I just got through most of this series. I thought it was very well done and educational. It went through in detail what the component sounds of a piano are, and how Manny implemented each one of them.

But it left me with a question:

How did Manny figure out what the components of the piano sound were? How did he know which sounds he had to implement (string vibrations, string overtones, hammer sounds, etc)? How did he take a pre-existing sound and do that analysis?

How did Manny figure out what the components of the piano sound were? How did he know which sounds he had to implement (string vibrations, string overtones, hammer sounds, etc)? How did he take a pre-existing sound and do that analysis?

A quick google search on 'Piano acoustics', as well as knowing how a piano works & what components it is comprised of, etc., and you could piece together what various sounds need to be layered to sound like a piano.

Hey Michael,

I'm glad you found the article and video series interesting and useful.

To your question, my approach to doing this was a blend of a couple of things - First, as Darryl notes, is having studied acoustics and knowing how a piano is built and works. Then I assembled the needed harmonic components from the ground up, as described the the articles and videos.

As to how I specifically made those components in FM, choosing which Ratios, Modulator depths, algorithms to use etc., in my case that comes from the 'ear experience' of programming FM for over 35 years and is kind of a long story. Short version - upon getting my original DX7 in 1983 while still in college, I deep-dove into John Chowning's original academic articles as well as his book with Dave Bristow, "FM Theory and Applications" to understand the math and phase/amplitude behaviors of Modulataion Index. And, I programmed constantly.

Back in those original DX7 days ready access to FFT's and spectral anaysis of acoustic sounds was quite limited compared to what's available today. So basically I'd listen closely to my acoustic piano (or a recording of some other instrument/sound) and tweaking my DX7 right next to it, matching the timbres and transients by ear until it sounded right to me. Similar to people with "Perfect Pitch" (which I don't have) I am able to hear timbres extremely well. Combined with the years of experience, I've built up a large mental library of how FM behaves.

I'm not exaggerating that before culminating in this Montage resynthesis example I've probably well over 1000 variants of FM synthesized Piano simulations beginning with the original DX7, through the TX816, DX7II, SY77/99, and FS1r. And while it works for me, doing it all by ear isn't for everyone.

Thor Zollinger has put together a contemporary guide to acoustic sound emulations at javelinart.com that is a good place to start with a pragmatic, methodical approach to FM using spectral analysis. Check it out !

Manny

Thanks, Manny.

I am currently going through Thor Zollinger's guide for the second time, learning more as I go. It sounds like I have a large learning curve to climb. Fair enough, and I enjoy that.

Maybe someday I will be able to play my Roli Rise 25 and sound at least somewhat like Jerry Garcia… 😀

(Any tips you have on FM-X synthesis of electric guitar sounds will be gratefully appreciated.)

Probably a MIDI guitar is a better controller if you really aspire to sound like any given guitarist. You can play chords using keyboard keys that follow guitar rules and simulate what more-or-less naturally happens with a guitar - but the MIDI guitar controller will make string-to-string timing (offsets) happen automatically without as much work.

That said - I'm sure you've seen this video - but it's worth being redundant. https://www.sweetwater.com/insync/how-to-play-guitar-parts-on-keyboards/ It covers a good collection of things one would need to develop as a keyboard player to emulate the guitar sound.

Although focused on the acoustic sound - there's not much difference to the electric guitar for most things except for the core sound. But the "physics" of the instrument and what you need to do to make a good acoustic sound will apply more-or-less equally to the electric guitar.

That said - I'm sure you've seen this video - but it's worth being redundant. https://www.sweetwater.com/insync/how-to-play-guitar-parts-on-keyboards/ It covers a good collection of things one would need to develop as a keyboard player to emulate the guitar sound.

i may have seen something similar a long time ago. Thanks for the link.

The key to successful & convincing emulative synthesis is proper idiomatic playing.

Alternate i.e. non-keyboard controllers can help really help a lot, and very often a 'weaker' sound played well to the articulations and dynamics of the sound/instrument you're simulating will sound more real than a 'better' sound played out of idiom.

Even using a keyboard, effective use of Mod Wheels, Aftertouch, Foot Pedals and Breath Controller will make a huge difference.

