It is currently Tue Jul 16, 2019 6:21 pm

All times are UTC - 7 hours





Post new topic Reply to topic  [ 15 posts ] 
Author Message
PostPosted: Sat Mar 23, 2019 3:28 am 
Offline
Formerly Espozo
User avatar

Joined: Mon Sep 15, 2014 4:35 pm
Posts: 3501
Location: Richmond, Virginia
Basically just the thread title; I'm strictly curious (I'm not sure anyone here would have an idea though; I don't). Every single FM synth IC that I can think of used for video games, arcade games, vintage computers (YM2151, YM2612, YM2610, etc.) is only capable of 4 op FM synthesis, while (correct me if I'm wrong) many synthesizers from about this era, like the DX7, are capable of 6 op FM synthesis. Was it that 6 op FM ICs were too expensive, (just look at what these old synthesizers retailed for) or was it that they were never sold individually?


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 4:27 am 
Offline
Formerly WheelInventor
User avatar

Joined: Thu Apr 14, 2016 2:55 am
Posts: 2024
Location: Gothenburg, Sweden
The benefits of FM synthesis are that the technique were
-cheap on computational power if implemented in software
-cheap on logic if implemented in hardware.

Extra pins for control may cost more, but honestly i think it's because studio synthesizers needed to be marketed as feature-rich, versatile, omnipurpose. The hardware casing, keys and PSU means you might as well put all the versatility you realistically can on the pcb to justify the price. If it's in the 80's or early 90s, you buy one or a couple synthesizers for your band or studio; they're expensive, you want excel at some special niche or be as broad as possible.

But you're well past the point of diminishing returns after 4 operators. It's rare for me to make use of more than that, not to mention the tedium. A restriction on operators per voice (either set by the hardware or by yourself) might help you get to the results quicker.

Yamaha realized this too. By the late half of the 80s, their studio oriented synthesizers included waveform playback mixed in with the fm synthesis to help musicians/composers/studio technicians get results quicker - even if they still kept the 6 op voices.

I mean, i can make most sounds i want to make with 2 operators, some envelopes, vca:s and a dedicated LFO (doesn't need to be a full fledged operator unit).

_________________
http://www.frankengraphics.com - personal NES blog


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 5:21 am 
Online

Joined: Sun Sep 19, 2004 11:12 pm
Posts: 21505
Location: NE Indiana, USA (NTSC)
I'm curious as to what actually needs more than two operators, in order to map the exact difference in capability between an OPL (such as the NSF scene's use of VRC7) and an OPN (such as the YM2612 in the Genesis). Understanding what to do with the extra two operators, other than a pair of 2-op voices tuned a fifth (3:2) or fourth (4:3) apart to improve perceived polyphony, might help composers using OPN for the first time.

_________________
Pin Eight | Twitter | GitHub | Patreon


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 6:44 am 
Offline
Formerly WheelInventor
User avatar

Joined: Thu Apr 14, 2016 2:55 am
Posts: 2024
Location: Gothenburg, Sweden
Yeah, introducing partials some ratio above or below (ie, additive synthesis) is a common use. Or you can cross-fade two distinctive sounds, for example one attack sound and one body of tone sound, or to sweep in overtones at aftertouch. Or control depth of chorus between two detuned but similar operator pairs.

For a straight, dry organ sound you might route all four operators to audio out directly.

Another common trick is to let the output operator be an LFO and have other operators modulate that with hearing range pitches, as opposed to modulating a hearing range operator with an LFO range operator. This is beneficial for designs where an operator can only have one output route but you want several operators to sound lfo modulated.

Just some uses on top of my head.

On a related note, if an NES mapper could time an irq every n:th scanline, you could use it to introduce rudimentary FM synthesis to any generated channel... probably most useful on the tri channel.

