MSF Physical Modeling FR

Official support for: meldaproduction.com
Chandlerhimself
KVRian
1111 posts since 19 Dec, 2013 from Japan

Post Thu Jan 10, 2019 7:29 am

I decided to make a new thread so as not to derail the other one. I'll email Vojtech about this later if he doesn't read this, but I though I'd post here to get others opinions on this. I have few requests, but I'll start with the most interesting and then move to the more mundane.

1). I was thinking earlier today about modal synthesis and new synthesis types and came up with an idea. What if you used the technology available in the OSC and wavetables with the modal filter? I'm specifically talking about the ability to analyze a WAV file and extract the harmonic levels. What I'm proposing is a new type of "modal filter". Instead of being based on the harmonic series, the harmonics could be any multiple of the root. The filter could analyze a WAV file and get the volume of each harmonic as well as the pitch(not the real pitch, but the relationship between the fundamental and the harmonic). It would set this for you automatically. Perhaps a higher number of harmonics than are currently available would be good, of course with the ability to reduce them to save CPU(64 or 50 seems good). With this you could accurately model different sounds like glass, wood, metal, etc and then use different resonators to excite them. Here is a video to give you and idea of what I'm talking about.

https://www.youtube.com/watch?v=Qafl_LGLDKg

The 2nd part of this module would be like a wavetable. Once you got the parameters from the analysis you could arrange them just like wavetables and scan through them and change them timbre in real time using whatever modulators you wanted. You could morph between a Xylophone and a Marimba or between a piece of glass and plastic. I've never heard of or seen this type of synthesis before and I think it could give some interesting results. It should also be possible with the current Melda tech without causing sparks to fly out of anyone's CPU. Call it Modaltables, MTables, PMTables or something. If it works the way I think it will, it should be simple, fun and unique.

2). I think an exciter module is needed. I've been using the Drum Synth4NN, but the sine sweep only goes to 10khz, so I'd like something that goes all the way from 20hz to 20khz or at least close. Another limitation with DS4NN is the lack of modulation. The OSC module is better in better in this regard, but the range of pitch modulation is too low. I think an exciter module with a sine sweep, noise and filters would be a useful addition. The filters would allow more noise colors such as red noise or blue noise. It would nice to modulate the sine sweep upper frequency by velocity or key-tracking also. Having 4 filters that could be set to LP, HP, Peak, Lowshelf, etc would be more than enough I think. That would also allow people to save noise/exciter sounds because everything would be in 1 module. Instead of making a new module this could also be added the the noise module. There is a lot of room left on the GUI, so these additions might make it more interesting.

3). Please add filter slope/poles to the modal filter. At first I thought "Order" did this, but now I realize that seems to add BP filters in series, which sounds cool, but can kill the transient. Increasing the BP filters from BP12 to BP48 can make a big difference in sound, while leaving the transient intact. I'm sure this will use more CPU, but IMO its worth it. The order control is good though, so I don't want to see it go away, just add slope controls next to it(Thanks Whywhy for this suggestion).

4). Please increase the range of the modal filter detune. Right now they only go down/up 1 semitone. IMHO an octave would be better.

5). Consolidate the bandpass filter and peak filter into 1 module. Both of them have a lot of extra space, so why not make something called "multifilter" that features a HP,LP and 2 peak filters. The filter and peak filter modules can do the same thing, so I think I don't think the peak filter module is really needed. This is a small thing, but it will save time and get rid of redundancy.

6). A simple EQ with perhaps 6 bands that has the normal stuff and the high/low-slope. The most important thing is that it has frequencies displayed. I think this will be more important when samples come into play later. If #5 from above happens this will be a small thing, but it might save some headache.

Ok, I'm sure that's more than enough and I'm getting sleepy. I think all these things could be done fairly easily and would make PM in MSF more fun and less painful. Tell me what you think of these ideas.

Whywhy
KVRist
334 posts since 11 Aug, 2018

Re: MSF Physical Modeling FR

Post Thu Jan 10, 2019 2:04 pm

Hi Chandler, I have read your post 3 time, to be sure to understand all :

1 and 2 : Melda Modal Table is a great idea !
I add something:
A multi band graphics, like a frequency splitter in Mps and be able to modulate the volume of each band . A band representing a series of harmonics.
It's a pain to modulate the volume of each harmonics into the modal filter, this system, couple with the mod graph could be more fast and easy, like some windows editing.

