The best of FM Chip-Music: YM3812 - ADLIB - OPL2
Vložit
- čas přidán 25. 08. 2012
- ym3812 is a fm synth chip, used in many systems during the end of 1980's and early 1990's.
Also known as OPL2 and "ADLIB" for PC, it was used as a poor midi player.
Fortunatelly, many people created some adlib trackers to use the true power of this chip.
These tunes used to be at: www.chiptune.de, the page no longer exists, look for "adlib music archive" or similar
The creator of Loudness tracker is called "Andras Molnar" (not Molmar). - Věda a technologie
![Best of Chiptune [8 bit music, retro visuals]](/img/n.gif)
Nice. I don't think the Adlib was ever really pushed to its limits like the SID chip in the C64 was. I like the sound of the SID chip better, but the adlib has a cleaner sound to it than the SID and more voices. So I think ultimately the Adlib could have done a lot better. But yeah, most of the music that we heard in DOS games was designed to be platform independent (MIDI) so never really took advantage of what the chip could do.
The iBookGuy True! The music in most games didn't push the limits of the adlib standard. The only guys who took it to the limit are the vibrants crew, with their edlib tracker. They made wonders!
+The iBookGuy Same can be said with the POKEY chip in the Atari computers.
+The 8-Bit Guy check out diode milliampere ;)
Totally. Some of these demos, and a few of the better OPL3 MIDI patches are actually quite nice, unlike the rather boring OPL2 patch sets found in most MIDI compatible drivers.
I don't think FM will ever be as musically versatile as simple waveforms w/ or w/o analog filtering, but there are some cool flavors to explore here.
Guys, 1 question, where are the voices and instruments stored? in the FM chip itself? or the FM Chip only process the sound, thanks
Genesis fanboy here but the Adlib was awesome too, love and respect to all FM
Nice to see you here.
Hey nice to see you too :)
oh hi hows it going Genesis (Megadrive) Fanboy? i know we are searching if they're any yamaha synth chips left.
:)
Didn't Genesis use the same Yamaha OPL2 chip by Yamaha that Adlib was based on? I'm pretty sure they're technically the same...
OPL-2 was used in a lot of arcade games in the late 80's and early 90's as well. Toaplan used in in their 68000 based arcade boards up until around the time they did Truxton II.
At 1:44, that meant to say EDLIB Tracker, not D00 Tracker. Anyway, this video goes to show the true power of FM chiptunes which I believe is underappreciated. Great showcase, Osystem!
Other good YM3812 music:
Wacky Wheels
S.P.Y.: Special Projecy Y
Fury of the Furries
MegaRace
Genpei War
The Legend of Sword and Fairy
CyberSphere
Tyrian
Cobi Comi
Empire of the Angel 2
Zone 66
Princess Maker 2
Eol's Adventure
Olmang Jolmang Paradise
Ys II Special
Haunted Castle
Zeliard
Metal & Lace: TBotRB
First Samurai
TFX: Tactical Fighter Experiment
Dune
KGB
Hired Guns
Laser Squad
MiG-29
Lollypop
HMI Sound Setup Utility
Knights of Xentar
System Shock
Alianger megarace
I love fm synthesis. I have a theory on making sugar free drinks based on fm synthesis. Love this video
???
YM3812 was much better than most people thought. In Adlib Trackers still is possible to create good music for this chip. I thought that YM2612 (used is Sega Mega Drive and Fujitsu FM-Towns) was weak sound chip but sound quality depends also on sound driver and music composer skills. Most Mega Drive games use GEMS which is poor sound driver comparable to MIDI played by OPL FM sound chip in MS-DOS. Composing music for Sega Mega Drive was much more difficult due to hardware limitations but most games using later versions of SMPS 68k or some custom sound drivers have great music. About the types of music composed for various systems I do not have any bias.
Old sound synthesizer chips sounded the best when the composer pushed the chip to its limits. And that would almost always require the composer to also be a good programmer, like Rob Hubbard with the SID chip. GEMS on Genesis and MIDI on AdLib/SoundBlaster cards just doesn't push the hardware to its outer limits.
