mbandalauk wrote on 2025-10-29, 17:19:

Yeah, I was lucky to find my PCM-3370 SBC — it was actually headed for the skip when I spotted it in a pile of scrap boards. These things are quite expensive nowadays! I hope you manage to repair your board and get your adapter tested soon.

By the way, could I know the value of the L2 component on your backplane? While I was testing, a metal connector accidentally touched the gameport connector, and it blew the L2 inductor. Everything else seems to be working fine, but I assume the joystick will not function without that component.

Ouch! Ya, be careful! The way I had to organize the layers meant the top copper pour is all 5V. This is the main reason you need to use the isolation washers on top. L2 is the joystick 5V source and is needed to function. You could use a zero ohm jumper if you had nothing else, but the part is this ->https://www.digikey.com/en/products/detail/ta … 5HS250NT/957833, (25 Ohms @ 100 MHz 1 Power Line Ferrite Bead 0805 (2012 Metric) 6A 4mOhm).

Hi Tevian – Thank you. I will source that ferrite bead and repair my board.

I have been experiencing inconsistent behaviour with the PC/104+ adapter. My Voodoo 3 has always been slightly temperamental and sometimes freezes the system at random, so at first I assumed that was the cause. However, after testing several other graphics cards, they all show similar random freezes when used with the PC/104+ adapter. Even my FX 5500 freezes at times. All of these cards (except the Voodoo) work flawlessly when I use the PCISA-3716E2V with your backplane, which leads me to believe that the issue may relate to how I am using your PCI mapper card with the PC/104+ adapter.

To be fully transparent, I have not used the spare unpopulated PCI mapper board you sent me, because the populated one you provided worked perfectly with the PCISA-3716E2V setup. I have been using this same mapper with the PC/104+ adapter and my PCM-3370 SBC. Sometimes it works for a few minutes, and sometimes it freezes almost immediately. I am beginning to suspect that I may not be using the PCI mapper correctly in this configuration, particularly in relation to the IDSEL numbers I shared with you a few weeks ago. Could this be the root cause of the instability?

That's unfortunate... Since my working PCM module is slightly different in routing, this may need some trouble shooting on your part. I'm typing this mobile, so forgive me...

From what I can tell, the PCM-3370 (based on your probing), is another unique case of differing IDSEL and PCI INIT routing.

For the PCI card to work, the IDSEL must be present. That is unless the card needs IRQ functions. In that case the INIT pins must be routed properly. In my testing, I had strange behavior with some graphics cards needing this IRQ function. And, in the case my ATI card, it would only work in PCI 1 and and not 0...

Regardless, looking at the numbers, the PCISA config is close but different. It's possible your cards can work in one slot but not the other. For instance, they both use IDSEL 20, but not for the same PCI slot, throwing off the INIT routing...

This is the config the mapper card has unless you changed it. https://github.com/MagicPhase/Backwards-DIY-B … pci-mapper-card

I'll have to sit and examine later, but based on the numbers you gave, the correct PCI Mapper card config should be using the standard INIT routing side (Not the ALT routing!) With the IDSEL jumpers set for 20 (Slot 1) and 21 (Slot 2), instead of 19 (Slot 1) and 20 (Slot 2). Important to note, my board is marked (PCI 0 and 1) vs (1 and 2) just to make things complicated. So, you can try to modify the PCISA mapper by only changing the IDSEL pins to match the ones you discovered for your PCM module. Give that a try and let me know.

Oh! And it's important you set the PCI VI/O to 5V! It should be a jumper on you PCM module!

Hi Tevian — I followed all your suggestions from your previous two posts. After changing the configuration on the mapper card, my Voodoo 3 still freezes whenever I try to start a 3D game.

I’ve now managed to get hold of another Voodoo 3, this one a 3000 model, and I’m seeing exactly the same behaviour. I know both of my Voodoo cards are fully functional because I tested them on a regular Socket 370 mini-ATX motherboard using the exact same CPU, RAM, and HDD that I use with the PCISA-3716E2V SBC. On the motherboard everything works perfectly, but on the PCISA SBC both cards freeze as soon as 3D acceleration is invoked.

The Voodoo 3 2000 actually worked fine a few weeks ago on the backplane. The issues began right after I enabled the 3.3V jumper. Even after reverting that jumper, the problem persists. My other PCI cards (FX5500 and GT610) work without any issues, so at this point I’m starting to suspect I may have damaged the backplane during all the testing — which would be a real shame, and strangely it only seems to affect the Voodoo cards.

That's unfortunate to here. The PC104+ adapter board is still a prototype!

I doubt you hurt the backplane, your incident with blowing up the ferite beed not with standing...

The 3.3v jumper is legit and I'm assuming most PCI just ignore that, unless is was needed for some reason. It's in the PCI spec, so I doubt that's the issue.

The PCISA SBC working means the core of the electrical connections are fine.

I angry I bricked my PIII PC104 cards so I can do proper testing. I did get a 5200 GPU, however. I may try to get Windows running my my working AMD Geode board. What I haven't tested on the prototype adaptors is PCI signal integrity. I dealt with this on the ISA bus, but I didn't see any issues with PCI cards with my SBCs.

One fear I have is the PC104 PCI bus not accounting for long runs since it's meant to be stacked cards with very short traces. Extending the traces to +100mm might be causing some instability.

I won't leave you hanging, and I'm also curious why some cards work and not the Voodoo 3. Unfortunately I don't have a V3 to test with, and if there's instability, the V3 might be more suseptible than the other cards.

When I get some time, I'll try to get that 5200 that uses 3.3V on my backplane and get the scope on it.

✌️

Sorry, I was able to sit and read your post a bit closer.

