Update:

Got I2C communication code working, enabling configuration of the PWM-controlled fans.

Also did a bit of rework on the status LED situation, I'll now use an I2C-controlled MCU to drive a single RGB LED instead. That means we can setup standby- and power light intensity, patterns, etc in software / BIOS.

In addition, I thought about the HDD activity LED; my last prototype showed that only using the SD CMD signal didn't look particularly good, instead I'll try to hook into the BIOS control routines for reading and writing data to SD/USB/SATA and send activity signals to the MCU which will in turn control the RGB LED.
...and then it hit me, maybe I should try adding support for HDD clicking sounds as well, might be fun to try at least:

And here's the footprint for adding the D1 Mini - for Wifi/modem support:

Any other suggestions for improvements? I'm getting close to ordering the next batch.

Eivind wrote on 2023-07-04, 22:27:

Any other suggestions for improvements?

Is there a way to connect it to an external midi device over a gameport/db15? Or a joystick? 😀
Is the atx power 24 pin? That would be easier to use these days.

perhenden wrote on 2023-07-05, 06:28:

Eivind wrote on 2023-07-04, 22:27:

Any other suggestions for improvements?

Is there a way to connect it to an external midi device over a gameport/db15? Or a joystick? 😀
Is the atx power 24 pin? That would be easier to use these days.

Yes, yes and yes. 😉

Eivind wrote on 2023-07-05, 09:28:

perhenden wrote on 2023-07-05, 06:28:

Eivind wrote on 2023-07-04, 22:27:

Any other suggestions for improvements?

Is there a way to connect it to an external midi device over a gameport/db15? Or a joystick? 😀
Is the atx power 24 pin? That would be easier to use these days.

Yes, yes and yes. 😉

Just wondering if it is possible to have the USB-C option as well. Let's say you need the 12 volt, so you can use the smart chip to negotiate the 12 volt and use a bunch of VRs to provide the 5v and 3.3 (not sure which voltages are necessary).

Eivind wrote on 2023-07-04, 22:27:

Any other suggestions for improvements?

Optical S/PDIF output?
Header to allow addition of an external FM synth+DAC?

ahmadexp wrote on 2023-07-06, 02:19:

Just wondering if it is possible to have the USB-C option as well. Let's say you need the 12 volt, so you can use the smart chip to negotiate the 12 volt and use a bunch of VRs to provide the 5v and 3.3 (not sure which voltages are necessary).

I think I'm going to turn this one down. It'll add quite a bit of complexity and potential for breaking stuff.
For starters, here are the voltages needed: +12V, -12V, +5V standby, +5V, +3.3V. USB-C PD can only provide a single voltage at a time, and +12V isn't even one of them. We'd have to go up to 15V.
Then we'd need to convert that into the others. With a graphics card in the PCI slot, we'd need more current than can be converted using linear regulators, so we're talking switching buck converters with beefy capacitors for filtering.
Alternatively we could take 5V and boost and buck it to 12V and 3.3V respectively. In any case, lots of new components and I haven't even gone into the issue of safely combining all of this with the existing ATX connector, which I'll definitely want to keep. Also, to keep a standby voltage and soft-power-on, we'd need to introduce some kind of chunky power mosfet or relay to turn on and off the main rails.
I believe a much cleaner solution, if you really want to skip the ATX power supply, would be to use a picoPSU module. There you have all of what I just talked about combined in a quite small, tried and tested package.

640K!enough wrote on 2023-07-06, 05:23:

Eivind wrote on 2023-07-04, 22:27:

Any other suggestions for improvements?

Optical S/PDIF output?

Toslink or coax s/pdif: (I guess optical/toslink is the more usable of those?): Should be relatively simple to wire up. My main concern is that I would have to squeeze the existing backplane connectors together pretty aggressively in order to fit the new connector.
But it might actually be viable. I'll spend some time researching - thanks for the suggestion! 😀

640K!enough wrote on 2023-07-06, 05:23:

Header to allow addition of an external FM synth+DAC?

