best lowpass/hipass filter for quick, low latency sweeps?

[John Baylies]

Hi all,

I’ve just begun to dive into the H9000R and I’m seeking a lowpass/hipass filter with an adjustable slope (ideally from 6 dB/octave to 96 dB/octave) that has ultra low latency and responds well to very quick sweeps. I’ll be using Peter McCulloch‘s Max for Live device called XLFO to control the filter frequency, and it will be processing my tuba’s sound live, so that I can achive phat dubstep wobbles.

I ran brownian noise through a few filters earlier today to test them, but I haven’t found anything suitable yet.

Looking forward to any responces. Thanks in advance!

[macgee]

You could try these ones, I’ve tested them at some point but can’t recall

1426, 1427

3917, 3920

[John Baylies]

macgee wrote:

You could try these ones, I’ve tested them at some point but can’t recall

1426, 1427

3917, 3920

Thanks for the recomendations, macgee. 1427 looks like what I need, since it has eight filter orders to pick from. However, even though the cutoff dial moves immediately with the MIDI CC from XLFO, it sounds like the filter itself takes some extra time to sweep. Here is a recording of what I’m trying to describe: https://www.youtube.com/watch?v=m7UcyZyw6Xo

I don’t think there’s any internal MIDI CC smoothing going on, because I hear the same lag when I use my mouse to control the cutoff as well. Is there another lowpass filter with less lag between the cutoff dial and the actual filter frequency?

I’ll look into 3917 and 3920 as well, but that may take a few weeks, and I already have the LFO I need in Max for Live, but maybe 3917 and 3920 have added benefits that I’m not yet aware of.

[macgee]

Yea I see what you mean – It could be that this is an older algorithm and they’ve applied some smoothing to “DECLICK” the filter when modulating the frequency, not sure.

You could try offsetting when it’s triggered? Just trigger slightly earlier in the DAW unless you’re trying to do it in real time.

There is module in Vsig called MODFILTER that is ideal for filter sweeps and doesn’t suffer from clicking when changing the parameters and as such doesn’t need any smoothing. It’s VERY VERY likely there’s another algo that would suit your needs better (that uses this module) but finding can be a challenge.

I could make one for you in Vsig and share an algo you could try on your H9000 – for some reason when I shared a compiled algo to a user before, he couldn’t could get it work, not sure why so if that doesn’t work you could always load it using Vsig – not sure if you’ve explored the Vsig world but it’s excellent and takes the H9000 to the next level.

MODFILTER Group: Filter Modulatable Filter mfr

This module implements a classic state-variable audio filter. It provides simultaneous lowpass, bandpass, highpass, and notch outputs. It has variable Q (1/bandwidth) and frequency and has mod rate frequency and q factor modulation inputs.

This is the module to use to create any type of swept filter effects Audio inputs:

in

The delay input. Mod inputs:

fmod

Modulation input for cutoff frequency.

qmod

Modulation input for q factor. Audio outputs:

low

The lowpass filter output.

band

The bandpass filter output.

high

The highpass filter output.

notch

 

The notch filter output. Control inputs:

freq

Controls the cent frequency (or cutoff) of all filter outputs. Range: 0 to 20000 Hertz.

freqmodamt

Adjusts how much the freqmod input will modulate the filter center frequency. Range: – 20000 to 20000 Hertz.

q

qmodamt

Controls how much the qmod input modulate the filter q. Range: -1000 to 1000. Userobjects:

obj

Menupage of control inputs not connected to control signals. (collection) Order:

MODFILTER modulename in fmod qmod freq freqmodamt q qmodamt

[John Baylies]

I’m all about real-time low-latency audio effects. I was just looking at MODFILTER last night. I’ve only just begun to peruse Vsig. Thankfully I know someone who can give me a tutorial in person soon. I would be more than happy to try your Modfilter algo if it wouldn’t be too much trouble for you to make!

