THey do turn up on Ebay from time to time. They often sell for quite high prices, but a battered, non-working, one can be affordable. The main 'battered' state is a cracked or worse keyboard bezel. It's plastic, and it gets _very_ brittle with age. If you have the bits you cna stick them together iwth dichloromethane and then use the cotton fabric trick to streghten it, but it wil lstill be very brittle. Alas I doubt that repalcement bezels could be economically made. The 'non-working' is less of a problem. The only cusome ICs in there are the ROMs. These are HP mask-programmed ROMs, and they are odd in other ways. It is posisble to kludge in EPROMs with a bit of extra circuitry, and anyway the ROMs are not the most common failure point. The RAMs are 1103s, which are hard to find, but not impossivle. Jaut abotu everythng else is TTL, although some of the 74Hxxx parts are not trivial to get now. The main problem is fiding the fault. I feel there are 4 things you need to debug an HP9800. The schematics; the CPU microcrode listing; a logic analyser; a brain. It's a bit-serial machine, so you do need to be able to capture add examine the bitstreams in and out of the registers. And I find that one fo the best ways fo find out what hte machine is doing (so you can compare it with what it should eb doing0 is to conect the lgoic analyser to the microcode address liens (available on the test conenctor on top of the 09810-66513 CPU Control PCB [1]) and see what microcode routines are being eeecuted. You cna check it's fetching instructions, you can make sure it's not stuc in the I/O loop or addres indrection loop, things like that. [1] HP call it the 'microprocessor', but that term has a totally different meanign to the one we are all familiar with. as regards basic tropubleshoorting, what I do is : Tke everythign out out and look for burnt/misisng parts. Clean and oil the cooling fan, etc Power up the mains transofmrer only and check the AC output voltages. Fit the PSU PCB and check the DC voltages. Fit the CPU clock PCB (09810-66512) and check the master cloc, BitCLk and muCLk are all running correctly. Fit the rest of the CPU, the memroy address and data boards, ROM, RAM and display. Power up. If I am lucky (I never have been) I get the right display. Most likely I get a blank displane. The,,, Check fior DispStb on the display control edge connector. If it's present, then the most likely problem is inthe display logic. If it's absent, then check the CPU clocks are still running (if not, the darn thing might be stuck i na memroy refresh cycke), then check that the nemory controlelr state machine is running, that hte microcode is running, etc. Then trace the microcode. In genral you pick up the problem farily quickly if you un=derstnad the machine. Be warned that the first 200-or-so machines had a _very_ differnet memory system to later ones. It is very unlilkey you will see onme of these, in fact it's possible none exist (since if any part of thememyo system failed, the board-sapping repair involved replacing all the memroy and memroy control boards). But it's taht design that appears in the well-known patent. More likely you'll conme across a machine where the memory timing is handled by a state machine.v Bebugging that is 'fun'... Incidetnally, HPCC (Handheld and Portable COmputer CLub) ahs published a series of articles on repairign these fine machines. After minor correction and tidying up, it's likely they will be released as a single document. It's a fun emulaotr, but the real thing is much mroe interesting. -tony