minimum group delay filters

[xoxos]

i recently discerned, unsurprisingly, that the progress made in audio DSP has not been transferred to other fields. in point, the people who write betaflight (the software for multirotor UAVs) are still using biquads.

they need minimum phase group delay "especially around the cutoff" (which is typically... 100Hz lowpass at a sample rate of say 8kHz, newer boards are being released at 32k not that it makes much of a difference). it should go without saying these applications are realtime and all latency is abhorrent. stop keytracking me! you can read it when i'm finished.

had a look into it last night and got far enough to discern that cytomic's trapezoidally integrated state variable filter has half the group delay of a biquad. so, in one casual movement, i have markedly improved an entire field of human endeavour, just by opening my mouth.

i have yet to go through the assortment of open source 0dfs i've collated, anticipating an equal or better performance increase.


anyway. be forewarned - there is no profit here to this particular cretin, i am only being nice to human beings who are too incapable of cooperating amongst themselves to do this otherwise.

so, please tell me, oh nice persons of the audio dsp practice, if you have any insight into the matter! betaflight is freeware, and used by tens of thousands of freemasons around the world. please help these poor people realise the benefit of cooperation founded in mutual interest before they destroy all of themselves and the rest of us along with them. it's terrible watching people suffer from their own ignorance.

thank you for your assistance.

[xoxos]

i mean, dear god, when did andy open source that... a decade back?

i said, stop keytracking me you biscuits.

[matt42]

had a look into it last night and got far enough to discern that cytomic's trapezoidally integrated state variable filter has half the group delay of a biquad. so, in one casual movement, i have markedly improved an entire field of human endeavour, just by opening my mouth..

hmmm, the trapezoidal integrator are good, but AFAIK the group delay will be 100% the same as a biquiad (assuming bilinear transform) .

[xoxos]

official statement = aah nutbags.

i was up late in sweltering temps with broken ac and things crawling all over me. forgot to halve w before taking the tangent that's why it was half as much!

simper svar does have slightly better group delay in the passband and matching worse group delay in stopband, equilateral s-curve.

robin's 0df svar has twice the deviation from the biquad as the simper, so better passband worse stopband. of course all forms match at cutoff.

with my meagre understanding, poking around with IIR elliptic forms might gain *some* advantage in the passband with handpicked cutoff and parameters. i anticipate that better options exist but this isn't really my thing. ttbomk most filtering in this application (smoothing PID response for yaw/pitch/roll) is static though dynamic is discussed... i believe if you use a contemporary flight board or model bearing one you're getting static forms.

[xoxos]

don't ask why mag for the last two forms aren't visible but obviously improvements are able to be made. (probably have to view image on a new window to read it if it's rescaled.)