The main issue with these ADPCM samples is that they come out of an external memory, which you cannot easily update. Easily here means to play back a WAV file or some other type of data stream. There is no DMA mechanism or anything and the keyhole interface to update that sample memory through the chip is very slow so it will take a lot of CPU power just to send the data stream, and even more to compress it to Yamaha specific ADPCM format in case of wanting to play multiple samples simultaneously such as what a game like Doom might do.

Tiido wrote on 2024-03-19, 22:44:

The main issue with these ADPCM samples is that they come out of an external memory, which you cannot easily update. Easily here means to play back a WAV file or some other type of data stream. There is no DMA mechanism or anything and the keyhole interface to update that sample memory through the chip is very slow so it will take a lot of CPU power just to send the data stream, and even more to compress it to Yamaha specific ADPCM format in case of wanting to play multiple samples simultaneously such as what a game like Doom might do.

Well yes you'd want a steaming DAC for that. But as far as games programming is concerned if OPN had been the basis of PC sound cards instead of OPL then the way PC games approached sound would have been different anyway. OPN was widely used in Japanese PCs for example and they used it quite successfully. In that case game sound drivers would upload samples to the OPN's external memory ahead of time and the games would be designed around this. And 128k or 256k isn't as small as it sounds when you consider that ADPCM is 4 bits per sample. I'm not talking about a world where OPN and OPL coexisted I'm talking about a world where the Adlib went extinct alongside the Creative Game Blaster and IBM Music Feature Card and OPN based sound cards ruled MSDOS PC games. A YM2203 based card would have been much less expensive than the Adlib and might have been able to succeed based on the much lower cost (and assuming game developers supported it so marketing would have been important as well.) Adlib was imo way too focused on PC music and OPL on MSDOS succeeded mostly because it was the default.

Lack of a streaming DAC sounds like a deal breaker to us but for the majority of PC gamers back in the way it was a niche feature. Most people weren't into tracker music or anything like that they just wanted a card that would make music for their PC games.

But if you really did want a streaming DAC the best way would probably be to replace the OPNs companion DAC with something that could combine the audio stream from the OPN with a DMAed steaming audio sample. If youy could convince Yamaha there was money in it they might even have be willing to design something for that. If not there are other ways you could have added the stream from the OPN and the streaming DAC together before sending it to the companion DAC.

Check out the specs on the OPX chip.

https://minirevver.weebly.com/ymf271opx-music.html

It has user defined waveforms i.e. you can use PCM in external memory as operators. This thing is like a Yamaha TG-77 on a chip. Now THAT would have been the ultimate FM sound card. The TG-77 was the best FM synth module Yamaha ever made (some would argue the 6 operator FS1R was but it had a lot more limitations than the TG-77)

EDIT: OPX music example from youtube: https://www.youtube.com/watch?v=poOlOdZB6Vg&l … AKQ7WI4T0yOgZON

Atomic Skull wrote on 2024-03-22, 06:27:

A YM2203 based card would have been much less expensive than the Adlib and might have been able to succeed based on the much lower cost

How do you know YM2203 would have been less expensive? OPL was already a low-end product line. It's possible that the YM2203 is a smaller die (fewer operators overall) but it's also a larger (40-pin) package and requires external circuitry to mix the square wave channels output with the FM audio. Both chips use the same Y3014B DAC.

BTW, one other advantage of the YM2203 is that it doesn't require the huge delays that the OPL2 does when doing I/O.

Hm, by the way, does anybody know why the OPL4 wasn't being succesful in either hemisphere?
It could be paired with an external RAM and used like a wavetable synth (64KB sample limit?).
The MSX scene had adopted it (MoonSound?) and created awesome music for it.
Was it simply being too late, maybe? Let's imagine it did catch on PC or PC-98!

Jo22 wrote on 2024-03-22, 12:12:

Hm, by the way, does anybody know why the OPL4 wasn't being succesful in either hemisphere? It could be paired with an external […]
Show full quote

Hm, by the way, does anybody know why the OPL4 wasn't being succesful in either hemisphere?
It could be paired with an external RAM and used like a wavetable synth (64KB sample limit?).
The MSX scene had adopted it (MoonSound?) and created awesome music for it.
Was it simply being too late, maybe? Let's imagine it did catch on PC or PC-98!

