exxos's Atari, Amiga & retro forum

I've got Apple's first portable too... though it's 5 - 6 years newer in design and actually has a (lead-acid) battery.

Oh, and I think I have a plan to test whether it's the drive or the system.

I'll make up one of my floppy cable adapters and connect a Gotek with my disk images on it. That should check that the controller's working correctly.

The disk controller's fine as I can boot from the Gotek without issue. I don't have any Portable floppy images in IMG format at the moment though so I can't do a full test.

However, I can run my memory test program and it's showing some bad RAM in segment 0x6000.

There are three bytes with issues, 0x80b9 has 0xbc rather than 0xb9, 0x80f1 has 0xf4 rather than 0xf1 and 0x80fb has 0xfa rather than 0xfb.

The last one is a one bit error, which is understandable, but the other two seem to bits swapped!

They are at least consistent.

Well, here are the memory errors...


I managed to find an image with the portable LCD graphics driver, which allowed me to run the field test diagnostics RAM test, which passed. :-/

The "mouse" doesn't work. It's not dead but the output to the IR LEDs isn't very good. I'll have to look into that later.

As for the rest of the system, the LCD does have a few bad lines of pixels in the top left:


I've not yet worked how, or even if, the LCD contrast can be changed.

After removing a duff BC308P transistor, one of the three driving the IR LEDs, I now have a working "mouse" trackball thing!

You won't believe just how much brass went into building the tracking mechanism! Also, the encoder wheels are sheet aluminium with painted on lines so that the light from the LEDs bounces off the surface rather than going through slits. A very strange design! Almost Steampunk Victorian.


P.S. The transistor in question was leaking its collector voltage both to the base and the emitter, but not quite a dead short.

Out of curiosity I tried my F1 RAM expansion board in the Portable. Although it won't fit with the rear case on due to my design violating the specification in one direction it does so fine with that bit of plastic removed.

And it works!

There are still the three addresses in segment 0x6000 giving the same odd values, even though my memory is going over the top, so maybe the machine has some hardware there. Other than those locations my himemtst program shows no other problems and the field test disk RAM test shows no problems.

Looking good, and that Apple portable looks awesome.

Personally I'd like an Apple IIc but they seem to be like unicorns these days!

Well, well. The Apricot FP is actually significantly faster and more IBM-PC compatible than the F1/F2/F10 it seems. This is probably due to it having separate display RAM.

The F1/2/10 has a speed rating (when CheckIt! doesn't crash, which is most of the time) precisely the same as an IBM PC-XT. The FP is over 1.5 times this.


As for the system itself, I do think that there is a memory fault in at least one of the chips in the high bank. I wonder if any of the spare 41256 chips I have actually work. If I had enough I could upgrade the machine to the full 1MB by soldering in another 32 of them.

P.S. How did we cope with these early (1984 vintage) LCD displays? :-)

I'm trying to work out what sort of fault could cause the memory errors I'm getting.

My memory test program writes the word address into the words in memory, i.e. it writes a 16 bit address into the 16 bit word in memory its pointing to. It does this for all the segments and then reads it back and compares the result.

The system memory is two banks of 16 41256 chips.

Consistantly there are errors in the 7th 64K segment, always at address 0x80f1, which returns the value 0x80f4 (bits 1 and 3 swapped). Often address 0x80b9 returns 0x80bc (bits 1 and 3 swapped) and rarely 0x00f1, which returns 0x00f4 (bits 1 and 3 swapped).

When the machine is turned on from cold its power-on self test usually fails the memory test, giving 416K of RAM. If the machine is turned on, even for a short time, and then power cycled it will invariably pass its memory self-test.

If the problem is faulty bits in a chip then it would have to be two chips which have precisely the opposite fault at the same bit at the same time, which is unlikely. Similarly, if it's an on-chip address error it should only affect one chip.

It can't be the higher address decoding as that wouldn't show up as every segment would have the same pattern of addresses. It can't really be the lower address lines either as it's only one bank of RAM which is affected.

So, what failure mode could swap two bits in two chips at specific addresses? I'm stumped.

OK. I don't think it's the RAM per se as this morning I managed to run my memory test program very soon after a cold power-up and there were a huge number more errors in that bank and different bit showing bad. It can't be all the chips.

I think the first thing I'm going to do is recap the power supply and looks for any other electrolytics on the CPU board to change. I want to eliminate voltage levels.