• Please review our updated Terms and Rules here

80C287 install not working in Sharp PC-6200

Twospruces

Veteran Member
Joined
Dec 9, 2009
Messages
779
Location
Canada
Hi all,
I recently acquired and revived an old 286 laptop from Sharp. Quite a nice compact machine actually.
I thought I would try to add then 80C287.
I purchased a part off ebay : IIT 2C87 math coprocessor.


Didn't work out; maybe it was a fake? Hard to know.
What I observed:
* the installation is simply plug it in - nothing else to do apparently
* when installed, the laptop boots and runs, however
-- Checkit3 detects the 287
-- but it has trouble with any benchmarks. It fails to detect clock frequency. It shows incorrect values on the axes.
-- It fails to run the CPU or math co-pro benchmarks - meaning that the test completes and reports "00" for Dhrystones and Whetstones
-- checkit3 also seems to be unable to make calculations - which to me tells me the math is now "broken"
* PC-matlab now runs but it also can't compute any math.

I looked at the hardware signals and nothing seems off.
* the BUSY line on the 287 is toggling
* PREAQ and PEACK are toggling away
* all the signals appear to be correctly connected

Any thoughts? Did I get a bad chip?

It does appear that the CPU is requesting a calculation, because the system believes there is a 287.

The BUSY signal from the 287 goes to an LSI chip, which is then responsible to generate BUSY to the CPU.

Should I see BUSY toggling at the CPU as well? I would think so.. I assume that when the 287 is computing the 286 needs to wait.
Interestingly, BUSY input to the 286 processor is high through all the 287 activity caused by Checkit3.

I bet that's it.

Anyhow, comments?

thx
 
well well, the act of writing out my thoughts proved successful.
It is working now.
What I had to do:
* the signal from the "system LSI ASIC" which generates BUSY to the CPU based on BUSY from the 287 was NOT working
* I lifted the pin on the system LSI ASIC which is the BUSY output to CPU
* I bridged (meaning soldered a lead) to directly connect the 287 BUSY output to the 286 BUSY input

Instant success.
* Checkit3 now reports an IIT 287 chip
* at 12MHz, I get 1454 Dhrystones and 427.4 Whetstones on checkit3 Benchmarks
* I am now running PCMATLAB benchmarks as I type this out.

So, there you go. a hardware issue buried inside the system LSI. Maybe there is some register bit not being set by the BIOS correctly.

cheers
 
On the Intel iSBC 286-10/A CPU board there is a wire link to remove before fitting the 287 maths coprocessor.

Check your manuals (if you can find them). You may find a DIP switch?

Just a thought.

Dave
 
I do have the owner and service manuals..no switches associated with installing a math co. In any case its running great. Thanks. Just glad the effort panned out.
 
Hmmm. If the CPU busy line is routed to one of the ASICs - where is the 287 busy line routed to I wonder? Possibly the same ASIC?

In this case (internal routing within the ASIC) is there a BIOS setting?

I see you have got it working with a bit of circuit surgery, but I can’t imagine that Sharp has messed the ASIC up on something as fundamental as the 287.

Dave
 
yes, the circuit is like this:

80C287 BUSY --------> LSI ASIC -----------> 80C286 BUSY

the service manual says that the asic simply generates the BUSY signal output from the BUSY signal input, but I think there is probably a mux in the ASIC that is an "enable".
I have no alternative BIOS to pick from, but it could well be a BIOS bug to have not "enabled" the 80C287 functions in the LSI ASIC.

I agree with you ... hard to believe. But the schematic is clear that the above circuit is accurate, and there are no configuration options that need to be tweaked - no BIOS parameters to set, and no board jumpers to set. I traced out the signals and it seems like that LSI ASIC is not doing it's job.

I don't know how popular the PC-6200 laptop was, nor do I know how many of those people would have wanted the math coprocessor.
Strange, but like I said. I'm just glad my 25$ PC and 25$ math co processor have come together well enough!
 
Oh well, it wouldn’t be the first time a manufacturer has got it wrong - and it won’t be the last time either!

I was on an early Intel 286 course with a load of IBM engineers. During the course, the IBM engineers realised they had designed their very early 286 machine wrongly so that it wouldn’t switch correctly from real mode to protected mode.

There must be some very early IBM 286 machines that won’t run OS/2 as a result...

I wonder if those would be collectable machines and worth lots of money (like stamps) :)!

Glad you have got yours running...

Dave
 
Back
Top