DSP Robotics Support

Optimized linear phase reconstruction 5-band splitters, 3-pole and 4-pole versions.

Thanks, Martin. Such optimization condensate is very impressing. I've just realized that such transient-perfect quasi-synchronous band splitter family may be useful, when coupled to precise compressors-limiters, for maintaining under strict limits, across the whole audio spectrum, the error signal that a Widrow-Hoff LMS Machinery is processing. I've just completed a 64-tap Widrow-Hoff LMS Machinery. No bells and whistles yet. No optimization yet. Just a barebone Widrow-Hoff LMS Machinery, whose code and labels are "textbook" compliant.
See it here : http://www.dsprobotics.com/support/viewtopic.php?f=4&t=38857
It is cloning (identifying) the transfer function of a "plant" consisting in three generalized IIR Biquads in series. I designed the "plant", for resembling a medium speaker driver.

Have a nice day

Just for completeness I am posting a five-band splitter with linear phase for each band (and any combination thereof). It does have some (unavoidable) latency in order to make the filters causal. Slope is 24dB/octave, CPU load is not bad.

Have fun!

martinvicanek wrote:Just for completeness I am posting a five-band splitter with linear phase for each band (and any combination thereof). It does have some (unavoidable) latency in order to make the filters causal. Slope is 24dB/octave, CPU load is not bad.

Have fun!

Hi Martin, this is a great module, but it only correctly works in poly mode, not mono (blue).

In mono mode, there's only 2 usable bands, "highpass" and "bandpass3" : "bandpass3" contains sound of "bandpass2", "bandpass1" and "lowpass", but if i check others outputs than "highpass" and "bandpass3", there's no signal available. Is there a way to correct it? Thank for your great work.

Hi latour, it does work in blue here, see attachment. Can you post a schematic to demonstrate the issue that you encounter?

latour wrote:[...] it only correctly works in poly mode, not mono (blue).

I think I know what the problem was: The splitter module expected the input signal at all 4 SSE channels. This might not be the case in every usecase. Now fixed, and included more bandsplitter variants.

It now works perfectly. Thank you Martin.