Great job, I love the YMF chipset family.

After sourcing all of the parts the card finally works! Although I bought the wrong audio connectors and they don't fit with the bracket. I've found the right connector and updated the project repository.

The card works beautifully with no noise when using line out. Everything just works!

Excellent effort ~

Update: 3D printable model of the bracket is now in the repository! I've tested and it worked perfectly.

Some clarification needed about this paragraph... regarding 33.8688MHz crystal.

C38, C39, R29, X2: If using a 3rd overtone crystal, which most probably is for 33.8688MHz in HC-49U package, then go with the values above. If using a fundamental crystal, then replace C38, C39 and R29 according to the crystal's datasheet. If the card doesn't work correctly then I suggest measuring the crystal output with an oscilloscope.

I'm interested in assembling one and am looking for materials, but when it comes to crystals I'm a bit confused upon reading this, as not many sellers seem to explain this aspect clearly enough. I wonder which type of crystal of this frequency is more common (fundamental or 3rd overtone), and is there any known RC value combinations in case of a fundamental crystal?

And by the way, can I use smaller (HC-49S) crystals instead of HC-49U? After searching online I couldn't find any clear difference between the two packaging other than physical dimensions, and HC-49S seems to be more common at present.

I've got the original here and a very similar but Opti-based card that does have the QS1000/QS1001A wavetable portion that is left unpopulated on both the real Audician 32 plus as well as the Replica.

Can I do something to help? It would be so cool if you could have the Aucician but with that wavetable. It is considered low quality in parts of the retro scene, but I believe that is largely like the bad reputation of the ESS Audiodrive cards back in the day - they were almost exclusively used for cheap cards where the component quality led to muffled sound output or noise, it's not the fault of the chipset itself.

Congrats.

th1r5bvn23 wrote on 2026-07-31, 14:34:

Here are some interesting findings during the process:

I get that it’s a replica, a "retro sound" and all that.But using the TEA2025 in its datasheet configuration is an insanely bad idea.
It has a gain of around 45dB and is designed for a 20mV input from a phono stage.To start with, this setup wastes gain that could have been used to reduce distortion.
And finally, the non-inverting input is pulled to ground inside the IC, so its offset is determined solely by the base current of the PNP transistor. Consequently, it cannot handle signals as low as 1–2V. Moreover, using a resistor-biased differential pair (rather than a constant current source) means it struggles to process common-mode signals (the classic non-inverting configuration from the datasheet).
Switch to an inverting configuration with 10dB gain.

It will make an excellent headphone amp.
Though back in the day, that kind of concept (plugging headphones into a computer) wasn't really a thing. 😀

LSS10999 wrote on 2026-08-14, 03:25:

Some clarification needed about this paragraph... regarding 33.8688MHz crystal. […]
Show full quote

Some clarification needed about this paragraph... regarding 33.8688MHz crystal.

...

I'm interested in assembling one and am looking for materials, but when it comes to crystals I'm a bit confused upon reading this, as not many sellers seem to explain this aspect clearly enough. I wonder which type of crystal of this frequency is more common (fundamental or 3rd overtone), and is there any known RC value combinations in case of a fundamental crystal?

And by the way, can I use smaller (HC-49S) crystals instead of HC-49U? After searching online I couldn't find any clear difference between the two packaging other than physical dimensions, and HC-49S seems to be more common at present.

That problem occurred to me as well. I asked the seller of the 33.8688MHz crystal about whether it's fundamental or 3rd overtone and he replied me it's fundamental, while after I installed the surrounding components as fundamental only to find that it run at 11.2896MHz and the sound was distorted. There is no good way to know that. I would assume that 33.8688MHz 3rd overtone is more common. I would suggest you install fundamental RCs and try it first.

HC-49S crystals could be installed as it shares the same footprint with HC-49U on the PCB. If the legs are long enough, you can install it the same way as HC-49U. Anyway, be sure to not short the adjecent U8 pads.

NorsteinB wrote on 2026-08-15, 06:31:

I've got the original here and a very similar but Opti-based card that does have the QS1000/QS1001A wavetable portion that is left unpopulated on both the real Audician 32 plus as well as the Replica.

Can I do something to help? It would be so cool if you could have the Aucician but with that wavetable. It is considered low quality in parts of the retro scene, but I believe that is largely like the bad reputation of the ESS Audiodrive cards back in the day - they were almost exclusively used for cheap cards where the component quality led to muffled sound output or noise, it's not the fault of the chipset itself.

