33.868800 will produce a slighly off MIDI UART freq which can result in no MIDI signal getting received by the chip. Proper MIDI freq is 31250 baud but 33.8688MHz will give 32072 baud.

There is an ES689F based card with ES981P wavetable ROM which uses 33.8688MHz crystal. I don't own such a card but I have found a picture of it. So I assumed it would work fine. Of course it doesn't mean the same applies to ES690F but I do have ES689F based card with 33.000MHz crystal. So I think ES690F have same specs as ES689F but adds reverb. Unfortunately I cannot find detailed data sheets for ES689 to confirm it.

I have Serdaco's ESS WB card here with 33MHz crystal (attached to a later chip). I can try it with 33.8688MHz and if it works with it then yours should too. I will proceed doing so

Great, I'm looking forward to your findings.

BTW, here's a picture of another ES689F card with 33.8688Mhz crystal.

I just got 33.8688MHz tested and the card works fine with it, so the problem must be elsewhere...

Things look like they are connected properly as far as the new signals go. It could be that the dump you made failed because of a badly connected pin or something. I have had to fight that a lot, taking many reads (with removal of socket between them) and comparing them to make sure I have matching dumps. More often than not I have differing reads...

EDIT: Where is the !CE line of the ROM going ? There's no separate !CE signal on ES690 datasheet that I could see, only on ES689. I would image it needs to be also connected to GND for proper operation so that data is always available.

I actually have one of these cards, so color me interested (even though I have no experience with SMD stuff). I'm planning to use it in a 286 build and I'm planning to hunt down a wavetable board for it as well.

Checked the card and it seems the CE pin of the ROM chip is connected to A20 pin of the ES690F. Perhaps a mistake in datasheet as there is no need for such a pin? Checked another card which has a place for both ES689 and ES690 and the ES690's A20 pin is connected to ES689's CEB pin.
Tried to swap crystal for another one (in fact tried two others), then tried to replace ES690 chip for another one. At first I couldn't hear a thing when playing MIDI music,even though the tested software behaved like something is playing. So I tried to replace MX29F1610 with the real ES982A MASK ROM and the result was the same. Did a reflow of both ES690F and ES982A and after that crackling occurred again after initializing the card. It's weird.
Concerning the ES982A chip - at first I made a dump with TL866 programmer. Since TL866 cannot be used for programming MX29F1610 chips I had to build a new one designated for those. Tried to dump the ES982A again with this new programmer and the dump was the same. So either the dump is good or the chip has been damaged somehow during desoldering process. Maybe it would be a good idea to get hold of another one but finding it is a real pain. I have one possible source though which I can try. Is there a way to validate the dump? I know someboy on Vogons unpacked the 1MB dump, maybe we could try to do the same with the 2MB one.

EDIT: Checked the schematics and it's true one has to connect ROM's CEB pin to ES690's A20. It's a weird solution nonetheless.

I am watching this project with interest as I have quite a few cards that I can upgrade with this wavetable..

Hopefully I'll be able to finish the project, after that I plan to share my work with others so everyone can modify such a card. Currently I own three different types and I would like to successfully modify all of them, with documentation.

A20 being the !CE makes sense, it is the topmost address line and can be used as a !CE for lower 2MB and when inverted, !CE for higher 2MB if there ever was a need.

About verifying a dump, there's no good way without having something to compare to. I made a procedure to get new dumps by doing many dumps with device removal between dumps. When I get several matching dumps I know I have good data. Very often one or two pins don't connect properly and create garbage here and there. Since it is a Sample ROM you can open it in some audio editor and give a listen, most samples should be in same raw PCM format and should sound clean etc. There may be a block of noise in the beginning or end, which probably is the the program code for the internal CPU that controls the sound hardware to turn MIDI into music.

