Make Noise QPAS

clone

#1

Anyone wanna help make a stereo multimode filter with me?

From the Website:

"The Story of the QPAS by Tony Rolando

When I worked at Moog calibrating Voyagers, I monitored with headphones, and one of my favorite sound sensations was that of the stereo low pass filter. You might not even recall that the Voyager had a stereo Dual LPF. It is under utilized since most folks are trying to achieve classic Model D goodness, and the OG MiniMoog was a very mono instrument.

The binaural effect of a stereo Dual LPF is simple: one channel of the sound is brighter than the other. However, the effect is heightened when the filter has an exponential response and the cutoff is modulated. At lower cutoff frequencies the two channels appear to move at similar velocities, but as the cutoff frequency increases, one channel will smoothly increase in velocity, becoming brighter faster. The other channel lags behind, but eventually reaches the same velocity, and finally at high cutoff frequencies the two channels coalesce into full brightness. Where the mono LPF smashes you in the forehead with simple heavy sound, the stereo LPF envelops you, putting you inside the heavy sound.

When I developed the QMMG I learned about multi-peak filtering. Another under-utilized technique, this is processing a single channel of sound, with multiple filter channels, which are modulated, and then summed together to create a new version of the original sound. On the QMMG all four channels of filters were summed to the MIX OUT. There were a few normalizations, but they did not really ease or encourage multi-peak use. The QMMG UI was awkward for use as a multi-peak filter.

A common use for multi-peak filtering is the creation of formants, where multiple band pass filters are carefully programmed to generate human vowel sounds. This can be fun but it requires meticulous programming. If you forget about formants and just modulate until it sounds good, a multi-peak filter can be gorgeous fun. When the multiple peaks are offset from each other and modulated, the effect can be dizzying, as the sound appears to be dancing with itself, moving in many directions and yet it is still a mono sound. It is an auditory illusion.

I began thinking about what it would be like to have BOTH effects simultaneously… the binaural enveloping of the stereo spaced peaks AND the animation of two or more peaks in a single channel dancing around each other. Obviously some primitive formants could occur as well if some inventive modulation destinations were offered.

QPAS is short for Quad Peak Animation System. Inside the QPAS there are four identical state variable filter cores with a control system devoted to utilizing them in a stereo multi-peak configuration amongst other things.

I wanted to design a dramatic stereo filter that could also be dynamically animated, primitively vocal and also capable of mono to stereo, stereo to mono, mono to mono and stereo to stereo utilities. Something that was more functional than just linking multiple filter modules.

We needed at least two filter cores to achieve stereo or multi peak animation and so we had to first design a single filter core where the character and repeatability was not so defined by the vactrol utilized in other MN filters. I enjoy the QMMG and MMG, but they require many trims and specialized selection of the vactrol parts to work well and even then there is a great deal of variation from one unit to the next. That might not work well for a stereo filter. We developed several completely different cores before finding the right one.

Once we had a good filter core designed, we began prototyping a multi-core, stereo version. We attempted 2, 3 and 4 core versions. It was not until we reached 4 cores that we achieved the animation and stereo image I had been imagining.

Finally we had to come up with a way to control 4 filter cores simultaneously that was powerful but simple enough to encourage system integration rather than system domination.

The Radiate Left and Right parameters define the stereo image and peak animation of the final sound. I believe they are an elegant way to handle animation of multiple filter cores within a modular synthesizer. The normalization between the two Radiate parameter further eases this handling. Radiate offsets the Left and Right sides from the Center Cutoff Frequency creating the stereo image while simultaneously moving the two peaks in the associated channel away from each other, or back together again, creating the peak animation. Radiate could also be used to program primitive formants.

The Cutoff and Q (or resonance) parameters are linked to all four filter cores. Adjusting these parameters may result in pronounced changes in the sound because they are changing the center Cutoff Frequency and Q for all four filter cores simultaneously. But remember, since all 4 cores may be modulated and/ or offset to different cutoff frequencies, and the response of this filter is highly exponential, and the perceived strength of resonance varies with cutoff frequency, modulating these macro parameters is capable of animating the animation and shifting the entire stereo image. This is why Cutoff Frequency has such a large knob.

The !!¡¡ inputs are intentionally mysterious inputs that may be used with just about any signal. We wanted to encourage blind patch experimentation. Audio rate modulations create sidebands that may aid in the creation of those primitive vocaloid sounds or work to decimate the peaks with quasi aliasing, resulting in low digital tech crunch. Slower modulation, such as a pattern of gates, may serve to accent particular moments in time or even cause spurious damped oscillations.

