If we're talking about boards from 286 to Pentium era, then yes, boards without PS2 port and with a KBC situated next to ISA slots are in minority. Then let's say you're interested only in Socket 3 baords and want to build some nice 486 system. Suddenly boards which fall into those criteria are in majority. Almost all boards with VLB/PCI connectors and 3V CPU support have KBC next to ISA slots. Check UltimateRetro site and see for yourself.

Anyway, I already have some ideas how to deal with it. The customized adapter board could be soldered under a socket (only with precision pins since those are partially naked and have enough room for 0.8mm board). That way you could exchange KBC freely which is a nice bonus. If that solution is still too high, I could try to solder the adapter to two rows of precision pins which should free a milimeter or two. That means no socket unfortunately. The PS2 pins would need to be routed out with a cable in both cases to be sure they don't interfere with an ISA slot.

Paar wrote on 2022-03-28, 17:01:

In that case it could be better to place all the components on the bottom of the board to get another mm or so. Will probably need to modify it to be usable with my board. The prospect of getting native PS2 support with 486 is a strong one.

I just did a version like that. Passes DRC but hasn't been prototyped yet, so caveat emptor

Paar wrote on 2022-07-12, 06:37:

If we're talking about boards from 286 to Pentium era, then yes, boards without PS2 port and with a KBC situated next to ISA slots are in minority. Then let's say you're interested only in Socket 3 baords and want to build some nice 486 system. Suddenly boards which fall into those criteria are in majority. Almost all boards with VLB/PCI connectors and 3V CPU support have KBC next to ISA slots. Check UltimateRetro site and see for yourself.

Anyway, I already have some ideas how to deal with it. The customized adapter board could be soldered under a socket (only with precision pins since those are partially naked and have enough room for 0.8mm board). That way you could exchange KBC freely which is a nice bonus. If that solution is still too high, I could try to solder the adapter to two rows of precision pins which should free a milimeter or two. That means no socket unfortunately. The PS2 pins would need to be routed out with a cable in both cases to be sure they don't interfere with an ISA slot.

I was able to do a successful "behind the slot" installation using the rev 1.5 adapter that I don't think I ever released here and I've unfortunately lost the design files for. It has the same PCB dimensions as rev 2 and 3 but still uses discrete transistors on the top of the PCB. Maximum height 1mm under the KBC, i.e. there is a 1mm gap between KBC IC and interposer PCB.

No sockets. KBC soldered to interposer, interposer soldered to pin headers soldered to motherboard. Fits under any card I can find.

Looks promising. What is the height of the inverter IC used in the newer revisions?

Paar wrote on 2023-05-12, 19:46:

Looks promising. What is the height of the inverter IC used in the newer revisions?

1.75mm, the tallest thing is the inductors at 2mm. The revision 3 I posted above moves all SMD components to the bottom. Should in theory allow the shortest stack height possible.

Can someone give me a bit of advice with my conundrum please? I built the interposer, but even with only the original chip installed on it (and with or without IRQ 12 connected to it), it slows down my 386 system significantly so that booting from a floppy becomes excruciatingly slow. I tested with both a MEGA-KB-F and a VIA VT82C42N and they both behave the same.

I was able to load PS2SUPPC successfully and then load the mouse driver with the mouse connected. But the issue is with the slowing down of the system with the interposer installed.

Thanks

mockingbird wrote on 2023-08-15, 17:05:

Can someone give me a bit of advice with my conundrum please? I built the interposer, but even with only the original chip installed on it (and with or without IRQ 12 connected to it), it slows down my 386 system significantly so that booting from a floppy becomes excruciatingly slow. I tested with both a MEGA-KB-F and a VIA VT82C42N and they both behave the same.

I was able to load PS2SUPPC successfully and then load the mouse driver with the mouse connected. But the issue is with the slowing down of the system with the interposer installed.

Thanks

Sounds very much like one of the GPIO pins on the 8042 is controlling turbo on your board and the interposer either doesn't pass that pin through or it was one of the mouse pins.

Despite the speed, does the mouse function properly?

Who's interposer design did you build?

maxtherabbit wrote on 2023-08-16, 04:51:

Sounds very much like one of the GPIO pins on the 8042 is controlling turbo on your board and the interposer either doesn't pass that pin through or it was one of the mouse pins.

Despite the speed, does the mouse function properly?

Who's interposer design did you build?

Thanks. Yes, the mouse works perfectly (tested with ps2suppc, ctmouse3, then MS-DOS' edit.com for the cursor test). I am using your design from this post. The "slowness" affects the floppy seek at post, it will not complete (it's as if it is struggling to seek), and the memory count will also now go through the entire thing slowly, as opposed to only the first few pc speaker clicks being slow during the count and then speeding up shortly thereafter.

I'm at a bit of a loss then, de-turboing the system shouldn't prevent the floppy from seeking

maxtherabbit wrote on 2023-08-16, 15:10:

I'm at a bit of a loss then, de-turboing the system shouldn't prevent the floppy from seeking

I should articulate: It doesn't complain about anything, it seeks fine after boot, but the seek test at POST doesn't sound or look right.