Could you take a high-quality photo of the card and possibly measure the connections and values of surrounding components? Although measurement of MLCC capacitors without desoldering them are not possible, I could guess from the datasheets. However, I can't test if the wavetable works.

shevalier wrote on 2026-08-15, 12:43:

I get that it’s a replica, a "retro sound" and all that.But using the TEA2025 in its datasheet configuration is an insanely bad idea.

...

To be honest, I'm not fond of the TEA2025. Speaker amplifier on sound cards is generally a bad idea by today's standard and a modern headphone amplifier should be used instead. In fact, some early PCI sound cards like the Sound Blaster Live! have already utilized a proper headphone amplifier. And the Sound Blaster Live! (1998) is only ONE year younger than these OPL3-SAx cards (1997). But the goal of this project is just to replicate that card and validate the reverse-engineered schematic. I'm indeed planning on making a properly mordernized version of OPL3-SAx sound card. As always, a bigger project is started with small steps.

th1r5bvn23 wrote on 2026-08-15, 14:58:

That problem occurred to me as well. I asked the seller of the 33.8688MHz crystal about whether it's fundamental or 3rd overtone and he replied me it's fundamental, while after I installed the surrounding components as fundamental only to find that it run at 11.2896MHz and the sound was distorted. There is no good way to know that. I would assume that 33.8688MHz 3rd overtone is more common. I would suggest you install fundamental RCs and try it first.

HC-49S crystals could be installed as it shares the same footprint with HC-49U on the PCB. If the legs are long enough, you can install it the same way as HC-49U. Anyway, be sure to not short the adjecent U8 pads.

Thanks for the reply. Just noticed the presence of a U8 pad which seems to be reserved for a specialized clock generator (MK1420S). Guess it will not look very pretty when using HC-49S crystals with this PCB layout...

From what I could find online about the MC1420 clock generator it seems to require a 14.318MHz clock, just I don't know where it should be sourcing the input clock from. The schematic suggests BUSCLK but the resistor in between is left DNP.

th1r5bvn23 wrote on 2026-08-15, 14:58:

To be honest, I'm not fond of the TEA2025. Speaker amplifier on sound cards is generally a bad idea by today's standard and a modern headphone amplifier should be used instead. In fact, some early PCI sound cards like the Sound Blaster Live! have already utilized a proper headphone amplifier. And the Sound Blaster Live! (1998) is only ONE year younger than these OPL3-SAx cards (1997). But the goal of this project is just to replicate that card and validate the reverse-engineered schematic. I'm indeed planning on making a properly mordernized version of OPL3-SAx sound card. As always, a bigger project is started with small steps.

Actually, my experience with HKZLab's ES1868_ISA8 card seems to suggest a similar issue with amplifiers. The card's SPKR_OUT which went through a LM1877 amplifier feels rather distorted, but its LINE_OUT (without the amp) is clean and loud enough even when testing with a simple headset. As such, I'm using the card's LINE_OUT exclusively and haven't any issue with volume and such.

Not sure how the output currently sounds at present (would need a properly assembled card to verify), but it may be a good idea to add a non-amplified output in the future.

EDIT: Just noted where the amplifier/line out switch (JP1) is...

th1r5bvn23 wrote on 2026-08-15, 14:58:

To be honest, I'm not fond of the TEA2025. Speaker amplifier on sound cards is generally a bad idea by today's standard and a modern headphone amplifier should be used instead. In fact, some early PCI sound cards like the Sound Blaster Live! have already utilized a proper headphone amplifier. And the Sound Blaster Live! (1998) is only ONE year younger than these OPL3-SAx cards (1997). But the goal of this project is just to replicate that card and validate the reverse-engineered schematic. I'm indeed planning on making a properly mordernized version of OPL3-SAx sound card. As always, a bigger project is started with small steps.

/ooftopic
To be honest, the NJM4556 as a headphone amplifier is no better than the TEA2025B.
Something resembling a proper headphone amplifier only starts to happen with the AD8397, thanks to its high output current capability. Unless you are catering to Sennheiser owners, whose headphones traditionally have an impedance of 150 Ohms, a single-supply amplifier with 3V RMS is more than enough for 32-Ohm headphones.And the TEA in an inverting configuration with a gain of 6–10 dB is actually more than good for this application because its output current is very decent.
Plus, it keeps that retro feel.

As far as crystals go things at 30MHz and beyond are almost always 3rd overtone, it is difficult to make a fundamental mode crystal at such freqs.

LSS10999 wrote on 2026-08-15, 15:55:

Actually, my experience with HKZLab's ES1868_ISA8 card seems to suggest a similar issue with amplifiers. The card's SPKR_OUT which went through a LM1877 amplifier feels rather distorted, but its LINE_OUT (without the amp) is clean and loud enough even when testing with a simple headset. As such, I'm using the card's LINE_OUT exclusively and haven't any issue with volume and such.