I checked out the first video and looking at how multiple PARTs are used to create the pieces that make up the sound, it peaked my interest for several reasons, so I will definitely be diving into the entire series at some point soon...Great stuff! 😉

However, for right now it got me thinking more about an FM-X sound that I am struggling with, whereby I am kinda trying to do the opposite and use effects to achieve the sound by using just 1 single PART instead of multiple PARTs (mainly because PART realestate is limited on the Montage and I want to be able to play more than just a couple of 4 PART sounds via KBD CTRL in a Performance). So I was hoping you might have some advice and a few things to try..!?

Here's my issue. I have an awesome 'Orchestral Strings' sound on my DX7II-FD synth that I exported and imported into the Montage. Since on the DX it has 'Unison' enabled & detuned, it converted into the Montage as a 4 PART performance with each part detuned differently(-5,-2.5,+2.5,+5) to get the same Unison type sound.

To try & emulate this using only 1 part, I added two "Pitch Change" effects to the InsA & B on that PART (as per Jason's suggestion in a previous thread, which worked really well for a different Unison based sound that I had also converted from the DX), then I modified the 2 Pitches in each effect, thus spreading out the pitches 4 ways, similar to how the 4 PART converted Performance did. Actually 5 ways because the original sound of that PART is there & not detuned, which I think is what is causing my issue. (Note that the Dry/Wet Balance on both effects is set to most wet value "D<W63")
... This single PART sound is fairly close to the 4 PART, however there is a subtle 'wave' in the original PART sound (prior to effects), that is not exaggerated/noticable on the DX7 using Unison, nor on the Montage's 4 PART version; but using the single PART with the two effects, the wave sound is greatly exaggerated and I think it's because of the main (5th) sound is there & is not detuned. If there were a way to mute the main sound and only have the Pitch Change sounds, I think there would be no noticable wave sound and it would be the same as the 4 PART version, but that is not an option.

So, working with what I've got, setting the Initial Delay 1 & 2, as well as the Feedback Level 1 & 2 on both Insert effects, helps to mask the wave a fair bit, but not fully.
Any ideas on how I could possibly minimize this wave sound even more (or completely) in my Orchestral Strings PART?

One thought that just came to me is maybe I could try using 2 PARTs, tuning PART 1 to -2.5 and PART 2 to +2.5, then add the Pitch Change Insert effect on each, detuning PART 1 using negative values and PART 2 with + values, thereby eliminating the main 'untuned' middle sound..!?

I’ve been experimenting some more with the examples in Thor’s excellent guide to FM synthesis. I set up a simple situation on my Montage:

Algorithm 71
Op 3, Ratio 5, Modulator, Sine, Level 53
Op 4, Ratio 1, Carrier, Sine, Level 99.
(All other operators set to Level 0.)

Page 16 of Thor’s guide says I should see peaks at 1, 6, and 11. I do, but I also see peaks at 4 and 9. Why I am seeing those extra peaks? (I am looking at them in Visual Analyzer.)

Another question: given that I have 8 operators per Montage Part and 8 keyboard-controlled Parts per Performance, it seems to me that I can have 64 carriers under keyboard control at once. So can I just set each peak I want with an individual carrier and not bother with modulators? (I can have 128 carriers if I want by doing Cubase tricks.) Does this seem like a promising approach?

Darryl, re:

So, working with what I've got, setting the Initial Delay 1 & 2, as well as the Feedback Level 1 & 2 on both Insert effects, helps to mask the wave a fair bit, but not fully.
Any ideas on how I could possibly minimize this wave sound even more (or completely) in my Orchestral Strings PART?

The artifact you hear is a byproduct of any static, delay based effect, which would include the Ptich Change in the Montage. The first thing I would try is do not use feedback at all, and make sure all the right and left delay times are different in th two effect instances. In general I set the less detuned portion to a longer delay than the greater detuned one, and play around in the range of 8-20 msec, though sometimes values of up to 30-40 msec will alsow be usefull.

The second thing I'd try is use one Pitch Change as the first effect, then a Chorus as the second effect -- the fact the Chorus isn't static, it won't create the 'tone' thing you mentioned. You may also want to try 2 Chorus effects. Play around !

