Cheapest of all is solder bridge jumper.
Not as easily reversed, but I suppose you could put holes for jumpers for those who want that option.

Tevian wrote on 2025-08-04, 07:09:

I could still do a "standard" orientation but routing might get tricky. The ISA block would be fine with a 2-layer PCB, but addi […]
Show full quote

mbandalauk wrote on 2025-08-03, 23:25:

Tevian wrote on 2025-08-03, 11:48:

Okay... I'm looking into it. It shouldn't be too complicated to make an adaptor card for PC/104+. The only issue, I don't have a […]
Show full quote

Okay... I'm looking into it. It shouldn't be too complicated to make an adaptor card for PC/104+. The only issue, I don't have any PC/104 cards on hand. The technical drawings in the spec sheet look good, so it "should" work. Although, for testing, I'd probably have to get one. 😅

The attachment Screenshot 2025-08-03 042145.png is no longer available

There is an issue of orientation. In the standard versions, with the ISA connector along the bottom edge, all the pins are crossed.

The attachment Screenshot 2025-08-03 043610.png is no longer available

A rotation of 90 degrees clockwise makes the connections a bit more straightforward. This makes the CF card slot point up and should be fine

The attachment Screenshot 2025-08-03 043754.png is no longer available

The next issue is the baffle. A custom PC CARD baffle might not be worth it, so I'm looking for an off-the-shelf one from AliExpress, but I'm not finding one that has all the ports I'd like. A 3D printed one would technically work, but a metal bracket off-the-shelf would be preferred.

The attachment Screenshot 2025-08-03 044450.png is no longer available

The attachment Screenshot 2025-08-03 044513.png is no longer available

Another fantastic contribution to the community! You're right—you’ll likely need a PC/104+ board for proper testing. The 90-degree rotated version looks excellent. I believe my PCM-3370 would fit nicely, and conveniently, the CF slot would face upwards. However, there might be limited clearance between the heatsink and the EISA connector—see the PCM-3370 board below for reference.

I'd be happy to send you mine for testing if that helps. I’m sure we could work something out, though I’m based in the UK, and unfortunately, shipping costs and potential import duties do make things a bit tricky.

I could still do a "standard" orientation but routing might get tricky. The ISA block would be fine with a 2-layer PCB, but adding the PCI would probably need a 4-layer PCB minimum...

As far as clearance, I'd probably have to get an Advantech PCM PC/104 SBC. And, if I use the PCM-3370 as a base, I'd need to verify the dimensions and pin pitch on some of those connectors.

From the pic you posted, although it's not an isometric image, it "appears" to clear the bottom edge from the heatsink while just sticking out over the top. Right now, there's about 24mm from the bottom mounting hole to the edge of the connector pins.

The attachment Screenshot 2025-08-03 234100.png is no longer available

Thank you, BTW, for the offer of sending me your SBC! The shipping and responsibility to send it back might not be worth it. While the shipping of the backplane to you was about $35, I didn't take into account the tax, which pushed that over $50. Don't worry, I'll eat that cost, as I'm still new to shipping internationally. 😅😎

I've purchased a cheap PCM-3353 from eBay that's probably cheaper than the two-way shipping for your SBC. While it has a different set of external connections, it is a PC/104+ card that I could verify works on the adaptor card.

This is getting extremely interesting! The adapter will likely consist of just connectors, yet it might require a 4-layer PCB. The PCM-3353 board looks excellent—very interesting that it uses an AMD chipset and offers four USB ports, compared to only two on the PCM-3370. In your design, the CF port would be facing upwards, which seems like the ideal orientation. Just make sure the PCM-3353 has both the ISA and PCI connectors physically populated. Some SBCs don’t include both on the PCB by default—specifically, you’ll need the long header pins on the underside.

I've been thinking about the baffle—it’s going to be tricky to accommodate all the ports you might want there. Considering that people (myself included) may need to make their own custom cables from the SBC to the adapter, it’s probably best to prioritise the most essential connectors. I’d say a VGA and a PS/2 port would be the most universally useful, similar to that AliExpress ISA adapter I shared in a previous post. All other ports can be routed to an external bracket, although finding suitable off-the-shelf options can still be challenging.

Sorry to hear you had to pay extra on taxes—I really appreciate it, and I’ll make sure to compensate you next time. Thanks again!

mbandalauk wrote on 2025-08-04, 17:14:

This is getting extremely interesting! The adapter will likely consist of just connectors, yet it might require a 4-layer PCB. T […]
Show full quote

Tevian wrote on 2025-08-04, 07:09:

I could still do a "standard" orientation but routing might get tricky. The ISA block would be fine with a 2-layer PCB, but addi […]
Show full quote

mbandalauk wrote on 2025-08-03, 23:25:

Another fantastic contribution to the community! You're right—you’ll likely need a PC/104+ board for proper testing. The 90-degree rotated version looks excellent. I believe my PCM-3370 would fit nicely, and conveniently, the CF slot would face upwards. However, there might be limited clearance between the heatsink and the EISA connector—see the PCM-3370 board below for reference.

I'd be happy to send you mine for testing if that helps. I’m sure we could work something out, though I’m based in the UK, and unfortunately, shipping costs and potential import duties do make things a bit tricky.

