rs-met

Music Engineer · Post by **Music Engineer** » Thu Feb 28, 2008 1:57 pm

mystran wrote:I've been pleased by how stable it tends to be in comparison to anything else I've tried.

this is good news for me. now, as for further improving it...

.. how about this: either using FFT or autocorrelation or whatever, estimate the fundamental pitch, then use lowpass-zero-crossing to find the exact crossing, aiming for a previously estimated period? For extra fun, estimate multiple fundamentals (say two oscillators tuned in a fifth) and aim for their periods multiple (or GCD).

oh i don't think, i want to do FFT in the waveform mode, but looking for the 'true' zero crossing (of the unfiltered signal) with the zero crossing of the filtered signal as point of departure might be worth a try. maybe i should also replace the lowpass by a bandpass with lower cutoff frequency somwhere in the subsonic range.

Oh and I'll report a bug as well (other than the randomly bugging lines that one sometimes sees in the spectrum view): open any drop down menu. Then move the plugin window.

mmhh, i'm using a GUI library (JUCE) so these dropdown-menu related things are not really in my hands, i think. as for the random lines - might this be the not-in-time GUI-redrawing thing which was already mentioned or is this something else?

Music Engineer · Post by **Music Engineer** » Thu Feb 28, 2008 2:01 pm

cnicky wrote: With some wild abuse I did get the waveform display to disappear.

wot? disappear?! errm, the whole display disappeared or only the waveform itself? did it come back? what kind of 'abuse' was it?

This is one of the best tools I've seen in a vst.
Thank you.

Func Shaper is fanastic too.

much thanks

cnicky · Post by **cnicky** » Thu Feb 28, 2008 3:30 pm

The abuse was holding down a note and violently modulating the frequency & amplitude of an FM OP Modulator using my midi keyboard joystick.

The resultant noise is unpleasant. This, on one occasion, caused the wave display to disappear, just the display. Clicking to spectrum and back made it return.

Not that repeatable during normal use.

I'll try a video capture if you'd like?

Music Engineer · Post by **Music Engineer** » Thu Feb 28, 2008 9:41 pm

cnicky wrote:The abuse was holding down a note and violently modulating the frequency & amplitude of an FM OP Modulator using my midi keyboard joystick.

The resultant noise is unpleasant. This, on one occasion, caused the wave display to disappear, just the display. Clicking to spectrum and back made it return.

Not that repeatable during normal use.

I'll try a video capture if you'd like?

well, if it is somewhat reproducable and you can capture it, i'd like to see it. this appears really strange. could it be, that the synth itself creates an invalid (out of reasonable range) sample value? i do not do any bounds-checking for the signal before drawing it, so maybe the display tries to draw a line to a point at infinity or something like that...just an idea

Optomadic · Post by **Optomadic** » Wed Apr 23, 2008 1:22 am

Robin any chance that Signal Analyzer can run as light as Multi Mod?

Thanks for any helpp.

Music Engineer · Post by **Music Engineer** » Wed Apr 23, 2008 10:58 am

Optomadic wrote:Robin any chance that Signal Analyzer can run as light as Multi Mod?

i see some potential for optimization for the oscilloscope view when it is zoomed out (i think, my buffering technique is suboptimal here). however for the spectrum analyzer i can't see that much optimization potential at the moment. as light as MultiMod? whoa...gimme an FFT algo as light as a multiplication...well, not exactly the same case but close.

Optomadic · Post by **Optomadic** » Wed Apr 23, 2008 4:07 pm

As I know nothing about this stuff it kinda of confuses me when thinking of more effects is less processing. Or maybe it just seems that way when the hood is closed down over the engine

..

Music Engineer · Post by **Music Engineer** » Wed Apr 23, 2008 4:34 pm

Optomadic wrote:As I know nothing about this stuff it kinda of confuses me when thinking of more effects is less processing.

errrmm...you mean like: the more audible a process is, the more cpu-load you expect? the notion of 'more effect' is rather ambigous. i could - as a simple example - just hard-clip some incoming audio and drive the input of the clipper heavily. that would certainly be a VERY audible effect, yet take next to no cpu. and an FFT analyzer is not suppsoed change the incoming audio stream at all - its sole purpose is visualization. but in order to visualize the signal, it has to do some serious processing. on the bright side, it takes cpu only when the GUI is open (as opposed to effects which modify the sound).

atomota · Post by **atomota** » Wed Apr 23, 2008 4:37 pm

mystran wrote:For scoping audio, the biggest problem is getting a waveform to stay still when you want it to see how it changes in response to some parameter tweaking or whatever.

This is where the hold off control on my analog scope really helps. Contrary to what was said earlier in this thread, it doesn't set the delay between the trigger and the subsequent sweep (there is a different function for that), it holds off the trigger (ignores trigger conditions) for some adjustable time after the end of a sweep, e.g., until approximately the start of the next period. Alternatively, you can put a wave with the same period but a more stable shape on the other channel and trigger from that.

Optomadic · Post by **Optomadic** » Wed Apr 23, 2008 4:39 pm

No no I was thinking the more feaures a plugin has the more CPU it would consume.

I just figured a plugin that is being used as a visual aid would be easier to optimize than one with filters etc. When you mention FFT processes going on under the hood in the analyzer it makes much more sense though.

One more question, does just the frequency analysis portion use FFT? Or the amplitude anaylsis also?

Music Engineer · Post by **Music Engineer** » Wed Apr 23, 2008 4:56 pm

Optomadic wrote: No no I was thinking the more feaures a plugin has the more CPU it would consume.

well, there might be some correlation but it certainly depends on the specific algorithm. many knobs = high CPU ... nah, that's too simple. in fact, i have a vocoder in mind that would not strictly require any controls at all, yet eat CPU for breakfast (and sound great).

I just figured a plugin that is being used as a visual aid would be easier to optimize than one with filters etc.

why that? in particular filters can be made pretty fast - as long as don't design them a sample-rate.

When you mention FFT processes going on under the hood in the analyzer it makes much more sense though. One more question, does just the frequency analysis portion use FFT? Or the amplitude anaylsis also?

no, the waveform display does not need FFT. but it needs buffering and peak-to-peak extraction for larger 'zoomed out' views - here is where i could probably squeeze out some more efficiency.

Music Engineer · Post by **Music Engineer** » Wed Apr 23, 2008 4:59 pm

atomota wrote:Alternatively, you can put a wave with the same period but a more stable shape on the other channel and trigger from that.

which is exactly what i do: the waveform with the same period but 'more stable' shape is just a lowpassed version of the incoming signal. that works really well for mono-pitched signals.

Optomadic · Post by **Optomadic** » Wed Apr 23, 2008 5:18 pm

Robin from www.rs-met.com wrote:in fact, i have a vocoder in mind that would not strictly require any controls at all, yet eat CPU for breakfast (and sound great)..

Sounds interesting.. I'd be willing to feed it some CPU if its hungry

...

Music Engineer · Post by **Music Engineer** » Wed Apr 23, 2008 5:34 pm

Optomadic wrote:I'd be willing to feed it some CPU if its hungry ...

yeah, but it doesn't exist - and if it will exist someday, then with a modified algorithm (...which has some controls (and less cpu hunger)

)

Optomadic · Post by **Optomadic** » Wed Apr 23, 2008 5:38 pm

Robin from www.rs-met.com wrote:no, the waveform display does not need FFT. but it needs buffering and peak-to-peak extraction for larger 'zoomed out' views - here is where i could probably squeeze out some more efficiency.

Not to sound ungrateful but may you could make 2 separate analyzers?