AFM conversion to DX

The Yamaha SY99 is a synthesiser combining frequency modulation synthesis (branded as Advanced FM) and sample-based synthesis (branded as Advanced Wave Memory 2) and the direct successor to Yamaha's SY77/TG77

Moderators: Derek, parametric, Saul

Monte
Member
Member
Posts: 84
Joined: Tue Oct 31, 2017 9:50 am
Japan

AFM conversion to DX

Unread post by Monte » Fri Nov 03, 2017 3:01 am

Hi,

I'd like to have the "Nitehwk" sound from the SY99 on the Montage.
Other interesting FM rhodes as well for that matter!

Conversion is not a possibility, right?

Are there any recreations whatsoever?

Would be great regardless if anyone could spare some cool FM rhodes sounds in DX format!
((i))
User avatar
Derek
Global Moderator
Global Moderator
Posts: 4249
Joined: Fri Dec 07, 2007 1:00 am
Contact:
Wales

Re: AFM conversion to DX

Unread post by Derek » Fri Nov 03, 2017 8:58 am

NiteHawks is indeed one of the signature AFM sounds in the SY.

Whether or not it could be converted depends on what features of the AFM engine it is using. I am speaking from the experience of building DX FM converters into my fs.factory and sy.factory librarians, so I spent rather a lot of time poring over the implemention of FM in the DX7 and AFM in the SY, less so on the FS1r as the FS1r has a built in ability to receive DX7 SYSEX (the only Yamaha synth with the native ability to do that, I believe), which I used for the conversion. Like all Yamaha synths that seem to have to go through a "quirks department" for something daft/limiting to be in the feature set before release, the self inflicted own goal on the FS1r is: yes it can receive DX7 SYSEX, but only single voice edit buffer dumps, not 32 voice Bulk Dump format. So fs.factory makes up the difference :) and does bulk dump conversion by opening a bulk dump file and sending single DX voices to an FS1r edit buffer and then reading it back in FS1r format. Why do I say all of that? Basically Yamaha did not care too much for backwards compatibility and when they tried in the FS1r they didn't do a complete job. ;) I am not sure how good their DX to Montage converter is, and I am also guessing they felt obliged to provide one given that the Kronos can DIRECTLY read DX7 data (i.e. no converter software needed). ;)

The Montage FM-X engine is (possibly a simplistic view) "half an FS1r" which is quite a different engine to AFM. Both FM-X and AFM have their roots in the DX7, but the FS1r and thus FM-X is more closely aligned and easier to do a conversion as it was designed from the outset with that in mind. The DX7 6 OP engine drops directly onto the FS1r engine in that all the parameters are identical (esp the operator rate scaling method), all the algorithms are recognisably there, etc.

It is a different problem in AFM as whilst it is a six op engine, it is radically different in many ways. Bizarrely, Yamaha did not capitalise on the DX7 heritage and provide a patch converter at the time, which sort of suggested that they thought it was tricky as well (or wanted people to buy new patch libraries). Several people did write converters of course, including myself (as when I came to the SY range in 2007 the DOS based existing tools were old and unsupported).

So what are the potential conversion issues from AFM to FM-X? One of the big ones is operator rate scaling. The DX7 had simple linear or logarithmic curves, positive or negative, either side of a breakpoint that you had little control over. You can set break point, Lin/log plus or minus and the end level for each slope. AFM has four break points that you can set the key and their levels. So it's possible to get an approximation of the DX7 if all the responses were set to linear, but if the responses were logarithmic and the sound depends on that log response, then forget it. Having said that, most of the DX7 conversions came through my FM converter with a good rate of success, but some were way off, and when you looked it was usually always the rate scaling that caused the deviation.