PS Any static delay based effect with feedback becomes a crude 'resonator' that will emphasize a particular 'tone' in a sound

Michael, re:

Page 16 of Thor’s guide says I should see peaks at 1, 6, and 11. I do, but I also see peaks at 4 and 9. Why I am seeing those extra peaks? (I am looking at them in Visual Analyzer.)

This is due to the negative harmonics created with "through zero' FM, which means the harmonics you create are both the carrier plus (multiples) of the modulator -- that's why you get 1 (carrier) then 6 (1+5) and 11 (1+5+5) etc... and the carrier MINUS the modulator 1-5 = -4, 1-5-5 = -9 etc. The 'negative' is just a math thing that means the phase is inverted, it sounds the same... I go over this here: Manny's FM-Xpert #3 "It's Just a Phase, Man,,,". Also check out the video linked to that article as well: FM-Xpert Video 3

The negative harmonics are why you can make a square wave two ways in FM. The carrier = 1.00, modulator = 2.00, as well as carrier =1.00 and modulator = 4.00 They will sound slightly different with the same modulator level, but if you decrease that in the 1 to 4 version they'll get very close.

re:

Another question: given that I have 8 operators per Montage Part and 8 keyboard-controlled Parts per Performance, it seems to me that I can have 64 carriers under keyboard control at once. So can I just set each peak I want with an individual carrier and not bother with modulators? (I can have 128 carriers if I want by doing Cubase tricks.) Does this seem like a promising approach?

You could to this, that's called additive synthesis, and is very very time consuming, However, that concept is useful for certain circumstances, and that's why many of the FM-X algorithms are the "Classic" DX 6 Operator algorithms with the extrx two Ops 1 & 2 as separate Carriers

The artifact you hear is a byproduct of any static, delay based effect, which would include the Ptich Change in the Montage. The first thing I would try is do not use feedback at all, and make sure all the right and left delay times are different in th two effect instances. In general I set the less detuned portion to a longer delay than the greater detuned one, and play around in the range of 8-20 msec, though sometimes values of up to 30-40 msec will alsow be usefull.

I have tried no feedback, but that actually makes it more noticeable.
I will definitely try setting the less detuned portion to a longer delay as you suggest, because I currently have it the other way around, whereby currently it is the greater detuned ones that have the longer delay times ... Thanks!
I can't remember off hand what the exact numbers are, but after playing around with various delays, the best I could find to help mask the issue is 0 & 19 on the less detuned portion, and 38 & 57 on the greater detuned ones. These 4 settings are equally notched delay times from each other.

The second thing I'd try is use one Pitch Change as the first effect, then a Chorus as the second effect -- the fact the Chorus isn't static, it won't create the 'tone' thing you mentioned. You may also want to try 2 Chorus effects. Play around !
PS Any static delay based effect with feedback becomes a crude 'resonator' that will emphasize a particular 'tone' in a sound

I'll give that a try as well, thanks!

Now for my update on what I found last night after a deeper dive. I fired up the DX7II and compared it to the Montage 4 PART converted version. The DX was quite a bit smoother & more realistic (as orchestral strings) than the Montage 4 PART converted version. I figured out why quite quickly...

The DX7 has two modes for LFO waves. Single and Multi. When Multi is used in conjunction with Unison (I have this sound set to detune of 4), it basically has 4 voices with LFO waves at different times/rates!?, which makes the strings sound like 'Huge' & 'realistic' orchestral strings. So basically there are two things that DX7 has that the Montage does not, which is (A) Unison and (B) LFO 'Multi' mode waves. After scratching my head, I dove into the Montage and dialed back the pitch detune a bit on the 4 PARTs because the conversion spread them too far (-5,-2.5,+2.5,+5). I set them to (-3.2,-1.6,+1.6,+3.2) and it was right on par with the DX7 Unison set to 4. But the Montage still had more of the wave sound and I knew it had to do with the LFO 'Multi' setting, because as soon as I set the DX7 to 'Single' LFO wave, it had that wave sound pronounced the same as on the Montage. So I set the Axis on the 2nd LFO differently on all 4 PARTs (0,90,180,270)...then I changed the 'per element' LFO settings from the main one of 8 and each element set at 1, to the main one down to 4, but each element set at 2. This made the Montage sound nearly identical to the DX7 'Multi' wave mode.