I could still do a "standard" orientation but routing might get tricky. The ISA block would be fine with a 2-layer PCB, but adding the PCI would probably need a 4-layer PCB minimum...

As far as clearance, I'd probably have to get an Advantech PCM PC/104 SBC. And, if I use the PCM-3370 as a base, I'd need to verify the dimensions and pin pitch on some of those connectors.

From the pic you posted, although it's not an isometric image, it "appears" to clear the bottom edge from the heatsink while just sticking out over the top. Right now, there's about 24mm from the bottom mounting hole to the edge of the connector pins.

The attachment Screenshot 2025-08-03 234100.png is no longer available

Thank you, BTW, for the offer of sending me your SBC! The shipping and responsibility to send it back might not be worth it. While the shipping of the backplane to you was about $35, I didn't take into account the tax, which pushed that over $50. Don't worry, I'll eat that cost, as I'm still new to shipping internationally. 😅😎

I've purchased a cheap PCM-3353 from eBay that's probably cheaper than the two-way shipping for your SBC. While it has a different set of external connections, it is a PC/104+ card that I could verify works on the adaptor card.

This is getting extremely interesting! The adapter will likely consist of just connectors, yet it might require a 4-layer PCB. The PCM-3353 board looks excellent—very interesting that it uses an AMD chipset and offers four USB ports, compared to only two on the PCM-3370. In your design, the CF port would be facing upwards, which seems like the ideal orientation. Just make sure the PCM-3353 has both the ISA and PCI connectors physically populated. Some SBCs don’t include both on the PCB by default—specifically, you’ll need the long header pins on the underside.

I've been thinking about the baffle—it’s going to be tricky to accommodate all the ports you might want there. Considering that people (myself included) may need to make their own custom cables from the SBC to the adapter, it’s probably best to prioritise the most essential connectors. I’d say a VGA and a PS/2 port would be the most universally useful, similar to that AliExpress ISA adapter I shared in a previous post. All other ports can be routed to an external bracket, although finding suitable off-the-shelf options can still be challenging.

Sorry to hear you had to pay extra on taxes—I really appreciate it, and I’ll make sure to compensate you next time. Thanks again!

No biggy about the shipping cost. I need to learn about it so I know for future shipments. 😅

I did a preliminary route of the ISA block and I'm still leaning towards 4 layer. The stack up will probably be; signal/power-ground-signal/power-signal/power. Like I said, 2 layer would work for ISA only, but the added roughly 100 traces for the PCI would get very tight. Having an inner ground layer means I can route easier on 3 layers.

Baffle aside, the key ports will be needed. VGA, keyboard mouse, USB, and Ethernet. There should be room for those on the standard 100mm baffle. I can see now why the Allen Bradley cards have the 110mm non-standard height for extra ports. Anything else could easily be used on a side bracket. Pinouts for the ports might be tricky as I don't know if there's an industry standard for something a 2 row 2.54mm VGA header... the PCM 3353 I bought doesn't have the same ones on the 3370. I have the stuff to make custom cables, so I guess I'll just use a standard 2.54mm pin header with pinouts according to the 3370 datasheet.

I'll do more research on this when I get a minute.

I bought another 3353 from ebay that has a bag of all the header cables. It's claims to be not working, but for $35, I will have two PC104+ cards for testing. 😄

I can share hundreds of Asus boards in kicad format.
Almost any footprint you desire can be found.
Vga is a pretty standard footprint.

Re: RDRAM vs SDRAM suprisingly not much faster

myne wrote on 2025-08-05, 03:04:

I can share hundreds of Asus boards in kicad format. Almost any footprint you desire can be found. Vga is a pretty standard foo […]
Show full quote

I can share hundreds of Asus boards in kicad format.
Almost any footprint you desire can be found.
Vga is a pretty standard footprint.

Re: RDRAM vs SDRAM suprisingly not much faster

When I saw industry standard, I'm saying, "Is there an industry standard pin header for VGA pigtails?" The different manufacturers of these PC104 boards probably run the gamut as far as pinout, pin pitch, ribbon or flex cable, etc... If Advantech for instance had a standard "PCB connector" for there VGA, I could use that as a base. However, even the ones I just bought (PCM-3353) have a different PCB connector for VGA vs (PCM-3370). Meaning I'd need a custom cable regardless of how I implement this on this adapter card. Stuff like serial, keyboard mouse, and USB seem much more "standard". I'm going to do my best to avoid custom cabling, but it might be difficult with the absences of an "industry standard" PCM VGA cable to PCB header for VGA brake out...😉

Oh, the pin header?
I thought it was pretty standard - albeit there might be different pitch pinouts.

I received one of the PCM-3353 today! This was a steal for $35, as not only does it work as far as I can tell, but it came with all the cabling, which is a must for these PC104 boards.

This has the same relative footprint as the PCM-3370 with different headers. However, now I can see that the vertical mounting would just be ugly! So, I've begun to route for the "normal" horizontal mount PC104... This is a bit more complicated, but will make the PCB about 20mm shorter and much more appealing with the CF card on the left-hand side.

I'm going to focus on this as the port arrangement. VGA, Ethernet, USB, and Keyboard/Mouse. The baffle would need to be custom, possibly printed or laser cut from a blank, but it's all within the 100mm standard. The need to double route between the pin header was a must. On further examination of my PCISA 400C board, there are even triple routing in some cases, with some very small pitch traces, so double routing seems fine.