The next big one would be algorithm. DX FM has 32 algorithms, AFM has 45. So if NiteHwks is using one of the algorithms not present in the DX, that would be a problem unless you can find the same algorithm in FM-X

Other issues then are if an AFM patch is using multiple feedback loops. Conventional DX FM (and I believe the FS1r and thus FM-X) have a fixed feedback loop in each algorithm You can't change it, only the amount of feedback. AFM allows you to have three feedback loops and you can put them where you want in terms of source operator and destination operator. Then you can set the level of a modulator going into a carrier. In DX FM the level is fixed and the only level control you have is the Envelope. So in AFM you can make the modulation more subtle by controlling the input level.

AFM also has a "design your own algorithm mode" which is a "hidden feature" not accessible by the front panel editing. Unlikely anything is using it, but you never know.

Then you have looping envelopes (but I doubt a Rhodes sound would be using that). Parameter scaling is different. 0-63 on the DX and 0-127 on AFM, but that is a simple scaling calculation.

Then you have the filter, if being used, that is possibly quite different to the filter on the Montage, but there may be some mapping.

And finally the effects units and architectures. If you are relying on effects for the sound character then that can make a profound difference. It is one reason why when I migrated from an SY77 to an SY99 that I also had to get a TG77 to keep the SY77 sound, which may sound a bit daft if I have just traded up to an SY99. The SY77 and SY99 have the same AFM engine but very different effects units (and better/bigger wave ROM on the AWM side). The SY77 effects are quite "lo-fi" and grungy, but that is part of the synth's character. The SY99 is much cleaner and brighter (sometimes too clean), but as a result some signature SY77 patches sound completely wrong on the SY99 as they are too bright or lose some of the effects as there is sometimes no direct mapping between an SY77 effect and an SY99 effect!

So quite a long story to a simple question :) but it is not going to be a simple converting any SY AFM patch that is going to be making use of AFM features not present in FM-X. If I get five minutes (hah!) I might dig out Nitehwks and take a look and let you know. If it i feasible I may even have a go a programming it to get to know FM-X.....
Regards
Derek Cook

http://www.carregddu.co.uk
http://www.echoes-music.co.uk
http://www.xfactory-librarians.co.uk
http://www.ex5tech.co.uk
User avatar
Derek
Global Moderator
Global Moderator
Posts: 4249
Joined: Fri Dec 07, 2007 1:00 am
Contact:
Wales

Re: AFM conversion to DX

Unread post by Derek » Fri Nov 03, 2017 9:04 am

PS. Easy Sounds has just released an FM-X specific library for Montage (no AWM anywhere in sight), so that may be worth checking out to see what he has done on the EP front. Peter has been active in FM programming for a long time (and Yamaha synths in general).

http://easysounds.eshop.t-online.de/epa ... ts/014110D
Regards
Derek Cook

http://www.carregddu.co.uk
http://www.echoes-music.co.uk
http://www.xfactory-librarians.co.uk
http://www.ex5tech.co.uk
Monte
Member
Member
Posts: 84
Joined: Tue Oct 31, 2017 9:50 am
Japan

Re: AFM conversion to DX

Unread post by Monte » Fri Nov 03, 2017 9:16 am

I know, I already have it. I buy everything they make! but unfortunately no, that was not the purpose of that library.
((i))
User avatar
Saul
Site Admin
Site Admin
Posts: 10383
Joined: Sun Feb 01, 2004 1:00 am
Contact:
Ireland

Re: AFM conversion to DX

Unread post by Saul » Fri Nov 03, 2017 12:09 pm

Couldn't Yamaha have just stuck a broader based FM engine inside Montage? I know nothing about it so I assume it would either be incredibly difficult to do or for some marketing reason they decided not to but surely they understand that many people are eager to import older FM sounds and find it incredibly frustrating not being able to do so?
Saul
Site Admin
Subscribe To Our YouTube Channel!
Review Yamahamusicians.com on Trustpilot
Sign up to the Newsletter
Follow us on Twitter @YamahaMusicians
Follow Us On Facebook
Dowina Acoustic Guitars: Dowina Cabernet DC
Dowina Chardonnay GAC DS - Dowina Rustica GAC - Dowina Marus - Dowina Puella

