Pedals are always a fun source of inspiration. They tend to be focused, have all sorts of design constraints, and focus on real-time playability. This creates a vibrant and diverse set of outcomes from a few basic building blocks.
A few years ago I was looking at Pladask Elektrisk pedals, likely because I was considering the FABRIKAT in my currently unsatisfied search for a granular pedal. I noticed the FEBER pedal which contains the following description:
An exploration of the intersection of single sideband modulation, ring modulation, phasing and tremolo. FEBER combines a ring modulator pair, a Hilbert transform approximation, feedback and a random value sample/hold generator into a feature rich tool for exploring amplitude- and phase-based modulation at both high and low frequencies.
That sounds awesome and definitely a fun realm for all sorts of sounds both lovely gentle warbles and clangly, feedback mayhem. I set it aside to come back to create a prototype based on this and keep an eye out for purchase.
A few months ago, a similar topic came up during a weekly Kata call and that led to the creation of a bidirectional pitch shifter.
One of the issues with digital pitch shifters is that they ‘reflect’ the waveform as the signal hits the boundary at the top (half sample-rate ) and bottom (negative frequencies). Aliasing.
Alan & Pete, along with some others I forget (Jim, Andres, Simon,…?), worked through how to compensate for that. Here is the signal flow:
A couple notes, especially for those who may not be familiar with Kyma concepts:
- Kyma Hilbert provides the 90 degrees signal on the left signal output and the orginal on the right.
- The Hilbert has latency, defined by the NumberTaps setting. The default of 1051 is recommended. If you need greater accuracy especially at sub-audio rates, you can increase the value and the expense of increased latency. The default works great for our needs.
So that is the core bidi-shifter. I created an encapsulated version to reduce clutter for the next step. The encapsulated version is included in the link below. It has a couple tweaks with sample rate controls for band mix, frequency, and dry signal mix. The dry signal is latency compensated.
From the bidi shifter, the rest of the main signal path for the FEBER is easy: signal > shifter > filter and wrap that in a feedback loop. The modulation is then sent to the frequency.
Modulation is the first area where the designer choices really start to appear and there is no ‘right’ choice, just opinions and preferences, though admittedly those opinions tend to cluster in patterns.
The FEBER description is noise into a sample-and-hold with an interpolate option and cycle rate between 0 and 30 hz. I’d guess the noise in the original is not pure white but pink or is white but then transformed so that the modulation is distributed equally across each octave when at max. But white noise could be fun too, especially at lower settings. So let’s have both. Our control rate is low enough we could do this all in Capytalk easily but I am doing them here in signal flow because I find it easier to understand a glance and because I tend to want to try different processes to control signals and it faster for me to test ideas out in the signal editor and then go back and tidy up with Capytalk if desired.
The interpolation method isn’t defined in the docs, but it makes sense that it is not a straight line and is synched to the rate. I am using a control filter that works below 4 hz (the native filters don’t work properly if their cutoff is below 4 hz or so). Because controls are cheap in Kyma (unlike the limited surface of a pedal), we can add a variable smoothing rather than an on/off switch. At full smoothing, the cutoff frequency, and hence the interpolation curve between values, is synched to the sample-and-hold rate.
So we have a modulation source in the form of a fluctuating random voltages circuit but with some extra bits. In the future, I will likely add a feedback loop (more likely a Capytalk equivalent) here to create a correlated signal which could be handy for creating a brownian walk to the sound.
So at this point, we have already gone afield of the FEBER through our choices within the modulation section (and our ignorance of what choices Pladask made here outside the some guesses and what is documented in the manual), but we have the core FEBER design and it can be tweaked and shaped in many different directions. It is capable of so much from gentle tremolo sounds and glitchy feedback squeaks and squalls. The pedal must be a ton of fun, especially when paired with other effects and feedback. It is easy to see why they are popular on the secondhand market and sell out quickly when available new.
From here, we can leverage the fact that we have DSP to burn and few limitations on the control interface other than those we self-impose.
Add a high-pass filter and saturation to the tone. At this point, I started inserting envelope followers to modulate the filter and side-band mix. The envelope followers really add a nice lively dimension to the sound. I added a control to change the response rate in real-time.
Next was to add a delay to the feedback path. I played with a normal fixed delay and that was nice but I wanted something ‘integrated’ with the rest of the overall design. I pulled in the delay Pete built that has tapehead like qualities and synced its delay time to the frequency of the pitch shifter. This means at lower frequency values the delay time can get long (this one has a fixed maximum of 120 seconds). I’ve been meaning to do more with this delay. It is so nice. I know there is an immense gravity pull to go down a certain paths with a tapehead core, I feel them myself regularly, but they are useful in many additional contexts.
But synching the delay to the frequency is not the only choice. Another option is the sample-and-hold rate. The goal was to constrain the available controls and these are the two most natural as they both share a time based component with the delay. A switch could easily be added as well as a clock divider or a probability chance so the sample hold skips around out of step. But that is getting ahead of ourselves.
And a VCS that looks like this:
I arranged it so that it was readable and to hopefully be functional to interact on screen but also auto-mapped to my Motormix in a useful way for live playing (the first 8 of the top row, hence the filter section dropped a bit lower). I think another live control to add would be an ‘external’ audio sidechain path into the envelope follower.
This shift > filter > feedback delay combo is as good a place as any to wrap up here. The sound is linked below. No presets to start with. There are many ways to continue to expand and explore. Lots of modulation options, especially with shift registers and other stepped noise. Add multiple delay heads to the feedback loop or a granular delay. Wavefolders are always a good time but can be difficult to tame. Or a scattering of particles. And it should be stereo.
Enfebre.zip (793.8 KB)
