Data or manual for this 286 motherboard (PA286-SA1)

[Megatron-uk]

I've just bought a replacement 286 motherboard for a generic Headland HT12 board which I've owned since new for the past 30 years and which has finally stopped working

The new board is almost exactly the same as a Peaktron PA286-SMT board, as described here:

https://stason.org/TULARC/pc/motherboards/P/PEAKTRON-COMPUTER-INC-286-PA286-SMT-12-16-20-PA286.html

However, mine has subtle differences, such as a different number of jumpers on the main 'SW1' block: it starts from jumper 2 and runs to 8, rather than 1 to 8 (jumper 1 which, on the PA286-SMT switches the ISA bus from 1/2 to 1/3 clock) - as well as having no async/sync FPU jumper like the Peaktron board.

The main problem I'm having is trying to decipher the memory configuration on the board. The Peaktron board supports up to 16MB (in 4x 4MB SIMMs) and the configuration of the various sizes is easy as per the details on that site, but, mine doesn't work with the jumpers set for 16MB as per the Peaktron data. Nor does it work configured as 4MB (in 4x 1MB SIMMs). It seems that the memory configuration is set differently. I really can't figure out from the printing on the board alone what the configuration should be.

Long shot, but does anyone have any data on this? The silkscreen identifies the board as a 'Micro Systems' MICROCOMPUTER SYSTEM PA286-SM1 Rev D3A. Other than those couple of jumpers it's identical to the PA286-SMT from Peaktron.

The only thing I've been able to find was a Czech site who said it was "very fast" (faster than a 386SX-25, even with a slower clocked Harris 20MHz), and a single forum post on a Hungarian site with someone trying to get a SIPP version of the board working.

Otherwise, I'm facing trial-and-error, and I'm getting rather fed up of the 3-beep-memory-error already....

 

[eswan]

Oooh, that _is_ a nice board. I remember it as being very fast as well. Looks like it has 4 banks of 2x256x4 chips on board for 1 meg. That might be conflicting with the simm slots. I think I'd try to get it working without any simms first, mtype1&0 off, size1&0 on?

 

[Megatron-uk]

It does indeed have DIP ram as well, it actually came with 8 chips (without parity) fitted for a total of 1MB.

 

[Megatron-uk]

Hmm, this doesn't look good at all, does it?

That's with just the 8 DIP chips fitted back in, the jumpers set as per the way it arrived (fortunately I took pictures of the board when unboxing), and a basic Trident 8900 VGA card and keyboard.

No memory error beeps, but it just powers on to a flashing (somewhat corrupted) green cursor in the top left corner of the screen, nothing further, even leaving it for a minute or two.

Any other changes to the jumper blocks or swapping to the SIMM models gets me the memory error beep code again.

 

[eswan]

VGA card fully seated? Looks like it might be hitting the oscillator at the back of the slot.
Might check without a video card, see if it gives a different (no video card) error beep.

Mono/Color jumper? If I remember correctly, set to mono will give something like that.

 

[Megatron-uk]

Just tried removing it - it definitely reports a video card error; one long beep and then 8 shorter beeps. Tried the card in all slots - no difference with it plugged in, just get the flashing green cursor.

Unfortunately, I only have the one ISA VGA card at the moment. I had a lovely CL-GD5428 previously, but it seems to have mysteriously disappeared... although we have moved house twice since I last used it.

I have a nice Tseng ET4000AX coming, but not sure when it will arrive (Covid and some severe winter weather in the UK is disrupting a lot of mail).

 

[eswan]

Sorry, I keep editing my post instead of putting on a new one...
Mono/Color jumper? If I remember correctly, set to mono will give something like that.

 

[eswan]

I'm going to take a wild educated guess and say the MBSA B0/B2 & MBSB B1/B3 control whether the simm slots are mapped to the lower ram banks (use if no dip ram) or upper ram banks (use if dip ram populated). In the picture they're set to B0 and B1. Could set them to B2 and B3 and use both onboard ram and simms I think. Or leave at B0 & B1 and pull the dip ram.