Next, I tried to tackle being able to come closer via using just 1 PART with the 2 'Pitch Change' Insertion effects. I first dialed down the detuning on the effects to be closer to the DX7 Unison 4 amount. Then I modified the 'per element' LFO settings setting the main one to 3 (4 was too much) and each element set at 2. This helped to mask that wave sound better. Not perfect, but better and reasonably close.

I'll try reversing the longer/shorter delay times on the effects, as well as try 1 or both using Chorus...

The ultimate best scenario would be if Yamaha were to implement my idea that I put up on Ideascale to add Unison functionality as it is on the DX7II, as with the LFO axis and per element changes, it would then be identical to the DX7 sound. (see my ideascale idea listed below)

'Montage-MODX UNISON Layering Functionality'
https://yamahasynth.ideascale.com/a/dtd/Montage-MODX-UNISON-Layering-Functionality/214572-45978

Realistically, I am not going to bank on Unison being added to the Montage and so I will get it as close as I can with what I currently have to work with, which is great ... just not 'super-awesome' like the DX7's sound is. Most people would likely think the Montage version I currently have is awesome, unless they heard the DX7's version right after. It is unfortunate though that Yamaha didn't include both 'Unison' and 'Multi' LFO wave mode in the Montage's FM-X engine, because it is the combination of both of those features on the DX that make it sound so huge, smooth, realistic and almost identical to actual movie score orchestral strings ... almost all of my best sounds on the DX7II use both 'Unison' and 'Multi' LFO wave mode!

On the plus side, the Montage allows me to use some great reverb & possibly other effects to mask the issue a bit, as well as stereoize, EQ and enhance the DX7 sounds!

Realistically, I am not going to bank on Unison being added to the Montage and get it as close as I can with what I currently to work with. It is unforuntate though that Yamaha didn't include both 'Unison' and 'Multi' LFO wave mode in the Montage's FM-X engine, because it is the combination of both of those features on the DX that make it sound so huge, smooth and almost identical to actual movie score orchestral strings ... almost all of my best sounds on the DX7II use both 'Unison' and 'Multi' LFO wave mode!

Looking back to the original DX7 (1983), the LFO, there was only one LFO applied to all Notes played, it was applied exactly the same. One movement fit all...
In the DX7IIFD (1986), there are 16 LFOs, one for each voice, in other words, each Note-On will have its own phase. And even though all still have the same settings, they can operate (begin) independently when “Multi” is selected. If “Single” was selected the LFO will behave as the original DX7.
“Key On Reset” is the parameter in the current generation FM-X engine.

The Unison modes found in the DX7IIFD (second generation units) was an attempt to thicken the sound, (make it phatter) - those were the days when to make certain FM sounds richer you simply linked another module of the TX816, and detuned each module... you didn’t have any effects in synthesizers, yet! ...in a single unit like the DX7IIFD Key Mode “Unison” folded the tone engine on itself, halving the polyphony with each fold. You had Union Poly where, folded twice, the polyphony was reduced to 4 thicker notes, Unison Mono, naturally was one very fat note.

One way that you can accomplish this sound is by copying your current FM-X sound to an adjacent Part. Select the Part: [SHIFT] + [EDIT] to COPY.
If you do nothing to it, you will have basically succeeded in making the sound a little more than a dB or two louder. However, if you Delay one of them, ever so slightly, the “sonic magic” starts to happen... as you’ve started to discover.

Pitch Change, (and Detune) has what you need but maybe overkill... (with certain drawbacks, as noted)

May I recommend using the “Symphonic” Insertion Effect Type (It is found among the Chorus Category). This rather unique Yamaha Chorus (from back in the SPX90 days) has the ability to remove all movement — you can actually set the LFO SPEED = 0.00Hz, you also have access to straight DELAY OFFSET (0.0ms~50.0ms. If you don’t want to “spend” a Part to duplicate the Tone, use “Symphonic” with Dry = Wet (D=W) this way you hear both the original FM-X and the delayed clone.

Pick a fixed “Delay Offset” (more than 30ms you’ll begin to hear two distinct events when you strike a key... you just want to thicken the sound, so stay below where the two sounds are separately distinguishable... less than 30ms. Then to maximize that ‘tubular’ type sound quality that Unison Poly was so popular for.... increment the LFO SPEED... one click from 0.00Hz to 0.04Hz is profound. It is the introduction of this very, very small movement that is the serious ear-candy... Experiment....
Set the LFO DEPTH to taste (0 Depth, means zero effect).