I think the main issue with OPL4 with computers is that Creative had no interest in it. Creative focused on creating their own CQM and sample based cards so that they could cut-out Yamaha and make more money. Since the OPL4 FM is identical to OPL3 FM, and it also has GM compatible wave table, budget sound card companies liked it. But nobody took full advantage of it, until Sunrise created the original MSX Moonsound cartridge. The OPL4 can actually use 2MB of sample memory when wavetable is enabled. When wavetable is disabled, it can use 4MB. (Majority of Moonsound carts today have 2MB, though there is one Korean made cartridge that does support 4MB mode). As for NEC PC-98, YM2608 OPNA made more sense, as the chip is fully backwards compatible with the YM2203 OPN, so it was a simple way to go from mono to stereo sound.

The OPL4 did get some professional use in the arcade world. The Sega MultiPCM chip is so similar, that music from games like Daytona USA have been converted to run on OPL4.

SuperDeadite wrote on 2024-03-22, 13:00:

I think the main issue with OPL4 with computers is that Creative had no interest in it.

Let's not give Creative more credit than they deserve. Had the OPL 4 arrived earlier, I suspect it would have dominated; I think Yamaha was just too complacent in raking in the profits from the YMF262 and derivatives, and didn't realise that others would eat their lunch until it was too late.

Had the OPL 4 become well established, with software using its interface directly, Creative would have had no choice but to adopt it or create a sufficiently-compatible part. It was the same with GM on the PC, where everyone and his dog tried to emulate the Roland instrument characteristics. Creative never got that quite right, and it showed. Their hardware was out-performed in the MIDI world (as used by DOS titles) by everything from the GUS PnP to cheap Taiwanese-made cards with Fat-certified sound sets. SoundFonts didn't even make a dent in the world of PC game music.

Also, the OPL 4, as originally introduced, was an expensive solution for typical PC applications, typically requiring a processor, instrument ROM (typically YRW801) and/or SRAM (up to 2 MiB of each!) and glue logic (or dedicated controller ASIC), in addition to the YMF278. It wasn't until the cost-reduced, less capable, integrated versions became available, incorporating simplified 1-MiB instrument bank, processor and MIDI interface, in addition to the synth, that it took off on cheap cards. Before that, there were only a few notable cards that used it; namely Yamaha's own SW20 line, Logitech's SoundMan Wave, a few Miro designs and, probably the best specimen, the Audiotrix Pro from Quebec's Mediatrix Peripherals (a quality design that, technically, put the AWE32 to shame).

ROM samples was the worst thing to ever happen to PC gaming imo.

640K!enough wrote on 2024-03-22, 16:10:

Before that, there were only a few notable cards that used it; namely Yamaha's own SW20 line,

I didn't realize Yamaha eventually brought out a PC sound card with RAM samples. Back when they released the SW-60XG there was a "coming soon" SW-70XG version that was supposed to have 256K of RAM for user uploadable samples but it never came out. Apparently there was an internal fight between Yamaha's PC sound division and their professional audio division due to the DB50XG and SW-60XG using the same MIDI sound chipset as the QS300 and they almost didn't come out for that reason. The SW-70XG was a bridge too far for the professional audio side and they managed to get it cancelled.

EDIT and I just recalled that the DB50XG and SW-60XG had a hidden QS300 edit mode that let you create sounds with an undocumented user bank using ROM samples. Though the performance mode is absent so you can only make sounds with up to two layers.

So the DB50XG isn't an MU-50 on a daughtercard it's really a QS300 on a daughtercard. That's why certain of it's synth lead sounds are actually more expressive sounding and better than the same sounds on an MU-100 or even the MU-2000. "waspysyn" is particular is way better than it's equivalent on the MU line of modules.

Atomic Skull wrote on 2024-03-22, 06:27:

Well yes you'd want a steaming DAC for that. But as far as games programming is concerned if OPN had been the basis of PC sound […]
Show full quote

Tiido wrote on 2024-03-19, 22:44:

The main issue with these ADPCM samples is that they come out of an external memory, which you cannot easily update. Easily here means to play back a WAV file or some other type of data stream. There is no DMA mechanism or anything and the keyhole interface to update that sample memory through the chip is very slow so it will take a lot of CPU power just to send the data stream, and even more to compress it to Yamaha specific ADPCM format in case of wanting to play multiple samples simultaneously such as what a game like Doom might do.