_________________
http://www.frankengraphics.com - personal NES blog


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 11:36 am 
Offline
User avatar

Joined: Sun Jan 22, 2012 12:03 pm
Posts: 7526
Location: Canada
tepples wrote:
I'm curious as to what actually needs more than two operators, in order to map the exact difference in capability between an OPL (such as the NSF scene's use of VRC7) and an OPN (such as the YM2612 in the Genesis).

The main way I think 4-op provides a massive upgrade:

A 2-op pair is very good at letting you select one aribtrary timbre / waveshape, but the modulator envelope only really lets you vary the strength of that timbre from 0 / dull / sine to full strength.

A 4-op pair lets you have two of these. Thus letting you have two parts of the sound with two different timbres each on their own envelope, rather than just a single timbre for the whole sound.

Where this matters a lot is trying to model the "attack" part of a sound differently from its sustain/decay portions. In real instrument sounds, or just in practical synthesized sounds, the onset of a note generally has very different timbral needs than the sustain. With 2-op you're trying to do double duty, either the attack or sustain timbre needs to (severely) compromise for the sake of the other.

In this respect I think the parallel 2-op algorithm (FM-AM) is the most immediately and easily useful configuration of 4-op. The other configurations are useful too once you know your way around this stuff, but this algorithm is definitely my "default" choice.


So... once you get to 6-op there's less advantage to be had. Two stage sounds correspond very well to modelling natural or practical instrument sounds. Finding a use for that third stage becomes more subtle. One way that's somewhat intuitive is the "multitimbral" sound of tying an extra voice to velocity or one of the expression controls (mod wheel, breath, etc.) which lets you get a component that adjust across a range of use. However, that particular kind of thing is probably most useful for live performance instruments; when the entire playback is computer controlled, you can kinda substitute for this by just controlling parameters directly.

A 6-op instrument is diminishing returns in this respect. I think the DX7 etc. had some very good uses for live performance in this way, but for video games it wasn't that helpful. More voices was probably a better idea than more complex voices... which is more or less what happened...

tepples wrote:
Understanding what to do with the extra two operators, other than a pair of 2-op voices tuned a fifth (3:2) or fourth (4:3) apart to improve perceived polyphony, might help composers using OPN for the first time.

Well, with a lot of notes at once, the need to have precise timbres isn't so great, so in that respect it is sometimes better to have twice as many 2-op voices. Trying to do it with a 4-op in FM-AM mode is really limited though. I don't actually recommend that; trying to write good contrapuntal melody under restrictions like that is a weird hell. Sometimes it's good to have a pad instrument that just plays a 5th (the General Midi sound set had one or two of these included), but it's very situational.

On OPL3 you had the option of just using 18 independent 2-op channels, though even its 4-op mode went with a 6 x 4-op + 6 x 2-op configuration, i.e. 6 "complex sound" channels, and 6 "simple sound" channels, which is a reasonable trade, though was also frustrating you couldn't just get 9 x 4-op for a uniform synthesizer. (I'm currently doing research on ESFM which kinda scratches that itch.)

If you're comparing to OPN2, yes this is stuck with 6 x 4-op, though even there one of its channels could be split as 2 x 2-op. Then it also has a DAC, and the Genesis had a separate PSG to back that up with a few extra "simple sound" channels... so there's a lot of extra stuff on top of the 4-op voices here.

The more limited OPN had 3 x 4-op + PSG, as well. There's a lot of variations on this combination of 4-op with other simpler generators.


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 12:40 pm 
Online

Joined: Sun Sep 19, 2004 11:12 pm
Posts: 21505
Location: NE Indiana, USA (NTSC)
These are the OPN2 algorithms, where operator "a" is capable of feedback. (Sources: gendev; hacking-cult; GameDev)

Code:
I.    [a]->[b]->[c]->[d]->out

II.   [a]-+->[c]->[d]->out
          |
      [b]-'

III.  [a]------+->[d]->out
               |
      [b]->[c]-'

IV.   [c]------+->[d]->out
               |
      [a]->[b]-'

(IV is identical to III except for which side does feedback)

V.    [a]->[b]-+->out
               |
      [c]->[d]-'

VI.   [a]-+->[b]-+->out
          |      |
          +->[c]-+
          |      |
          `->[d]-'

VII.  [a]->[b]-+->out
               |
      [c]------+
               |
      [d]------'

VIII. [a]-+->out
          |
      [b]-+
          |
      [c]-+
          |
      [d]-'


The same article on GameDev.net lists no fewer than thirty-two 6-op algorithms.