One interesting potential issue was the PCI IDSEL pins... These IDSEL pins on the PCISA backplane are only present on the PCI slots and are not carried over from the SBC. These IDSEL pins are grounded to a specific PCI address pin, setting that specific card's PCI ID. On my backplane, I accomplish the cross-compatibility with the use of a configurable card for the IDSEL and INT pins since these are specific to the manufacturer of the SBC.

The INT routing is in the PC/104+ spec doc, but it's oddly absent on the IDSEL to address pin configuration. I ran into this issue with both Allen Bradley and PISA/PCISA SBC. You can have non-interrupt PCI function with the IDSEL pin alone, but you need the 4 INT# pins for proper interrupts. But, if the IDSEL pin is grounded to the wrong assumed address pin by the BIOS, there is no PCI function at all! The PISA and PCISA have the same IDSEL to ADDR pins with different INT# arrangements, and the Allen Bradley I had to reverse engineer.

The PC/104+ card actually has the 4 IDSEL pins accessible on the 120-pin PCI header. This differs from the PISA backplane as the backplane doesn't carry these pins to the PCI slots in favor of those slots having the IDSEL pins "hardwired" on the backplane. Or, my PCI Mapper card in the case of my backplane. The PC/104+ is the backplane in this case, as it's designed to have cards plugged directly onto it. So, the IDSEL pins are present in the connector and hardwired in the module! They will need to be left floating for the adaptor card, while all standard PCI pins will be routed just like a PCISA SBC. That means, in the case of my backplane, a card can be made to use the first two PCI slot addresses. I've reverse-engineered the PCM-3353 and found this configuration:

1IDSEL0 = AD20
2IDSEL1 = AD21
3IDSEL2 = AD27
4IDSEL3 = AD23

This is an odd numbering and is not compatible with the PCISA backplane, which is hardwired for this configuration:

1IDSEL0 = AD19
2IDSEL1 = AD20
3IDSEL2 = AD21
4IDSEL3 = AD22

Since these IDSEL pins are prewired on the PC/104+ CPU module, full compatibility with PCISA backplanes may not be possible. The only common address is AD20, so it's possible that a non-interrupt PCI card could be used. Since my backplane is fully configurable, a PCI mapper card with the right config should work fine. This will need to be tested once a prototype adaptor card is made, though.

Luckily, the PC/104+ ISA and PCI headers were cheap, so I ordered a small number to assemble some prototypes once I feel comfortable with the routing. No timeline yet, but I'll keep at it. 😅😎

Tevian wrote on 2025-08-07, 11:10:

I received one of the PCM-3353 today! This was a steal for $35, as not only does it work as far as I can tell, but it came with […]
Show full quote

I received one of the PCM-3353 today! This was a steal for $35, as not only does it work as far as I can tell, but it came with all the cabling, which is a must for these PC104 boards.

The attachment 1000004321.jpg is no longer available

The attachment 1000004320.jpg is no longer available

This has the same relative footprint as the PCM-3370 with different headers. However, now I can see that the vertical mounting would just be ugly! So, I've begun to route for the "normal" horizontal mount PC104... This is a bit more complicated, but will make the PCB about 20mm shorter and much more appealing with the CF card on the left-hand side.

The attachment Screenshot 2025-08-07 032130.png is no longer available

The attachment Screenshot 2025-08-07 032151.png is no longer available

I'm going to focus on this as the port arrangement. VGA, Ethernet, USB, and Keyboard/Mouse. The baffle would need to be custom, possibly printed or laser cut from a blank, but it's all within the 100mm standard. The need to double route between the pin header was a must. On further examination of my PCISA 400C board, there are even triple routing in some cases, with some very small pitch traces, so double routing seems fine.

One interesting potential issue was the PCI IDSEL pins... These IDSEL pins on the PCISA backplane are only present on the PCI slots and are not carried over from the SBC. These IDSEL pins are grounded to a specific PCI address pin, setting that specific card's PCI ID. On my backplane, I accomplish the cross-compatibility with the use of a configurable card for the IDSEL and INT pins since these are specific to the manufacturer of the SBC.

The INT routing is in the PC/104+ spec doc, but it's oddly absent on the IDSEL to address pin configuration. I ran into this issue with both Allen Bradley and PISA/PCISA SBC. You can have non-interrupt PCI function with the IDSEL pin alone, but you need the 4 INT# pins for proper interrupts. But, if the IDSEL pin is grounded to the wrong assumed address pin by the BIOS, there is no PCI function at all! The PISA and PCISA have the same IDSEL to ADDR pins with different INT# arrangements, and the Allen Bradley I had to reverse engineer.

