extenders -- I wish I did. One problem is that the CPU boards all have 15 and 18 pin DS connectors, but at different positions on the 4 boards. This is great for failed circus, who;d find it difficult to get the PCBs in the wrong slots (I won't say impossible, I've met failed servoids..). It does mean you either need a different extender for each PCB, or you need to use 2 little boards, one for each connector, which is not mechanically too rigid.
I guess if you want to work on boards in the memory cage with extenders, you'd put the entire cage on an extender (you only _need_ the rear connecotr, the front one is for the expansion ROM cartridge signals only), and then use another set of extenders to put the boards outside the left side of the cage. If I was going to do this, I'd make up some metal rods that would screw into the memory cage mounting holes, aand which would then support the cage on the extender.
Anyway, what I actually do is to either use the wonderfully slim probes of the LogicDart, which will get to wquite a few IC pins without extenders (!), or to pul lthe card in question, solder wires onto useful testpoints, put the card back in, and test the free ends of the wires. It's a bit slower than using extenders, but it does work.
Tracing a 'no display' fault on a 98x0 is non-trivial. The display is scanned in software (well, firmware), so for the display to come up, the entire CPU, memory controller, ROM, RAM, I/O controller and display circuitry have to be working. I am trying to work out ways to really trace faults -- HP kindly provided a test connector on the CPU control card (09810-66513) which carries the microcode address signals, and connectors o nthe memory box/memory control cards (depending on the model) that carry the address/data bus. To use the latter, you'd want to trigger a logic analyser on a particular micrcode address -- say just after the instruction fetch.
All this is possible (and I have a lot of notes I've made on how it works), but it'll have to wait until after the HPCC mini-conference.