Ensemble Circuit Configuration Questions

[Sean Costello]

"Paul R. Higgins" wrote:

> Is it possible for someone on the list to post this 120 degree LFO circuit (from
> the Korg Polysix)?  I'd love to see the schematic.

Go to http://users.bart.nl/~brun/p6schem.htm. KLM368 is the ensemble
section of the Polysix.

> Has anyone tried using quadrature LFOs with phasers or analog delays?  

I'll give it a try in Csound, and post the results, when I have time -
I'm pretty busy assembling the site for the instruments and compositions
I have. I could put a few sound files on the "ensemble" page I am
working on, as well as the relevant orchestras/scores. I'll also try out
the Dimension D chorus algorithm, and any others that seem interesting.

> I had no idea that the Dimension D was such a simple design--Roland really makes
> it sound like there's all kinds of weird stuff going on inside. 

Marketing. Roger Meyers described the workings of his Octavia in an old
Guitar Player magazine; he made it sound like he had put a Moog 55 in
there. It's a full wave rectifier, for crying out loud.
> 
> Would there be any advantages to mixing in the dry signal with the delayed ones?
> It seems to me that you would get even more of an ensemble effect that way, with
> more complex phase cancellations.  I imagine that sound quality would also
> improve; I've been building analog delays for years, and none of the ICs I've
> worked with (mostly the EG&G/Reticon BBDs) had superb fidelity even with short
> delays.

I tried mixing the dry signal in with Csound, and the result was a far
more boring sound. With only the delay outputs mixed together, every
signal will have a nice vibrato on it; mixing in the straight signal
will add in a boring static component. The phase cancellations of the
delay lines with the straight signals will not be nearly as interesting
as the cancellations of the delay lines with each other. As the delay
lines will be moving back and forth around a set delay time (around 9
msec, wavering from 3 msec to 15 msec), there will be lots of complex
"through-zero" flanging as the delay lines move past each other. With a
straight signal mixed with a delay line, there will never be
"through-zero" flanging, as the delay line will never be "perfectly" in
phase (or out of phase) with the straight signal.

Of course, working in Csound eliminates most of the noise and aliasing
problems with analog delay lines, although the linear interpolation used
by delay lines in Csound causes its own problems. In the analog devices
I use that feature ensemble effects (Moog Opus 3 and Korg Polysix),
there is a certain amount of noise added by the effect. Still, I find
that this is not that big of a problem in the ensemble setting, as long
as the noise is controlled by a compander-type situation.

In my opinion, the main disadvantage of leaving out the straight signal
is the loss of high frequency definition.  In the Polysix, switching
from "Chorus" or "Phaser" to "Ensemble" results in a noticable loss of
high frequencies. This is probably due to the antialiasing filters on
the input of the BBD lines. Using longer delay chips (say, 1024 stages
instead of 512) would allow you to double the sampling rate of the BBD
chips for the same delay time, allowing for more of the high frequencies
to be present in the signal. The price difference between 512-stage BBDs
and 1024-stage BBDs seems to be small enough (less than $1.00 from
Digikey in single quantities) that it would be worth using the longer
BBDs in order to get higher sampling rates, and more high frequencies in
the output. This might add more noise, though (opinions, Synth-DIY
gurus?).


> Using time delay techniques would, I think, give a less electronic "swooshy"
> kind of effect, and a more convincing ensemble quality.  I'm sure that many list
> members have heard this before, vis a vis the phaser vs. flanger debate.  (Uh
> oh, I feel some flames coming on...).  Of course, using VC allpass stages based
> around an OTA is no doubt much simpler and cheaper.

I personally prefer phasers to flangers, as I feel that the
comb-filtering of flanging (where every notch is a multiple of some
frequency) is less interesting sounding than phasing (where there are
less notches, but they are not simple multiples of each other in
frequency). However, this is based on hearing a single delay line mixed
with the original signal, versus hearing a single allpass filter mixed
with the original signal.  The allpass-based vibrato units I have heard
before do not sound nearly as smooth as the delay-based vibrato units.
As I feel that a nice vibrato in the delayed signal is the first key to
fine-tuning the ensemble sound, this may mean that allpass based
ensembles aren't that good of an idea. Still, I will work on an
allpass-based version in Csound. The tricky part of this is getting more
precise control over the phase shifting frequency of the allpass filter;
I'll be working on this soon, as having linear and/or exponential
control over the phase shift frequency is essential to getting a nice
vibrato sound. 

Sean Costello