Seagull S6 Original/Vintage VE2000GG Gordon Giltrap/Ashbury AG-160/Roland FA-07/Pianoteq 6 Pro/Amplitube 3.0/Apple iMac 27"/Logic Pro X/Cubase AI 9/Absynth 5/Ableton Live 9/KRK Rokit RP5 G3 Monitors
Monte
Member
Member
Posts: 84
Joined: Tue Oct 31, 2017 9:50 am
Japan

Re: AFM conversion to DX

Unread post by Monte » Fri Nov 03, 2017 12:52 pm

I only saw your second comment earlier regarding FM-Xperience and totally missed that one. Thank you very much for the info. I have to agree. I hate the fact that I might have to resort to getting a SY99 just for some more FM sounds! but it could happen.. when the time is right and the moon is aligned with my window.
I'll add your information to the SY99 / Montage comparison?
Derek wrote:
Fri Nov 03, 2017 8:58 am
NiteHawks is indeed one of the signature AFM sounds in the SY.

Whether or not it could be converted depends on what features of the AFM engine it is using. I am speaking from the experience of building DX FM converters into my fs.factory and sy.factory librarians, so I spent rather a lot of time poring over the implemention of FM in the DX7 and AFM in the SY, less so on the FS1r as the FS1r has a built in ability to receive DX7 SYSEX (the only Yamaha synth with the native ability to do that, I believe), which I used for the conversion. Like all Yamaha synths that seem to have to go through a "quirks department" for something daft/limiting to be in the feature set before release, the self inflicted own goal on the FS1r is: yes it can receive DX7 SYSEX, but only single voice edit buffer dumps, not 32 voice Bulk Dump format. So fs.factory makes up the difference :) and does bulk dump conversion by opening a bulk dump file and sending single DX voices to an FS1r edit buffer and then reading it back in FS1r format. Why do I say all of that? Basically Yamaha did not care too much for backwards compatibility and when they tried in the FS1r they didn't do a complete job. ;) I am not sure how good their DX to Montage converter is, and I am also guessing they felt obliged to provide one given that the Kronos can DIRECTLY read DX7 data (i.e. no converter software needed). ;)

The Montage FM-X engine is (possibly a simplistic view) "half an FS1r" which is quite a different engine to AFM. Both FM-X and AFM have their roots in the DX7, but the FS1r and thus FM-X is more closely aligned and easier to do a conversion as it was designed from the outset with that in mind. The DX7 6 OP engine drops directly onto the FS1r engine in that all the parameters are identical (esp the operator rate scaling method), all the algorithms are recognisably there, etc.

It is a different problem in AFM as whilst it is a six op engine, it is radically different in many ways. Bizarrely, Yamaha did not capitalise on the DX7 heritage and provide a patch converter at the time, which sort of suggested that they thought it was tricky as well (or wanted people to buy new patch libraries). Several people did write converters of course, including myself (as when I came to the SY range in 2007 the DOS based existing tools were old and unsupported).

So what are the potential conversion issues from AFM to FM-X? One of the big ones is operator rate scaling. The DX7 had simple linear or logarithmic curves, positive or negative, either side of a breakpoint that you had little control over. You can set break point, Lin/log plus or minus and the end level for each slope. AFM has four break points that you can set the key and their levels. So it's possible to get an approximation of the DX7 if all the responses were set to linear, but if the responses were logarithmic and the sound depends on that log response, then forget it. Having said that, most of the DX7 conversions came through my FM converter with a good rate of success, but some were way off, and when you looked it was usually always the rate scaling that caused the deviation.

