Personal web pages of

Tim Stinchcombe

[Whilst I lumped this page in with the others under 'modifications', there are no modifications here, just some observations.]

Plan B M11 Evil Twin Bandpass filter

The M11 page at the EAR Group website contained a lot of typically exaggerated jargon to talk up the filter (the original website went down years ago, but the Wayback Machine still has copies of the page). Some of the more flowery expressions are fair enough, being as the terms are subjective, but some of those having a more technical meaning are at best misleading, if not just plainly erroneous—here's an extract, my emphasis:

Discrete Component Core

Utilizing an all-discrete core comprising twice the vactrols found in the mellifluous Model 12, the Evil Twin's open architecture takes advantage of circuit nodes which would be buried inside op-amps of conventional filters: a capacitance-coupled Twin Feedback path which engages an aggressive yet stable distortion mode not available anywhere but the 11. Brought into the circuit via a switch, the Twin Feedback generates timbres normally reserved for tube-based distortion units

On tracing out the circuit of the M11 its roots in the Buchla 291 are very evident—the filter core is pretty much a straight copy, with a handful of component value changes, and one extra component, empirically chosen to seemingly 'pep it up'. It is also plain to see that the above statement about the 'twin feedback path' is erroneous, as the 'single/twin' switch merely connects an attenuated amount of the input audio signal through to the main frequency controlling sub-circuit: it is tempting to call this 'feed forward', but even the use of that term may not be completely rigorous, as the signal concerned doesn't actually feed in directly again to the downstream version of the audio signal—however, feedback it most certainly isn't. At the time I thought that this being the case, it might be interesting to see if one could sensibly 'emulate' this feature of the circuit by external patching, and recently I have made just such an attempt.

Simple calculation showed that when the 'level' pot is set at max and the switch is at 'twin', the input audio signal is attenuated by about ×0.25 ≡ −12dB at the point where it enters the frequency CV circuit; furthermore there is a capacitor in the path, giving a highpass effect with a 3dB frequency of about 25Hz, so any input signals in my attempt (which was not going to incorporate such a capacitor) ought to be significantly above this frequency in order to not fall foul of unwanted attenuation. Signals input at the 'FM in' input are subjected to a small gain in the CV circuit (when the associated pot is set to maximum), and so the attenuation of any external signal would need to be in the region of ×0.17 to replicate the full internal attenuation. (All of these figures were double-checked via SPICE simulation.)

This diagram shows the equivalence, that an M11 operating in 'twin' mode is essentially the same as running in 'single' with an attenuated version of the audio signal feeding the 'FM in':

( The '×0.17' gain triangle could be implemented by the likes of a Doepfer A-138 mixer; the 'of suitable amplitude' is just to emphasize that the 'level' input pot needs to be fully clockwise, and so any desired attenuation of the input signal, to prevent distortion etc., needs to be external to the M11.)

I decided to use a patch which had a striking difference whilst flipping the switch between 'single' and 'twin', so as to be sure that the so-called 'twin feedback path' was having a significant effect—what I came up with was a PWMed square wave fed into the filter, with a low frequency cut-off. Here is the sound file of the clip I recorded—the sequence in the file is:

In the patch the filter seems fairly tame when run in 'single', but when in 'twin' it is much more lively (mainly because both the bandwidth and resonance pots are almost at full), and as can be heard, and as expected, the change between the proper 'twin' operation and the 'emulated' one isn't that great, effectively proving the point about what the 'twin' operation really is—the defining feature of the module is no more than audio-rate modulation of the cut-off frequency (albeit with the extra tweaks to the filter—over the 291—'maxed out'). (The actual attenuation needed was a little less than 0.17—measured at 0.22—but I later realised that I hadn't kept the input frequency as high as I needed to avoid the attenuation from the cap in the 'FM in' input path, and this could well account for the difference.)

(The full patch was:

[Page last updated: 03 Jan 2015]