[sdiy] Triple oscillator cores?

[cheater cheater]

Hey guys,

Following my previous thoughts on the dual oscillator cores, I started
considering what would happen if I wanted to do a multi-integrator
design of a triangle core. I think what would be needed is three
oscillator cores. Here's what I see could be happening:

Call the cores A, B and C. They all go from 0 to 1. A starts out at 0,
B starts out at 1, C starts out at 0. They all start out inactive
(i.e. their level is constant)

The topology looks like this:

pos current source ->
neg current source -> (selector -> VC Resistor JFET -> integrator ->
VC Resistor JFET)x3 -> mix -> out

Each of the cores has a selector which flips from the positive to the
negative current source. This is our slow element.

Core A starts out with the selector on positive current, B starts with
negative current, C starts with positive current. Of course the
selected current does not do anything because the resistor is
essentially open.

1. Core A is activated and it starts integrating towards 1.
2. When A reaches 1, it is deactivated by removing current from its
input by turning up the VC resistor.
3. Core B starts integrating with negative current. This is the part
most important to timing. Once this is ensured, core A's output is
switched off and core B's output is switched on at the same time. At
the same time A's selector is flipped to negative.
4. Core B reaches 0. It is frozen by opening off the input resistor
and C is started. Then the output of B is attenuated while the output
of C is mixed in using the output VC resistors. Then B's selector is
flipped to positive current.
5. Core C reaches 1 and is stopped, Core A starts integrating towards 0.

and so on.

Again the selector can be slow, it doesn't matter. Essentially the
important part of switching happens on the output of the core instead
of on the input of the core, which means that the switching mechanism
is not at risk of shorting a negative and positive voltage rail
together, therefore - my intuition suggests - it can reach higher
speeds.

I thought that this sort of design shouldn't have a problem with FM,
because the integrator used previously to the current one would always
be available to take over when the current integrator reaches the
discontinuity - however the problem is that at this point that
previous integrator will have already switched directions.

This brings me to the question of why not do this with a saw oscillator core?

Again we have integrators A, B and C going from 0 to 1. They all start
out at 0 with the input resistor open.
This is the topology:

current source -> (VC input resistor -> integrator -> VC output
resistor)x3 -> mix->out

1. Core A's input resistor is shorted and the core starts integrating.
2. Core A reaches 1 and its input resistor is switched open. Core B's
input resistor is shorted and integration starts. The output of core A
is attenuated and core B is mixed in. At that point we don't want to
reset core A yet.
3. Core B reaches 1 and its input resistor is switched open. Core C's
input resistor is shorted and the core is started. The output of core
B is attenuated and core C is mixed in. At this point core A is still
frozen at level=1. Now we reset it.
4. Core C reaches 1 and its input resistor is switched open. Core A's
input resistor is shorted and the core is started. The output of core
C is attenuated and core A is mixed in. At this point core B is still
frozen at 1, and now we reset it.
GOTO 2

This sort of workflow seems more suited for through-zero FM, since the
'previous core' is always there ready to start being emptied once the
current core becomes completely empty and reaches the discontinuity.
It would seem that this could work much better than trying
through-zero with a single-core triangle oscillator, because the
traversal of the discontinuity could be, as a guess, much quicker.

What do you guys think about this sort of design? What would be the
biggest problems with this? Do you think that coordinating all those
VC switches could be a bit difficult?

Also: Essentially, maybe this sort of triple-core design for the saw
oscillator could be reflected for the triangle oscillator in a
hex-core design.

Cheers
D.