Robin Whittle schrijft:
I just measured the four capacitors in the TT-303 filter. The first one
is 0.018uF (micro Farads) and the other three are 0.033uF. These are
the same values as in the TB-303.
The TT-303 uses transistors for the diode functions - an exact copy of
the TB-303 arrangement. It doesn't matter what the transistors or
diodes are, since they all have the same (or similar enough) voltage to
current curve at any given temperature. Diodes would have been
marginally less expensive and would have worked just as well. Maybe the
choice of transistors over diodes in the TB-303 was to get around some
patents involving actual diodes in the filter.
I haven't traced out the circuit of the TT-303 filter, but these crucial
components look the same as in the TB-303. There are dual transistors
at the driving and receiving ends too, as in the TB-303 circuit.
As far as I know the filter and the rest of the audio circuitry is the
same as the TB-303. Other people with more patience then me will
probably do A/B sound comparisons.
Whether or not this is exactly a 24dB per octave filter is something
which could be debated. The filter probably wouldn't be 24dB per octave
even if all the four capacitors were the same, since these stages, with
their resistor equivalents in the small signal transfer function of the
diodes, are coupled together in a way which means their input and output
impedances are not quite what would be required for them to operate at
6dB per octave.
David Bulog pointed to David Stinchcombe's treatise on diode ladder filters:
http://www.timstinchcombe.co.uk/index.php?pge=diode
which links to another page with this new URL:
http://fa.utfs.org/diy/roland_filters/
You can see the x0xb0x version of the filter at the "snapshot" image from:
http://www.ladyada.net/make/x0xb0x/download.html
There are three dual transistors - Q12 to drive the ladder, Q22 to
terminate it and Q21 as a differential amplifier receiving the signal
from the top. This follows exactly the TB-303 arrangement:
http://machines.hyperreal.org/manufacturers/Roland/TB-303/schematics/roland.TB-303.schem-5.gif
You can think of the capacitors as reasonably heavy wooden rungs on a
rope ladder. I am up in a tree holding the two ropes at the top, only
interested in the difference in tension between the two ropes, not
caring about the total or average tension. You are on the ground
differentially wiggling the two ropes at the end with your "VCO" signal,
which contains a bunch of harmonics. If the VCO signal is positive, you
pull more on the left rope and less on the right. When it is negative,
you do the opposite.
If you don't place general overall average tension on the ropes, they
are hanging loose and each pair of ropes between one rung and the other
is quite stretchy. This means they have a "high impedance". In
electrical terms their resistance or impedance to small AC signals (your
wiggling) is high, meaning not much of the wiggling from one rung can
affect the next rung above. Since the rungs are made of heavy wood,
they have a fixed inertia. All that matters to my perception is their
inertia in terms of one end being raised while the other is pulled down
- which is what you are attempting to do with your wiggling. The wooden
rungs are the exact equivalents of the fixed capacitors in the diode
ladder filter. After a few sets of slack rope and heavy rungs, with
each stage forming a low-pass filter with a low cut-off frequency, I
either feel none of your "VCO" wiggling, or at most just the very lowest
harmonics.
If you pull down tight on both ropes, while maintaining your wiggling,
this tightens up each section of rope and reduces its impedance to small
changes in tension. The rope behaves much like a diode or a transistor
with its base and collector connected, as is done in these filters.
(Using chain, steel cable or thick solid nylon would not behave this
way, because their impedance for small changes is much the same, quite
low, no matter how much tension is on them - assuming the chain has at
least some tension.)
Now each stage of the filter has a higher cut-off frequency, since there
is less impedance in the ropes, and so a stronger drive to move each
rung above with your wiggling signal. I will feel more of your wiggling
signal in general. In particular I will feel much more of the higher
frequencies, since each stage of the filter now has a higher cut-off
frequency and is not attenuating these higher frequencies as much as
when the rope had little tension on it.
Tim Stinchcombe has some daunting maths concerning this.
Returning to the TB-303, the important thing for me is not so much
precisely what the frequency response of the filter is, but how the
whole circuit behaves at higher resonances, and in the Devil Fish, how
it behaves with very low currents (very low tension on both ropes)
and/or with very high drive signals. In the Devil Fish I drive the
filter, in theory, 66.6 times harder than usual, if the Overdrive pot is
fully clockwise. This would be hard to measure, but it is the
theoretical ratio between the drive resistances of the unmodified TB-303
and the Devil Fish with Overdrive fully clockwise.
These "slack tension" (very low cut-off frequencies) and/or heavy
overdrive characteristics of the diode ladder filter are musically
really fruitful, and are not as accurate or mathematically perfect as a
"properly designed" 4 pole LPF, in which each resistor-capacitor stage
is used in a manner which is unaffected by the frequency varying
impedances of the other stages, by way of op-amps or transistors:
http://fa.utfs.org/diy/roland_filters/SH1.jpg
http://fa.utfs.org/diy/roland_filters/JP8.jpg
I would be surprised if the choice of 0.018uF instead of 0.033uF for the
first ring in the TB-303 made a clearly audible difference and I don't
know why this choice was made. It is equivalent to using a lighter
piece of wood for the first rung of your ladder, which means that in
theory the first stage has a cut-off frequency 0.033 / 0.018 = 1.833
times that of the other stage.
- Robin
http://www.firstpr.com.au/rwi/dfish/TT-303/
mss is deze wel authentieker dan de x0x
weeral een gemiste kans, ze slaan dus zoals bij elke clone so far weer de bal mis met de seq