The PC/104+ card actually has the 4 IDSEL pins accessible on the 120-pin PCI header. This differs from the PISA backplane as the backplane doesn't carry these pins to the PCI slots in favor of those slots having the IDSEL pins "hardwired" on the backplane. Or, my PCI Mapper card in the case of my backplane. The PC/104+ is the backplane in this case, as it's designed to have cards plugged directly onto it. So, the IDSEL pins are present in the connector and hardwired in the module! They will need to be left floating for the adaptor card, while all standard PCI pins will be routed just like a PCISA SBC. That means, in the case of my backplane, a card can be made to use the first two PCI slot addresses. I've reverse-engineered the PCM-3353 and found this configuration:

Copy code to clipboard

1IDSEL0 = AD20
2IDSEL1 = AD21
3IDSEL2 = AD27
4IDSEL3 = AD23

This is an odd numbering and is not compatible with the PCISA backplane, which is hardwired for this configuration:

Copy code to clipboard

1IDSEL0 = AD19
2IDSEL1 = AD20
3IDSEL2 = AD21
4IDSEL3 = AD22

Since these IDSEL pins are prewired on the PC/104+ CPU module, full compatibility with PCISA backplanes may not be possible. The only common address is AD20, so it's possible that a non-interrupt PCI card could be used. Since my backplane is fully configurable, a PCI mapper card with the right config should work fine. This will need to be tested once a prototype adaptor card is made, though.

Luckily, the PC/104+ ISA and PCI headers were cheap, so I ordered a small number to assemble some prototypes once I feel comfortable with the routing. No timeline yet, but I'll keep at it. 😅😎

First of all, what a fantastic find! Scoring a PCM-3353 for just $35 is a great deal. I hadn’t realized how much thought had to go into the PCI slot layout—maybe that’s part of the reason why we don’t see any PC/104+ to EISA edge adapters out there.

Sorry to circle back to the topic of baffles and brackets, but the ports you’ve chosen are genuinely useful. It’s a bit of a shame that they still require a custom bracket to be fully functional.

Speaking of brackets, I noticed your backplane has a pin header for joysticks. Could that be converted into a standard game port? Maybe by using a cable like the one in the picture below—or even better, something that could be mounted directly on a bracket?

mbandalauk wrote on 2025-08-08, 00:01:

First of all, what a fantastic find! Scoring a PCM-3353 for just $35 is a great deal. I hadn’t realized how much thought had to […]
Show full quote

Tevian wrote on 2025-08-07, 11:10:

I received one of the PCM-3353 today! This was a steal for $35, as not only does it work as far as I can tell, but it came with […]
Show full quote

I received one of the PCM-3353 today! This was a steal for $35, as not only does it work as far as I can tell, but it came with all the cabling, which is a must for these PC104 boards.

The attachment 1000004321.jpg is no longer available

The attachment 1000004320.jpg is no longer available

This has the same relative footprint as the PCM-3370 with different headers. However, now I can see that the vertical mounting would just be ugly! So, I've begun to route for the "normal" horizontal mount PC104... This is a bit more complicated, but will make the PCB about 20mm shorter and much more appealing with the CF card on the left-hand side.

The attachment Screenshot 2025-08-07 032130.png is no longer available

The attachment Screenshot 2025-08-07 032151.png is no longer available

I'm going to focus on this as the port arrangement. VGA, Ethernet, USB, and Keyboard/Mouse. The baffle would need to be custom, possibly printed or laser cut from a blank, but it's all within the 100mm standard. The need to double route between the pin header was a must. On further examination of my PCISA 400C board, there are even triple routing in some cases, with some very small pitch traces, so double routing seems fine.

One interesting potential issue was the PCI IDSEL pins... These IDSEL pins on the PCISA backplane are only present on the PCI slots and are not carried over from the SBC. These IDSEL pins are grounded to a specific PCI address pin, setting that specific card's PCI ID. On my backplane, I accomplish the cross-compatibility with the use of a configurable card for the IDSEL and INT pins since these are specific to the manufacturer of the SBC.

The INT routing is in the PC/104+ spec doc, but it's oddly absent on the IDSEL to address pin configuration. I ran into this issue with both Allen Bradley and PISA/PCISA SBC. You can have non-interrupt PCI function with the IDSEL pin alone, but you need the 4 INT# pins for proper interrupts. But, if the IDSEL pin is grounded to the wrong assumed address pin by the BIOS, there is no PCI function at all! The PISA and PCISA have the same IDSEL to ADDR pins with different INT# arrangements, and the Allen Bradley I had to reverse engineer.

The PC/104+ card actually has the 4 IDSEL pins accessible on the 120-pin PCI header. This differs from the PISA backplane as the backplane doesn't carry these pins to the PCI slots in favor of those slots having the IDSEL pins "hardwired" on the backplane. Or, my PCI Mapper card in the case of my backplane. The PC/104+ is the backplane in this case, as it's designed to have cards plugged directly onto it. So, the IDSEL pins are present in the connector and hardwired in the module! They will need to be left floating for the adaptor card, while all standard PCI pins will be routed just like a PCISA SBC. That means, in the case of my backplane, a card can be made to use the first two PCI slot addresses. I've reverse-engineered the PCM-3353 and found this configuration:

Copy code to clipboard

1IDSEL0 = AD20
2IDSEL1 = AD21
3IDSEL2 = AD27
4IDSEL3 = AD23

This is an odd numbering and is not compatible with the PCISA backplane, which is hardwired for this configuration:

Copy code to clipboard

1IDSEL0 = AD19
2IDSEL1 = AD20
3IDSEL2 = AD21
4IDSEL3 = AD22

Since these IDSEL pins are prewired on the PC/104+ CPU module, full compatibility with PCISA backplanes may not be possible. The only common address is AD20, so it's possible that a non-interrupt PCI card could be used. Since my backplane is fully configurable, a PCI mapper card with the right config should work fine. This will need to be tested once a prototype adaptor card is made, though.

