Gregsbrain dSpec & XVOX modules

Gregsbrain LLC is a boutique modular synthesizer company based in Portland, Oregon. Their focus is innovative digital Eurorack modules, including xVox - a 4 voice pitch-shifter and dSpec - a unique Spectral VCO. We chatted to maker Greg Burns about his modular creations:

Thanks for talking to us Greg - so I'm really excited to hear all about your modular innovations in areas like pitch shifting and additive synthesis. But I thought we'd back up a moment and talk about your beginnings. Your background is in embedded systems & DSP audio processing; with a career in software development, right?

I have been actively working in embedded systems since 2001 but my software career goes back a lot further than that. I spent my earlier years working on compilers and code optimizers. In the early 90's I was at Microsoft and in 2000, feeling the need to do something different. I quit to join a Bluetooth startup where I got to be a very hands-on VP of engineering and CTO. This is where I first started doing DSP and audio processing work. We did a lot of work on Bluetooth audio including developing the world's first lossless audio codec for Bluetooth. We licensed our "BlueMagic" (that name was NOT my idea) software to Apple for Bluetooth support in the original iPhone. Last time I looked, BlueMagic is still listed on the iPhone legal page. 

So what attracted you to the world of modular?

It feels weird to say it, but I built my first modular synthesizer fifty years ago while studying Computer Science and Cybernetics (yes seriously CYBERNETICS) at the University of Reading, in the UK. I was learning about analog computers which, just like analog synthesizers, are built out of op-amps. I don't remember all the details, but my modular system had a couple of 555 based LFOs, an 8-step sequencer based on 8 pots and a shift register. An envelope generator and a VCA. It only had one VCO, based I think on an 8038 that cost a week's worth of student beer money. I also cobbled together a spring reverb out of a mini loudspeaker, a ceramic phono cartridge, and a bed spring.

Things like work and family got in the way so I didn't get back into modular synthesis until I bought a Moog Mother32 sometime in 2016.  A year later I had hand built a couple of utility modules and filled up a TipTop Mantis case.

So there you were, having fun with patch cables and blinkenlights. What inspired your first product, Xvox - a pitch shifter that creates chords out of a monophonic input? Did you feel something was missing in the modular world?

Yes, I found I could make lots of interesting noises but wasn't having much success making music. The main elements of music are melody, rhythm, and harmony, and it seemed to me I was missing harmony. I wanted to play chords, so this was the main motivation for developing xVox. With my DSP experience it was relatively easy to test out a proof-of-concept pitch shifter. To get the decent audio quality I decided to go with a frequency domain implementation. It took a good year to turn that proof-of-concept into working code running on an embedded CPU. There is a video from 2021 on my YouTube channel of a prototype running on an Electrosmith Daisy development board.

What do you think Xvox's standout features are?

xVox takes input from a single oscillator and turns it into four voices. It does this in real-time so modulating the input changes all the voices simultaneously. Each voice has its own volt-per-octave inputs and trigger inputs for a built-in envelope generator. There is a quantizer with three modes: pitch which constrains the voices to a specific scale, chord to fit a selected chord structure, and progression for chord progressions based on the circle-of-fifths. There are some bonus features too. xVox has two audio inputs that cross-modulate for vocoder effects and work nicely with voice. An option for the envelope generator slightly staggers the trigger starts times creating a strum effect.

Very neat stuff - experimental for sure, but also very musical! So onto your new module, dSpec. I'm a huge fan of additive synthesis - owning a beautiful Kawai K5m. But even I will admit - additive can be unwieldy! What inspired you to create this new take on these old-school concepts?

I am unfamiliar with the Kawai K5m but have read that it is a bear to program, I used to have a DX7 and some say the K5m is similarly hard to program. The thing about dSpec is you don't program it, you select an algorithm which generates the core audio spectrum, then apply modulation using knobs or CV inputs. The changes from modulation are typically incremental and easy to understand.

I should note that dSpec is a dual oscillator. Both oscillators share the same CV inputs but have volt-separate per-octave inputs.

In many respects dSpec is a child of xVox, they both do their internal processing in the frequency domain, meaning they manipulate the audio spectrum rather than the shape of a waveform. xVox takes an input waveform, turns it into an audio spectrum, processes the audio spectrum to shift the pitch four ways, then turns it back into waveforms. dSpec directly constructs an audio spectrum from an algorithm, modulates the spectrum based on CV inputs, pitch shifts it two ways before turning it into waveforms.

The selectable core spectral algorithms seem to promote "tweaking", giving the casual user a starting point to create additive tones. What was your philosophy in balancing accessibility with depth/flexibility?

dSpec is very much designed to be, as you put it, tweaked. What I refer to in the user documentation as the core spectrum is a collection of harmonics (and in-harmonic tones) generated from some familiar integer series. The harmonic series is simply all the odd and even multiples of a fundamental frequency; the Octaves series add harmonics that are powers of two above the fundamental frequency; the Primes series is based on prime numbers; the Fibonacci series is another well known series. There are eight series in total. On its own, each series gives a distinct and different timbre, but you can explore a whole range or sounds just by tweaking the low pass and comb-filter knobs. Except for the Octaves series, which has only even harmonics, tweaking the odd/even knob pans the sound from harsh to smooth. Of course, you get much more dynamic and less predictable results when dSpec is modulated with external CVs from LFOs, sequencers etc. If you want to go deeper, dSpec has two audio frequency modulation inputs, one for each of the oscillators. dSpec does a spectrum analysis of these audio inputs and uses the spectral information to modulate the corresponding core spectrum.

Those inputs sound wild, I'd love to feed drums, speech, whatever through them and then tweak away! The optional Quantizer & keyboard display is a nice touch - very musical. Are you from a piano background?

I have been around keyboards for most of my life, but I have never been very serious about learning to play well. It was while working on xVox that I realized that if I wanted to generate chords, and chord progressions, that I would need to learn a little music theory. It then seemed natural to display the notes of the selected scale or chord on a piano keyboard. dSpec borrows this idea, and some of the code from xVox.

Fantastic stuff - how can our readers purchase your modules?

You can purchase my modules directly from my website: www.gregsbrain.co

xVox is currently only available as DIY kit but it is quite an easy build. dSpec is available full assembled. I should also note that both modules are available as free VCVRack plugins.

Finally, any plans for the future? Any more designs in the works?

Right now, I am working on a few firmware improvements for dSpec. I will most likely continue working in the frequency domain because I enjoy the technical challenge, and it is a product differentiator. I have been thinking a little about how a frequency domain-based sampler might work but it is too early to know if that will go anywhere.

Thanks for chatting to us Greg, and best of luck with Gregsbrain LLC!