The TB-303 confusion cleared up once and for all...

[tripledot]

OK OK OK...

People talk about 3-pole filters, 4-pole filters, the accent twanging the envelope, "mystery circuits", the slide pre-empting forthcoming notes, etc.

Robin Whittle has already dispelled (quite comprehensively) many TB-303 myths in his writings on the subject, but obviously people haven't taken it on board and there remains a great deal of confusion. So here it is...

1. The TB-303's filter is a 4-pole design. However, the first pole of the filter (the first stage of attenuation that the signal from the VCO sees) is tuned one octave higher than the three subsequent poles. The capacitor responsible for this is C18. The result is that the effect of the three "correctly tunes" poles is damped somewhat by the rogue pole.

2. The filter does not have "extreme amounts of resonance". It does not self resonate at all, and has much less of the aggressive bite heard on, say, the MS-20's filter, which has a much steeper attenuation curve. The 303's filter is actually quite gentle. The pronounced squelch is not a result of especially powerful resonance, it is more to do with the non-linear frequency response of the resonance. The 303's filter resonates much less at lower cutoff frequencies, again due to C18's "damping" effect on the attenuation curve. This means that unlike "conventional" synths, who lose low frequency definition at high resonance levels, the 303 can maintain an unmuddied low and lower midrange while still screeching at the top of its filter.

3. The TB-303 has two envelope generators. The Main Envelope Generator (MEG) is best described as the filter EG. The decay pot controls the MEG's decay rate, NOT the VCA envelope. The VCA envelope is fixed.

4. The MEG is indeed bipolar. This is not uncommon in subtractive synthesis, but what is uncommon is its seemingly accidental implementation in the 303. The voltage output by the MEG affects the VCF's bias voltage, which is normally around 5.2 volts (I think!) - so during the decay phase of the envelope the bias drops below 5V and the cutoff drops below the "normal" setting as dictated by the cutoff pot.

5. The accent seems to be a subject of much confusion. Accented notes' trigger signals short out the MEG decay to its minimum, regardless of the accent pot's setting.

5a. The accent pot controls the amount of boost applied to the VCA input gain on accented notes.

5b. The accent pot controls the amount of boost applied to the filter cutoff (and to a lesser extent, the resonance) on accented notes.

5c. The resonance knob determines the amount of signal sent to a "lag" circuit before the VCF. The higher the resonance, the higher the proportion of signal that is affected by the filter lag. At high MEG settings, this is not really audible, but with the env. mod. pot turned down, the filter is (mostly) only affected by the accent boost (which runs through the lag circuit) and consequently the filter goes "wow" instead of "pow".

5d. Interesting to note is the "reverb halo" effect that the accent causes: with the accent and decay pots set to minimum, accented notes sound almost indistinguishable from un-accented ones. However, with the accent pot turned up full, the cutoff frequency of the filter is boosted sufficiently for the filter to remain slightly open (despite the MEG's bipolar effect) while the VCA struggles to attenuate the louder input signal from the accents. The result is a pseudo-reverb effect that sounds somewhat like an increase in the VCA's release time.

6. There is nothing mysterious about the slide. It is fixed time portamento. The slide doesn't start until the beginning of the note being slid into. Robin Whittle is credited with confirming this beyond all doubt with a logic analyser hooked up to the TB's CPU.

7. Accents in quick succession prevent complete discharge of a cap in the accent filter boost chain. Consequently the filter rises higher with each new accent, providing that excitable TB sound (again, thanks to Robin Whittle for writing an excellent treatise on this).

More important than any of these design oddities is the way in which we interact with the TB-303. It is not an accessible instrument, but it has personality, charm, and oftentimes attacks of Alzeimer's. But it is FUN!

Hopefully we will see the end of misinformed mythology someday. The better we all understand this thing, the better our chances of producing more synths with similar charm.

Peace, td.

[mosc]

What a great post to start with... Thanks...

[tripledot]

Thanks! Immediately after posting it I realised that there were more clued-up people on here than I thought, and I was worried that I might look like I was just angrily dictating shit that everyone already knew.

So apologies to anyone who thought that my rant was uncalled for!

I know there are many topics to do with certain synths, mikes, compressors etc etc that are regularly done to death, so I guess I was attempting to start some kind of "go-to" meta thread for the 303.

I'm trying desperately not to fail my degree right now, seeing as most of my time seems to go on designing a standalone devilfish-type instrument - PCBs might be available in the quasi-distant future!

nice to find this forum, there's invaluable info here.. I'm off to learn about oscillators! thanks!

[blue hell]

[Tim Stinchcombe]

(I've only recently signed up here, so apologies for being nearly 3 years after the event...maybe some are still actively watching this...)

I applaud tripledot's attempt to clarify some aspects of this filter. However I do have a few things to say: point 1, first sentence, I wholeheartedly agree - it is a 4 pole filter, period. Second sentence, that C18 makes the 'first pole one octave higher': in my paper

http://www.timstinchcombe.co.uk/synth/Moog_ladder_tf.pdf

I show on pages 46-48 that this idea is misplaced - unlike a Moog transistor ladder there is no 'isolation' between the poles, so changing the value of the lower cap affects all the other pole positions.

I have wandered back into this subject more or less by chance - about 6 months ago I acquired a (then non-working) Analogue Solutions TBX-303 clone, and spent some time looking at the circuit. From the notes I made it seems I worked out an expression for how both the frequency pot and the envelope voltage determine the current down the ladder (and hence the cut-off frequency) - it looks as though without too much effort I can extend this to include the effects of 'accent' via the second gang of the resonance pot. It will not be a pleasant expression, and have an exponential or two, and I have no idea if it will be any help in producing a digital simulation or not (there seems a lot of interest in that in the Nord G2 forum). In any case, I'll see where I get with it.

Tim

[Lorenzo]