3,4 : agree !

5,6 could be link : hp,lp,peak with a frequency display.

Chandlerhimself
KVRian
1111 posts since 19 Dec, 2013 from Japan

Re: MSF Physical Modeling FR

Post Fri Jan 11, 2019 5:34 am

Whywhy - I like that idea of faster editing of the modal filter. I was thinking about it and maybe we could have an advanced button that would have functions for adjusting all the harmonics at once. For example

Odd level - Adjusts the level for all the odd harmonics
Even Level - Adjusts the level for all the odd harmonics
Tone - It would slant the harmonics up and down to make things have more bass or treble
Random - moves things around randomly

Of course you could have the same controls for the detune and perhaps something at the top that let you select the range of harmonics, so that you could adjust only harmonics 5-20 for example.

The idea to make 5 and 6 the same thing with a frequency display is brilliant. Vojtech said he couldn't have a frequency analyzer in the generator, but for me that's fine. Just having a frequency display like what is displayed in MComb would be enough. In normal mode have the node greyed out(or completely gone) and then in fixed more show it on the frequency display. Having 4 filters that could be set to HP, LP, peak, low shelf and high shelf would be enough I think.

User avatar
MeldaProduction
KVRAF
11256 posts since 15 Mar, 2008 from Czech republic

Re: MSF Physical Modeling FR

Post Fri Jan 11, 2019 12:49 pm

Hey guys, well, what apparently seems simple to you, isn't simple at all in a modular multivoice environment :). Ok so:

1) This is basically an additive synthesis, we'll get to that. (if you don't swamp me with stuff :D ). We could also change the modal filter actually - the original point was about harmonics, but we can simply add an override parameter to each harmonics, which would control the multiply of the original frequency. Analyzing a wav would then be pretty easy probably.

2) I don't think I understand the term exciter here :), at least in my head it means a few things, none of which would make sense here :). Could you elaborate?

3) Actually that's what order is ;). Duplicating the filter is essentially multiplying its order.

4) Ok, but note that from that it will be much harder to use. Maybe I'll rather add a switch.

5) It's a performance issue. I understand our point, but doing that would increase the CPU requirements, so...

6) The EQ is not a problem, but the visualisation is. It's always a lot of work to do visual stuff, and here especially, since it will depend on the voice, then how about if multiple voices are played, not mentioning that many technical difficulties...
Vojtech - MeldaProduction
Image

Whywhy
KVRist
334 posts since 11 Aug, 2018

Re: MSF Physical Modeling FR

Post Sat Jan 12, 2019 1:08 am

Hi Vojtech,

We know ! Having some ideas is not so complicated , but make them a reality is another story, it's why we put all our synthesis dream in your hands :wink:

1 : a lot of audio dsp concepts are linked,
Additive, wavetable....like delay, chorus...
Betweens additive synthesis and modal synthesis, I see some differences, correct me if I'm wrong :
A workflow difference, in modal synthesis, the overtones, harmonics, are time frequencies dependent: low registers have more sustained like in the real life.
I'm agree with you, we can mimic modal with additive, but we need to assign a lot of volume envelopes to be able to translate a modal patch into an additive patch. I think, it's why modal is used for physical modeling, due to his natural sounding, high overtones have a faster decay.
I will post a preset example, demonstrate it, later today, when I will be back in front of my daw.

2 : it just for make one generator, an exiter, instead of using a modular module with a noise,an osc, an eq.

3,4,5 : ok !

6 : about the display, no need to have a representation per voices, one voices is enough .
If you let me choose between : developing new modules or add an eq display : simple : more modules.

Chandlerhimself
KVRian
1111 posts since 19 Dec, 2013 from Japan

Re: MSF Physical Modeling FR

Post Sat Jan 12, 2019 5:50 am

Fair enough it might not be as simple as I'm making it out to be. Hopefully some of the things will be possible, but I understand somethings might not be possible.

