How about a summary of the results here so we don't need to follow links and watch films?

well, it's very subjective, you have to hear for yourself to decide... :p

something I wonder is if you have compared other OPL cards, as some of the differences could be down to filtering etc. maybe?

dionb wrote on 2020-09-01, 21:49:

How about a summary of the results here so we don't need to follow links and watch films?

Good point. So here is my personal opinion/conclusion:

I was surprised how good / how close the CMI-FM and the ALS-FM were. Good choice!

Crystal-FM is not very accurate, I would avoid them if you have alternatives. (Without direct comparison it might be acceptable)

AnalogDevices-FM is one of the worst clones but I think it is interesting because of its uniqueness.

I like ESFM, it's close to the original but with its own character and definitely a good alternative.

Creatives CQM was a big surprise to me in a positive way. I can't share all the criticism they got for it. It's definitely one of the best clones for me.

OptI is above Crystal but beyond ESFM/CQM. A clone with its own character - neither very good nor very bad I think.

Aureal-FM was a disappointment to me as I heard it should be good. In my opinion it is one of the worst clones.

Ensoniq-FM emulation: No comment.

Dream-FM: Not very accurate and speed sensitive but barely acceptable.

FortéMedia-FM: Big surprise, very accurate (maybe a copy?)

S3-FM: Not very accurate but interessring and unique sound that may fit the mood in some titles.

imi wrote on 2020-09-01, 22:02:

well, it's very subjective, you have to hear for yourself to decide... :p

something I wonder is if you have compared other OPL cards, as some of the differences could be down to filtering etc. maybe?

I do totally agree. Overall I would say that many of the clones are good to go.

I haven't compare different Yamaha OPLs, this might be stuff for a further comparison. But yes, there is a difference in the filtering. Older cards mostly have a warmer/softer tone and using low-pass-filtering that is uncommon on newer cards like the YMf71X and later ones as far as I know.

a thing that might make comparing them easier would be to use shorter snippets and play back the exact same part on each other card for direct comparison.
sure that breaks up the song a lot, but listening to the song is not the primary goal here I guess ^^ could also make that into one video... like play few second snippet on OPL... card1...OPL...card2....etc.

This is good info to add to the collective knowledge, thanks.

I have read that it is tough to nail down the *exact* characteristics of FM chips without a complex database of fourier comparisons across many different cards. That is, to truly isolate what is the FM chip, and what is "the rest of the soundcard's influence" such as DACs and filters.

I listened to a few of the recordings (Crystal, CMI, CQM, and S3). I still like the Yamaha. In general it seems to have more "character", maybe because the composer is targeting the Yamaha with whatever trickery they're using to make interesting noises happen. The clones seem to sound watered down probably because they don't respond the same way.

A big issue with games in the synthesis era is that since every card sounds different, you can run into trouble if your card isn't a close match to what the programmer was using. So having the theoretically "best" card isn't even necessarily an advantage if it's always playing music that wasn't composed for it.
With mainstream games you can probably assume that most were tested on Yamaha OPL2 or OPL3 (depending on age of the game), because that was the standard.
With MIDI files it's more random because whoever composed those could have had anything, but they probably had something upscale like a Roland in many cases.

I remember seeing a segment on some computer TV show back in the 90s where that "Fat Man" guy (who did the music for a million popular DOS games) was going around endorsing sound cards based on how well they played his music. Later I read an article where he supported the use of redbook audio CD tracks because it would solve the consistency problem.

Do you have an OPL2 card?
I remember noticing the difference when I upgraded from a Sound Blaster Pro v1 (which uses OPL2) to an AWE32 with CQM. I was generally disappointed with that part of the upgrade but since the point of AWE32 was wavetable, I wasn't feeling very bothered about it.
It's been a very long time since I used that SBPro. The last time I ran it was for a few minutes like 15 years ago.

I was surprised how good / how close the CMI-FM and the ALS-FM were

Judging by recent ALS100+ discovery of true OPL3 clone, I won't be surprised to know that ALS120 just integrated that clone into main chip, similar to some Aztech and Creative OPL3 chips.

that OPL thing is so overrated, appart from those more modern PCI horrors and maybe CQM. But Crystal, ESS, ALS, they all perform pretty good, You do not need a real OPL or that ADPCM dumb OPL-SAx.

