Try to bypass the amp chip. Anyway its legs look badly soldered.

If you don't mind me asking, how would I do that? Remove the TDA entirely?

Yep, remove the amp chip and then connect its left channel input pad to left channel output pad and do the same for the right channel. I really don't know what else you can try if there are no jumpers on the PCB for bypassing the amp.

P.S. Now looking at TEA2025B datasheet, I think you can probably try to reduce the gain through Rf (R41 and R42 on the PCB) and C1 (I can't see those markings on your pictures) connected in series to feedback pins (6 and 11):

I already got the TEA2025B out, and bridged both OUT 2 and OUT 1 to IN2 and IN1 respectively.

Do I also bridge the BOOT pins?

Why you have to bridge those boot pins? On the similar soundcards there are only 2 jumpers for bypassing left and right channels.

I didn't, just tied the OUT2 ant OUT1 to IN2 and IN1, and saw there's BOOT 2 (pin3) and BOOT 1 (pin14) on the pinout and figured I'd ask if those need to be tied together as well.

Did you remove the amp chip?

Yes, entirely removed.

Ok, then you can now try if the card's output stereo jack will work in something like line-out (non amplified) mode with externally amplified speakers.

Same clipping, and only one channel appears to work. (right channel)

So, the sound clipping is not caused by excessive amplification from TEA2025B chip. I really don't know what else can cause such a problem with this cheapo soundcard. As for the only right working channel - I see that some of the capacitors have been replaced, but it seems like the soldering job was not very clean. Maybe now there is a short to ground on the left channel circuit. What's the history of this card?

Ever since I got it (standalone, from some scrapper, it wasn't in any PC I'd got from him) it had this issue. It was present even before the recap.

EDIT: Fixed it finally. Had to tie BOOT2 and BOOT1 together (pin3 and pin 14), as well as the FEEDBACK (pin6 and pin 11) pins. My only assumption is either the TEA2025B had an issue, or some of these weren't wired together thru the PCB. I can confirm that connecting them together got rid of the horrid clipping!

I did check the new caps and redid all joints - no shorts.

I don't understand why have to bridge those feedback and boot pads, If the amp chip is removed from the PCB, but if it works - it works 😀

Is it worth buying a replacement amp chip for this card?

Given it sounds good now, I'll run it as it is, using the chip's own amplifier. (IIRC it has an internal amp?)

As for the reason of connecting BOOT and FEEDBACK together - the diagram I found in the TEA2025B datasheet shows them connected together with the IN and OUT pins (not counting the resistors).

I found some statement that BOOT1/2 pins are connected to some capacitors which reducing signal distortion (probably have relation to clipping too).

Do you mean bridging FEED1 pad to OUT1 pad and FEED2 to OUT2? I couldn't understand what pads you've connected to make it finally work with amp chip removed.

Can you write out which pads are connected after the amp removal, so the circuits are now working fine i.e. pin2 (OUT2) - pin7 (IN2); pin10 (IN1) - pin15 (OUT1)... etc.?

A bit rough but here's the wiring. Blue is the IN and OUT wires, orange are the BOOT and FEED wires.

BOOT goes under the positive lead of that cap because one of the traces were likely torn when I had to remove the TEA2025B.

As for pins connected:
- pin 2 goes to pin 7 (OUT 2 connected to IN 2)
- pin 15 goes to pin 10 (OUT 1 connected to IN 1)
- pin 3 goes to pin 14 (sorta, the trace from 14 goes under the cap and since that trace broke when I removed the TEA, I just ran a wire under the positive lead of the said cap - these ard BOOT 2 and BOOT 1 tied together, as per the reference TEA2025B diagram posted earlier)
- pin 6 goes to pin 11 (FEED pins, same as with the BOOT - tied together since they are part of the same circuit BOOT and I/O wires were tied together to)

Thanks for the details, PcBytes.

The original circuit was this:

And basically now the output circuit looks like this:

I have no explanation why the connections in red color are needed while amp chip is missing (the bridge between FEEDBACK pads 6 and 11 still seems absolutely redundant to me), except the presence of the residual connected capacitors and resistors and probably improperly selected IN-OUT bridge points. On the soundcard PCBs without installed "internal" amp out of the factory, If not all, most of these passive elements are omitted.

P.S. This topic for bypassing of the built-in amp on such old soundcards is very interesting for me, as I also have a couple of noisy ESS ISA cards with no option for bypassing of their amp chips with jumpers and I can't find on the internet any clear explanation or schematic/diagram for similar output circuit modification.