1). Its not exactly additive synthesis because it wouldn't really be working with sine waves. I think Whywhy explained above how it is different. The wavetable/modaltable thing is somewhat for convenience, but the more I think about it the more I realize it would work better than the current way things are set up. Rather than modulating 32 individual harmonic levels, it would be much easier to set up "states/tables" and modulate between them using a "table index". That would also allow you to set up something like the harmonics graph to control the level, tuning, and width if you wanted to save space. This would be a unique feature and it might make the modal filter easier to use at the same time.

2). By exciter I mean something to trigger the sound of the resonator or the modal filter. In this case noise or a quick sine sweep. Before the OSC semitone range was limited, so it couldn't do a full spectrum sine sweep. I guess this has been changed in the latest beta, which is great. I'd still like to see a dedicated exciter just to make things easier. Basically it would be DrumSynth4NN with a greater pitch range and modulate-able controls. A sine sweep, noise and 2 filters are enough. This isn't really a "must have thing", but it would make life easier since you could save everything together. If you have free time or you can do it quickly it would be nice, but there are work arounds if you can't do it.

3). There seems to be some sort of difference in attack or delay. There is some sort of delay happening when using the order or putting bandpass filters in series. I thought this was intended, but it might be a bug. I tested it using a simple BP filter with resonance all the way up. I had a noise burst going into the BP48 filter and it sounded as I expected. Then I set the filter to BP12 and put 3 copies of it after the 1st. I expected it to sound like the BP48, but it doesn't. There is a long attack and no transient. The series of 4 BP12 filters sounds very different from the single BP48 filter. I'll leave an example below. In that preset you can hear there difference between the modal filter on "Order 4" and the BP48 filters in parallel.

Code: Select all

$eNq9WEtz4jgQvvMrVOKaAUt+psr21iRDZqkNgQEmkz0qtgKuCImS5VTYw-72LcmAHQjG2d3KBdztrx9ft+yWHP72umLghco8EzyCqGdBQHki0owvIliopy8B-C3uhKOZKHh6QxIl5CanSmV8kYNuBBEE3e8k4+NCRfCL17tEWGuYeCQsn5NHg7gWjJF1TtNptliqCeGURdCC4LqQknI1kTSnavgtgtdLwlNG5d8rkRIGHgl-BhiC7z+Hs+wviiOIXM+9CKzLvW5A8o3WY3er--3Hz3VOVmtmGFgQ3E5na5YpRaXGowgi5PuHahxB7NoQjMjr7c34ionkWasjaGMI7nfl8YNL7AQwDmfF44QVi4xvKxKHpTh7W5lz-JTcAA9cTU7EnVFGE0XT0ncErQsH7iLluwtwUzB2R1Y0gnciyylYUE4lUUICpP3mz0Nu6lDDQPBgVH-q3xO5d+8FK7SFBUH3V5ZSbgr6BZ1MlqYRdLCNncDzPQ-G4Wj8besFXIuCK+P3G12rpb6Kw6lV0+cysSI44C8AwX4c9ivjOOy-zXCveKcGI5EWjEg85Jr+Wx0E40KtCzXkKX1FJZd9IZxThfifm4gu7MYm3mRMUVklX8q1RO1TiXa-rtYDTh6ZboS1fS5187FTf0iR1wsuPQi6P0qr+WZNI3g1AU7wPoV+U7236WIIhnxXWsuE10vPBMSHVGpdsEwOD+b3Q9xuqlfPJ1G0Gyg6zRSxoYj-BcWeG3wuS6eBZdDM0jYs7Y+zxJ9L0W2giLxmjo7h6Hyco-25HL0GjvaZB9I1HN2Pc3R6unyfQnOUvZYvScNyO0XeIV3JyIStZGzWXSXbpkeV7Jg2V7JrSmLkmaKEWZU-I6MDGR-I9oHsHMiukXfz2rN31+ZmqahRf2+U1TT4SGMfaZwjjXswEt2TzR8ViqLqEm9zb25ifz-0zs7z+sp6M9sJO5yPO2Utb+9k3gN7++9u-71qjETQdnr6bTuWKZWm+90pzQUnPKH-meDRjrFT20Q+nMj3vYBHGwrrotrD5f3m2A9xJxzwFwS+KkWS5whaPctCnh-4gedb2GzPErLReuxZPkK+ZV86NgSzIlemRjrWXNKVYGK2JGsKZvPBZDIdzAZzfVPH1wHKOLgNFpdYuw3WLrFOG6xTYt02WLfEem2wXon122D9Ehu0wQZxJyzPTXoTXG8Qdi3Ptx0Pe75bawU634q9v7gT3t6MEbilfKG333odHaANAp9F2GcRzlmEexbhnUX4ZxFBM2JKeCpWSNeQ8PxJyBVRmeBgcHe-AwamiDvgzgS3NcF7E7utib03cdqa6HX+B93kCWG0BZsKWpnh9maa0z1lIslU65AH+AMH+IMO6hkcQjuVK31eywtJmyCZSpZXlKcNmJFIfy0pZQ2QebZ6lNR3GiBXkhK1tHADZPC61hlngiPUALsTio6fno4R31mW0vI9cHTva6KyF1qeYPcfY+aSlqf7QZqZI3gcXhNFF0JuhoquhvxJ9ONwe-o9tiqnr2lP3b2OJtdgShTVz52OPqKKyjwOf2WpWo7icCYKmVAwkfTJzPLthxrtaQfpb206Yf+9L0953PkHDEy4iA==
4). If you add the ability to override the harmonics and enter the ratio directly this isn't really needed. That would be more than enough. Consider this feature request withdrawn.

