[Solved] MSF Modular32 + Crossover works on 31 out crashes on 32

Official support for: meldaproduction.com
Post Reply New Topic
RELATED
PRODUCTS

Post

I add a Modular32 effect container and insert Crossover.
Then I set Crossover output to 31 and it works fine.
But if I set it to 32 MSF crashes.
You do not have the required permissions to view the files attached to this post.
Last edited by arachnaut on Tue May 21, 2019 6:55 am, edited 1 time in total.
Jim Hurley - experimental music
Windows 10 Pro (20H2 19042.662); i9-9900K@5.1GHz;
Cakewalk; Adam Audio A8X; Axiom 61

Post

Yes, already fixed, please wait for the next version ;)
Vojtech
MeldaProduction MSoundFactory MDrummer MCompleteBundle The best plugins in the world :D

Post

arachnaut wrote: Sat May 18, 2019 12:22 pm I add a Modular32 …
Whoa, interesting:) Can't WAIT to see what YOU can get MSF to do, Jim—hope you'll share a little!

Post

Definitely will share when I figure what I am doing.
This is the beginning of an acoustic dispersion preset.
MSF seems able to handle 30 filter bands and 10 delay stages quite easily - much more than Realtor by a big margin if my initial tests are to be believed
Jim Hurley - experimental music
Windows 10 Pro (20H2 19042.662); i9-9900K@5.1GHz;
Cakewalk; Adam Audio A8X; Axiom 61

Post

arachnaut wrote: Sat May 18, 2019 6:25 pm
MSF seems able to handle 30 filter bands and 10 delay stages quite easily - much more than Realtor by a big margin if my initial tests are to be believed
What do you mean much more then reaktor can handle ?
Here I made an effect that splits the incoming signal in 1024 bands , and reconstructs the signal with 1024 sine oscilators .
Do you know how to use the voice module , because that is essential
Running at 44 khz , using 46% cpu on a 8 year old desktop
audio example
https://app.box.com/s/ds2sv92zkxnoh5ehy9dumweu254sx44j
First sentence is dry , the rest are 1024 sine waves reconstructing the signal and playing with the partials picth and envelope follower decay

Image
Eyeball exchanging
Soul calibrating ..frequencies

Post

I have only started to use MSF a few days ago, so I said "Initial tests".

Here I compare two similar instruments - one in Reaktor which uses 121 voices. It has 11 filter band and 11 stages (11x11=121 voices).

The spectrum is split 20-20K into 11 bands called a stage, each stage needs to have the previous 11 filters summed to form a new signal for the next stage.

It is this voice combining between stages that takes some CPU. The filters need to be butterworth for this to work without phase issues.

In MSF, I did 10 stages of 31 banks which would be 310 voices. The filters are done with crossover and the 31 delays are linear-phase.

Reaktor took 44% CPU; MSF took 27% CPU. The Reaktor version is monophonic, the MSF is stereo, so it is actually 620 voices.

This is not working yet in MSF, but it is mostly sketched out.

When I finish I will be able to have a more definitive answer.

While the Reaktor version is fully tested and I know how to use Reaktor moderately well, it took me about a week to make that ensemble.

I don't have a clue how to take advantage of MSF yet but I made that initial mock-up in about an hour. I think I spent most of the time trying to find the right modules.

I have no right to give any opinion yet on the merits or strengths of MSF, but I like what I see so far.

For your vocoder example to be similar you might have to do something like this:

Make 2 100 band vocoders, use one as a carrier and another as a modulator into another 100 band vocoder and so on for 5 vocoder stages. That would be 1000 voices - and there are no delay lines involved.

So voices 1 to 100 get added to be carrier1 and voices 101 to 200 get added to form modulator1; these are sent to voices 201 - 300 and 301 -400 for stage 2; these are summed and sent to voices 401 -500 and 501-600, etc.
Then while doing that why not put in a delay line for each of those voices?

You will get a multi-stage - delay modulated vocoder which will not sound anything like the thing you made,
and probably would not be recognizable as a vocoder at all.
You do not have the required permissions to view the files attached to this post.
Jim Hurley - experimental music
Windows 10 Pro (20H2 19042.662); i9-9900K@5.1GHz;
Cakewalk; Adam Audio A8X; Axiom 61

Post

The reason I do these things in stages is because that is the only convincing way I have found to do acoustic dispersion effects. If you have a long metalbar or cable and you strike it on one end and then listen at various lengths over the cable you will find that sound smears out and the frequencies shift in a unique way. I use the stage as a unit of distance - more stages is farther away.
These give me sci-fi ray-gun blast effects for want of a better description. Acoustic dispersion requires delay time to be inversely proportional to the square root of the frequency, (or something like that I forget the details) so it ideally needs to done continuously over each frequency - not 1/3 octave bands, but the effect is convincing enough.
But I like what happens when you modulate the delay times in different ways, you get mind-boggling unpredictable results. One of my favorite effects is the superformula-modulated delay lines in one of my Reaktor instruments.
But that takes more CPU than most people want to pay for to sound right.

It was poorly described here in one of NI's blogs:
https://blog.native-instruments.com/10- ... r-library/

I don't think that would be possible to do in MSF -the number of math blocks would be quite significant - it is a very hairy equation:
You do not have the required permissions to view the files attached to this post.
Jim Hurley - experimental music
Windows 10 Pro (20H2 19042.662); i9-9900K@5.1GHz;
Cakewalk; Adam Audio A8X; Axiom 61

Post

Interesting!

gentleclockdivider: I'm no expert in Reaktor, but my quess is that you are not really using a crossover, but FFT (which is of course many many times faster). Handling 1024 bands crossover isn't imho possible even with the optimizations we have there. Plus the separation between bands would be quite poor.
Vojtech
MeldaProduction MSoundFactory MDrummer MCompleteBundle The best plugins in the world :D

Post

TRue , I am just using reaktor primary bandpass filters and sine osc's , reaktor core filters use way more cpu ( and sound considerably better )
I must have misinterpred (which happens a lot ) what arachnaut said .
My apologies
Eyeball exchanging
Soul calibrating ..frequencies

Post

gentleclockdivider wrote: Sun May 19, 2019 11:06 am TRue , I am just using reaktor primary bandpass filters and sine osc's , reaktor core filters use way more cpu ( and sound considerably better )
I must have misinterpred (which happens a lot ) what arachnaut said .
My apologies
No problem. It would actually be immensely cool if we could use a crossover with like 1024 bands with extreme separation, because that would then possibly allow minimum phase pitch shifting for example. Though I fear the phase changes may make the whole process pointless :D, it would be cool to try anyways.
Vojtech
MeldaProduction MSoundFactory MDrummer MCompleteBundle The best plugins in the world :D

Post

New version just went up on the website, includes fix for this. I just tested it and it's behaving itself now.

Post

Yes. It looks fine.
Jim Hurley - experimental music
Windows 10 Pro (20H2 19042.662); i9-9900K@5.1GHz;
Cakewalk; Adam Audio A8X; Axiom 61

Post

arachnaut wrote: Tue May 21, 2019 2:42 am Yes. It looks fine.
Perhaps, it's time to add SOLVED to the thread's name.

With the proliferation of threads lately, it could help onlookers.
Although counting the high bug-squashing rate, this kind of threads will soon be lost in the past...
However, I'd rather have an informative thread list, if possible.

Edit//Thank you arachnaut

Post Reply

Return to “MeldaProduction”