Luckily, the PC/104+ ISA and PCI headers were cheap, so I ordered a small number to assemble some prototypes once I feel comfortable with the routing. No timeline yet, but I'll keep at it. 😅😎

First of all, what a fantastic find! Scoring a PCM-3353 for just $35 is a great deal. I hadn’t realized how much thought had to go into the PCI slot layout—maybe that’s part of the reason why we don’t see any PC/104+ to EISA edge adapters out there.

Sorry to circle back to the topic of baffles and brackets, but the ports you’ve chosen are genuinely useful. It’s a bit of a shame that they still require a custom bracket to be fully functional.

Speaking of brackets, I noticed your backplane has a pin header for joysticks. Could that be converted into a standard game port? Maybe by using a cable like the one in the picture below—or even better, something that could be mounted directly on a bracket?

Did you open the package I sent? If not, there's a surprise inside. 😜 Hint, it looks much like the pics in your post.

Better yet, here's the pic of what I sent you. 😁Notice it now?

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PCI combo backplanes and these PC/104+ cards.

When I researched my backplane, I'm still far from an expert, I discovered this PCI ID scheme and the issues with cross-compatibility. The issue is these IDSEL configurations seem to.be BIOS dependent. Meaning this configuration is internal and not adjustable unless a BIOS specifically allows for changing the PCI config order.

The backplanes seem to have this configuration hardwired and thus, a BIOS is configured with that PCI order. For instance, this is why there's an alternate configuration on my backplane for PISA and PCISA. In that case the PCISA in name only used the PISA specification but altered the Interrupted order for the second PCI slot which caused me some headaches.

For the PC/104+ to backplane operation, all.that needs to happen is to emulate the IDSEL configuration that's hardwired into the PC/104+ CPU module. Simple with my backplane and the PCI Mapper Card, but on the "real" PCISA backplanes, these are hardwired to match their SBC cards and BIOS. It could work if you can cut the traces to the PCI IDSEL pins and connect them to the corresponding ADDR pins, but that's unlikely and destructive.

So, it should work on my backplane fine, and the ISA should work regardless on the other backplanes since it has no configuration pins like PCI.

Tevian wrote on 2025-08-08, 01:14:

Better yet, here's the pic of what I sent you. 😁Notice it now?

The attachment 1000004287.jpg is no longer available

An image is worth more than a thousand words! Thanks.

Tevian wrote on 2025-08-08, 01:43:

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PC […]
Show full quote

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PCI combo backplanes and these PC/104+ cards.

When I researched my backplane, I'm still far from an expert, I discovered this PCI ID scheme and the issues with cross-compatibility. The issue is these IDSEL configurations seem to.be BIOS dependent. Meaning this configuration is internal and not adjustable unless a BIOS specifically allows for changing the PCI config order.

The backplanes seem to have this configuration hardwired and thus, a BIOS is configured with that PCI order. For instance, this is why there's an alternate configuration on my backplane for PISA and PCISA. In that case the PCISA in name only used the PISA specification but altered the Interrupted order for the second PCI slot which caused me some headaches.

For the PC/104+ to backplane operation, all.that needs to happen is to emulate the IDSEL configuration that's hardwired into the PC/104+ CPU module. Simple with my backplane and the PCI Mapper Card, but on the "real" PCISA backplanes, these are hardwired to match their SBC cards and BIOS. It could work if you can cut the traces to the PCI IDSEL pins and connect them to the corresponding ADDR pins, but that's unlikely and destructive.

So, it should work on my backplane fine, and the ISA should work regardless on the other backplanes since it has no configuration pins like PCI.

Does that mean you’ll need to make another mapper board for that purpose? It seems like a straightforward solution, but it could complicate things once you have more components—especially with the risk of losing or misplacing them if you're not careful. I wonder if version 1.1 could include a small microcontroller to handle all the mapping options instead.

mbandalauk wrote on 2025-08-08, 16:22:

Tevian wrote on 2025-08-08, 01:43:

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PC […]
Show full quote

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PCI combo backplanes and these PC/104+ cards.

When I researched my backplane, I'm still far from an expert, I discovered this PCI ID scheme and the issues with cross-compatibility. The issue is these IDSEL configurations seem to.be BIOS dependent. Meaning this configuration is internal and not adjustable unless a BIOS specifically allows for changing the PCI config order.

The backplanes seem to have this configuration hardwired and thus, a BIOS is configured with that PCI order. For instance, this is why there's an alternate configuration on my backplane for PISA and PCISA. In that case the PCISA in name only used the PISA specification but altered the Interrupted order for the second PCI slot which caused me some headaches.

For the PC/104+ to backplane operation, all.that needs to happen is to emulate the IDSEL configuration that's hardwired into the PC/104+ CPU module. Simple with my backplane and the PCI Mapper Card, but on the "real" PCISA backplanes, these are hardwired to match their SBC cards and BIOS. It could work if you can cut the traces to the PCI IDSEL pins and connect them to the corresponding ADDR pins, but that's unlikely and destructive.

So, it should work on my backplane fine, and the ISA should work regardless on the other backplanes since it has no configuration pins like PCI.