I certainly feel that OPL3 is quickly overrated when it is just used to listen to an OPL3 MIDI driver playing MIDI music. There are occasions that a game has music composed for the OPL3 chip that programs the OPL3 chip directly, but it is rare. Most games used some sort of MIDI engine and an OPL3 driver to play that MIDI data. In my opinion, if you want "authentic", you go for the intended MIDI module that was used to compose that music. If you go for "nostalgic", just buy what you had back in the day that gives you that special memory of the past.
There is quite a lot of Adlib music (hence OPL2) that was specifically created for the Adlib card. I would say, just get such an Adlib card or a good clone and call it a day. 😉

shamino wrote on 2020-09-03, 01:59:

I listened to a few of the recordings (Crystal, CMI, CQM, and S3). I still like the Yamaha. In general it seems to have more " […]
Show full quote

I listened to a few of the recordings (Crystal, CMI, CQM, and S3). I still like the Yamaha. In general it seems to have more "character", maybe because the composer is targeting the Yamaha with whatever trickery they're using to make interesting noises happen. The clones seem to sound watered down probably because they don't respond the same way.

A big issue with games in the synthesis era is that since every card sounds different, you can run into trouble if your card isn't a close match to what the programmer was using. So having the theoretically "best" card isn't even necessarily an advantage if it's always playing music that wasn't composed for it.
With mainstream games you can probably assume that most were tested on Yamaha OPL2 or OPL3 (depending on age of the game), because that was the standard.
With MIDI files it's more random because whoever composed those could have had anything, but they probably had something upscale like a Roland in many cases.

I remember seeing a segment on some computer TV show back in the 90s where that "Fat Man" guy (who did the music for a million popular DOS games) was going around endorsing sound cards based on how well they played his music. Later I read an article where he supported the use of redbook audio CD tracks because it would solve the consistency problem.

Do you have an OPL2 card?
I remember noticing the difference when I upgraded from a Sound Blaster Pro v1 (which uses OPL2) to an AWE32 with CQM. I was generally disappointed with that part of the upgrade but since the point of AWE32 was wavetable, I wasn't feeling very bothered about it.
It's been a very long time since I used that SBPro. The last time I ran it was for a few minutes like 15 years ago.

Even if the composers work with the original, their compositions could sound very interesting/nice with the different sounding clones. I really like some Doom II tracks with the S3-FM, and that is far from perfect:-) But sure I am not the typical person as I have the choice which card I am using, and I can understand that most users want to listen to the music in that way they used to listen it back the days or the way it was composed for the original. To be honest I really like the fact that there were different clones so you can have different results how a track sounds.

I also have opl2 cards, but I could not hear a difference to the yamaha OPL3 that is fully backwards compatible to opl2 as far as I know.

digistorm wrote on 2020-09-03, 09:20:

I certainly feel that OPL3 is quickly overrated when it is just used to listen to an OPL3 MIDI driver playing MIDI music. There are occasions that a game has music composed for the OPL3 chip that programs the OPL3 chip directly, but it is rare. Most games used some sort of MIDI engine and an OPL3 driver to play that MIDI data. In my opinion, if you want "authentic", you go for the intended MIDI module that was used to compose that music. If you go for "nostalgic", just buy what you had back in the day that gives you that special memory of the past.
There is quite a lot of Adlib music (hence OPL2) that was specifically created for the Adlib card. I would say, just get such an Adlib card or a good clone and call it a day. 😉

I can understand your point of view. My soundcard back the days was the Soundblaster pro, so the real deal for me is FM even if I like my wavetable cards and the external midi expanders today. But yes, I prefere my 386 Acermate with the Soundblaster 2.0 for the old Adlib stuff.

Cyberdyne wrote on 2020-09-03, 05:49:

that OPL thing is so overrated, appart from those more modern PCI horrors and maybe CQM. But Crystal, ESS, ALS, they all perform pretty good, You do not need a real OPL or that ADPCM dumb OPL-SAx.

I would prefere CQM over Crystal

The Serpent Rider wrote on 2020-09-03, 02:25:

I was surprised how good / how close the CMI-FM and the ALS-FM were

Judging by recent ALS100+ discovery of true OPL3 clone, I won't be surprised to know that ALS120 just integrated that clone into main chip, similar to some Aztech and Creative OPL3 chips.