The next big one would be algorithm. DX FM has 32 algorithms, AFM has 45. So if NiteHwks is using one of the algorithms not present in the DX, that would be a problem unless you can find the same algorithm in FM-X

Other issues then are if an AFM patch is using multiple feedback loops. Conventional DX FM (and I believe the FS1r and thus FM-X) have a fixed feedback loop in each algorithm You can't change it, only the amount of feedback. AFM allows you to have three feedback loops and you can put them where you want in terms of source operator and destination operator. Then you can set the level of a modulator going into a carrier. In DX FM the level is fixed and the only level control you have is the Envelope. So in AFM you can make the modulation more subtle by controlling the input level.

AFM also has a "design your own algorithm mode" which is a "hidden feature" not accessible by the front panel editing. Unlikely anything is using it, but you never know.

Then you have looping envelopes (but I doubt a Rhodes sound would be using that). Parameter scaling is different. 0-63 on the DX and 0-127 on AFM, but that is a simple scaling calculation.

Then you have the filter, if being used, that is possibly quite different to the filter on the Montage, but there may be some mapping.

And finally the effects units and architectures. If you are relying on effects for the sound character then that can make a profound difference. It is one reason why when I migrated from an SY77 to an SY99 that I also had to get a TG77 to keep the SY77 sound, which may sound a bit daft if I have just traded up to an SY99. The SY77 and SY99 have the same AFM engine but very different effects units (and better/bigger wave ROM on the AWM side). The SY77 effects are quite "lo-fi" and grungy, but that is part of the synth's character. The SY99 is much cleaner and brighter (sometimes too clean), but as a result some signature SY77 patches sound completely wrong on the SY99 as they are too bright or lose some of the effects as there is sometimes no direct mapping between an SY77 effect and an SY99 effect!

So quite a long story to a simple question :) but it is not going to be a simple converting any SY AFM patch that is going to be making use of AFM features not present in FM-X. If I get five minutes (hah!) I might dig out Nitehwks and take a look and let you know. If it i feasible I may even have a go a programming it to get to know FM-X.....
((i))
User avatar
Derek
Global Moderator
Global Moderator
Posts: 4249
Joined: Fri Dec 07, 2007 1:00 am
Contact:
Wales

Re: AFM conversion to DX

Unread post by Derek » Fri Nov 03, 2017 1:33 pm

Saul wrote:
Fri Nov 03, 2017 12:09 pm
Couldn't Yamaha have just stuck a broader based FM engine inside Montage? I know nothing about it so I assume it would either be incredibly difficult to do or for some marketing reason they decided not to but surely they understand that many people are eager to import older FM sounds and find it incredibly frustrating not being able to do so?
I guess they had to pick one. From the three choices, Vanilla DX FM, AFM and FS1R FM, I think they probably made the right choice in the terms of picking a more advanced engine and then the case of the best backward compatibility path back to the DX range. But it still puzzles me why they only took half of the FS1r? Arguably the format sequencing side was probably one of the least used features, but lack of a decent editor did not help there (and the FS1r user interface is like trying to paint the Cistene chapel through a letter box).

It is a shame though that they did not design FM-X to also be compatible with AFM as well, as I think AFM was the better implementation in terms of additional features that make the SY range special.
Regards
Derek Cook

http://www.carregddu.co.uk
http://www.echoes-music.co.uk
http://www.xfactory-librarians.co.uk
http://www.ex5tech.co.uk
Monte
Member
Member
Posts: 84
Joined: Tue Oct 31, 2017 9:50 am
Japan

Re: AFM conversion to DX

Unread post by Monte » Fri Nov 03, 2017 11:02 pm

Dammit. Now I want to get a SY99.
On top of the EX5 maybe?
((i))
User avatar
MdMax
Member
Member
Posts: 14
Joined: Sat Oct 21, 2017 12:57 pm
France

Re: AFM conversion to DX

Unread post by MdMax » Fri Nov 03, 2017 11:03 pm

Hi all !
Derek wrote:
Fri Nov 03, 2017 1:33 pm
But it still puzzles me why they only took half of the FS1r?
I don't have the answer but this is my guess: the Montage has real hardware tone generators, it's not a single CPU emulating soft synths. I think the FM-X engine has dedicated circuits and transistors for each FM operator or element and note, and it has a real polyphony of 128 notes. This is a huge number compared to the 16/32 notes of the FS1r or the 16 notes of previous FM or AFM synthesizers.