I agree, do not hold out for Unison mode coming back, it was necessary back then, due to less options... basically it’s a workaround that’s no longer necessary. And “Multi” concerning the LFO is implement by allowing each Note-On to either reset the LFO wave or when Off it just runs free and additional notes join the movement.

Hope that helps. There is no one “movie score orchestral string” sound... for every movie made, there is a slightly different ‘orchestral string sound’... once you have the FM-X tone, don’t limit yourself to just that technology. By adding a bit of AWM2 to your Performance you may make a new standard for your own “movie score orchestral strings”. And if you are so inclined, share it on Soundmondo!

I’ll post some examples.... soon...

In the DX7IIFD (1986), there are 16 LFOs, one for each voice, in other words, each Note-On will have its own phase. And even though all still have the same settings, they can operate (begin) independently when “Multi” is selected. If “Single” was selected the LFO will behave as the original DX7.
“Key On Reset” is the parameter in the current generation FM-X engine.

Wow! Lots of great history and information here, thanks BM!! 😉
I will start with setting the “Key On Reset” parameter.

The Unison modes found in the DX7IIFD ...in a single unit like the DX7IIFD Key Mode “Unison” folded the tone engine on itself, halving the polyphony with each fold. You had Union Poly where, folded twice, the polyphony was reduced to 4 thicker notes, Unison Mono, naturally was one very fat note.

One way that you can accomplish this sound is by copying your current FM-X sound to an adjacent Part. Select the Part: [SHIFT] + [EDIT] to COPY.
If you do nothing to it, you will have basically succeeded in making the sound a little more than a dB or two louder. However, if you Delay one of them, ever so slightly, the “sonic magic” starts to happen... as you’ve started to discover.

I was planning to use two PARTs anyway to add some stereo separation, so this will work.

Pitch Change, (and Detune) has what you need but maybe overkill... (with certain drawbacks, as noted)

May I recommend using the “Symphonic” Insertion Effect Type (It is found among the Chorus Category). This rather unique Yamaha Chorus (from back in the SPX90 days) has the ability to remove all movement — you can actually set the LFO SPEED = 0.00Hz, you also have access to straight DELAY OFFSET (0.0ms~50.0ms. If you don’t want to “spend” a Part to duplicate the Tone, use “Symphonic” with Dry = Wet (D=W) this way you hear both the original FM-X and the delayed clone.

Pick a fixed “Delay Offset” (more than 30ms you’ll begin to hear two distinct events when you strike a key... you just want to thicken the sound, so stay below where the two sounds are separately distinguishable... less than 30ms. Then to maximize that ‘tubular’ type sound quality that Unison Poly was so popular for.... increment the LFO SPEED... one click from 0.00Hz to 0.04Hz is profound. It is the introduction of this very, very small movement that is the serious ear-candy... Experiment....
Set the LFO DEPTH to taste (0 Depth, means zero effect).

I will definitely try the “Symphonic” Insertion Effect

What if I use the “Symphonic” Insertion Effect in both Ins A & Ins B? Will this allow me to thicken 1 PART to a 4 PART Unison like on the DX7II?

Hope that helps. There is no one “movie score orchestral string” sound... for every movie made, there is a slightly different ‘orchestral string sound’... once you have the FM-X tone, don’t limit yourself to just that technology. By adding a bit of AWM2 to your Performance you may make a new standard for your own “movie score orchestral strings”. And if you are so inclined, share it on Soundmondo!

I definitely plan to do multiple layers off this FM-X orchestral strings and AWM2 strings. One of the reasons I want to try and get the FM-X orch strings as good as I have them on the DX7 is for instances where I am using up a lot of AWM2 polyphony via sequencing & playing piano or other sounds. I can use the FM-X engines polyphony on certain sounds like this so that I don't completely run out. I may have all 16 PARTs doing something at certain times during a song, so I am trying to introduce some of my own best DX7II sounds, as well as ones that are Prests on the Montage as well ... I'll likely program some new ones too. Thanks!

Now the fun is to take all this info and see if I can get 1 or 2 PARTs to sound the same as the DX7II's Unison/Multi!