BrainModular

hi, thought i would take some wise advise and try posting here in the patching questions. i'm getting help in my general topic thread about my smartmixer project and need to invert the phase of some lowpass filters. I see there is phase invert in the delay module so i'm using that with zero delay:

here is my progress so far but it doesn't seem to be working, any ideas?



edit i should say i'm just doing the left channel for now just to test

yep better here betniks. gonna wach in one hour or two, but the top part seems good to me, just need one filter, the out is LP, ok you got your lp sig, you made it go to delay taht reverse the phase that is summed with dry in vol module. try to listen the top right vol audio out, is it worhing, are you hearing some HP ?

no i don't think i am hearing HP, it sounds like the dry does.

you're right i, here i ve tested two signals, onedry and same with inverted phase and on the delay ive got volume lowering with a few sample 20-25 delay, but can't get total cancelation. either need to find other way to invert phase, but feel usine filters arn't the best to do that job, there must be better vst filters/crossovers (sonalksis/sonnox/waves/tc..). Ideally for no coloring would need a linear phase filter but that would add latency. im searching a way to make a simple phase inverter cause pb is maybe the fact that the delay phase invert takes a small amount of time to process, but fail for the moment...don't understand, at a certain nb of samples there should be in theorical a total phase cancelation....mmmm
investigation

yeah i was getting that vol lowering as well. i'll have to look into the vsts that you listed.

that's what i was thinking about why this deley phase invert method isn't working (processing time)

i would try the 1-input or similiar but i'm not sure how to relate phase to either db or 0 to 1, seems like if you did 1- either of those you would dealing with vol.

yup i tried as well it's more complicate than I thought, not 1-,....
think need to find a good 3rd party vst plugin for filter and phase inverter...there might exist some direct crossover as well...im interested too so i ll return results if i find a good setup... but im sure it works, tested on ableton, the phase trick is working.

thank you for the help with this. anything you find would be great. yeah in ableton it works, that's how bentosan's smartmixer works in ableton.

from first test this seems a way to make a phase inverter, have not total cancelation but quasi, seems to give good results with usine filters.. have to put a delay of a few samples on the dry signal before summing with inverted phase from LP to compensate processing latency (here about 17samples), but results seems good, crossfading between dry and recomposed HPLP gives a quite transparent result (i tried Q at 0,5 crossFreq around 1k)

excellent! can't wait to try this when i'm in front of usine.

so i can just use a delay module on the wire between dry and audio out?

yup here is the pach to compare dry/recomposed.

dry is going to L1, recomposed to L2 of crossfade. here my source is a sampler to replace.

yes the delay is on the dry before summing the inverted LP.
in fact you ll need to separate what goes to L2, one is the LP direct out from LP, the two other wires compose HP.

thanks seems to work with 17 samples delay for me too. thank you so very much for figuring this out. i'll still look around at your vst suggestions but i'd like to switch my model around to this method in the mean time.

do you think it would be possible to measure the amount of samples delay you need and then enter that into num samples? how likely is that number to change?

usine can report latency estimation in samples for vst,but not for such internal patching. hopefully i think/hope this latency should remain quite stable/fixed. the best mesurement seems to be the old school method, by ears. here as it concerns HP, i tried to disable LP wire for a while and adjust the samples till I hear only hig freq, in fact here it's more around 10samples with fine adjust,
but with carefully hearing I think we can find the best sample setting, then should remain the same if global setup (block executions/cpu speed don't change... but might worth testing sending a little impulse and count elapsed time, will try, but the couting system might have itself process latency... gonna try.

edit: so im totally wrong!! try bsork's tech

to have complete threads for future reference, i am posting the solution that bsork provided in my other thread:
"if you take the output from the filter module and subtract it from the unfiltered audio, you get what you want. If the filter is in LP mode, the "original audio - low passed" output will be a high passed version of the audio. If you want to check the result, add the low and high passed and you should get the original. To do a further check, you can substract that audio from the original and get silence."

"In this slightly silly example, you'll hear the original audio. Remove the connection between the filter and the audio out, and you get a high passed version (if the filter is set to LP). If you keep the connection between the filter and audio out, and instead add a connection between -X.out to audio out, you should get total cancellation."

Really interesting thread guys! Bsork solution seems so simple and so efficient!

yep
interesting job !
will add this one to utils compilation 2...

yeah it works great and it was very easy to setup.

just so people can see what it looks like, i took a snapshot of my test design.

when the crossfader is 0 you have HP, at 1 you have LP, at .5 you have combo of both.

edit ignore that i have it plugged into audio out R

I'd just like to add that inversion iis quite useful for stereo/mono manipulations, as found in the stereo spread (or whatever it's called) various places in the demo patches. With a bit of simple math you can easily divide a stereo signal into left, right and center, and you don't have to treat the result as stereo+mono, although that is the most common usage.

yup m/s have lot applications too. i love to use m/s to isolate S parts of some tracks when mixing , a lot of general "dance" stull have rythmic parts and bass strong centered, while fx,synth,amb reverb are more panned. extend a lot mixing, funny to some time quasi solo a synth part of a track when killing M, Im a bit astonished it's not that used in general virtual mixer, (and even on traditionnal mixers) like traktor and others, don't remember about deckadance...but might be an extend function for your smartmixer beatnix;)

very interesting idea 23fx23, i'll have to look into that more. i don't think deckadance has anything like that stock.

I might also begin working on a 5-band eq filter setup instead of just low and hi, using a few LP filters and this inverse trick.