That can be very tricky, I had a case recently where a suspicious trace had full continuity when measuring due to the slight pressure of the tips of the measuring probes, but in normal operation it was interrupted. Took me some time also.

For sealing repaired traces I use a lacquer that is used in model making. The brand is "Life Colours", it´s guaranteed to last 15+ years 😉
It can be bought in small amounts like 15 or 30 ml so there´s no risk most of what you bought dries out befoe you can use it.

BTW, this mainboard is extremely picky when it comes to keyboard controllers. The original (defective) chip was a Mitsubishi "M5L8042-146P", I tried several Intel, AMIKEY and NEC chips, but none of them worked 100% correctly. Some reported a keyboard error, some complained about VGA RAM, some would not start witch certain AT-PSUs , some would not work at all.
Finally I tried a "Jetkey 5.0" chip that did not work on countless boards I had tried it on in the past so I thought it was somehow broken, but this one works perfectly on the Modula Tech mainboard.

I have been testing and optimizing this mainboard for a while now, because there still were / are some instabilities.
I found a defective cache chip that randomly pevented the board from starting, then I observed keyboard errors that seem to have been caused by a (cheap) modern keyboard with integrated USB-PS2 converter.
Still there seems to be one issue left: Sometimes the beep during the BIOS summary screen is of a higher pitch, then the system continues to boot but the CPU is running at 8.33 MHz instead of 25 MHz. The oscillator frequency seems to be divided by 6 instead of 2 in these cases although the "Turbo" connector/jumper is in the "normal" position.
When the jumper is tet to "Turbo", the CPU runs at 12.5 MHz so oscillator frequency is divided by 4 in this case.
Moving the video or I/O card to a different ISA slot or changing a jumper restores normal operation - until the next event.
The frequency of this issue seems to be related to the video adapter I use, its rarely happens under my Cirrus Logic 5422 adapter, but with my ELSA Winner 1000 (2MB) it´s happening quite often.
I also tried different BIOSes and KB-controllers - to no avail.
I´m quite sure I encountered similar issues decades ago, but can´t remember the reasons or solutions...

Maybe the jumper/switch for turbo makes poor contact? A high-resistance connection could be keeping the logic level between 0 and 1, and that can cause all kinds of issues.

And speaking of issues, I was testing old mechanical AT/XT keyboard on that mobo with bad via I mentioned and it didn't want to boot. Nothing on POST card either. Lifting the whole mobo PCB by one side makes it work again - for a time. I think I have another bad via or micro crack... welp, that's part of the hobby I guess. I just don't have the time to dive into it right now.

This is my kind of 386 board. Full-sized for no apparent reason. It's a nice shade of green too.

The SiS Rabbit chipset came in A/B/C revisions. I'm pretty sure this is a Rev A. I don't think those support more than 16MB memory or 128kb cache.
I have rev A rabbit board from CTI. It also only supports 1MB SIMMs...and I have been unsuccessful getting any replacement BIOSes working.
What kind of AMIBIOS does your board use? Is it the Color/Hi-Flex BIOS, or the earlier one from '89-'90? Mine uses an older AMI BIOS that is not very user friendly. It lets you adjust register bits, but it's kind of cryptic.

One other quirk with my board is that it only works with original 386s, not DLC type upgrades.

The jumpers have all been replaced and the pins cleaned, so jumper contact can´t be an issue. I also double (and triple-) checked all the traces in this area and couldn´t find any interruptions.
I also added a socket for the oscillator and found that the board runs perfectly at 33 MHz - at least under DOS...

The oscillator signal passes several 74F74 flip flops for reducing the frequency in turbo mode and maybe additional configurations. There are also some dividers in one of the SIS chips that are controlled by BIOS registers. It seems somwhere in this interaction there´s a problem that is triggered by - whatever.

This board will only accept 1MB RAM sticks, it´s running the old black and white AMI BIOS from 89/90 and here the BIOS date is June, 13th 1990. (will add a dumpfile soon). Cache is probably limited to 64K or 128K. The 4 jumpers near the cache chips are for enabling "internal cache", a cheap alternative when no external cache / TAG chips are provided.

For whom it may concern- the original BIOS file:

Anonymous Coward wrote on 2023-02-20, 16:26:

This is my kind of 386 board. Full-sized for no apparent reason. It's a nice shade of green too. […]
Show full quote

This is my kind of 386 board. Full-sized for no apparent reason. It's a nice shade of green too.

The SiS Rabbit chipset came in A/B/C revisions. I'm pretty sure this is a Rev A. I don't think those support more than 16MB memory or 128kb cache.
I have rev A rabbit board from CTI. It also only supports 1MB SIMMs...and I have been unsuccessful getting any replacement BIOSes working.
What kind of AMIBIOS does your board use? Is it the Color/Hi-Flex BIOS, or the earlier one from '89-'90? Mine uses an older AMI BIOS that is not very user friendly. It lets you adjust register bits, but it's kind of cryptic.

One other quirk with my board is that it only works with original 386s, not DLC type upgrades.

I have a rabbit chipset motherboard but how do I tell of this revision that I have?

Cheers,

Good news, it seems I finally found the cause of the instability.
First i noticed that the "too low CPU frequency" problem kind of diminished during my troubleshooting. After a couple of hours or so everything seemed fine, the system would always start at the correct frequency. The next morning the same old issue was back again with a vengeance.

I also noticed that during starts that resulted in a 10.7 MHz CPU frequency, the tick-sound during memory count was faster and the POST beep sound had a higher pitch. So some oscillator frequency was different in these cases. Probably the frequency was too high and the system starts in some "fail safe" reduced frequency mode so BIOS setup can be entered.
Because the issue occured when the system was cold I warmed single chips with a soldering iron before startup and ended up at a HEX-inverter that operates as a (ring-) osciillator (or two of them) with the 14. 31818 MHz crystal. When warm, the system would start at 33 MHz CPU frequency, when hot it would not start at all! So I removed the old "SN74F04N", fit a DIL14 socket in and populated it with a "SN74HC04N" - the only version I have on stock at the moment.

So far 100% of the starts have been successful, but this is going to be tested thoroughly.

majestyk, thanks for the interesting story and bright colorful photos of the interesting mainboard. Good job!

You´re welcome, Roman555!

I´d also like to add another observation. Before the oscillator was fixed I tried an ISA graphics card with a Cirrus Logic 54xx chip that doesn´t have an oscillator of it´s own (at least I cannot find one). This card failed to sync with the monitor quite often and the monitor reported "vert sync 134Hz". It was probably irritated by the faulty bus clock speed.