Well yes you'd want a steaming DAC for that. But as far as games programming is concerned if OPN had been the basis of PC sound cards instead of OPL then the way PC games approached sound would have been different anyway. OPN was widely used in Japanese PCs for example and they used it quite successfully. In that case game sound drivers would upload samples to the OPN's external memory ahead of time and the games would be designed around this. And 128k or 256k isn't as small as it sounds when you consider that ADPCM is 4 bits per sample. I'm not talking about a world where OPN and OPL coexisted I'm talking about a world where the Adlib went extinct alongside the Creative Game Blaster and IBM Music Feature Card and OPN based sound cards ruled MSDOS PC games. A YM2203 based card would have been much less expensive than the Adlib and might have been able to succeed based on the much lower cost (and assuming game developers supported it so marketing would have been important as well.) Adlib was imo way too focused on PC music and OPL on MSDOS succeeded mostly because it was the default.

Lack of a streaming DAC sounds like a deal breaker to us but for the majority of PC gamers back in the way it was a niche feature. Most people weren't into tracker music or anything like that they just wanted a card that would make music for their PC games.

But if you really did want a streaming DAC the best way would probably be to replace the OPNs companion DAC with something that could combine the audio stream from the OPN with a DMAed steaming audio sample. If youy could convince Yamaha there was money in it they might even have be willing to design something for that. If not there are other ways you could have added the stream from the OPN and the streaming DAC together before sending it to the companion DAC.

I am intimately familiar with most Yamaha FM chips, in terms of making music with them, writing software to utilize them, using them in my own custom hardware and also emulating them. I am very well aware how the chips can be used in every conceivable way pretty much (but I am most familiar with OPN family). It would have certainly been interesting to see a 4op chip be standard in PC world but I don't think YM2203 or 2608 would have been it, particularly the latter with its external memory requirements and much less 2610(B) which needs two kinds even (although later NeoGeo games used a chip that could join the two sound buses into one so single memory could be used for both types), not in the time when memory was still very expensive. The few sound channels were definitely gonna be a limiting factor as far as 2203 goes, it would not be able to compete with OPL2 or OPL2+CMS... I would say you really want the 6 channel chips and if there was gonna be one it was gonna be 2612/3438 instead of 2608 or 2610, but even then I don't see almost any non-japanese composer make good use of it.

You can see this on Mega Drive/Genesis games, most of the actually good sounding games were all japanese ones and vast majority of the western made games had pretty poor music although not always poor compositions. There were a few composers who really could make the chip sing like Matt Furniss but in general it seems that people didn't know how to actually use these chips or perhaps even cared... Even OPL2/3 can be made to produce excellent sound, stuff that OPN or OPM would really struggle to do due to less channels but that's not really what was done when these chips were still relevant either so I am really not sure if a different chip would have made any real difference in what we got musicwise anyway... Potential certainly would change but all that potential seems to always get utilized after the fact, this guy's OPL3 stuff is basically undoable on OPN/M family for example : https://www.modules.pl/index.php?id=modules&aid=1691

I very much meant streaming, a DMA based method will be necessary for that for sure so you could sort of end up with SB deluxe, depending on what kind of FM chip was there 🤣. Samples were the thing that really mattered it seems, and sooner or later the limited memory of 2608 etc. would have been a bottleneck, which you were not going to having with a stream device. That is probably one reason why PC sound cards remained quite boring compared to stuff you saw in some of the other computers etc and perhaps why all those other computers ended up dying out in the end, as sad as it may be...

As far as OPL4 goes, I think it came to the scene too late and probably was too expensive too. It also could not use DRAM (although there's reference to OPL4D chip that could but I have never seen one in the wild and no documentation of it seems to exist either), only SRAM or ROM and that makes it immediately unviable by the time it came around. Yamaha SW20PC card has a 128KB SRAM on it, but you cannot do a whole lot with that little memory, not in context of sound effects and especially when they had no compression at all. If it was around in like 1990 then perhaps it could have made some difference but not in 1995 where games already were using many hundreds of KB for sound data. You also cannot actually update the sound memory on the card without turning off all sound generation, the sound rendering needs to be disabled before the RAM/ROM bus becomes available so it is basically not possible to update samples on the fly, definitely not when music has to play at the same time...

