Alesis Andromeda A6 Tips And Tricks Manual page 87

The audio input is applied to one of the transistors at the bottom pair and the audio output is
differentiated off the collectors of the top pair.
Resonance is obtained by feeding back attenuated levels of the output signal back to the transistor
opposite the one receiving the input signal.
Two things happen with increasing resonance; first, the harmonics close to the cutoff frequency are
added to those in the original signal. This happens because the harmonics close to the cutoff
frequency have a phase shift approaching 180 degrees and the output is DIFFERENTIATED
(subtracted) off the top transistor pairs of the ladder. Off one of the transistors of the top pair you
see the original harmonics, off the other one you see the harmonics with a phase shift of 180
degrees aka INVERSION. Subtract plus one (original) and minus one (inverted) and you get two.
Therefore the harmonics closest to the cutoff frequency are increased.
Second: all the unfiltered harmonics that are at least 1/4 octave below the cutoff frequency (IE bass
frequencies) are (ideally) in phase with the original input signal. As you increase the resonance and
add more of the original in phase harmonics, the differentiating action at the output of the filter
DECREASES these harmonics. Off one of the transistors of the top pair you see the original
harmonics and off the other one you see the original harmonics in phase. Subtract plus one and
plus one and you get zero.
I'm not going to touch on why the filter breaks into oscillation at high resonance, so don't attempt to
relate adding numbers to oscillation. It's more complex than that. I used the numerical operation to
illustrate the action of differentiating signals in an electrical circuit.
At zero resonance there is no very little output signal fed back to the ladder and therefore minimum
attenuation of bass. But as you increase resonance and increase the signal back to the ladder, the
unfiltered harmonics (IE bass frequencies) gradually cancel out.
That is why the Moog filter (and A6's 24dB filter2) attenuates bass frequencies with increasing
resonance.
Some may consider that an inferior trait because of trying to balance the volume between sounds,
but you can easily get around that by planning ahead by optimizing your patch volume levels.
Optimizing means to program your minimum resonance patches with lower volume levels and
using higher volume levels for those patches that use increasing resonance. That way all your
patches sound even as you select one from the other. I had to do this for years on my Moog
equipment.
Since the Moog filter was patented during the analog synth heydays of the 70s and 80s, other filters
achieved their action by methods other than the transistor ladder and the resonance/bass
attenuation behaves differently, as seen in the Roland System 700 24dB filter, Juno/Jupiter/MKS
24dB filters, and the ARP 4023 12dB filter in early Odysseys.
What is unique about the Moog ladder filter is that the even ordered harmonic distortion is
minimized by the balancing operation of the transistor pairs, a trait that has yet to be duplicated in
subsequent designs. That is why the filter sounds pleasing to the ear when it is overdriven.
Everyone has their tastes and preferences. The quality of a filter is very subjective. Alesis chose
the Moog filter because of its popularity and the Oberheim filter because of its sound and relative
inaccessibility, owing to the small number of SEM systems in the market compared to Moog
systems. As Mike Peake stated, others were considered but rejected for market reasons and for IP
reasons.
On a related note: why do the Moog 904A VCF module and the Minimoog VCF have non constant
Q or resonance? The answer is in the differential amplifer at the output of the filter, where it
amplifies the output pairs off the top of the ladder. I examined the schematic of a Minimoog
prototype (the Model C) and found that Moog had attempted to correct the constant Q problem by
using JFET transistors in the differential amplifier. But when I played that prototype Minimoog at
Audities, the resonance was cold and uninteresting. The 904A and the production
Minimoog use *BJT* transistors which sounded MUCH better but their lower input impedance
loaded down the ladder at low frequencies and caused the resonance to diminish.
So why does the A6 24dB filter's level drop with higher cutoff frequencies? This is another mystery
that occurs when reducing discrete circuits to the VLSI micron level. I don't have an definite answer
to that one. Speculation suggests that the microscopic traces on an IC substrate cannot carry the
current density of lifesize PC board traces, and there's also the inductive/capacitive issues that crop
up with closely paralleled substrate traces. I'm just guessing from my EE knowledge, I didn't study
IC fabrication in depth.