The 2612, used correctly and alongside the other audio devices in the system (at least, in the Megadrive - I don't know what, if any, the FMT had), should be a highly capable device, and this has been shown off more in recent years by dedicated chiptuners. When each piece is added together, you've essentially got something very nearly the same as a 2608, which you could say also gives output in a similar general class to the OPL3, or at least much closer to it than the OPL2... but with quite a bit of added intelligence thanks to the Z80 sitting alongside the PSG part.
Or in other words, you've got 9 tone channels overall (6 FM, 3 squarewave), and possibly more if you can learn to grok the CSR / individual sinewave mode on FM channel 3 (= 5 FM, 4 sinewave, 3 squarewave, so 12 total), a dedicated white or periodic noise channel, and PSG envelope which can be manipulated to produce a bass buzzer noise (or, the whole thing can be given over to producing samples with somewhere between 4 to 10 bit precision "depending") and an 8-bit PCM (taking up one of the FMs? I can't remember... nor can I remember if it's ADPCM capable, but the Z80 should be more than powerful enough to deal with the decoding needed there, so long as there's the right interconnects in place to allow it to get the data from ROM, access enough RAM to work on it, and then shunt it into the 2602/separate PCM channel).
...ahem, sorry. 9 to 12 tone channels of which, at any given time, at least 4 and upto 6 are 4-op FM, a dual-mode noise channel, and 8-bit PCM possibly with ADPCM capability either onboard or via the sub-processor.
Compare against 2608 with 6x 4-op FMs (one of which can also do the weirdo 4x sine thing), the same 3x squarewave PSG (albeit with a wider frequency range) and a similar but not quite identical noise channel and envelope, plus a functionally identical PCM with onboard ADPCM decode. Or even the 2203 in fact, with no PCM and half as many FMs (but still the PSG, and 4x sine capability)...
Or OPL3 with 36 FM operators that can be combined in various ways (all of them can be paired, with one of the 2-op modes allowing something close to individual sine generation, 24 can be quadded together instead, and a further 6 can be sacrificed to provide improved quality percussion), all with more basic waveforms than just the regular sine, giving 9 to 18 tone-or-noise channels plus optionally 5 percussion voices. OPL2 essentially being a subset with fewer (but still multiple) base waveforms, half the operators and the only options being 2-op modulated or pseudo-individual-sine, so basically just 9 tone-or-noise voices. PCM and PSG not being relevant for them as in the latter case they can be persuaded to generate reasonably good squarewaves and pseudo-whitenoise anyway, and in the former they are almost universally (except in the early Adlib cards) installed on a daughterboard with a separate buffer register and DAC forming a simple dedicated channel anyway. Though I have seen, and tried out, an audio driver supplied with a particular old DOS soundtracker that enabled successful, if of course rather noisy 6-bit PCM output through the OPL2 or a compatible/emulating rival chip alone without needing a dedicated DAC. No idea how the dickens that worked (using a combination of operator and LFO settings that kept one or more of the channels at maximum and/or minimum amplitude vs the "zero" centre line, and both swapped the polarity and adjusted the master volume for either the channel or the whole card very rapidly, similar to the techniques used to get samples out of PSGs?), but it was surprisingly effective. Probably couldn't play any actual FM alongside it, mind.
Long story short they're all actually rather similar to each other, particularly in their installed configurations, and if any of them in particular sounds exceedingly "bad", then, barring physical malfunction of the hardware, it's probably a software problem. In other words, either whoever wrote the software to allow composing and playback on that hardware, or whoever composed, arranged, and programmed-in the music made with that software to play on that hardware, did it really badly. Because it's pretty easy to find examples for each of those listed systems that sound **really** good... and if it can be demonstrated as actually being capable in the right hands, the only conclusion you can come to, if it's got a reputation for sounding like shit, is that most of the time it's actually been in the _wrong_ hands.
