You are here

Filter (Form)

Package icon Paths.zip2.64 KB

the type (form) filter consumes much cpu.
in some configurations (poly) reaches 90% in jack; (mono) up to 20%.
Here are some examples:

rncbc's picture

well that figures do depend on several factors, including current system configuration: cpu power, jackd frames/period, etc.

for instance, on my system (intel core 2 quad, jackd -p128) the provided presets just tops 35% on jackd dsp load.

it may also depend on the build target and compiler optimizations in place (release is ok; debug not so much).

and obviously the more voices are playing simultaneously (mono vs. poly) the greater the load.

however i must give you an optimization hint:- don't set LFO Reso to anything but 0% when filter type is Form(ant): it won't make up much of an audible difference, especially if the LFO BPM and/or Rate are relatively high, but will do a lot of a computational difference as the formant filter coefficients will have to get recalculated all the time, which is just overkill.


kobraji's picture

Ok, will be more careful with the filter Form (ant), I will not exaggerations ..

I quite curious because as a developer optimizes your system.
in my case (intel i3 3.07Ghz 4nucleos, 4ram, 64bit) and you have better performance?

jackd is in 48000Hz, 1024, with buffer 3. (rtirq, and jackd = open limits.conf)

yeah, usually poly sounds, use four voices.

thanks'll be more careful ...


lfo have reason (Reso) greatly affects performance.
I put Bss_Ld-Oeegrgr with lfo (Reso) = 0, and all was normal wing
Dsp = 3.8% and sounds the same.
lfo = (cutoff), it is the culprit, clear with Filter (Form).
I put Ks_Pd-Sirium with lfo (cutoff) = 0 and have obtained a normal DSP about 12% (4voces) .no sounds the same but ...

Thank you, you made me enterder the root problem.

Greetings! I love the sound of this synth, but the modulation factor of the filter envelope is so small that at 100% of its influence it is almost inaudible :(.

rncbc's picture

well, yes, if this is about the "Formant" filter, then again yes, modulation (or automation) isn't really a feature you can take good advantage of, sorry to tell :)

as previously stated, modulating (or automating) the "cutoff" or "reso" of the "Formant" filter is not cheap in cpu usage anyway; and add to that that those parameters are not quite true to their labels or names: cutoff just maps to a discrete "vowel" frequency peak (a=1 to u=5 in range as also though human vocal range as tenor, soprano, etc.) and "reso" maps to peak bandwidth (or Q factor if you prefer); continuously changing those parameters do make more arm (high cpu load) than good, if any an audible effect at all.

so yes. again. please don't touch LFO Cutoff or Reso when the DCF is set to "Formant" type: you better off keeping those at 0.0 ;)


I'm afraid it's about the classic LP and HP filter. I use Synthv1 from the KX-Studio repositories, and I can't modulate the cutoff frequency of the filter to get the punchy and dynamic sound because the cutoff frequency does not respond to the ADSR-knob.

rncbc's picture

you should play and balance between DCF base settings (Envelope, Cuttoff and Reso) and the LFO corresponding ones (Cutoff and Reso when not zero 0.0)...

also, higher LFO rates tend to have close to no audible effects but cpu intensive ones :)


ps. may you provide a screenshot of some kind as for letting us see what is not really effective?

Add new comment