Does that mean you’ll need to make another mapper board for that purpose? It seems like a straightforward solution, but it could complicate things once you have more components—especially with the risk of losing or misplacing them if you're not careful. I wonder if version 1.1 could include a small microcontroller to handle all the mapping options instead.

The mapper cards are reconfigureable by moving the jumper resistors. It takes a soldering iron, but the cards are open to any configuration of INT pins and IDSEL to AD pins from AD19 - 31.

This covers "most" possible ranges. For the PCM PC/104+ boards, it's a simple matter of emulating the hardwired IDSEL configuration to one side of the cards. I included two cards for you, Allen Bradley / PISA and PCISA / Blank. The blank side can be populated for the PCM config.

The card idea was done for ease of use, and I assumed people would settle on one type of SBC to use and not need to swap out the card. However, if you testing or using different types, the cards would need to be swapped for proper PCI function.

The card slot was a compromise for space on the board vs a forest of jumper blocks to allow for reconfiguration. But, if you lost the cards, it would be an issue. I haven't put the Gerbers online for those yet, but will soon.

The number of blocks is great, nearly 60! But I might reassess a jumper block for V1.1. I'll think about it. 🫣😄

mbandalauk wrote on 2025-08-08, 16:22:

Tevian wrote on 2025-08-08, 01:43:

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PC […]
Show full quote

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PCI combo backplanes and these PC/104+ cards.

When I researched my backplane, I'm still far from an expert, I discovered this PCI ID scheme and the issues with cross-compatibility. The issue is these IDSEL configurations seem to.be BIOS dependent. Meaning this configuration is internal and not adjustable unless a BIOS specifically allows for changing the PCI config order.

The backplanes seem to have this configuration hardwired and thus, a BIOS is configured with that PCI order. For instance, this is why there's an alternate configuration on my backplane for PISA and PCISA. In that case the PCISA in name only used the PISA specification but altered the Interrupted order for the second PCI slot which caused me some headaches.

For the PC/104+ to backplane operation, all.that needs to happen is to emulate the IDSEL configuration that's hardwired into the PC/104+ CPU module. Simple with my backplane and the PCI Mapper Card, but on the "real" PCISA backplanes, these are hardwired to match their SBC cards and BIOS. It could work if you can cut the traces to the PCI IDSEL pins and connect them to the corresponding ADDR pins, but that's unlikely and destructive.

So, it should work on my backplane fine, and the ISA should work regardless on the other backplanes since it has no configuration pins like PCI.

Does that mean you’ll need to make another mapper board for that purpose? It seems like a straightforward solution, but it could complicate things once you have more components—especially with the risk of losing or misplacing them if you're not careful. I wonder if version 1.1 could include a small microcontroller to handle all the mapping options instead.

mbandalauk, could you fill in the blanks for me? The IDSEL configuration for these boards could be different, and I'm wondering if you could probe and find the 4 IDSEL connections to the relevant PCI AD pins. For instance, my PCM-3353 has this configuration:

1(PCM-3353)
2IDSEL0 = AD20
3IDSEL1 = AD21
4IDSEL2 = AD27
5IDSEL3 = AD23

These IDSEL# pins are connected to each AD# pins directly and can be discovered using a simple continuity meter (0 ohm connections). This is what that looks like on my board from the top down:

If you can, please fill in these numbers:

1(PCM-3370)
2IDSEL0 = ?
3IDSEL1 = ?
4IDSEL2 = ?
5IDSEL3 = ?

Tevian wrote on 2025-08-11, 11:10:

mbandalauk, could you fill in the blanks for me? The IDSEL configuration for these boards could be different, and I'm wondering […]
Show full quote

mbandalauk wrote on 2025-08-08, 16:22:

Tevian wrote on 2025-08-08, 01:43:

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PC […]
Show full quote

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PCI combo backplanes and these PC/104+ cards.

When I researched my backplane, I'm still far from an expert, I discovered this PCI ID scheme and the issues with cross-compatibility. The issue is these IDSEL configurations seem to.be BIOS dependent. Meaning this configuration is internal and not adjustable unless a BIOS specifically allows for changing the PCI config order.

The backplanes seem to have this configuration hardwired and thus, a BIOS is configured with that PCI order. For instance, this is why there's an alternate configuration on my backplane for PISA and PCISA. In that case the PCISA in name only used the PISA specification but altered the Interrupted order for the second PCI slot which caused me some headaches.

For the PC/104+ to backplane operation, all.that needs to happen is to emulate the IDSEL configuration that's hardwired into the PC/104+ CPU module. Simple with my backplane and the PCI Mapper Card, but on the "real" PCISA backplanes, these are hardwired to match their SBC cards and BIOS. It could work if you can cut the traces to the PCI IDSEL pins and connect them to the corresponding ADDR pins, but that's unlikely and destructive.

So, it should work on my backplane fine, and the ISA should work regardless on the other backplanes since it has no configuration pins like PCI.