YM2612 aka the sound chip in the sega mega drive had more operators,but less sound channels than YM3812
@Blaze Rhadon
I know your comment is 2 years old, but the GEMS driver wasn't really the problem. I watched a video a long time ago that showed what it was and what it allowed people to do with the Mega Drive's audio, and it turns out what most composers ended up doing was use the default settings while composing and not bother to experiment with the features that allowed one to alter or create different fm sounds and instruments.
I think there are many OPL3 tracks mixed here. Or at least soms have double YM3812 (because of the stereo sound). Some have complex waves that wouldn't be really possible without the extra modulation modes that the YMF262 offered.
Correct me if I'm wrong, as I'm from the 2001, so AdLib wasn't really a thing anymore then xD
Either way, this is a really nice compilation of OPLx music ❤️
I recorded them with emulators, they are all opl2 tunes, but emulators added stereo by using 2 emulated opl2 chips. Anyway it is not very different from a mono opl2 output. You can check the programs (or trackers) used to create these tunes, were programmed for opl2 (loudness, edlib, rad). I recently played most of these tunes on a real opl2 (a little adlib clone connected to an arduino).
@@Oosystem Mmm... Interesting, i knew OPL2 is powerful (I've been playing with a few chips to see how hackable they are), but this is way over my expectations. Also nice to know you actually tested these tracks! Thanks for clarifying that
@@rhodexa I created a little OPL2 player for msdos. it has bugs, but it works ok for the most part: github.com/mills32/Adplug-for-8088-86. OPL2 is kind of powerful, but to make great sounds you have to use a lot of CPU, so it is not the best sound chip for games or demos, unless you use a "fast" CPU (386 or later). Loudness format is the best in my opinion, because it did not require a very fast cpu and it produces very cool sounds, the bad thing is, there is no "tracker" available to create more loudness songs. Edlib tracker is available, and it is the most powerful, but it needs a fast cpu. The easiest program to use for OPL2 music is RAD (the first version), it can produce convincing instruments and does not require a lot of CPU power.
Very interesting.
The OPL2/3 music wasn't killed by Ultrasound etc, but by lazy coders of music, who used poor GM like sound patches/presets...
0:06 Someone has never heard the soundtrack for MegaRace...
Are these actual Adlib (OPL2) or Adlib Gold (OPL3)? If genuinely the former, those composers were *gooooood*. The sophistication of the instruments seems too much, and certainly "twintracker flutes" sounds almost like it's got some kind of 8-bit PCM sampling going on.
Any time I've heard 2-op synthesis before, it's been rather more simplistic and thinner than this, though of course the OPL2 has the advantage of having more than just the plain sinewave to rely on as the base or modulating wave, and can combine the simple genius of its modified-sine waves together in 15 extra ways than a more straightforward pure sine chip could.
Nice little medley to listen to either way, so thanks for putting it together :)
They are all OPL2 tunes, not OPL3.
Go to chiptune.de, download adlib archive, the module is inside twintracker folder. You can play it with some ms dos programs. I couldn't find a twintraker editor, just a player.
check out the Yamaha PSS-470 keyboard. it has the YM3812 chip inside and actually allows you access to the chip's synth engine to sculpt your own sounds. you can find em for cheap, too!
I'll check !!
At the end of the video i say it was recorded using adplug plugin/emulator, ( it simulates stereo).
It is in_adlib plugin, with emulated chip
Adlib was completely wasted in the era it was used. I can't believe it's actually - more or less - superior to SID.
And here's me just wondering what kind of memory and CPU power this would take to implement these songs in games. Honestly speaking if someone knows, id love to hear!
Great compilation. Personally, I love the first bit of music in DOS 'Lemmings' (you can find it on www.vgmpf.com). I think that's 'AdLib' (though you'd think it'd be PCM, having been written by DMA Designs ;-) ).
www.chiptune.de no longer seem to exist. Did it move somewhere?
I had the same feelings about it too, figured the hardware was decent but music from DOS games was either good or terrible with not much in the middle. Tough gig though, many hardware vendors and support requirements, so much easier with one sound chip per platform.