The MSX Moonsound expansion seems to have cost 399 or 499NLG in 1995, I wonder much money was that...

Tiido wrote on 2024-03-23, 02:40:

I am intimately familiar with most Yamaha FM chips, in terms of making music with them, writing software to utilize them, using […]
Show full quote

Atomic Skull wrote on 2024-03-22, 06:27:

Well yes you'd want a steaming DAC for that. But as far as games programming is concerned if OPN had been the basis of PC sound […]
Show full quote

Tiido wrote on 2024-03-19, 22:44:

The main issue with these ADPCM samples is that they come out of an external memory, which you cannot easily update. Easily here means to play back a WAV file or some other type of data stream. There is no DMA mechanism or anything and the keyhole interface to update that sample memory through the chip is very slow so it will take a lot of CPU power just to send the data stream, and even more to compress it to Yamaha specific ADPCM format in case of wanting to play multiple samples simultaneously such as what a game like Doom might do.

Well yes you'd want a steaming DAC for that. But as far as games programming is concerned if OPN had been the basis of PC sound cards instead of OPL then the way PC games approached sound would have been different anyway. OPN was widely used in Japanese PCs for example and they used it quite successfully. In that case game sound drivers would upload samples to the OPN's external memory ahead of time and the games would be designed around this. And 128k or 256k isn't as small as it sounds when you consider that ADPCM is 4 bits per sample. I'm not talking about a world where OPN and OPL coexisted I'm talking about a world where the Adlib went extinct alongside the Creative Game Blaster and IBM Music Feature Card and OPN based sound cards ruled MSDOS PC games. A YM2203 based card would have been much less expensive than the Adlib and might have been able to succeed based on the much lower cost (and assuming game developers supported it so marketing would have been important as well.) Adlib was imo way too focused on PC music and OPL on MSDOS succeeded mostly because it was the default.

Lack of a streaming DAC sounds like a deal breaker to us but for the majority of PC gamers back in the way it was a niche feature. Most people weren't into tracker music or anything like that they just wanted a card that would make music for their PC games.

But if you really did want a streaming DAC the best way would probably be to replace the OPNs companion DAC with something that could combine the audio stream from the OPN with a DMAed steaming audio sample. If youy could convince Yamaha there was money in it they might even have be willing to design something for that. If not there are other ways you could have added the stream from the OPN and the streaming DAC together before sending it to the companion DAC.

I am intimately familiar with most Yamaha FM chips, in terms of making music with them, writing software to utilize them, using them in my own custom hardware and also emulating them. I am very well aware how the chips can be used in every conceivable way pretty much (but I am most familiar with OPN family). It would have certainly been interesting to see a 4op chip be standard in PC world but I don't think YM2203 or 2608 would have been it, particularly the latter with its external memory requirements and much less 2610(B) which needs two kinds even (although later NeoGeo games used a chip that could join the two sound buses into one so single memory could be used for both types), not in the time when memory was still very expensive. The few sound channels were definitely gonna be a limiting factor as far as 2203 goes, it would not be able to compete with OPL2 or OPL2+CMS... I would say you really want the 6 channel chips and if there was gonna be one it was gonna be 2612/3438 instead of 2608 or 2610, but even then I don't see almost any non-japanese composer make good use of it.

You can see this on Mega Drive/Genesis games, most of the actually good sounding games were all japanese ones and vast majority of the western made games had pretty poor music although not always poor compositions. There were a few composers who really could make the chip sing like Matt Furniss but in general it seems that people didn't know how to actually use these chips or perhaps even cared... Even OPL2/3 can be made to produce excellent sound, stuff that OPN or OPM would really struggle to do due to less channels but that's not really what was done when these chips were still relevant either so I am really not sure if a different chip would have made any real difference in what we got musicwise anyway... Potential certainly would change but all that potential seems to always get utilized after the fact, this guy's OPL3 stuff is basically undoable on OPN/M family for example : https://www.modules.pl/index.php?id=modules&aid=1691

I very much meant streaming, a DMA based method will be necessary for that for sure so you could sort of end up with SB deluxe, depending on what kind of FM chip was there 🤣. Samples were the thing that really mattered it seems, and sooner or later the limited memory of 2608 etc. would have been a bottleneck, which you were not going to having with a stream device. That is probably one reason why PC sound cards remained quite boring compared to stuff you saw in some of the other computers etc and perhaps why all those other computers ended up dying out in the end, as sad as it may be...