"I’ve now managed to get hold of another Voodoo 3, this one a 3000 model, and I’m seeing exactly the same behaviour. I know both of my Voodoo cards are fully functional because I tested them on a regular Socket 370 mini-ATX motherboard using the exact same CPU, RAM, and HDD that I use with the PCISA-3716E2V SBC. On the motherboard, everything works perfectly, but on the PCISA SBC, both cards freeze as soon as 3D acceleration is invoked."

So, the Voodoo 3 won't work in the blackplane now? That's very strange. Can you send a close-up pic of what config your PCI mapper card is? I want to verify that is matches my PCISA Celeron 400 which is by the same manufacturer, I believe.

It would be hard to tell, but can you measure the current draw on the 5V while the Voodoo 3 card is running? Maybe verify that you aren't getting a voltage drop. I suffered a drop in the 5V when I was using my SLI Voodoo 2 on an IEI manufactured backplane. In the end, I had to solder an extra 5V wire near the PCI slots. Maybe try a different power supply?

I'm sorry it's not working as it should... I'll do my best to troubleshoot with you. I can't justify buying a genuine Voodoo 3 right now to test with. Those PCI versions are insanely expensive! Please treat yours with care!

Also, carefully examine the SBC slots! I've mentioned before, I damaged a lower SBC slot pin by constantly reinserting the cards. I must have done so and not kept the card straight once, but I didn't notice. If you damage any of those pins, it's almost impossible to recover! It's very hard to see, and you may need a loop to see the bottom rows.

I don't have a good explanation right now why it would freeze, other than some Voodoo 3 specific conditions that make it unstable... Do you have another PCISA card you can test with?

Hi — I have two PCISA SBC setups I can test with: a PCISA-3716E2V and a PCM-3370 which workes fine with your PC/104+ adapter. When I first received the backplane, both systems ran the Voodoo 3 2000 without any problems.

Then I started testing other graphics cards that require the 3.3V line. After enabling 3.3V, both the FX 5500 and GT 610 worked perfectly — but the Voodoo began behaving intermittently, eventually reaching the point where it no longer works for 3D acceleration.

Both SBC setups still boot normally with the Voodoo installed, but they now always crash as soon as a game tries to initialise 3D. Since the behaviour is identical across both SBCs, it makes me think your suspicion about the SBC slot might be correct.

I have just tested the SBC slot closer to the edge of the backplane (since using the inner slot leaves no room for a PCI card, even in PCI slot 0). I can rearrange the large heatsink on the 3716E2V to make the fit workable. I swapped SBCs in and out of that slot quite a lot while testing — I tried to be careful, but it’s certainly possible something was damaged.

A damaged slot does make sense, especially because my PSU is a high-quality unit, and the Voodoo works flawlessly with it on another Socket 370 motherboard. I haven’t checked the 5V rails under load yet, so I’ll test that next. Here is a picture of the mapper I am using.

Ok, I’ve now finished testing the SBC slots. I managed to fit my PCISA SBC by repositioning the heatsink fan to the side instead of on top of the heatsink. This allowed me to install the PCISA card in slot SBC2 with the Voodoo card in PCI0. Since that worked mechanically, I also tested the SBC in SBC1 with the Voodoo in PCI1.

Unfortunately, neither configuration worked. This suggests that the slots themselves are physically fine, and that the issue is likely due to some kind of electrical instability rather than a damaged connector (they all look fine under my magnifying glass).

I also checked the 5V rail at multiple points, including directly on the Voodoo card, and consistently measured a solid 5.00 V. So the power line seems stable.

Given this, what else could be causing the issue?

Can you send a picture of the PCI Mapper card you're using,

Tevian wrote on 2025-11-18, 20:49:

Can you send a picture of the PCI Mapper card you're using,

Two posts above.

Oh... duh! BTW, I sent a message to you. 😅

For clarification, here's the two PISA/PCISA mapper configurations. Note the difference is only on PCI1 (second slot closer to the SBC slots on my backplane).

Some GPUs do require the IRQ lines to work properly. However, the system will boot and potentially get into Windows with only the IDSEL populated. The issues can arise when the card becomes activated and needs IRQ (Possibly for the 3D acceleration, but I'm speculating here!).

Also, I see what looks like some solder mask scuffing near the IDSEL lines on the card. Can you confirm the traces are intact!

You can verify which mapper config is needed by trying a PCI card that requires an IRQ, like a network interface card. Install a card into PCI slot 1, install the drivers under Windows, and see if you can get an IP over DHCP. If yes, the right mapper card is in; if not, you'll get hanging in Windows, even though the device manager says it's a working card.

Please, read the message I sent, and I'll keep thinking about what the issue could be. Did you try a fresh install of Windows? Maybe a weird driver conflict when you've tried multiple graphics cards?

Hi – I don’t have a PCI network card, but I’ll try to find one so I can confirm which mapping is needed. During my testing, I’ve tried both mapper cards, and in both cases the system boots normally with the Voodoo installed, but it still crashes as soon as I start a game.

I checked the traces on that mapper card with a multimeter and they’re fine. The scuffing on some lines was caused by me changing the IDSEL positions during my testing with the PC/104 adapter. I also tried completely fresh Windows installations — on different CF cards and even on an SSD — but the behaviour remains exactly the same.

Hi — I have tested the replacement backplane, and it worked with no problems using the exact same setup in which the other backplane failed. I still do not know what could have caused the original backplane to fail. The Voodoo 3 2000 seems to be working fine now.

Also, the PC/104+ is working perfectly. Have you been able to sort out the baffle plate for it?

Wow! This is good news! The other backplane suffered some kind of electrical or wiring failure obviously. I'd love to get that back for testing. 😊 Go easy on this one!😜🤣

Haven't yet looked into getting a handful of brackets made... Got busy but still need to look into that.