These are certainly better than that most games did with the hardware, but the hardware was definitely bottom of the barrel in synthesis with much better hardware existing before it. But the chip was olf the shelf and inexpensive, and so it got chosen.
What's the name of the very first song in the video? It didn't appear to be named in the video.
I'm looking for it, :)
The file is "DRX_BUS.D00" the song is called "bus" it is surelly inside the adlib archive, (if it is around the web)
Thanks! Found it by googling the title you gave.
It's also one of the Edlib tracker default songs
1:46 Does that one (and the others that are in stereo) use Dual OPL2?
They are all using the dual opl2 emulator.
can you please upload dos game music with this yamaha 3812 chip? Like king's quest & stuff.
You'll have to search.. There are some adlib music archives around.
Can you share the source for the song at 01:30?
I think of Nichibutsu and Tyrian 2000.
Guys, 1 question, where are the voices and instruments stored? in the FM chip itself? or the FM Chip only process the sound, thanks
The instruments are stored in files, they are some instructions to be read by the chip, and the music sequence is also stored in a file, not in the chip. You can use Reality Adlib Taker (RAD) inside dosbox, to create music like that, and you'll se how the music is created. The chip just plays the instructions from a file and generates the sounds.
Well, the FM chip *makes* the sounds, so you could say the voices are stored within it :D
But no, other than the 2151 and its close siblings, no common Yamaha chip has any real "presets" inside. You power it up, all of the internal registers (master volume, panning data, note on/off for each channel, note value, volume, oscillator type and relative frequency, operator connection mode, various individual settings for each operator itself, feedback, ADSR envelope specs, LFO, detune, and a bunch of other even more esoteric things) are zeroed out, and just to get a simple A440 sine wave out of the thing you're going to have enough logic "offboard" of the chip to be able to write probably at least a dozen bytes into various different register addresses. There's not even any place to cache the settings for a particular instrument so you can just say "make channel 3 use defined instrument 7", the registers have to be rewritten every time. Though luckily once set, they stay set, so playing a simple tune on that channel you just make play a middle A gets a lot easier - a "note off" command to make it shut up for a moment, then a bunch of note value, note-on, and note-off commands, which will make it peep out a sinewave with the same instrument settings but different notes.
(one of their smartest ideas was building the ADSR envelope into the chip and having it as another register-controlled thing, otherwise the bulk of commands sent to it for anything but the most rudimentary output would be endless volume adjustments; even using the PSG style envelope to produce simpler attack and release slides with coarse manual decay/sustain would mean a lot more traffic for each channel than having the envelope programmed in and sending note on/note off)
Think of it a bit like a complex pipe organ in a church... there's a whole load of different sounds that can be made by pulling out different stops and pressing the keys on the manuals, but you still have to pull the stops and press the keys. And it's not even one of those fancy pipe organs with certain stops that make preset combinations happen, you have to set each one individually. The tech to automatically make a certain sound in response to a certain code being sent to the chip, or a certain combination of stops being pulled on the organ, is in place, but it doesn't have the ROM inside (or with the organ, whatever contrived set of cams etc would be needed to automatically pop out certain stops at the push of a single button) to allow easy one-byte selection of a certain instrument.
Which is a bit odd, really. Given the number of different things Yamaha DID use on-chip ROM for in their soundchips, the existence of the 2151 (which had a limited set of registers for user control of a single voice, then fifteen other voices that were baked into a tiny ROM (easily less than 1/10th the size of what you'd need for a General MIDI chip) and selectable with a single command), and how General MIDI L1 is pretty much perfectly set up for use with a chip that can pick between any of 256 different voices (128 main patches, 64 or so percussion pieces, and then another 64 of whatever you like) using a one-byte value, you'd think that there'd be at least _one_ such thing amongst their product line. But, there isn't, and even if it later turns out I missed one and there _is_ , it ain't any of the OPLs (...well, except the OPLL, as that IS the 2151).