Would you please consider altering your design to pass through all the pins? I do rather like this motherboard.

I have a 486 board I'll be testing this on to rule out the adapter, but I'm waiting on a 40-pin socket to solder in.

Can you remove the board, disable turbo, then run the machine?

If so, is the system still slow with the post doing weird things? When my turbo is off my post beep and ram tick sound flat and sad... It's a reminder for me to hit the button!

ChrisXF wrote on 2023-08-16, 17:12:

Can you remove the board, disable turbo, then run the machine?

If so, is the system still slow with the post doing weird things? When my turbo is off my post beep and ram tick sound flat and sad... It's a reminder for me to hit the button!

Thanks, that was helpful - thanks. Helpful in that it confirms what @maxtherabbit theorized -- namely that the turbo mechanism of this system is relying on the KBC somehow. I plugged in an LED indicator into the motherboard and lo and behold, with the interposer, turbo is inactive and the system is slow, with the turbo pins having no effect. With the KBC controller plugged in without the interposer, turbo works as it should (on this motherboard open is turbo and closed is slow).

So if maxtherabbit would implement a workaround for this I would be grateful. It's the CH-386-33H/40H baby AT motherboard which is pretty nice, and I already fitted it with the not-a-Varta mod and a Cyrix coprocessor.

I was unfamiliar with the effects of Turbo because the last time I used a 386 was perhaps in the late 80s.

EDIT:
I checked which pins are not connected and noted them:

They all show some continuity to the PS/2 header. Is it safe to jumper them straight through the pins around the socket with them already connected in circuit?

mockingbird wrote on 2023-08-16, 15:32:

I should articulate: It doesn't complain about anything, it seeks fine after boot, but the seek test at POST doesn't sound or l […]
Show full quote

maxtherabbit wrote on 2023-08-16, 15:10:

I'm at a bit of a loss then, de-turboing the system shouldn't prevent the floppy from seeking

I should articulate: It doesn't complain about anything, it seeks fine after boot, but the seek test at POST doesn't sound or look right.

Would you please consider altering your design to pass through all the pins? I do rather like this motherboard.

I have a 486 board I'll be testing this on to rule out the adapter, but I'm waiting on a 40-pin socket to solder in.

My design already passes through all the GPIO pins except for P10, 11, 22, and 23 because those are required for the mouse.

IOW what you're asking for is impossible.

mockingbird wrote on 2023-08-16, 19:16:

I checked which pins are not connected and noted them:

ps2.png

They all show some continuity to the PS/2 header. Is it safe to jumper them straight through the pins around the socket with them already connected in circuit?

Every single one of those is required for the proper operation of the mouse. If you start messing with them bad things will happen

Occasionally the motherboard will have a jumper block to select which one of the GPIO pins it gets the turbo signal from. You can try changing this if present, and hope that your KBC rom is firing turbo on all of them.

Alternatively you can figure out which pin the motherboard has wired to turbo and force it to the other logic state using a bodge wire and resistor (at the motherboard socket, not the KBC!). This will disable the soft turbo function and keep the system "fast". If you have a hardware turbo button it may or may not still work. Every motherboard implements turbo differently

Yeah I was about to say, looking at Max's schematic everything that's not directly related to the mouse support is passed through perfectly. You'll need to have a dig around the turbo circuitry and try and figure out exactly how the keyboard controller is flipping turbo on and off.

Could it be using one of those mouse outputs to do it... That would be weird but possible maybe? It might be possible to watch the pins with a multi-meter and see if any go solid high/low with triggering the turbo keyboard shortcut (usually CTL+ALT+- or ++).

The bodge wire and resistor idea sounds like it'll sort it out though, I'd probably look at that and delete the hotkey support for turbo.

Thanks, I'll try the bodge method. Forgive my ignorance, will a standard 4.7kOhm to 10kOhm resistor do the trick?

I'll be happy to talk you through the resistor thing but first could you please post a photo of the motherboard in the area around the keyboard controller?

maxtherabbit wrote on 2023-08-17, 22:09:

I'll be happy to talk you through the resistor thing but first could you please post a photo of the motherboard in the area around the keyboard controller?

Thanks, that's very gracious of you.

And this photo is from The Retro Web:

EDIT:

Here is a photo of the top from The Retro Web (sorry about that, I'm a bit slow today):

Thank you, it appears pin 24 controls turbo on your board. In your first photo you can see the tiny trace jumpers (the little triangles) that select between pins 23, 24 and 27 with the one for pin 24 connected.

Unfortunately for you, all of those pins are required for the mouse so it's not as simple a matter as changing the jumper. What I would recommend is taking the mod board out and putting the original KBC back in the board. Then, with the system running at full speed, measure the logic level of pin 24. You can also use the keyboard shortcut to de-turbo the machine and verify that pin changed state as a sanity check.

Once you know whether we need to pull pin 24 high or low we can select an appropriate resistor value.