Does that mean you’ll need to make another mapper board for that purpose? It seems like a straightforward solution, but it could complicate things once you have more components—especially with the risk of losing or misplacing them if you're not careful. I wonder if version 1.1 could include a small microcontroller to handle all the mapping options instead.

mbandalauk, could you fill in the blanks for me? The IDSEL configuration for these boards could be different, and I'm wondering if you could probe and find the 4 IDSEL connections to the relevant PCI AD pins. For instance, my PCM-3353 has this configuration:

Copy code to clipboard

1(PCM-3353)
2IDSEL0 = AD20
3IDSEL1 = AD21
4IDSEL2 = AD27
5IDSEL3 = AD23

These IDSEL# pins are connected to each AD# pins directly and can be discovered using a simple continuity meter (0 ohm connections). This is what that looks like on my board from the top down:

The attachment Screenshot 2025-08-11 040202.png is no longer available

The attachment Screenshot 2025-08-02 025303.png is no longer available

If you can, please fill in these numbers:

Copy code to clipboard

1(PCM-3370)
2IDSEL0 = ?
3IDSEL1 = ?
4IDSEL2 = ?
5IDSEL3 = ?

Hi! No problem—I can probe those pins on my board and share the results here. I’m still abroad visiting family, but I’ll be back in the UK next week and can run the tests then.

Getting close to a potential first version of the adapter card. I only have a few questions left about implementation, but I'm reading through the datasheet for answers. In fact, I've found a few errors in the documentation about header pinout! In addition to the headers, I've decided to include a back-mounted ATX power connector. This might be useful for some modules, and in the case of the PCM-3370, I included the ATX power switch, which includes the PWON and 5VSB, so the module can control the power supply. In case you want the module to be able to wake on LAN, for instance. Minor, but it could also be used to power on the module on the bench without the PCM power cabling also.

Update on the "Backwards PC104+ Adaptor Card"

Good news and not so good news (for me anyway)...😋 And, if you're reading this and maybe want to try one, feel free to DM. However!!!! As you can see, this card is special and will not fully work in a genuine PISA/PCISA backplane. At least not the PCI function, as these backplanes are hardwired for IDESL# and INT#. This card is truly made for my Backwards Backplane. The power and ISA passthrough should. Meaning the ATX power connector on the card itself can pass power back to the backplane, and I've tested this on my IEI PCISA-compatible backplane. Oh, and most importantly, there is no slot bracket yet! Still looking into that one!

I haven't tested every part, but the V1.0 board seems to be functional! The PC104 module can draw power from the PCISA slot or its onboard 20-pin ATX power connector. I still need to test the USB and Ethernet passthrough. I found a few mistakes in the PCM-3353 datasheet! The pinout order doesn't match the real pinout order on the physical board on some ports. 😬 Glad I double checked with a meter.

I hit a snag on first power-up. I was happy to see it powered up, meaning I routed it correctly. However, I got no ISA function! I mean zero! It took a few hours of troubleshooting and probing to see if all the PC104 ISA pins had continuity. Everything looked good, so I moved to check the signal on the scope. The rise and fall times all looked good. So, why wouldn't it see any ISA cards, including my ESS sound chip built into my backplane?

Upon probing, the ISA (MEMCS16) and (IOCS16) stayed HIGH all the time, even during boot or using something like UNISOUND to poke the sound cards. Both of my recently acquired PCM-3353 modules behaved the same way, so I was left assuming a signal or routing issue. Or even my termination, possibly interfering with the PC104 ISA bus. The BIOS on these cards has all the usual options, except that the resource allocation menu is lacking a few key things. Specifically, the ability to reserve DMA channels.

These PCM-3353 use an AMD Geode LX800. This is the clue, and during my search, I ran into how these AMD boards support the legacy IO. The companion chip AMD Geode CS5536 is the SouthBridge chip that "emulates" the ISA bus, or uses some other bridge IC that may not fully support legacy ISA! 🤦‍♂️🤦‍♂️ I didn't dig into the datasheet for this chip, but I did find another whitepaper that mentions something important.

---------------------

Issue:
PC/104 (ISA) expansion cards may exhibit problems or not work on some AMD Geode-based
CPU boards.
Background:
A number of VersaLogic SBCs use AMD Geode™ processors; namely, the GX 500 and LX 800
processors. The architecture of these chipsets does not include a PC/104 (ISA) interface. So
VersaLogic implemented a PLD that functions as an LPC-to-ISA bridge to accommodate the
use of ISA devices with these products.
Because the Low Pin Count (LPC) bus is an 8-signal interface, certain limitations apply to ISA
transfers. Memory reads and writes are limited to eight bits and DMA transfers are not
supported. Early revisions of the EPIC-2 (Gecko) and EPM-5 (Puma) also limited I/O transfers
to eight bits, but current versions support 16-bit transfers.
The LPC-to-ISA bridge used on VersaLogic Geode-based SBCs has the following
characteristics:
 The memory range addressable by the bridge is C8000h to DBFFFh.
 Generally, the I/O range supported is 0x100 to 0x3FF (see the SBC's reference manual
for exclusions).
 Subtractive decoding is used to route traffic from the PCI bus to the LPC bus; however,
the LPC-to-ISA bridge operates in a positive decode mode.
 The bridge typically supports a limited number of interrupts (see reference manual).
 The bridge does not support:
o DMA
o zero wait state cycles
o memory refresh cycles
o bus master mode
o 16-bit memory transfers
Recommendations:
Given the architecture of Geode processors and the VersaLogic LPC-to-ISA bridge, a PC/104
(ISA) board will work on Geode-based SBCs provided the PC/104 board:
 Requires only 8-bit memory transfers.
 Does not require an interrupt other than those supported by the board (see reference
manual).
 Does not require zero wait states.
 Is not a bus master.
 Does not use DMA.