5). Understood. It would be nice to have, but it isn't essential.

6). Although visualization would be nice, but I don't really need that. I'd just like a way to filter things using "Q" and "Hz" instead of octaves and resonance. If it were possible to switch to HZ and Q in constant mode with the filters, that would be enough, but if not, a simple EQ(like MTurboEQ, but not vintage models) with no visualization is fine. It doesn't need keytracking, modulation, multi-voices or any of that other stuff.

As always thanks for all the work you're putting in on MSF. I don't want to bog you down with stuff, so if it takes a long time to get to these its not really a problem. #3 is the only one that really hurts preset creation or makes certain things impossible. The rest of the stuff I could live without if I had to(although I'd always wonder what could have been without #1).

User avatar
operator
KVRist
248 posts since 6 Jan, 2015 from Austria

Re: MSF Physical Modeling FR

Post Sun Jan 13, 2019 8:17 am

https://soundcloud.com/aweful-noise

Please Vojtech introduce some Saturation into the Feedback loop of the String Modul, and maybe some additonal options like in Zebra (Diffused Comb, ...)

Here i used Zebra with only one Comb with the "Diffused Comb" option... And what gives it the realness is the DISTORTION in the Feedback (and the Vibrato and a very tiny bit of Reverb... )
Try it out for yourself, when changing the Flavor Knob you will get almost close to real Trumpets, Brasses and Strings and a bunch other very cool disonant/atonal sounds... a screenshot of the Zebra Preset is attached.

By the way Urs copied Logic´s own PM synth "Sculpture" for the Comb Module in Zebra.. He stated that in an interview or somewhere in the U-he Forum...
You do not have the required permissions to view the files attached to this post.
Everyones better than me...

Whywhy
KVRist
334 posts since 11 Aug, 2018

Re: MSF Physical Modeling FR

Post Sun Jan 13, 2019 9:00 am

I add : the zebra dissonant comb, a 4 delays network, is so useful and beautiful !

User avatar
MeldaProduction
KVRAF
11256 posts since 15 Mar, 2008 from Czech republic

Re: MSF Physical Modeling FR

Post Wed Jan 16, 2019 3:12 am

1) I already added a 4 slot XY pad with analysis etc... I also created an Additive synthesis module based on that, seems pretty cool, a bit experimental, but still :)

2) I'd probably avoid that since you can do that with envelopes. There is like infinite number of potential excitation signals like that.

3) Hmmm, indeed! I need to check it out!

- Saturation for the feedback loop - I'll check! As for more complex combs, I'd probably keep that for later, I actually plan MTurboFilter :).
Vojtech - MeldaProduction
Image

User avatar
MeldaProduction
KVRAF
11256 posts since 15 Mar, 2008 from Czech republic

Re: MSF Physical Modeling FR

Post Wed Jan 16, 2019 3:20 am

3) So I checked and there is a different distribution of resonances for each BP in Filter compared to ModalFilter. Now, I can certainly do the same thing, but the question is of course if that's a good idea... I'll probably make it switchable.
Vojtech - MeldaProduction
Image

jmg8
KVRAF
2074 posts since 9 Jul, 2015 from UK

Re: MSF Physical Modeling FR

Post Wed Jan 16, 2019 8:13 am

MeldaProduction wrote:
Wed Jan 16, 2019 3:12 am

- Saturation for the feedback loop - I'll check! As for more complex combs, I'd probably keep that for later, I actually plan MTurboFilter :).
Also please, a really basic 1 knob compressor and an allpass filter.
These are things that are often placed in the feedback loop for physical modelling.

Also, is there any chance of having a few algorithms in there too? There are a few different ways of routing the feedback that create different sound. One for example is that the feedback is not added rather subtracted, this is good for bowed instruments.
If possible it would be great to get some very basic algorithms like from MTurboReverb, like:
FDNs and CCs etc

Just some very basic delay matrix with about 4 delays and various feedback routings with allpass etc.
It would add loads more flexibility to the resonator.
i7/16gb/win10,64bit/Live10/MTotal. + Surface Pro 4 (i5)

dmbaer
KVRian
1300 posts since 11 Nov, 2009 from Northern CA

Re: MSF Physical Modeling FR

Post Wed Jan 16, 2019 1:33 pm

jmg8 wrote:
Wed Jan 16, 2019 8:13 am
Also please, a really basic 1 knob compressor and an allpass filter.
These are things that are often placed in the feedback loop for physical modelling.

Also, is there any chance of having a few algorithms in there too? There are a few different ways of routing the feedback that create different sound. One for example is that the feedback is not added rather subtracted, this is good for bowed instruments.

. . .
Can I ask where you are getting your information about these things? I would love to discover a web site that started with Physical Modelling 101 and allowed the student to progress from there. I'd even be interested in a full book if one even exists.

Chandlerhimself
KVRian
1111 posts since 19 Dec, 2013 from Japan

Re: MSF Physical Modeling FR

Post Wed Jan 16, 2019 6:43 pm

MeldaProduction wrote:
Wed Jan 16, 2019 3:12 am
1) I already added a 4 slot XY pad with analysis etc... I also created an Additive synthesis module based on that, seems pretty cool, a bit experimental, but still :)

2) I'd probably avoid that since you can do that with envelopes. There is like infinite number of potential excitation signals like that.

3) Hmmm, indeed! I need to check it out!

- Saturation for the feedback loop - I'll check! As for more complex combs, I'd probably keep that for later, I actually plan MTurboFilter :).
1). Sounds great. I think an XY is a great alternative to a wavetable setup and might even be better for MPE if it works with modulators.

2). Understood.

3). The option to make it switchable would be great!

Chandlerhimself
KVRian
1111 posts since 19 Dec, 2013 from Japan

Re: MSF Physical Modeling FR

Post Wed Jan 16, 2019 6:55 pm

jmg8 wrote:
Wed Jan 16, 2019 8:13 am

Also please, a really basic 1 knob compressor and an allpass filter.
These are things that are often placed in the feedback loop for physical modelling.

Also, is there any chance of having a few algorithms in there too? There are a few different ways of routing the feedback that create different sound. One for example is that the feedback is not added rather subtracted, this is good for bowed instruments.
If possible it would be great to get some very basic algorithms like from MTurboReverb, like:
FDNs and CCs etc

Just some very basic delay matrix with about 4 delays and various feedback routings with allpass etc.
It would add loads more flexibility to the resonator.
I would like to see an allpass filter added, but the rest of the stuff I’d like to see in an all new module. It would be great to have a “wave guide” synthesis module that let you link resonators together and generate feedback. I’d love to see this in the future because I don’t think there is any VST currently that can do this, but IMHO it should come after things like samples.

Whywhy
KVRist
334 posts since 11 Aug, 2018

Re: MSF Physical Modeling FR

Post Thu Jan 17, 2019 11:49 am

:love: The new modal table module is just freaking awesome ! :love:
Fantastic idea and development :hug:
Just one thing, if it's possible, a little x,y, abcd, square could be enough, I find the x,y pad a little bigger compare to the others parameters.

Return to “MeldaProduction”