You could make something of an argument that the PSG counts, because you only have to set frequency and volume to get a sound out of it, but that's a bit of a cheat because it only has one "instrument patch" - plain old 50% duty cycle squarewave... there's nothing TO set up ;)
(would the SID count, similarly? it has multiple different waveforms, but, really, picking between them rates as nothing more than changing which of the four basic quasi-sines the OPL2 uses to make its sounds *from* ...)
So anyway, in this case, the instrument definitions live offboard of the chip, in the software. Even with soundcards that might have a Gen MIDI ROM on them, to make sending patch commands to the Yamaha chip a bit easier, that's still a separate thing and essentially just a different way of setting it up through software.
Ahem. Yes.
+Mark Penrice thanks dude
+Mark Penrice
The YM2151 has instrument no presets inside. The only Yamaha FM soundchips that have those are the YM2413 and its derivatives.
Cheers for the correction, fr_schmidlin - I did realise myself later on when double checking some things that I'd got some of the numbers mixed up, but there was no chance by that point of backtracking and correcting all the mistakes...
Yamaha just made so many very similar chips, a lot of them quite low capability, that it's hard to keep track of them all mentally. Maybe I should have taken to using their codenames instead, but then you start getting confused between OPN, OPNA, OPNB, OPL, OPLL, OPL2... etc :D
Is this emulated?
So the YM3812 chip is mono. How are some of the tunes in stereo?
well we know it's mono but it might be in stereo/surround sound.
Dual OPL2 setup
the opl3 is like a stereo opl2 or a dual opl2 setup
The remake of Zalza's My dirty old Kamel is probably better than the original. :)
Where can I download the midi from this video? Thanks
Not midi.. they are special formats for the chip.. look for the names in google, there are still some pages keeping them. You'll need adlib plugin for winamp or xmplay or any player you use, to play the files.
@@Oosystem there's one more option these days: vgm files, which just capture the registers and data sent to the synth chip. The are plenty vgmplayers out there, even ones with real hardware.
Is it possible to write heavy metal for OPL ?
Doom's soundtrack for 'Adlib' and compatible: Iconic. Maybe better on other synthesisers, but see for yourself.
Alright, having gone off for some Doom goodness [Adlib style], I guess it's soundtrack might be called 'Techno which evokes metal.'. It depends how broadly you want to define the genre.
thanks , so your telling me that the chip has 128 midi instrument, or only reproduces orders from the program files and outputs it with his own generator quality
The chip only follows orders, it is a "synthesizer", and it can theoricaly produce infinite instruments, not just 128, and it does not have to use the midi interface, most musics in this video can't be converted to midi because the files are using custom instruments.
The well known 128-patch (plus something like 64 percussion sounds) General MIDI instrument bank is actually nothing more than a set of recommendations for any device that claims compatibility to provide, and to set as active on a channel when receiving the matching GM command code. And indeed, that's only "layer 1" - the later L2 standard expanded it to something like 256 main patches (mostly variants on existing ones that originally encompassed multiple similar sounding instruments into a single compromise choice, but a couple new ones as well) and nearly 100 percussion sounds within each of several different drumkits.
GM L1 was drawn up so long ago now that it had to be deliberately limited and make some compromises on what instruments were or weren't included because of the difficulty of encoding that many different presets into an affordable amount of ROM at the time, and the need to keep the patch / percussion note signalling relatively compact to not overload the quite low bandwidth MIDI interconnects (a mere 31.25 kbit/s) if pushing things to the limit and regularly redefining what each of the 16 available channels is using at the time, going wild on the percussion, and using a lot of notes besides on the 14 or 15 channels not being used for drums. 3125 bytes per second (it uses 10-bit words for reliability...) may sound a lot if you're just encoding a fairly simple piece, like a piano played by an only moderately proficient soloist, but it divides away to very little as soon as you start doing anything complex, even with compact encoding, so there's no way verbose codes can be used.