Hmmm.... yeah maybe? YMF289 + YAC516 would be the obvious choice I guess? I don't have any more line-ins left though, would have to be shared with the external line-in?
Come to think about it, maybe it's better to let the the analog output from the Pi/wavetable share with the external line-in connector, since you might want to use an external MIDI device and the external FM synth at the same time.

Eivind wrote on 2023-07-06, 07:54:

I think I'm going to turn this one down. It'll add quite a bit of complexity and potential for breaking stuff. For starters, her […]
Show full quote

ahmadexp wrote on 2023-07-06, 02:19:

Just wondering if it is possible to have the USB-C option as well. Let's say you need the 12 volt, so you can use the smart chip to negotiate the 12 volt and use a bunch of VRs to provide the 5v and 3.3 (not sure which voltages are necessary).

I think I'm going to turn this one down. It'll add quite a bit of complexity and potential for breaking stuff.
For starters, here are the voltages needed: +12V, -12V, +5V standby, +5V, +3.3V. USB-C PD can only provide a single voltage at a time, and +12V isn't even one of them. We'd have to go up to 15V.
Then we'd need to convert that into the others. With a graphics card in the PCI slot, we'd need more current than can be converted using linear regulators, so we're talking switching buck converters with beefy capacitors for filtering.
Alternatively we could take 5V and boost and buck it to 12V and 3.3V respectively. In any case, lots of new components and I haven't even gone into the issue of safely combining all of this with the existing ATX connector, which I'll definitely want to keep. Also, to keep a standby voltage and soft-power-on, we'd need to introduce some kind of chunky power mosfet or relay to turn on and off the main rails.
I believe a much cleaner solution, if you really want to skip the ATX power supply, would be to use a picoPSU module. There you have all of what I just talked about combined in a quite small, tried and tested package.

Makes total sense. Thank you for all the great work.

@640K!enough:
I've looked at the CS datasheet again, and there's one thing I can't quite figure out, maybe you can help me?
It looks like one should (or could) use the XA0, XA1, X D[7:0], /XIOR and /XIOW pins/functions (in addition to the /SCS, /SINT and /BRESET, of course) when connecting to the YMF262/289, but I don't really understand why not use the ISA bus directly for these? Is it because the CS will act as a buffer/directional driver (74244, 74245), to save on external components...?

Oh, and prices for the YMF289 are insane now, from what I could find. ~$38 per chip, as compared to $1-2 for the YMF262. Ouch. Guess I'll go for the YMF262 + YAC512 then. Seems like the TL074 is a commonly used opamp for the YAC512.

Quoting on my antique phone wouldn't be fun, so please forgive me for just babbling at you.

From what I've been told, prices for the YMF289 have gone nuts. The good news is that there's no compelling reason to choose it over the YMF262 for such an application.

I have a preference for optical digital output, mainly because it is less expensive and easier to do properly than coaxial. The last time I checked, the necessary transformers cost more than all of the hardware to do optical, as long as you don't insist on genuine Toshiba (or similar) transmitters.

The interfacing can get somewhat complicated, depending on what else you're doing. If you're also planning on supporting digital output, you can't use the CS4237 to drive the bus during reads, and still need a '245 or similar. You can, however, still leverage it for configurable address decoding, somewhat as you've described (which I would recommend). You not only cut down on external glue logic, but take advantage of Plug and Play configurability for both the Ad Lib and Sound Blaster FM address ranges. So far, I can't think of a reason not to use it.

Since you only have one ISA IRQ line, you can ignore SINT# entirely.

If I forgot to address anything, or if you were looking for more detail, please feel free to ask. If we get any deeper into the trenches, privately may be better, so that we don't completely pollute your thread with the details.

640K!enough wrote on 2023-07-07, 17:06:

Quoting on my antique phone wouldn't be fun, so please forgive me for just babbling at you. […]
Show full quote

Quoting on my antique phone wouldn't be fun, so please forgive me for just babbling at you.

From what I've been told, prices for the YMF289 have gone nuts. The good news is that there's no compelling reason to choose it over the YMF262 for such an application.