As far as OPL4 goes, I think it came to the scene too late and probably was too expensive too. It also could not use DRAM (although there's reference to OPL4D chip that could but I have never seen one in the wild and no documentation of it seems to exist either), only SRAM or ROM and that makes it immediately unviable by the time it came around. Yamaha SW20PC card has a 128KB SRAM on it, but you cannot do a whole lot with that little memory, not in context of sound effects and especially when they had no compression at all. If it was around in like 1990 then perhaps it could have made some difference but not in 1995 where games already were using many hundreds of KB for sound data. You also cannot actually update the sound memory on the card without turning off all sound generation, the sound rendering needs to be disabled before the RAM/ROM bus becomes available so it is basically not possible to update samples on the fly, definitely not when music has to play at the same time... Closest thing is GUS here, and while GUS didn't need to be silenced to update the memory, it seems to have been pain in the ass enough that games simply didn't use the chip the "intended" way and just turned it into a poor streaming device, totally wasting the chip...

The MSX Moonsound expansion seems to have cost 399 or 499NLG in 1995, I wonder much money was that...

It was being sold? It never crossed my mind that such things were available as anything else than a DIY fashion. I always seem to forget that not all computer hobbysists were tinkerers. My bad.

This gave that info : https://www.msx.org/wiki/Sunrise_MoonSound
But it isn't the only one around it seems, there are a few flavors.

I now notice there's this strange double post of mine... I wonder how that came to be...

Tiido wrote on 2024-03-23, 14:17:

This gave that info : https://www.msx.org/wiki/Sunrise_MoonSound
But it isn't the only one around it seems, there are a few flavors.

I now notice there's this strange double post of mine... I wonder how that came to be...

OPL4 carts have never been cheap. The DalSoRi R2 was the best one ever made and the final batch was $120 USD. But it's been retired due to parts shortage. https://www.msx.org/wiki/JunSoft_DalSoRi_R2.0

For anyone after a cart, this is your best option: https://theretrohacker.com/product/msx-opl4-w … rtridge-ymf278/

Those are new ones, the originals from 1994 or 1995 are an actual reference point in time when those chips were new and possibly mainstream in some way which is why they were brought up here. I'm not sure how much money those original ones were, the numbers seem big but they're in currency of another time and from a country I know basically nothing about.

I'm surprised nobody has mentioned the SAAYM ISA card (CMS/GB/2x SAA1099 + YM2151/OPM) from TexElec. This could basically reproduce any OPN sound we need since it's just a mildly feature-stripped version of OPM.
I wonder what the chances are of getting a DOS Genesis/MD emulator like Genecyst to work natively with the SAAYM. And the SAA1099 chips could be used to reproduce the SN67489 channels to boot. (is the source for Genecyst available actually?)

I have forgotten that card even exists 🤣. But yeah, YM2151 will be able to do most stuff that YM2612 can except SSG-EG mode and PCM channel and luckily SSG-EG is not widely used but PCM channel is used, however. SAA1099 should be able to do very decent simulation of VDP PSG noise and tone capability although there is no support for 6.25% duty cycle mode of VDP PSG. There is an SN PSG only in SG-3000 and SG-1000, and it has some difference from VDP PSG which has one freq step less (freqs 0 and 1 are same on VDP, while on SN 0 is the lowest freq) and 6.25% duty cycle instead of 6.66% due to 16 vs 15bit LFSR, which will play a role in a number of SMS games which use that mode.

Only complication I can see from emulator standpoint is that OPM is actually really slow to access, requiring long access delays similar to OPL2, but on a 486 that won't matter since emulating the chip itself is gonna take even longer but on some faster hardware it will be the liming factor but definitely workable.