Open303 - open source 303 emulation project - collaborators wanted

DSP, Plugin and Host development discussion.
Post Reply New Topic
RELATED
PRODUCTS

Post

It's a sample being triggered, yes, with startoffset to get to the point of cutoff where I wanted. I just wanted to zoom in on an area where I needed that extra "hardness" in the filter.

I listened to your sound aswell. Is that kunn's filter, with the nonlinearities, with hp in the feedback path?

The dynamics of an overdriven transistor, looks very similar to a (simple onepole) feedback compressor. The effect I was after though, may be related to leakage, or other phenomena more though.

Post

Another sample using the current filter.

It's going to be interesting to tune this filter further :)

Post

Just out of curiosity, why so many tanh's in Kunns filter? From what I understand of transistors, (this is diode, but connected to work in a similar way?) (Btw some places I read that the 303 uses transistors in the filters aswell..??)

Conceptually, you need a control voltage (?) for the feedback, how's this done. Would that result in tanh(buf*amount), is that why there's so many of them(?).

Each places where there is some multiplication, results in nonlinear distortion, depending on the transistor/diode.

So, how many transistors for a one pole? two? three?

So, for instance, a tanh for each transistor.

buf = tanh(buf * (1-cut)) + tanh(in * cut)

Where's the plussing going on, inside a transistor?

Arg, this is turning to gibberish. Anyone who feel like correcting it is very welcome.

Post

well, the 303 filter is a diode ladder, but it uses transistors arranged to work like diodes
you don't need all the tanh() on each stage
just one nonlinearity on the feedback does the job quite well
It doesn't matter how it sounds..
..as long as it has BASS and it's LOUD!

irc.libera.chat >>> #kvr

Post

Yeah, that's my experience aswell. And that nonlinearity can be simplified to a clipper, it doesn't make much difference.

Post

You know if you want to improve your 303, some think a "nobassloss" tweak/mod is good. And particulary if you are doing resonance tuning, you don't unessecarily attenuate the filter.

This can be done by in - ((buf4 - in) * res), instead of in - (buf4 * res).

Post

I am very pleased with the sound of my filter, which is now a generic filter, with a "nobassloss" tweak, and clipping in the feedback path.

I think I am stopping there.

With an approximate accent feature, it sounds like this: http://www.paradoxuncreated.com/tmp/nobassloss.wav

The "nobassloss" tweak removes the attenuation from resonance tuning, and makes a completely perfect sweep over the frequency range, also due to the clipping in the feedback path.

Peace.

Post

what's this "nobassloss" thing you're talking about?
It doesn't matter how it sounds..
..as long as it has BASS and it's LOUD!

irc.libera.chat >>> #kvr

Post

Well, that is what the analog guys called it, when I first heard of it. It just that the attenuation that happens with increasing res, goes away.

Post

ParadoxUn wrote:Just out of curiosity, why so many tanh's in Kunns filter? From what I understand of transistors, (this is diode, but connected to work in a similar way?) (Btw some places I read that the 303 uses transistors in the filters aswell..??)
Actually you got exp() per diode (or transistor in the transistor ladder; see Diode in Wikipedia for Shockley diode equation) and it's the diffential diode-pair that simplifies to tanh().

Conceptually, you need a control voltage (?) for the feedback, how's this done. Would that result in tanh(buf*amount), is that why there's so many of them(?).
There is control voltage for the cutoff. There is none for the feedback (there is a real potentiometer for the feedback path instead), though there are versions of various ladders that do have control voltage for feedback too. It basically involves a VCA in the feedback path (or a multiply for the feedback signal in the simple discretisations).
Each places where there is some multiplication, results in nonlinear distortion, depending on the transistor/diode.

So, how many transistors for a one pole? two? three?
Actually one capacitor and one resistor. The trick is that from the point of view of the differential signal (ie the difference of voltages on the two sides of the capacitors) the transistor pair acts as a current controlled resistor (with the current being controlled by the control voltage). Since the signal itself is carried differentially, this control current is pretty much "invisible" as far as the signal itself is concerned.

The beauty of the original Moog design is that the transistors also act as buffers, so you really need 1 cap + 2 transistors per stage (and 2 more transistors to feed the thing). Various people (to dodge Moog's patents) wired the transistors as diodes (or just used real diodes) though. The diode pair's still work for the purpose of controlling the "resistances" but unfortunately you lose the buffering. Obviously the resulting transfer function is pretty much totally different, so it doesn't sound the same anymore (with or without consideration for what the non-linearities do to the sound).

So, for instance, a tanh for each transistor.

buf = tanh(buf * (1-cut)) + tanh(in * cut)

Where's the plussing going on, inside a transistor?

Arg, this is turning to gibberish. Anyone who feel like correcting it is very welcome.
The "plussing" is really Euler integration of the differential voltage over the caps, something like:

newVoltage = oldVoltage + differentialCurrentIn - differentialCurrentOut

The total current through a transistor pair is directly controlled by the cutoff CV (and is the same for all stages) and acts essentially as a voltage controlled resistance for the differential signal.

There is a few full derivations of this stuff around the web if you care for more, and a quick sketch or two somewhere in the forum archives too.

Post

ParadoxUn wrote:I am very pleased with the sound of my filter, which is now a generic filter, with a "nobassloss" tweak, and clipping in the feedback path.

I think I am stopping there.

With an approximate accent feature, it sounds like this: http://www.paradoxuncreated.com/tmp/nobassloss.wav

The "nobassloss" tweak removes the attenuation from resonance tuning, and makes a completely perfect sweep over the frequency range, also due to the clipping in the feedback path.

Peace.
i also think your filter sounds very nice; any chance of a little plugin
to play around with? :D

Post

The synth I am working on won't be ready for some time. And since it is just an experimental plugin I use to test DSP on, that time might be forever. However, someone might implement nobassloss and clipping in the feedback path filter into open 303. ? :)

Post

ahem, well...(cough) y'know, if you feel like releasing
a rrruff version, lol...i like the sound of this.
(forever?:cry:)

the loss of volume with res' was discussed at some point,
and was deemed to be 'authentic'.

Post

yes, the 303 filter looses _alot_ of it's output gain when you increase the resonance (even a little)
there is in fact something that looks like a "hack" in the circuit
the resonance pot is dual
the designers probably put an additional gain "boost" - when you increase the resonance, more feedback is fed to the filter, but also more of the output (via an additional path) is also mixed to the VCA (iirc see C22 C21 and how they are connected around the resonance pot)
It doesn't matter how it sounds..
..as long as it has BASS and it's LOUD!

irc.libera.chat >>> #kvr

Post

antto, the venom is sounding very nice now!
totally different from beta2. c'mon: the people
need beta3, hahaha...
-----------------------------------------------
and t'be fair, the x0xware could be trimmed down,
with stuff dropped, to fit on orig' cpu?

Post Reply

Return to “DSP and Plugin Development”