At the input I added a stereo VCA. The traditional location for a VCA is post-filter, so my hope is that this pre-filter VCA will encourage the less common signal path and be inspirational to folks. By design the QPAS will not self oscillate, but with high Q settings it may be excited by an input signal. Having a VCA pre-filter encourages exploration of the dampened ringing that results from the excitation of the QPAS filter cores.

We included the 3 most desired filter responses: Low Pass, Band Pass and Highpass, and were left enough space in the module to add another pair of outputs. I went back to the prototype and experimented with adding and subtracting different combinations of taps from the filter cores. We wanted to create a filter type that would allow for animation of resonant peaks without the loss of bottom or top end of the spectrum that must occur in a LP, BP or HP filter to give those types their sound. We also wanted to add something that could not be achieved by mixing the existing LP, BP and HP outputs. The SP, or Smile Pass, output is the subtle result of that experimentation."


#2

Yeah I saw that and how it sounded and thought we have to make this!!! Which filter model do you think would sound best?


#3

I think we can start with biquad filters and think up how all the panning and such work. It will all become easier when they upload the manual though.


#4

Sounds interesting! I just skimmed it tonight, I’ll have to do a closer read tomorrow.


#5

I think we should start with this filter:
27%20AM
SVF V1.1.audulus (23.9 KB)


#6

I’m just double checking, that R is for resonance? Because there is also a Radiate parameter.


#7

I think thats res - this is just the filter core that he’s sharing


#8

R = resonance. @biminiroad is correct, this is just the filter core for an SVF filter I built a while back. I thought it would make a good building block for the stereo unit. :cowboy_hat_face:


#9

QPAS Core proto v0.1.audulus (496.9 KB)

I worked it out eventually. Here’s my best guess as to how the QPAS filter core works, I’m still trying to decipher what radiate and smile pass filters actually mean.


#10

Smile pass sounds like an allpass maybe? Or like a multiple peak notch?

Radiate seems to leave the frequency center peak at center, but then moves the other two peaks both apart in frequency and to the edges of the stereo field - so it’s like you have 5 peaks if you spread left radiate to left and right radiate to right, and each are along the stereo spectrum.

My gut says you can accomplish this by making a knob that controls both the spread of the frequency and a pan, and then just duplicate that for two controls.


#11

Good explanation of Smile Pass from the demo at Perfect Circuit yesterday.


#12

Wow, I gotta watch that whole video now :wink:


#13

Does he actually say what kind of filter response it is? I just watched that part and he doesn’t really go deeper than the explanation that’s in the product description.


#14

I’m looking forward to fooling around with this myself, but at the moment I’m kind of on a random kick. I’m working on a random waveform oscillator at the moment. Random Waveform VCO (WIP) I think it will pair well with the QPA when we get it finished. (if anything in Audulus is ever really finished!:cowboy_hat_face:)


#15

K-PASA Proto 0.1 copy.audulus (691.4 KB)

Ok I got some semblance of the patch working. I think I need to work on the pitch scaling and may rebuild the whole thing from scratch as the next step, but I think it sounds pretty awesome running a single sawtooth wave though it and giving it some modulation.

edit: Also, obviously this patch benefits from a good stereo listening environment or headphones.


#16

Interesting break down of the QPAS with less of the poetic MN framing of the device.


#17

I seem to miss all the good threads!

The night of the annoucement I watched Tony’s presentation at Perfect Circuit and drafted up this module:

CutePAS-version-0.1.audulus (459.9 KB)

It’s pretty broken once you look at how I manage radiate controls and stuff. And the core filters I’m using don’t sound great for this job but I’ll worry about that later. I’m also not worrying about CPU usage at all. I got the 4 peaks dancing and that makes me happy! :slight_smile:

(No HP and SP outputs yet either. I’m treating SP as a Notch Filter for now)


#18

¡¡16 core filter!! :star_struck:


#19

Sounds great so far! Thick!


#20

okay, I cleaned up a lot of the mess in the previous version and pretty much started over. This one only has the LPF portion, since making the other ones is just a matter of copy-paste once I’ve got things ironed out. Still using the BiQuad filters, and still doing 16 cores because I want to keep the simultaneous outputs of all filter types.

CutiePAS-version-0.2.audulus (139.0 KB)

One thing I don’t understand about the BiQuad LPFs: When Q=0, they all start screaming in self-oscillation. I really don’t know why this happens when there are 4 filters but not when it’s just one on its own. For this reason I’m clamping Q so it doesn’t go below .01.

I also fixed the behavior of the Radiate controls after watching Loopop’s explanation.

The stereo effect now sounds quite good!