[macgee]

I’ve made something but need to test it and make sure it’s set up right and works as expected – I used the same filter cascading for my harmonic trem algo which worked well. Whether this style will work well for sweeping I don’t know yet.

This uses 3 filters, but could add more if needed – so 6 in total to support stereo signals

I’ll test it later in the week but feel free to test it now if you want, but I generally always miss something obvious initially haha

https://www.dropbox.com/s/52er6zu96ce341h/FilterSweeps.sig2?dl=0

 

[macgee]

I tested this out at the weekend and for some reason, at high freqencies – the one I tested had 5 in series and it was when there were 3 in series and higher, beyond 10k, but strange indeed! sounds like the kind of weird distortion I was getting when hit CPU threshold but that can’t be? with 5 x 2 modfilters and only at higher frequencies!

hmmm

[pmcculloch]

modiirq is probably going to work better for high frequency sweeps, though you can't control the Q at audio rate.

One thing that's important to remember is that when you stack up filters you're also stacking up any gain that those filters introduce.  (e.g. if you have a Q of 1, you're going to introduce some gain.  More sections = more gain)  If you want a higher order filter that doesn't introduce gain, you might look at something like a Butterworth filter (you would need to cascade multiple filters for that)

[macgee]

Thanks pmcculloch…
This was at the default Q of 0.5 so shouldn’t have been that? I should try reduce the gain and see if that makes a difference In case.
I’ll try the modiirq filter as well.
I’m not familiar with Butterworth filter design. Are you able to share a short description of how to achieve that in Vsig land? Everything I found was about the maths or electrical..
Sadly it’s nearly impossible to find Vsig style designs…

pmcculloch wrote:

modiirq is probably going to work better for high frequency sweeps, though you can't control the Q at audio rate.

One thing that's important to remember is that when you stack up filters you're also stacking up any gain that those filters introduce.  (e.g. if you have a Q of 1, you're going to introduce some gain.  More sections = more gain)  If you want a higher order filter that doesn't introduce gain, you might look at something like a Butterworth filter (you would need to cascade multiple filters for that)

[pmcculloch]

Butterworths are made up of second order sections (e.g., modfilter, modiirq) and in the case of odd order Butterworths only, and additional first order section.   

On the following Wikipedia page, you will find a table labelled: "Factors of Polynomial Bn(s)" that you can use to derive the needed Q values.  So, for example, for a 5th  order Butterworth with the factors (s + 1)(s² + 0.6180s + 1)(s² + 1.6180s + 1) we get the following information:

(s + 1) — this is a first order section, so use something like modhicut or hicut.  

(s² + 0.6180s + 1) — this is a second order section with a Q of 1/0.6180 (~1.6181)

(s² + 1.6180s + 1) — this is a second order section with a Q of 1/1.6181 (~0.6180)

https://en.wikipedia.org/wiki/Butterworth_filter

All of these filter sections will share the same cutoff frequency.  Butterworths don't have as steep a rolloff as some other filter shapes, but it's fairly easy to build higher-order filters with them once you know how to read the table.

 

macgee wrote:

Thanks pmcculloch… This was at the default Q of 0.5 so shouldn't have been that? I should try reduce the gain and see if that makes a difference In case. I'll try the modiirq filter as well. I'm not familiar with Butterworth filter design. Are you able to share a short description of how to achieve that in Vsig land? Everything I found was about the maths or electrical.. Sadly it's nearly impossible to find Vsig style designs…

pmcculloch wrote:

modiirq is probably going to work better for high frequency sweeps, though you can't control the Q at audio rate.

One thing that's important to remember is that when you stack up filters you're also stacking up any gain that those filters introduce.  (e.g. if you have a Q of 1, you're going to introduce some gain.  More sections = more gain)  If you want a higher order filter that doesn't introduce gain, you might look at something like a Butterworth filter (you would need to cascade multiple filters for that)

[macgee]

thanks pmcculloch, I’ll need to get my head around this – learning

[Author]

Posts