I have a preference for optical digital output, mainly because it is less expensive and easier to do properly than coaxial. The last time I checked, the necessary transformers cost more than all of the hardware to do optical, as long as you don't insist on genuine Toshiba (or similar) transmitters.

The interfacing can get somewhat complicated, depending on what else you're doing. If you're also planning on supporting digital output, you can't use the CS4237 to drive the bus during reads, and still need a '245 or similar. You can, however, still leverage it for configurable address decoding, somewhat as you've described (which I would recommend). You not only cut down on external glue logic, but take advantage of Plug and Play configurability for both the Ad Lib and Sound Blaster FM address ranges. So far, I can't think of a reason not to use it.

Since you only have one ISA IRQ line, you can ignore SINT# entirely.

If I forgot to address anything, or if you were looking for more detail, please feel free to ask. If we get any deeper into the trenches, privately may be better, so that we don't completely pollute your thread with the details.

Yeah, I intend to go a bit deeper into the trenches! 😁 I'll DM you, and thanks again for helping out!!

Awesome project, ITX is the perfect form factor for a build like this. Definitely following how it goes, and interested in picking one up if you have plans to sell it.

doublebuffer wrote on 2023-07-08, 10:14:

Awesome project, ITX is the perfect form factor for a build like this. Definitely following how it goes, and interested in picking one up if you have plans to sell it.

Thanks! 😀

Yeah I'll very likely try to sell a few of these. They're relatively expensive to make though, so I want to make sure everything's working 100% before being comfortable taking your money! 😉

@Eivind,

We need to start looking into building an I/O shield for the ITX-Llama.
Maybe we can use these shields available on amazon and get them cut to fit the I/O arrangement of the ITX-Llama.

ahmadexp wrote on 2023-07-08, 13:45:

We need to start looking into building an I/O shield for the ITX-Llama.
Maybe we can use these shields available on amazon and get them cut to fit the I/O arrangement of the ITX-Llama.

Yeah, that might be an option, sure.
I don't have a laser cutter myself, and I'm very hesitant being dependent on other people for my super-limited projects. I'll try making 3d-printed plastic shields first and we'll see how that goes!

Eivind wrote on 2023-07-08, 18:08:

ahmadexp wrote on 2023-07-08, 13:45:

We need to start looking into building an I/O shield for the ITX-Llama.
Maybe we can use these shields available on amazon and get them cut to fit the I/O arrangement of the ITX-Llama.

Yeah, that might be an option, sure.
I don't have a laser cutter myself, and I'm very hesitant being dependent on other people for my super-limited projects. I'll try making 3d-printed plastic shields first and we'll see how that goes!

You don't have to rely on anyone. We have been getting the PCIe bracket for the Time Card made by PCBWay's sheet metal service. I am pretty sure we can get the i/o panel as well. All we need is a step file and we can order it there.

Eivind wrote on 2023-07-08, 18:08:

ahmadexp wrote on 2023-07-08, 13:45:

We need to start looking into building an I/O shield for the ITX-Llama.
Maybe we can use these shields available on amazon and get them cut to fit the I/O arrangement of the ITX-Llama.

Yeah, that might be an option, sure.
I don't have a laser cutter myself, and I'm very hesitant being dependent on other people for my super-limited projects. I'll try making 3d-printed plastic shields first and we'll see how that goes!

Alternately you can use these panels that you can cut with a wire cutter.

What about using therasteri's method of getting a PCB made (like he did for the weeCee) ?
Make a PCB with the shape of an ATX I/O shield. You could even use the silkscreen to write what port does what 😁

Deksor wrote on 2023-07-09, 13:18:

What about using therasteri's method of getting a PCB made (like he did for the weeCee) ?
Make a PCB with the shape of an ATX I/O shield. You could even use the silkscreen to write what port does what 😁

Don't think that'll work, unfortunately. The IO shield has a thin, flexible lip for snapping in place. I still think a plastic 3d printed one is the easiest solution. Obviously it'll only be for the looks as plastic doesn't provide any EMI shielding. 😜

Update:

Following 640K!enough's suggestion of adding a digital output connector and an external OPL3 module, this is what I came up with:

Itching to put in another order... 😁