By "FM-AM", do you mean algorithm V?

_________________
Pin Eight | Twitter | GitHub | Patreon


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 1:04 pm 
Offline
User avatar

Joined: Sun Jan 22, 2012 12:03 pm
Posts: 7526
Location: Canada
tepples wrote:
By "FM-AM", do you mean algorithm V?

Yes. (That was an OPL3 convention I guess.)


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 1:19 pm 
Offline
User avatar

Joined: Sun Jan 22, 2012 12:03 pm
Posts: 7526
Location: Canada
Regarding algorithms:

Two modulators into one carrier can do a lot of things that a modulator-carrier pair can do, timbrally, and so can a modulator-modulator-carrier chain. They're not exactly the same, but they can interchange somewhat. The biggest difference is just which things have their own envelope.

That FM-AM mode is a good default way of thinking though, two independent parts of a sound. Easiest to get my head around.

The other algorithms might be seen as a way to trade just a little bit of flexibility there to squeeze out another sound component. Like I'd start with FM-AM, and after getting close to a sound I like, then I'd consider if I really need that other carrier's envelope. If both carriers have the same frequency multiplier, then likely I don't... so at that point I could shift to another algorithm that gives me another independent component. Picking and choosing which things need an envelope, which things can blend in a chain etc.


The real problem with FM synthesis IMO is just that it's really hard to make good instruments. It takes a lot of patience and experience to find a good sound with this stuff. Even if you know how everything works and have seen a lot of good instruments made, it's still a painstaking process making a new one. At least, that's how I've felt about it for a long time.


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 1:39 pm 
Offline
Formerly WheelInventor
User avatar

Joined: Thu Apr 14, 2016 2:55 am
Posts: 2024
Location: Gothenburg, Sweden
Of those, i believe algorithm II. to be the most general purpose. Keep (a) for attack timbre, (b) for sustain timbre, (c) for voice and (d) for vibrato and pitch bend. Or switch around the tasks of a & b.

III. and IV. can serve similar purposes.

V. is great for chorus & a simple chorused organ (set one pair a far interval away), or duotone voices.

VI. is a little nebulous to me. Do you have individual input control or just output-side?

VII. for complex but sharp, clear attack and chorused sustain.

VIII. is straightfoward. Use it for additive/register synthesis or as an autochord.

I. is maybe good for evolving textures. I find op stacks like these immensely tedious to work with and make the most of. Not sure if i ever succeeded.

Not a complete list of course; just how i’d approach it.

Often, i leave operators unused.

_________________
http://www.frankengraphics.com - personal NES blog


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 3:31 pm 
Offline
User avatar

Joined: Sun Jan 22, 2012 12:03 pm
Posts: 7526
Location: Canada
FrankenGraphics wrote:
VI. is a little nebulous to me. Do you have individual input control or just output-side?

A is modulator to all 3, but then B C D all independently add to the output mix.

If you're asking whether you can control the modulation strength individually for each input, no, there's no interface for that. They all get the same input from A.

(Probably one of the more difficult to work with algorithms, IMO.)


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 4:13 pm 
Offline
Formerly WheelInventor
User avatar

Joined: Thu Apr 14, 2016 2:55 am
Posts: 2024
Location: Gothenburg, Sweden
Yeah that answers ny question. Thanks.

In that case i guess i’d try to use it for for three near-unison components, or components that are perfect octaves apart, or a combination, so that the modulator/carrier ratio remains the same or about the same or stays harmonic. Could be good for big leads.

