Strange comment in wikipedia about TB303's square waves

[antto]

[valhallasound]

Now someone should edit the VCF section where it claims it has a 18 Db/oct, 3 pole filter where in fact it has a 4 pole, 24 Db/oct filter and reference this page: http://www.timstinchcombe.co.uk/synth/d ... diode.html

The latest analysis on that page seems to indicate that it is actually a 9 pole filter, due to the highpass filters formed by the coupling caps. The lower 4 poles apparently have resonance, but it is debatable if these poles will have much of an audible effect, compared to the 4 poles that are tuned by the diode ladder.

The highpass filtering will probably influence the depth of the cut below the resonant peak - I would like to see an analysis of this. The Minimoog and Moog Modular filter have similar highpass filtering, which results in the inability to self-oscillate below a certain frequency. My guess is that it also affects the amount of bass loss with high resonance.

The important thing I took away from the paper is that the gain is very high before self-oscillation. Which will have interesting effects in a nonlinear filter.

Sean Costello

[mystran]

The latest analysis on that page seems to indicate that it is actually a 9 pole filter, due to the highpass filters formed by the coupling caps. The lower 4 poles apparently have resonance, but it is debatable if these poles will have much of an audible effect, compared to the 4 poles that are tuned by the diode ladder.

From purely practical research into sticking stuff into the feedback loop, I can tell that whatever hipass you stick in there seems not to have much effect on the basic character of the filter, other than the mentioned loss of feedback gain at low frequencies (that kills resonance unless you boost the feedback to compensate) but when the filter is driven into non-linear behavior (which would be practically always) you'll get some DC offsets here and there, which have more or less no effect on the straight-feedback version, but which can (depending on the particular hipass filtering) cause some curious "DC offset modulation" which then interacts further with the non-linearities making things somewhat dirtier (or maybe unpredictable would be the word), and which can to some extend be controller by changing the hipass transfer function; at frequencies suitably above the hipass influence range the linear features seem to remain pretty much the same.

But no, didn't do any analysis, just poked around trying different things.