 

[Megatron-uk]

Not quite the colour jumper, but it does look like a relatively 'harmless' jumper was the cause of the problem. The board came with JP9 (labelled 'TBT' jumpered 2-3). Looking at the Peaktron details this is the turbo control (either via keyboard combo, or by case switch). I've taken the jumper off entirely (which I believe reverts it to front panel/motherboard header control) and it is powering on!

The suggestion to try the color jumper definitely set me on the right track (looking at non-memory related settings)!

Also, I found a high res image from that Czech site that benchmarked the board against the 386-SX25 and just about made out the jumper settings for their use of SIMM's instead of DIP's, so I put my 4x1MB modules in and it detected them fine.

Still trying to work out if I can get the 4x4MB modules installed, which according to the Peaktron manual, are supported. I wonder if it had a slightly different BIOS though. Tracking down that will be fun!

 

[Megatron-uk]

A few more updates on this board, in case anyone needs information in the future (ok, I'm kidding, how likely is it that anyone else is going to track down a board that no-one seems to have any information on???).

Anyway, here's a few things I've learned.

Zero wait state operation
The board has a jumper for 0-wait operation. It is located on the main jumper block, SW1. Jumper J3. Using relatively fast 70ns SIMM modules or the 70ns DIP chips this would not work reliably. I have since swapped to 60ns SIMM modules (of various sizes) and 0-wait seems reliable (read that as: completes the self-memory test every time) with those faster speeds.

Conclusion: 0-wait state on this board needs 60ns or faster DRAM at 16MHz CPU speed. I have yet to test 20MHz and higher.

Support for more then 4MB DRAM
The memory configuration for this board is odd. As evidenced previously. There's the additional problem that it's almost exactly the same as the Peaktron PA286-SMT, but just not quite. The Peaktron has loads of memory configurations, including a couple of combinations of DIP and SIMMs (for odd sizes like 1.5, 3 and 6MB). This board doesn't appear to have quite that flexibility.

However.

There is one combination that will let you have more than 4MB installed. You need two 4MB (4M x 9) SIMMs installed, in SIMM bank 0/2 (closest to the ISA slots) and the memory jumpers set as follows:

- SW1 J5 Closed
- SW1 J6 Closed
- SW1 J7 Open
- SW1 J8 Open

According to the details on the board itself, J5/J6 set like this to 4MB device sizes in bank 0 or 1. Setting J7/J8 configures the memory controller to use a single bank. This gives a total of 8MB, which is detected and counted by the self-memory test correctly.

The PA286-SMT has an additional configuration of two banks in this mode (with J5 to J9 set as Closed, Closed, Open, Closed), but this looks to be an unsupported combination for the PA286-SA1 and only leads to memory error beep codes if you try it. Still, 8MB is better than 4MB, which was better than the 1MB that the board was initially working with.

Conclusion: The PA286-SA1 board does support 4MB 30pin SIMM modules, but only to a maximum of 8MB, not 16MB like the PA286-SMT.

 

[Megatron-uk]

After a day spent testing and benchmarking this board I have several conclusions:

1. It is fast. Broadly speaking as fast or slightly faster than the late Headland HT12 boards.

2. But... you need faster-than-normal DRAM to enable 0-wait state mode. I had to step up to 60ns modules from my usual 70ns sticks in order for it to POST consistently.

3. At 0-wait state mode it is incredibly unstable; high disk IO can cause repeated hard lockups (I'm using XTIDE in a 3C509B, with a CF drive mounted on a standard multi-IO card - tested several CF cards and several multi-IO adapters). Doing a 'dir' of the Dos directory is enough to trigger a lockup. Backing off to 1-wait state and the system is rock solid. This wasn't apparent when doing my initial testing from a bootable floppy.

Unfortunately, I was wanting to build a very fast 286 system, so having to run this thing at 1-wait state at just 16MHz makes it slower than my old standard HT12 board (which admittedly, was a nice board). So it is completely unsuitable for trying 20 or 25MHz Harris processors.