So you have less (mostly unused ?) FS1r features, some new features, and a lot more freedom before reaching you polyphony maximum, even with multi-part layers, very long envelopes or when never releasing your sustain pedal. :wink: Polyphony may be a very important marketing advantage today.
Yamaha SY77
User avatar
Derek
Global Moderator
Global Moderator
Posts: 4249
Joined: Fri Dec 07, 2007 1:00 am
Contact:
Wales

Re: AFM conversion to DX

Unread post by Derek » Fri Nov 03, 2017 11:19 pm

Hi. I'm not convinced that would be the case. With what you can squeeze on a silicon die re transistor count now compared to twenty years ago, you could have all the features and the polyphony. One thing is for sure, we will never know why Yamaha make great tech and then drop it. FDSP for example.
Regards
Derek Cook

http://www.carregddu.co.uk
http://www.echoes-music.co.uk
http://www.xfactory-librarians.co.uk
http://www.ex5tech.co.uk
DrSynth
Member
Member
Posts: 55
Joined: Sat Jul 30, 2011 7:58 pm
United States of America

Re: AFM conversion to DX

Unread post by DrSynth » Sun Nov 05, 2017 6:49 am

Nice discussion of the architecture differences. I wasn't involved at the time so I can't give any insights on the how's and why's of Yamaha choosing an FS1r based implementation for FM-X. Obviously a lot of options were available from their legacy implementations -- 4 Ops or 6 Ops or 8 Ops; Sine wave or non-Sine wave Ops; 4 Stage or 6 Stage looping envelopes; single or multiple feedback loops; configurable algorithms etc. If I had been asked for input on a new FM implementation I would have recommended new/enhanced implementations of the AFM architecture. But FM-X is still pretty cool, and the timbral behavoir of the Operator "Res" waveforms is unique, especially with Motion Sequences

In the context of AFM -> FM-X there are potentially so many extreme variances in algorithm, Operator Waveform and Feedback loops in AFM that I would say the likelyhood of porting or converting a sound from the SY99/77 to Montage is negligible. Especially if the AFM patch uses any Operator phase shift or 'concentric' feedback loops with zero frequency Operators. BTW -- To clarify, I mean a 'native" SY77/99 patch, not a converted DX7 patch that was enhanced in the AFM engine -- those could probably be re-worked to get pretty close in FM-X.

A comment about RCM. IIRC the genesis of the implementation was an attempt at FM based re-synthesis. It wasn't supposed to use "harmonically complete" i.e. acoustic instrument samples as modulators. Imagine if you took a piano sample, applied some mathematical transforms to it to create an AWM waveform. Then use that wave as a modulator into the AFM engine with a specfic algorithm created in free algorithm edit, with precise Operator Ratio, phase and feedback settings that then would dynamically recreate the original sample -- that's where they were trying to take it. Think of the "proper" RCM modulator being a reduced overtone waveform with a complex overtone evolution over time.

I've had the Montage and SY99 side by side for about 6 months. In a broad stroke comparison, they are sort of a yin and yang to one another. There's really cool and unique things each one does that the other can't touch. FM-X is about building reasonably complex harmonic structures with the ability for extremely complex realtime control. The AFM is about building extremely unique complex harmonic structures with reasonably versatile control.

