SMF - Just Installed!

Main Menu

Atmegatron Filter LFO Issue

Started by Trapezoid, October 28, 2016, 12:12:05 AM

Previous topic - Next topic


On the original Atmegatron (V 2.4 I believe) the filter LFO tends to malfunction when the parameters are really pushed. If the resonance is high, the LFO rate is above 12 or so, and the cutoff is low (or high depending on the filter) the sound becomes distorted for a moment (relatively speaking lol) and then cuts out completely. Sometimes it can be brought back if the parameters are dialed down but sometimes it just will just stay silent unless you either restart the unit or load a preset.


Ah I think this may be down to the filter algorithm.  Are you using the more extreme dB filters - the 100dB parametric and the high and low shelves?  The maths can go a bit wrong if pushed too far with these ones.  Not yet found a solution, I think it may be to do with the 8-bit calculations of the floating point algorithms.  I won't go into any more detail before confirming we're talking about the same thing.  Next time it happens, goto filter type and switch up and down a value and let me know if sound comes back. thanks!


It definitely happens more often on the extreme DB filters but it happens nearly as often on the standard low pass filter. It typically occurs near the point where the cutoff silences the voice altogether or sooner if the resonance is high. It is also seems  to happen consistently with pulse LFO waveforms. I can see why all this would overload it but it handles almost everything else I throw at it like a champ.


I'll have a play and see if I can get it to do it as often.  I've definitely seen the problem before, but it only happens rarely, not at all with the lower numbered filters. But yes - low cutoff + high resonance is the situation that can initiate the problem. It's the scenario that requires the highest resolution of the floating point maths.  There's always a possibility that a change I made between versions has had a knock-on effect with the filter.

Interestingly I've just developed a fixed point maths version of the filter for the new Eurorack module.  I could do a version of that for the Atmegatron.  The maths won't crash, but the down side is there's some stepping in the sound at low cutoff freq values and the 100dB filter isn't possible.  The joy of 8-bit!