Modern 32-ohm headset has too little impedance for old amplifiers to work properly. The TEA2025B, LM1877N and another common amplifier TDA1517P are all designed for high-impedance passive speakers.

Line out on these old cards are not ideal for driving 32-ohm headsets either. Although chips like YMF71x and ES1868 have just enough RMS to drive 32-ohm headsets, their output capability is limited. 32-ohm headsets will almost certainly exceed the current limitation of the DAC. In long term, that would harm the lifespan of the DAC. In a proper way, there should be an headphone amplifier with large output current (like TI OPA1656/OPA1688).

th1r5bvn23 wrote on Today, 01:22:

LSS10999 wrote on 2026-08-15, 15:55:

Actually, my experience with HKZLab's ES1868_ISA8 card seems to suggest a similar issue with amplifiers. The card's SPKR_OUT which went through a LM1877 amplifier feels rather distorted, but its LINE_OUT (without the amp) is clean and loud enough even when testing with a simple headset. As such, I'm using the card's LINE_OUT exclusively and haven't any issue with volume and such.

Modern 32-ohm headset has too little impedance for old amplifiers to work properly. The TEA2025B, LM1877N and another common amplifier TDA1517P are all designed for high-impedance passive speakers.

Line out on these old cards are not ideal for driving 32-ohm headsets either. Although chips like YMF71x and ES1868 have just enough RMS to drive 32-ohm headsets, their output capability is limited. 32-ohm headsets will almost certainly exceed the current limitation of the DAC. In long term, that would harm the lifespan of the DAC. In a proper way, there should be an headphone amplifier with large output current (like TI OPA1656/OPA1688).

According to the datasheet, the TEA2025B delivers 2.5 W per channel into 16 Ω with a 12 V supply at 0.3% THD. This is in the standard datasheet configuration with a gain of 45 dB. In an inverting setup with a gain of +6 dB, the excess 40 dB of gain goes toward reducing harmonic distortion by a factor of 100.
So, with 16 Ω headphones you'll get 0.003% or -90 dB.
With 32 Ω headphones, it's 6 dB better.
This is better than what a Yamaha DAC will output, and more than enough for a retro sound card.

PS. Naturally, you need to properly lay out the PCB and use the right types of capacitors (no X-type ceramics, only C0G or film) and resistors (for cheap thick-film ones, the package size should be 1206).

This is better than what a Yamaha DAC will output, and more than enough for a retro sound card.

I’d recommend buying an ASUS D1 for audio measurements (not the PCIe DX—that one’s worse). It has decent real-world specs, even though it sounds terrible for listening. For measurements, it’s spot on: cheap and gets the job done.

th1r5bvn23 wrote on Today, 01:22:

Modern 32-ohm headset has too little impedance for old amplifiers to work properly. The TEA2025B, LM1877N and another common amplifier TDA1517P are all designed for high-impedance passive speakers.

Line out on these old cards are not ideal for driving 32-ohm headsets either. Although chips like YMF71x and ES1868 have just enough RMS to drive 32-ohm headsets, their output capability is limited. 32-ohm headsets will almost certainly exceed the current limitation of the DAC. In long term, that would harm the lifespan of the DAC. In a proper way, there should be an headphone amplifier with large output current (like TI OPA1656/OPA1688).

TBH I'm not using 32-ohm headsets extensively with my ES1868_ISA8 card. I only used such headsets during initial testing, and my primary headphone set is 250-ohm which also worked fine. It's just the amped SPKR_OUT can get distorted when connected to certain outputs (not all, actually).

Right now I'm connecting that card's LINE_OUT to a mixer which is meant to combine multiple inputs that will go to a single set of desktop stereo speakers. So far it's working without issues.

shevalier wrote on Today, 05:22:

PS. Naturally, you need to properly lay out the PCB and use the right types of capacitors (no X-type ceramics, only C0G or film) and resistors (for cheap thick-film ones, the package size should be 1206).

This depends on availability... and not all sellers state the three-character code accurately. They usually just list the capacitor's tolerance rating, like 5% (J), 10% (K) and 20% (M).

LSS10999 wrote on Today, 09:00:

This depends on availability... and not all sellers state the three-character code accurately. They usually just list the capacitor's tolerance rating, like 5% (J), 10% (K) and 20% (M).

Sheldon Cooper said it best regarding this in his quote about gravity.
A 220–1500pF X5/7 dielectric filter capacitor in signal circuits introduces -60 to -70 dB of distortion with a very long and unpleasant spectrum.
And there's nothing you can do about it other than replacing it with a C0G/NP0
Like I said, buy a Xonar D1 and measure it yourself.