First post, by th1r5bvn23
- Rank
- Newbie
I reversed engineered the Audician 32 Plus sound card (also known as LWHA151A00) and made a replica PCB of it.
TL; DR: the project is on GitHub.
I initially wanted to integrate the OPL3-SAx sound chip in another project, but only to find that no public schematic is available yet. The datasheet is pretty vague on hardware design, and other projects use this chip are not open-source. I always love the real OPL3 and clean output of this card, so I made an effort to reverse engineer the card. I desoldered every component on the card and took a photo of each side, used a multimeter to check every connection and tried my best making a replica PCB. The value of every SMD MLCC capacitor is measured with the multimeter and guessed according to my knowledge.
I also collected and read a lot of reference documents, including datasheets of every chip and ISA specification. I came by a wonderful post of modding this card and it helped a lot.
Here are some interesting findings during the process:
- YMF718 (OPL3-SA2C) shares the same pinout as YMF711 (OPL3-SA2), while YMF719 (OPL3-SA3C) shares the same pinout as YMF715 (OPL3-SA3). Some source claims that YMF719 is a custom-labeled version of YMF715. The difference between OPL3-SA2 and OPL3-SA3 is that OPL3-SA3 has some 3D support and has a little different pinout on some filter caps. I've marked their difference in the schematic.
- The onboard wavetable (QS1000, QS1001A and TDA1311 codec) are connected correctly while the wavetable header has left and right output channel reversed. Contrary to previous information, the "reverse channel" bug of OPL3-SAx's wavetable header origins from the PCB itself, not the OPL3-SAx chip. I also found that ES1868 datasheet has the wrong pinout of wavetable header which is reversed on output channels.
- The CDROM version of the card is only capable of 10-bit address decoding and potentially has problem in PCI systems. Although simply adding back the 74LS138 and change the jumpers will work fine.
Though I did my best, but the PCB itself is definitely not a 1:1 replica to the original card. In most aspects they are the same, however I made some modifications, either due to the limitation of modern software, or fixing bugs:
- Fixed the reversed output channel of the wavetable header.
- Different footprints and fonts on all components due to, obviously, I used KiCad to draw the board. I don't want to waste time on replicating old footprints and fonts and I think most reasonable persons won't, either.
- The PCB was made slightly longer by ~1.5mm to stick strictly within ISA spec. Also the gold finger is moved forward by ~0.5mm according to the ISA spec. I always had hard time inserting to and pulling out the card from the slot and I finally knew why. Therefore the layout of some traces and vias is different from the original.
- Silkscreen of some chips are changed. Labway even wrote "74SL138" instead of the correct "74LS138"... on all the revisions of the PCB...
- I think it's illegal to use other's FCCID and CE logo without certification so I replaced them.
There are still several problems to be solved in this project (not important, however):
- Many components that are not on any of these cards are unknown. The values of some resistors and SMD capacitors cannot be easily guessed. There is also a chip U7 that I know some function of it but can never know the model.
- No ISA bracket is provided at the moment.
- Full replica is a little more expensive than buying a used card currently. But the card is not easy to come across these days.
I made all of the schematic, PCB layout and many reference documents including PCB photos available on the GitHub repository. I've ordered PCB from JLCPCB and it will be arrived in 3 days. I don't plan on selling any of the cards.
Intel Pentium III 933MHz, ASUS CUBX-L, NVIDIA GeForce2 MX 400, Yamaha YMF719E-S