The ALS100 uses external 100% yamaha clones. From ALS100+ onward they integrate an own FM engine but maybe you are right and it is just a plain copy of the yamaha. I will take a closer look on that.
For the moment, a looked at CMI and I think they also just copy the yamaha, I could not hear any difference.

imi wrote on 2020-09-02, 00:05:

a thing that might make comparing them easier would be to use shorter snippets and play back the exact same part on each other card for direct comparison.
sure that breaks up the song a lot, but listening to the song is not the primary goal here I guess ^^ could also make that into one video... like play few second snippet on OPL... card1...OPL...card2....etc.

Good idea! I am working on a comparison CMI vs. Yamaha with small snippets. Due to this I come to conclusion that CMI just copy the yamaha because I couldn't notice any difference. (Video is coming soon)

dionb wrote on 2020-09-01, 21:49:

How about a summary of the results here so we don't need to follow links and watch films?

I think that would defeat the purpose of his plug.. (though I like his youtube channel 😁)

Here is how I relate to the whole OPL3 vs Rest debate: Fuck OPL3. Seriously. Aside from a couple chiptune music tracks that take advantage of specific bugs in producing sound, the differences are minor between different implementations (except the ones that are outright godawful of course, but FM shaming will come later) and claiming superiority based on those minor differences is absurd. Here is why: OPL3 cards themselves, which are held on a high pedestal and presented as what everything else should try to be, are mostly of wildly varying sound outputs. Some have brick wall filters (SB16), some have proper lowpass filters (SBPro), ome have no filtering and a metallic sound more often attributed to CQM (most ALS007 cards). Some have proper output filtering and a good bass response (YMF71x cards for example, which incidentally do not implement a filter unlike Sound Blaster cards..), some have incredibly shitty output filtering and a very thin bass response (Some OPTi cards I have but that is the manufacturer's fault not chipset makers). There is no one single true OPL3 sound (and even if there was, I would argue it is the YMF71x cards' sound and not the SBPro sound as is generally argued on the internet). The truth is, it's all over the place!

And honestly, and subjectively speaking, some FM implementations sound so much better to me in games. My favorite FM implementation is CrystalFM, particularly as found in CS4236 and CS4237 - this has an incredibly warm and organic sounding FM implementation that makes OPL3's synthetic sound appear almost a generation older in my view. You would be hard pressed to find anyone tell ESFM apart from proper OPL3 in a blind test if at all. Testament to that is the discovery that ALS100 cards that were FM-shamed for having "shitty OPL3" up until a few months ago have been elevated to a desirable status now that someone discovered they actually have genuine OPL3 clones on board. Excuse me but where are all these people who shat on ALS FM these past few years? I have been adamantly claiming that ForteMedia FM was near-perfect and incredibly compatible for a PCI card for years but that card was ostracized for having a shitty OPL3 clone because at some point some dude on the internet dissed it in a blog page for not being genuine OPL3. He probably never even properly compared it.

Anyway. At the end of the day, use what sounds best to you.

I've always liked my ALS007's FM. ESFM is really nice too. Crystal FM sux

appiah4 wrote on 2020-09-09, 07:40:

Testament to that is the discovery that ALS100 cards that were FM-shamed for having "shitty OPL3" up until a few months ago have been elevated to a desirable status now that someone discovered they actually have genuine OPL3 clones on board.

I think people were just dissing the FM of the ALS100 Plus cards.

From what I gather, it was known at least since 2015 that a non-plus ALS100 card has a perfect 1:1 OPL3 copy on board.

Joseph_Joestar wrote on 2020-09-09, 14:14:

appiah4 wrote on 2020-09-09, 07:40:

Testament to that is the discovery that ALS100 cards that were FM-shamed for having "shitty OPL3" up until a few months ago have been elevated to a desirable status now that someone discovered they actually have genuine OPL3 clones on board.

I think people were just dissing the FM of the ALS100 Plus cards.

From what I gather, it was known at least since 2015 that a non-plus ALS100 card has a perfect 1:1 OPL3 copy on board.

Yeah my bad. It was recently discovered that ALS100Plus actually has a 1:1 OPL3 copy as well. Like you said, how do those ALS 100Plus+ dissing threads look now: