mSF: Dial back default variation in physical modeling modules?

[Annabanna]

I've started several times now working on recreating some drum sounds with physical modeling in MSF and stopped for the same reason each time. I understand it's sometimes desirable to have a certain degree of variation on each hit, but...

... when the various modules get stacked, with everything from the modal filter to the resonance in the resonator, etc adding their own variations, those variations compound, and I keep ending up with unusable patches because the output varies wildly from one hit to the next playing the same note at the same velocity.

Wildly.... like 12+dB swings.

Obviously, I can't use a patch that's that unreliable from one hit to the next, so I need to dial back or eliminate the rng and limit it to items and ranges where it's helping and not hurting. I see tuts mentioning how to ADD more variation, but not how to reign in the default variations.

Is there a control or controls for this somewhere?

[Tagirijus]

Hey, maybe you should show us some preset you made (copy it with the copy button from the plugin out of MSF and paste it here into the code-tag)?

[Annabanna]

Hey, maybe you should show us some preset you made (copy it with the copy button from the plugin out of MSF and paste it here into the code-tag)?

It's any of them. I've tried a bunch of patches. Here's a very simple attempt at starting a tom patch. Not much going on, but the steep filter on the bandpass at the end emphasizes the bass (pretty common technique I use all the time for various bass drum type patches to give focus and edge to the low end.

Waveshaper is just there to WAY overdrive and clip the noise going in to eliminate the possibility that it's variations in the noise causing the issue.

Only other thing I can think of is dodgy automation issues, like the envelopes aren't firing reliably.

I've got just a simple one bar loop going with c1 on beat 1 and c2 on beat 3 just repeating with a velocity of 80 for both, and I'm getting 20+dB swings on output for the same note. The fluctuation gets even worse if stacked with other things like going into resonator.

The general impression I'm getting is that while some people may like the variations in isolation, they compound when stacked which really limits what can be done with them.

Pretty sure version here is out of date and I don't have admin access here to fix that, so I prefer to share patches from my home studio, but this might work to share:

Code: Select all

$eNrdWF1z27oRffevwMCvCQ0sgAUwI+lO4shJpnaiWL5J2jdGgm1OKNIlKdfur+8sSMiK7Vhq71tfNODq7MF+YbHk6I+7VcluQ9MWdTXmMhOchWpRL4vqaszX3eVrx-+YHIzO5vW6Wp7ki65u7tvQdUV11bLDMZecHb4v6x952V7kP+LzdFl0dTMv-h3GXKLSr9Dobek0b+-HXArlX1kAzj58+fOmzVc3ZdxTcHZ6Pr8pi64LES4Ja-CxGMYcUHF2ti67YpY3+Sp0oWmP67LMb9qwjExfk2POKCM0n4zm6x+zcn1VVIMvk1H-OP-Vp3kow6ILy-7PMRevkCdomxbsbV79PK7XVRd1TtZl+SlfhTH-VBdteB+q0ORd3TDJ2Vne-vxYRZu2IJx9j6K-0+9kRHTi6DcWHX6tyzVp0vpbsQxVDNfraCy5C15qY7wUnk9GZ5-fDQrswcB34aa7HnORGc6mVf6j7KM0GZ2LLVTbLMSYT6tbJvnRZHT0QDUZHf1q2kawMyLf8tswv85vQozGY9lWHOTOOAxe0PIqp9yA5uywvc5viupqVS-DmM--uc6bsOTseN00oepmTWhD9-EdhQtYO-w7GV01+c31ZDSri6pjJ2V+1b4dc6mj34+FgrO7GDshEIUR0jvg7L4P59FzJGy+quvu+ruMxt7F3x4vpLOgjZVaxRAPdvyjrlfsLL+jcshQaGcscnZWVFGgvJEadFT4X9NwUpTddgr6563wq53hPxnz10CHPi+qz+suhpSzwy+R4XCed+sm7-pmwtnhxf1NGPPTGdNuqMqT5wpSm32q8KRn+PKUQf5eHTbqX-5C4M7qZV4+jl4SbsUPd8ZvKkVayLSAtFBpodPCpAWmhU0LlxZ+WEBihsQMiRkSMyRmSMyQmCExQ2KGxJy0VdpBpR1U2iHRJbZElrgGqvf9kaVaOQ9tXeXVIlAFgKHq6Zr1ols3YauLx9ZHnonMpxraaD5XCXu1toeq2HBNRpvtBev381YoBGuV9pyJ01ipHp2wCOA8cCbHXGXSIyhADV5zFrMrMmtBSC+8iaKoKUAaD2hAgyFViiCQslTGWkuSHifRg-NeKgGOs1gckDnQynthvCHRADQGEZQy2pCUkqQzqcBJsBokiXqgcE57AxYdOsVZrC6dGem0k+gNIXWPVEIqIYUCJYjSRCB6o7QAoMtamoHSGzQgUTsvyUdKO2YGLICQwpDhmJBowRnrQHjPWaxfzLzwiIZiQqLBH+kckE2CLvZY35h5Z4VHr4Ei5AagUMajEjEFVFguk2CF0cq5KEpuG++NUWAsEGEv9dpbi9ICKscZUAatQIdaOCE9Z-EUuQw0aKMQPJJo4wpK9NY6IA-jMfOZtRqdAa0MiZIrEtFpQKMVZ3QMQWRktnLSKqKETWqEVAq1EJQxOqcAmVLSGW+tp11SEtEqkAjGKHKSDjJA5hxaq6RFTaIBacEKZbSS0hOnoTsyU0ZqA15FH1MWDaKwUkmnkTixR1onrZPa0e4piwq1VVZ556kMqFeAzujBaQHakmhAai28V8Z6STVEzQRMpgC8c8ZYMsglpFUgnNYY5zvqNoCZovtVIR05SImU6F2MlAUS91JLJ8lZTzHlTI25oZsVnPZWU9SpWQFVmjZKeodIoo2NmupSG0+VQd0MbKa0QucsWk+iFCEQXgoUEItyyIVCBQ6stgYtZ3rMMQNHMG2lBc6GREhS0soLQeOAIZhG44QSyqHjbMiCUk441JIaAWc45i4T3msDiAI5w9QWUEpAGh6QM0soNN5LJ0B6y9kQfrBGghYKKPhuzH08vtIZBdQ7htADKqvBIChqCD72LOmFEcoY5zkbAg-GCFQotEH1lwaPt3m1nOVt+3B5JsnW5Wl3Xp5xav4wSzPI8DjMHsPTvKxp5vgwYxI24tNZxDyZB9+Fy3xddlu+bRbtc+4+eX042Hqj+L73y4R4eJk4enmb75OD0bS6lexN1+WLn30lOCutVUit6XjddvUqjtFxr-NQhrwN-eU3rW4DRaPXTTPqIgIXd-3Lwy8Taz+unTQhsH5Ylb8ZbX8F3g3Ao8cbpufBKjb99HV2Pp1PL8i+o8noogmruqyj+Wx+MZ398i-Ryd5-2PJf-f97Db3Xah+s6rF6H6zusWYfrOmxuA8We6zdB2t7rNsH6yYHo-7rAg1tmxJ47iBPT978eXrBvn18N2Vv5nP2eTb9RFPhIr+PGpsSkQ8PDwU0X7ddnEzpTWKHXRuDJgej05PPkp2G6opmT+BP0BEBOxFqJ0LvRJidCNyJsDsR7mXEeV4t65WkGOZVe1k3q-gyuH0EXAxiAiYV2FcFNipqXxW1UdH7qtBB+Vu4bxd5Gfbw5gH6oAb7q5FPX0NZL4pu7y0f4R8RwH9JsG3BY+jBAxWdt3bdhJcgRbe4fhuq5QuYs3r57TqE8gXIRbH60QSrX4C8bULeXQt4ATK9uyGLi7qS8gXYcPpfQHyqu-D58vIp4n1ZLPs+8vS-47rqmrosQ9OyeVFdleH4Oq+qUKbT8mbRFbehb2IsvfNeNCFeZPGLad+QjvMuXNXN-ccurD5WlzWjwelfdbMcc0rl0AyfMvST1hbBvMu7MBmdBPpME9q4Bb0sp5vsEe5oYJ6M3sx6Kwm0bTU50dyw87wL1BHIqbP4HRYmo2-Fsrs+m4zm9bpZBDZrwmX8gDJ85iWqBDlKSgejo+c+NbeTg-8Ax9tGsw==

[Tagirijus]

Hm, so far I cannot see or hear any drasting volume fluctuations. Sounds quite normal, if you ask me. Maybe you could come up with a more extreme example?

By the way: if you decrease the main envelopes release value to something more instrument-suitable (now it is at 10000ms, but around 300ms are enough for this short percussive sound already), you can save 10x the CPU amount.

[jmg8]

The level variations are coming from the fact that noise is random. Try replacing the noise generator with a simple OSC and you will hear that every hit will have the same level.

[Annabanna]

The level variations are coming from the fact that noise is random. Try replacing the noise generator with a simple OSC and you will hear that every hit will have the same level.

Makes sense.

I had assumed this might be an issue which is why in this patch I tried SLAMMING the noise into a waveshaper that chops it off dramatically, so there should be no major variation in the peak level, though I suppose that doesn't necessarily guarantee that it isn't substantially greater or less in any particular frequency which may be where the issue is coming from since the bass is emphasized later in the chain... and rings out more with the resonance dialed down to mute the ringing of the highs.

Given that the point here is to create drums sounds, and that doing so with noise seems to just about always give a more suitable drum tone than triggering with oscillators, is there a better option than, say, recording a sample of the noise and replaying the sample each time for consistency?

[Annabanna]

Hm, so far I cannot see or hear any drasting volume fluctuations. Sounds quite normal, if you ask me. Maybe you could come up with a more extreme example?

By the way: if you decrease the main envelopes release value to something more instrument-suitable (now it is at 10000ms, but around 300ms are enough for this short percussive sound already), you can save 10x the CPU amount.

Wow. Good you mentioned it. I created a default that's wide open a few patches back because I just got irritated that things I was doing in the generator section were being overridden by the master envelope when I was using an inline enveloper within the generator anyway, so figured it's just easier to open the master wide so I don't have two envelopes to deal with.

Hadn't crossed my mind that I'm chewing up that much more CPU by doing it that way. Since I did that, I've jsut been putting together a few patches and haven't used them in a full mix yet, so I guess I just hadn't had that meltdown moment yet where the system can't handle it. Hmm... Going to have to rethink the topology of a few patches then.

[jmg8]

I think the best option is to use the sampler with a noise sample. I made like a lot of noise samples you can try. Texture/Synthetic are white noise captured through complex filters and resonators. Texture/Organic are hybrid mixes of various foley recordings. Then you can modulate sample start with a random mod to get full control over the variations. Another great thing about using noise samples is that you can also play with modulating the pitch with a very short attack.
I also noticed that you use the envelope in custom shape mode for a simple attack. In my opinion, it would be better to use the attack mods for this sort of stuff.

[Tagirijus]

I think the best option is to use the sampler with a noise sample.

Cool, good point, jmg. Thanks for the info and this idea!