_________________
http://www.frankengraphics.com - personal NES blog


Last edited by FrankenGraphics on Sat Mar 23, 2019 4:22 pm, edited 2 times in total.

Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 4:13 pm 
Offline
Site Admin
User avatar

Joined: Mon Sep 20, 2004 6:04 am
Posts: 3697
Location: Indianapolis
Drew Sebastino wrote:
Was it that 6 op FM ICs were too expensive, (just look at what these old synthesizers retailed for) or was it that they were never sold individually?


Probably the latter. We do know from Atari's internal memos (around 1984 or so) that Yamaha wouldn't let them put the YM2151 in consumer products, only arcade games. Presumably so it couldn't compete with the CX5M. Maybe they had a similar approach to the DX7, and it was more profitable to sell a whole instrument instead of a high-end chipset. With all the crazy arcade hardware there is, seems like somebody would have used it, if it was offered.


Top
 Profile  
 
PostPosted: Sat Mar 23, 2019 7:02 pm 
Offline

Joined: Thu Aug 20, 2015 3:09 am
Posts: 457
Well, pianos usually have up to 3 strings per key. 6-op FM lets you model three strings. So for an electric piano it makes a certain theoretical sense.

For games, not so much; price and polyphony are far more important.


Top
 Profile  
 
PostPosted: Mon Mar 25, 2019 1:04 am 
Offline
Formerly Espozo
User avatar

Joined: Mon Sep 15, 2014 4:35 pm
Posts: 3501
Location: Richmond, Virginia
rainwarrior wrote:
The real problem with FM synthesis IMO is just that it's really hard to make good instruments. It takes a lot of patience and experience to find a good sound with this stuff. Even if you know how everything works and have seen a lot of good instruments made, it's still a painstaking process making a new one. At least, that's how I've felt about it for a long time.

I think that's a pretty common sentiment. The few times I've dabbled with it in Deflemask were pretty much just trial and error, although I don't have a good enough understanding of how it works to really have a good opinion.

And about difficulty, it also appears to me that you can also create increasingly complex instruments using multiple channels by playing separate voices simultaneously (I think Comix Zone is great example of overlaying multiple voices; it's probably the fullest sounding thing on the Genesis: https://www.youtube.com/watch?v=o1QAKVfRFas) so there's potentially much more that you would have to worry about if you were to get the "best possible" sound, unless I'm mistaken.

Memblers wrote:
With all the crazy arcade hardware there is, seems like somebody would have used it, if it was offered.

Yeah, unless Yamaha were asking for a near infinite amount of money, you'd have figured that Sega or Namco or any other company known for ridiculously powerful (and thus also expensive) arcade hardware would have used it at some point. I'm sort of surprised that no one seemed to have designed their own FM synthesis solution, considering the number of custom PCM chips that were devised (even Irem made their own; that baffles me a bit). Although the OKI 6295 was also very popular. (unfortunately...) If production volume were to be an indicator, The YM2151 was probably just priced cheap enough that no one cared to create their own FM chip.


Top
 Profile  
 
PostPosted: Mon Mar 25, 2019 3:21 am 
Offline
User avatar

Joined: Wed Feb 13, 2008 9:10 am
Posts: 696
Location: Estonia, Rapla city (50 and 60Hz compatible :P)
No-one could create their own until 90's when patents ran out.

I have found algos 0, 1, 2 and 4 to be most useful when I go over the instruments I have created. First few allow for the most interesting sounds, 5, 6 and 7 are useful for simple sounds for the most part.
The higher the TL the mellower the sound, closer to zero and you'll be getting noises. This with MUL setting will create your tonality. If you don't want tonality to change as the instrument progresses give the modulating op(s) infinite sustain and slow release.
ADSR settings do not reset with new notes so slow instruments will lose their attacks as each ADSR envelope carries on exactly where old note/instrument left off.

_________________
http://www.tmeeco.eu


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 15 posts ] 

All times are UTC - 7 hours


Who is online

Users browsing this forum: No registered users and 3 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group