I agree, that they are not directly comparable, but not because AVRx is an hyprid synth (with the SwinSID/MIDIbox SID/AOUT combination you can add analog components as well), but mainly because it’s an “all-in-one” solution, and the synthesis concept allows waveform shaping.
I this is a bit of where I was coming from too. A good set of VCO’s in a modular are amazing - but on their own they are not a synth, I agree, SwinSID is an amazing component, but it is a component, not a complete synth.
Comparing SwinSID to AVRsynth ia little unfair too, AVRsynth is intended to be a simple, very cheap to build basic synth. Like AVRx the control surface load takes a lot away from the sound generating. A pair of AVR’s might have a chance, which is what you effectively have with SwinSID/MIDIbox SID
A dedicated sound generator,whether a real or modelled SID, will usually beat a single processor stand alone. Paul Maddox was able to model the whole of the MonoWave in a Chameleon, but that had a Coldfire for control and a DSP for the sound work.
My oppinion is mainly based on demos and specs (so far provided). E.g., there is no word about the sampling rate at AvrX site, but when I listen to the demo samples, and consider that the CPU is loaded with user interface and modulation tasks in addition, I guess that it cannot be so high.
I have to agree here - with the exception of the ‘dementor’ waveshaper stuff, a lot of it sounds a bit clean and simple. Well if it’s that limited, some lucky member of the group can get a ready built… We’ll see once I finish it up
And speaking about “expandable synthesizer platform”: I think that SwinSID approach has more potential compared to AvrSynth and AvrX - especially because alternative synthesis/filter/modulation routines can be integrated into the firmware on a straightforward way without taking care for multitasking and complex realtime requirements. You only need to take care, that the new sample word is calculated within 32 uS (=ca. 768 AVR instructions minus IRQ and DAC/Slave transfer overhead). And if 32 uS are too short for certain sound processing algorithms, you are free to reduce the number of oscillators, and/or the sampling rate of course 
No real argument here, apart from one factor I do know from personal experience. Filters. Doing filtering on an 8 bit controller is a bit of a lost cause. Not even limited recursion really,and only the more basic filter elgorithms. Now add and external filter, and straight away you’ve got a potential monster. I think qw’ew bith in solid agreement that the component approach wins.
I could change my oppinion about the “superiour project” once I read your review!
I’m not suggesting AVRx is superior to a MIDIbox SID - in fact I think a multi SID beats an AVRx hands down, but I do think it’s a good addition to the world of synth DIY. Review wise, I’m more interested in how it squares up to my MonoWave, which I still haven’t managed to chase up the rest of the weird pots for. If anyone here is building one, my suggestions are to either buy the whole front panel kit from Elby Designs, or to forget the front PCB and loose wire it. To me they are very similar devices: Lead/Bass synths with hardware generators and analogue processing. But I know need to reverse the old cliche and put my mouth (or ears!) where my money has been…
A minor thought. With MIDIbox, you’ve perhaps created a new class of synth. Somewhere between a full repatchable modular, and stand alone: “The Component Synth”.
Best wishes
MIke
(If this doesn’t make too much sense it’s because I’ve been laid low by that virus - I should be out running an event tonight!)