New interconnection standards have done away with the old low speed 5-pin DIN setup (...or at least, have a fast connection to a breakout box where you can use multiple individual cables, to the level of each channel getting its own 31.25kbit/s allocation) in favour of USB and other such much faster alternatives, so that's less of an issue now, likewise the ROM storage is far less problematic (though some super high fidelity PCM soundfonts can run to many gigabytes, never mind too many kilobytes!), so the cuffs are off somewhat and allow the industry standard to be far more comprehensive. Instead of it being "try to use these for intercompatibility, then use your own proprietary sysex codes for everything else".
Anyway, there's little magic in those GM codes. They are literally just numeric codes for instrument and percussion-type _names_ , nothing more. There's not even any recommendation on just what each should sound like, how they should be implemented, standard operator and envelope settings for particular types of synthesis, all that jazz. It just tells you the common name of a particular instrument, and you (as the manufacturer building up a sound bank, or a musician having to make your own patches) are free to - in fact, forced to - either use your own judgement when it comes to interpreting each one, or more likely find an existing version that you like and license, steal, or "be heavily inspired by" it.
So what you hear when throwing a MIDI file at any of these chips, or any synth in general in fact, is all down to the personal taste of whoever wrote the driver or built up that GM-corresponding sound bank... as well as somewhat how bothered they were about making a good job of it vs just getting it out the door as quickly as possible.
(prior to the GM standard, things were a bit more wild, as there was no actual defined "midi file" standard, it was just a connection method with a fairly basic set of protocols attached so that interconnected devices could understand each other - you may as well talk of a "serial file", "parallel file" or "USB file", to describe any set of data moving along those cables and being captured into permanent storage, rather than being interpreted and used immediately and then discarded. It's a bit mysterious where the idea of having an independent file that collected a load of MIDI note / patch / volume / etc commands together into an independent object came from, really, as until that appeared, you would instead have proprietary save files for sequencer programs - Cubase, for example - which would store the more detailed, master version of your work, and then when it was time for playback to happen, the program would deciper its own pet format into something to squirt down the cables, including sending target-device-specific commands to set a particular instrument to a particular channel, and maybe even transmitting the basic definitions for that instrument as part of the initial setup data... a few dozen bytes taking a matter of milliseconds for FM or other synths that built their waveforms up from raw oscillator signals, or maybe a few dozen kilobytes if it was transmitting actual samples (yes, actually a thing that happened, though more often it was preferable to use a more sophisticated sampler with its own disc drive that could pre-load the data instead of having to download it anew during the first playthrough of each studio session)... and each of them entirely unique to the file, studio, devices, musician in question. If you wanted to interchange the data with someone else, then they had to use the same software, or you had to use a converter program that could make it cross-compatible.
.MIDs and the GM standards have made SOME of that stuff a lot less of a hassle, but it does sort of stifle innovation in a way. If you want to do something truly unique, you still have to program it in yourself... and as that immediately breaks compatibility with everything else, demanding either a loss of capability or needing to run it through converters for other people to be able to use it, you end up back at square one. So a lot of people are probably discouraged from doing so. The rise of all-in-on, at-the-computer DAWs that pack an entire studio into a single program, however, do make it kind of moot)
What's the name of the track at 0:00-0:18?
"Bus" (drx_bus.d00) Look for Edlib tracker it is included there.
Andreas Molnar not Molmar
Sounds better than 2612?
no... just similar.
Osystem Thanks
😼
Why sega doesn't use this chip for genesis instead of ym2612. The ym2612 one soundes too thin & harsh.
They don't have much difference in terms of sound synthesis. And actually YM2612 is superior because it was capable of PCM playback which was often used for drum sounds.
@@stam1ska They are very different. Genesis did 4-operator FM with one sine wave and 6 channels, while ADLIB did 2-op FM with 4 sine waves and 9 channels.
@@RyumaXtheXKing it's still FM so there's no much audible difference in terms of conception of the sound. Also, Genesis had a channel which could do PCM, and also a PSG which had 4 channels of square waves.