So do I. I
much prefer to diagnose and rpeair faults logically, aomethign
you can do with the older machines.
Exactly! You can actually get at the individual parts! Everything
isn't 200% integrated. ;)
Not only can you get at the parts, you can also get the aprts. I have
foudn it much easier to get a replacement TTL IC for a PDP11 (or
wahtever) than an ASIC for a modern PC.
But as I said last night, the Qbus PDP11s have custom chip or chipset
processors and you can't 'get inside' them very easily. If you want to do
that you realyl want ot find oen of the older Unibus machines
I am currnetly having a conversation with a guy in the Netherlands who is
repairing a machine of a similar vintage (not a PDP11). He's borrowed a
logic analyser and we're goign thoguh it a the level of 'Well, in this
microisntrucionm, the program counter is fed to one inptu of the LAU, a
constand 1 is fed to the other intput, so as to add 1 to the PC ot
increment it. Now let's see if that's what's going on...' (OK, it's
not
quite that, the machine in question is bit-serial...)
HE seems ot be loving it. Being able to see how registers are selected,
how data flows from one to another, and so on .
Very interesting! I don't have too much low-level hardware knowledge
(yet. I intend to! I'm still young.) but that still sounds extremely fun.
The machine in question is an HP9830. Althoguh sold as a calcualtor, I
think that was jsut for marketting reasons. It has a QWERTY keyboard, a
(1 line) alphanumeric display and programs in BASIC. It is very hard to
claim that that is not a computer.
Although it's an all-in-one desktop machine, the processor is built from
lots of simple chips .There are 4 boards of TTL that are normally
considered to be the processor (these boards are also used in the HP9810,
9820 and 9821 machines) together with a couple of bords to interface the
bit-serial processr to RO mand RAM. of course there are boards to drive
the display, boards to oeprte the itnernal tape drive, a keyboard
encoder and a power suppyl too. But let's stick wit hthe processor for
the momebnt,
It's 16 bit, bit-serial. There are 2 accumualtors (A and B), the program
counter, a 4it extension register (E), and then some inte4nral registers
: M = Memroy address, T = Memroy data, Q ('Qualifier') = current machine
isntruction (normally) . The procerssor is controlled by 256 28 bit words
of mcirocode, stroed in ROMs on one of the PCBs.
My friend i nthe Netherlands has schematics, microcode source and memory
state machien listings. So far he's managed to fidn a problem wit hthe
memroy state machien (I think one of the flip-flops wasn't), and a
problem wit hte 'S input multipexer' -- a multiplexer circuit that feeds
the M, R, cosntant 0 or constant 1 itno one of the ALU inputs. We now
think there's a problem wit the 'R input multipleser' whcih feeds the
processor regsters or a constant 0 or 1 into the other input of the ALU.
So w'ere workign through that. We know the (complex) CPU clock is running
properly and that the microcode is following a possibly-correct sequence.
When you first start working on the internals of a processor at this sort
of level there is a lot to learn. But it does all make sense in the end,
or at least it did for me.
-tony
