TB 303 essay

[Tony Allgood]

Hi all,

>My guess on VCA cct is that you'll get similar results with any other
decent VCA with some right biasing and final discrete stage that
original uses

Yes, I think you are right. A 3080 with a darlington output could behave
just like a TB303. Notice there is no offset adjustment in the TB. I
don't know whether the BA662 was better behaved OTA than the 3080, but
my own measurements do not support that.

VCO Buffer >...and better opamps (I had very limited resorts) it should
be possible to eliminate FET.

Yes, I think so. The problem is the required output pd. It is required
to go to the +ve rail. The non zero Vgs of the FET follower acts as some
sort of waveclipper. Not a problem for the saw, but a major headeache
for the very fussy square wave shaper. This bit of the VCO is very fussy
on absolute level of the incoming saw. As for the 2SK30ATM-O, any FET
with a very low Idss of Vp will do. Select your BF244As and you should
be OK. Different pin out though. The 2SK30ATM-0 is a cheap part and
presented the designer with a good option. I wish I could get hold of
them here in the UK. Toshiba will sell to me, but I need to buy 3000 of
them. The other way is to run the FET of a higher supply rail. Perhaps
the unregulated, but smoothed volatge across the smoothing cap in your
PSU.

>VCF section is surely exception for Roland, they used OTA based filters
in those times.

Many other earlier Roland synths had this configuration. I think the
TB303 must have been the last. The SH3a, System-100/101 and others had
this diode arrangement. What is odd is the use of only four poles, the
others tended to use five poles. The change in the SH3 and SH3a (and
SH2000) from Moog ladder to unpatented diode ladder meant they needed a
24dB roll off. The four stage diode ladder didn't cut the mustard with
its messy slope, so they added another stage. But this extra stage also
lowers S/N in the diode ladder where the signal looses strength more
rapidly up the ladder. The TB303 with its tiny real estate and close
proximity to digital circuitry could afford to lose a bit of roll-off
for a trade in S/N. The real estate problem was probably the reasoning
against using four BA662s in an OTA based filter. And having heard the
SH09 and compared it against the System100-101, the diode ladder beats
it hands down.

>I was in doubt why they used transistors as diodes instead plain
diodes. My guess is that they needed _some_ matching for ladder diodes,
and those transistors had better specs than stock diodes...

I think this may be the case but I don't know for sure. Maybe the small
signal resistance is different and more suitable? In the SH2000 and
Sys100-101 they use hand matched transistors at the top of the ladder.
The bottom pair is a matched pair. The others are straight out of the
box. The TB303 was fortunate to use a single matched pair at the top.

But you haven't talked about the mixer. Load dependant distortion.
Assymetrical distortion that changes with output load. Bang your TB303
into a 1K input impedance and listen to it squeal. This mixer section
will compress the signal under accent conditions even on modest loading.

The weirdest thing about the TB303 for me, is the way that one of the
envelope signals is fed into a lag circuit that is controlled by the
resonance pot. Very odd, and very distinctive too.

Regards,

Tony Allgood  Penrith, Cumbria, England

Oakley Modular Synth and TB3030:
www.techrepairs.freeserve.co.uk/projects.htm
My music: www.mp3.com/taklamakan