---------------------

While this is for VersaLogic systems, the PCM modules may have the same or a similar architecture choice. If that's the case, there is no 16-bit ISA compatibility (even though the 16-bit extension connector is there!), and NO DMA support for ISA! This means no ISA sound cards! 🤦‍♂️🤦‍♂️ And, since I don't have any 8-bit ISA cards to test with, I left with the conclusion that my PCB routing is most likely correct, and an older PC104 board would work fine.

So far, the PCI function seems to work, and my PCI MAPPER cards indeed work as expected by emulating the IDSEL# and INT# hard-wired into the PCM module. Here's it working with a PCI graphics card.

This is my first PCB using a beveled edge. I'm pleased with JLCPCB's work on this!

Tevian wrote on 2025-08-11, 11:10:

mbandalauk, could you fill in the blanks for me? The IDSEL configuration for these boards could be different, and I'm wondering […]
Show full quote

mbandalauk wrote on 2025-08-08, 16:22:

Tevian wrote on 2025-08-08, 01:43:

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PC […]
Show full quote

You're not wrong about the lack of an EISA style adaptor card in the wild that's compatible with something like these old ISA/PCI combo backplanes and these PC/104+ cards.

When I researched my backplane, I'm still far from an expert, I discovered this PCI ID scheme and the issues with cross-compatibility. The issue is these IDSEL configurations seem to.be BIOS dependent. Meaning this configuration is internal and not adjustable unless a BIOS specifically allows for changing the PCI config order.

The backplanes seem to have this configuration hardwired and thus, a BIOS is configured with that PCI order. For instance, this is why there's an alternate configuration on my backplane for PISA and PCISA. In that case the PCISA in name only used the PISA specification but altered the Interrupted order for the second PCI slot which caused me some headaches.

For the PC/104+ to backplane operation, all.that needs to happen is to emulate the IDSEL configuration that's hardwired into the PC/104+ CPU module. Simple with my backplane and the PCI Mapper Card, but on the "real" PCISA backplanes, these are hardwired to match their SBC cards and BIOS. It could work if you can cut the traces to the PCI IDSEL pins and connect them to the corresponding ADDR pins, but that's unlikely and destructive.

So, it should work on my backplane fine, and the ISA should work regardless on the other backplanes since it has no configuration pins like PCI.

Does that mean you’ll need to make another mapper board for that purpose? It seems like a straightforward solution, but it could complicate things once you have more components—especially with the risk of losing or misplacing them if you're not careful. I wonder if version 1.1 could include a small microcontroller to handle all the mapping options instead.

mbandalauk, could you fill in the blanks for me? The IDSEL configuration for these boards could be different, and I'm wondering if you could probe and find the 4 IDSEL connections to the relevant PCI AD pins. For instance, my PCM-3353 has this configuration:

Copy code to clipboard

1(PCM-3353)
2IDSEL0 = AD20
3IDSEL1 = AD21
4IDSEL2 = AD27
5IDSEL3 = AD23

These IDSEL# pins are connected to each AD# pins directly and can be discovered using a simple continuity meter (0 ohm connections). This is what that looks like on my board from the top down:

The attachment Screenshot 2025-08-11 040202.png is no longer available

The attachment Screenshot 2025-08-02 025303.png is no longer available

If you can, please fill in these numbers:

Copy code to clipboard

1(PCM-3370)
2IDSEL0 = ?
3IDSEL1 = ?
4IDSEL2 = ?
5IDSEL3 = ?

Dear Tevian,

Apologies for the delay in getting this done. Last week I had just returned to work when I realized someone had borrowed my multimeter, and then I spent three days in rainy Cumbria for work.

In the end, the test was straightforward, and here are the results:

1(PCM-3370)
2IDSEL0 = AD20
3IDSEL1 = AD21
4IDSEL2 = AD22
5IDSEL3 = AD23

I hope this is useful. Thanks again.

Best regards,
Manuel

Hello again,

Now that I’m back, I can share my findings using the PCM-3370 on the backplane. In short — it works flawlessly!

I mounted the PCM-3370 on a PCA-104-ISA REV.A1 sourced from AliExpress. The SBC can be powered directly from the ISA connector, and everything seems to run fine. Both Windows 98 and MS-DOS detect the onboard ESS sound chip without issues. The audio is a little noisy, but nothing that really bothers me.

My only nitpick was with the PSU connector: mine has a 24-pin plug, and as you can see in the second picture, a capacitor is in the way. Because of that, the connector doesn’t fully seat. Fortunately, since it’s actually a 20-pin + 4-pin design, I was able to detach the extra 4 pins without any problem. (It might be worth considering a small layout adjustment in a future revision — if that capacitor were just 2 mm farther from the edge, the connector would fit properly.)

Performance-wise, the SBC is excellent. With its Celeron 400, it boots Windows 98 in under 10 seconds. Since it’s an Intel processor, your PC/104+ adapter should work fine with it, right? If you tried your PCM-3353 with the Chinese PCA-104-ISA, would you get the same results?

Best,
Manuel