I have opened the ROM dump in Audacity as Signed 16bit PCM data. There is noise in the beginning and at the end but otherwise, most of the file sounds legit. There are weird squeaks here and there as some samples are too short and are effectively merged together because of the fast play. But so far, it seems the dump is good even though I cannot be 100% sure. Maybe I could replace the ES982A chip with ES981P from my other known working ESS card and test it.

EDIT: Tried to play it again at 25% speed at must say that now I'm 99% sure the dump is ok. Every weird squeak from before sounded like proper instrument.

That seems like the dump is indeed proper. I guess you should sort things out using the original 1MB ROM first and then see further.

I suppose A19 is connected right ? I was thinking that if the card was never meant to have 2MB ROM it might not connect this address line to the ROM and that would cause problems.

Nice thread/project. I was about to start one asking if anyone knows what these cards were populated with and if we can retrofit it, what are the odds...

All of my cards have A19 pin properly connected so there shouldn't be an issue. One of them even says "ES981/ES982" on the silkscreen. I guess there is a problem either with the ES690F chip (which gets damaged while soldering, even though that seems weird to me as I'm applying heat only for couple of seconds) or in the crystal. I think I'll first try to swap the crystal from known working card and maybe retrofit it with the 33.8688MHz one. That should answer if there is problem with it or not. Will try it today, we'll see how that goes.

Status update: Tried to swap crystal from known working card with ES981P, the card worked. So the crystals are not a problem. Then tried to solder MX29F1610 onto it, swapping it with the ES981P chip and all I got is silence when playing MIDI. Tomorrow I'll try to make some new dumps of the ES982 ROM and perhaps solder it directly onto the known working card. In the meantime I'll contact a person who has a device with another of these chips so I can make a secondary dump, for comparison.

EDIT: Tried for fun to dump ES981P chip that I have removed from the working card. Took me probably around 20 tries to dump it properly (as there is already proper dump floating here on Vogons).

EDIT2: Tried to dump the ES982A chip around 10 times and apart from two cases (which differed from themselves) I got the same dump as originally. I'll try later to solder the ES982A onto known working card and we'll see if it works or not.

As the dump seems to be ok, I'm hesitating if I'll solder the ES982A chip on the working card. I have properly working MX29F1610 on there already and I think the result will be the same. Maybe I'll do it, to be 100% sure. Tried to solder CE pin of the MX29F1610 directly to ground but it didn't help. At the same time I have contacted a person which has another ES982 but yet without any answer. I'm missing something here and I don't know what.
If you're interested you can listen to the dump I've made, converted to OGG file. One is normal speed, the second is 25% speed.

Yeah, these sound fine, I cannot hear any regular artifacts or other things.

I think you need to try the original ROM chip, maybe there is some sort of timing problem with the flash chip. I'm not sure what the problem could be other than these sort of things...

Ok, here's my plan:

1. Solder ES982A onto known working card which previously had ES981P. If it works, then there is something wrong with the flash ROMs. If it doesn't, then:
2. Find second ES982 chip and dump it. Necessary for comparison that the first dump I made is correct.
3. If the second dump will be different, then I'll try to use it with the flash ROMs and modify my cards with them.
4. If all above fails, I'll release everything I came up with (together with the dump) with hope that someone more clever then me will get it working. I want my 2MB ESS wavetable!

Count me interested. I've just grabbed a ES981P card in a bad state, since it's going to be a bit of a Frankenstein once I fix it I'd be interested in testing stuff out if you need a volunteer.

Mine's identical PCB-wise, with slight differences - the BTC logo is nowhere to be found, the 1853/1855/1857/1859 print is replaced with some strange code. OEM perhaps, since it has an Optimus (legendary PC retailer in Poland back in the day) sticker on the back.

Mine has a 33.000KHz as well, though it's in a smaller, shorter package than the ones you have.

Some questions:
1) have there been any BTC cards that didn't have the wavetable area populated? Never seen one in the wild.
2) What Compaq laptops had these wavetable chips? I might start looking out for them and buy one if there's a chance. Then you'd have one more ES982 for dump comparisons.