Manny
FM Rig of DOOM!! Yamaha Montage 7, SY99's x6 , TG77's x4, Reface DX,FS1r, DX200, DX7II FD E!, DX7II FD, TX802
Yamaha VL1v2 & VL70m
Kurzweil K2000, Korg Wavestation EX, Kawai K1, Sequential Pro One, Arp Odyssey
What's a Roland ...?
YouTube FM Tutorials: https://www.youtube.com/channel/UCXY8GT ... R0TSw/feed
SoundCloud FM Examples: https://soundcloud.com/manny-fernandez-4856421
User avatar
Derek
Global Moderator
Global Moderator
Posts: 4249
Joined: Fri Dec 07, 2007 1:00 am
Contact:
Wales

Re: AFM conversion to DX

Unread post by Derek » Sun Nov 05, 2017 9:26 am

Hi, Manny.

Yes that was my view as well, but you've identified a few other details that would not be transportable. And you conclude with a good way of looking at the differences.
Regards
Derek Cook

http://www.carregddu.co.uk
http://www.echoes-music.co.uk
http://www.xfactory-librarians.co.uk
http://www.ex5tech.co.uk
DrSynth
Member
Member
Posts: 55
Joined: Sat Jul 30, 2011 7:58 pm
United States of America

Re: AFM conversion to DX

Unread post by DrSynth » Mon Nov 06, 2017 7:20 am

Monte,

I had some time this afternoon so reprogrammed "NiteHwk" for the Montage -- PM me and I'll send you the file. Do you know how to use Montage Connect?

Manny
FM Rig of DOOM!! Yamaha Montage 7, SY99's x6 , TG77's x4, Reface DX,FS1r, DX200, DX7II FD E!, DX7II FD, TX802
Yamaha VL1v2 & VL70m
Kurzweil K2000, Korg Wavestation EX, Kawai K1, Sequential Pro One, Arp Odyssey
What's a Roland ...?
YouTube FM Tutorials: https://www.youtube.com/channel/UCXY8GT ... R0TSw/feed
SoundCloud FM Examples: https://soundcloud.com/manny-fernandez-4856421
Monte
Member
Member
Posts: 84
Joined: Tue Oct 31, 2017 9:50 am
Japan

Re: AFM conversion to DX

Unread post by Monte » Mon Nov 06, 2017 8:35 am

DrSynth wrote:
Mon Nov 06, 2017 7:20 am
Monte,

I had some time this afternoon so reprogrammed "NiteHwk" for the Montage -- PM me and I'll send you the file. Do you know how to use Montage Connect?

Manny
Nice one DrSynth!
for sure.
((i))
DrSynth
Member
Member
Posts: 55
Joined: Sat Jul 30, 2011 7:58 pm
United States of America

Re: AFM conversion to DX

Unread post by DrSynth » Mon Nov 06, 2017 6:27 pm

Monte,

Here's some NiteHawk ports. One of them is layered with TX816 Strings. The one named "NiteHawk2E" is enhanced using the additional FM-X operators.

If one were to be extremely picky, they're not 'exactly identical', but you'd have to really listen extremely close to notice any difference. While the SY99 original does only use sine wave Operators, Element 2 does use an algorithm unique to the SY99/77, so these have a barely noticable difference in the lower end range when played at higher velocities

BTW, you may have to make some edits to taste of the effect wet/dry and the touch response for your playing style. Personally the original plays a little 'squishy' to me because of the chorus depth.

Enjoy!

Manny
NiteHawk.zip
(28.75 KiB) Downloaded 136 times
FM Rig of DOOM!! Yamaha Montage 7, SY99's x6 , TG77's x4, Reface DX,FS1r, DX200, DX7II FD E!, DX7II FD, TX802
Yamaha VL1v2 & VL70m
Kurzweil K2000, Korg Wavestation EX, Kawai K1, Sequential Pro One, Arp Odyssey
What's a Roland ...?
YouTube FM Tutorials: https://www.youtube.com/channel/UCXY8GT ... R0TSw/feed
SoundCloud FM Examples: https://soundcloud.com/manny-fernandez-4856421
Post Reply

Return to “Yamaha SY99 Forum”