[sdiy] multimode matrix filter question

[David G. Dixon]

> > I think you've misunderstood what I'm doing.  As you say, the LPF output
> > mixing must be a fairly precise thing in order to obtain the various
> derived
> > modes, but this will be based on hand-selected 1% resistors into
> individual
> > summing amps for each derived mode.
> 
> No, I understood what you have meant. Now think about these situations:
> 
> 1. someone wants to recreate those special modes outside of your box.
> So they would take a mixer, they would take say 2 or 3 outputs from
> your filter, and then they would just want to put all faders at the
> same level and have the expected mode at the mixer output. Then they
> would start playing with the faders, to animate the sound a little,
> always having the simple option of going back to all faders at the
> same level.

In that case, they'd have to build their own matrix interface.

Concerning the rest of your message, these are all very good and thoughtful
comments, and the short answer is this:  For the board I have designed, the
gain of each derived mode is fixed by a single summer feedback resistor.  To
put a switch on any or all of these would dramatically complicate the board
and defeat the purpose.  However, there are potentially good reasons to have
various gain levels for certain outputs, as you have pointed out.  In that
case, I would simply create those certain outputs twice, at two different
gain levels, and bring both options out to a separate jack, or to a switch
and a single jack.  By my reckoning, there are about 26 meaningful filter
modes (in addition to the original 4 LP modes) which can be derived from a
cascaded-stage 4P LPF, and only about 6 of those may require gain
compensation.

In any case, I'm obviously going to have to do some experimentation.
However, the nice thing about my board design is that one could choose
whichever modes one found to be the most useful and only put those on the
board -- the specific modes are not predetermined, but each board can
accommodate 18 different ones, including replicas of the same mode with
different output gains.


> 2. someone has two of your filter and he shifts one of them up a
> little in frequency from the other one. He sets up one filter which is
> something that's say a mix of low pass and band pass, and then sets up
> just band pass on the other one, and inverts it. Both filters have the
> same input. He wants this to happen: 1. set both faders to the same
> level. The filters are playing together and it's a nice sound etc.
> Then you reduce the offset of the second filter down to 0 and the
> mixer output should pretty much only have the low pass output. A nice
> phasing sound happens until the filters totally kill each other's BP
> outputs. This is a very common kind of trick that you do with
> multimode filters and it's difficult to do if the filter output gain
> has been manipulated with. What you want in this case is gain
> compensation OFF.
> 
> 3. someone has just one filter and he just wants to use some filter
> mode. He turns it on and he wants it to have the right gain, like all
> the other filters. What you want to do is you want gain compensation
> ON in this case.
> 
> 4. Someone has one of your filter. He takes the 2p LP + 2p HP mode (if
> there is such a mode, it's just for illustrative purposes) and the 2p
> HP mode. He inverts the 2P HP mode and sets it to gain 0dB on the
> mixer. The notch mode is set to -inf gain on the mixer. He starts
> turning up the notch mode until unity gain (0 dB) and he expects the
> HP mode to be cancelled out and only the LP mode to stay. In this case
> you want gain compensation OFF.
> 
> 5. Someone has one of your filter and he cycles through the modes with
> a round-robin. Say the JH interpolating scanner, but anything else
> will do. He puts a long-tail reverb on it, 80% wet. He expects there
> to be some animation but in general the spectral levels should be the
> same as the input, because the filter modes he selected should add
> together to the "whole sound". For example he cycles through 1p lp, 1p
> hp, 2p lp, 2p hp, 4pBP, 4p NP. They should all add up to 3x gain. What
> you want in this situation is gain compensation OFF.
> 
> 6. Someone has one ofy our filter and he's performing with his right
> hand and he's changing modes with the left hand as part of his
> performance. He doesn't want his volume to drop abruptly for no reason
> at all, that would destroy his performance quite badly. What you want
> in this case is gain compensation ON.
> 
> 
> Cheers,
> D.
> 
> 
> >> > If it's just a question of gain, a post-module mixer would do it.
> >>
> >> it won't be possible to set it easily at all. It's a delicate
> >> parameter and requires a lot of precision when mixing filters or
> >> filter modes together. I would go as far as to have a separate switch
> >> for each output, or two jacks for each output, but that's just me. If
> >> you're going to make this a separate PCB, then maybe having a special
> >> place for gain markup and a 2-way switch for each jack, and a
> >> pass-through, is a good option; the switches just supply a voltage,
> >> say 0 or 5V. This way you could have multiple configurations:
> >>
> >> 1. just outputs with no gain make up
> >> 2. just outputs with gain make up
> >> 3. one panel with gain make up, and using the passthrough, another
> >> panel without gain make up is connected
> >> 4. one panel with gain make up or not for each jack, with separate
> >> switches for each output
> >> 5. one panel with gain make up or not for all jacks at the same time,
> >> with one switch supplying voltage to all gain make up stages
> >> 6. one panel with gain make up or not for a few jacks at the same
> >> time, e.g. gain make up for 2p outputs, gain make up for 4p outputs,
> >> gain make up for 1p outputs, ...
> >>
> >> A simple thing like this can leave a lot of reconfigurability to the
> >> user, people love it when they can do stuff like that. I know I do!
> >>
> >> And then, the user can also do this extensibly. For example:
> >> 1. start out with just the 18 jack outputs and have all outputs set to
> >> no gain. Later drill a hole somewhere (either in the "outputs" panel
> >> or the "VCF control" panel) and add a single switch)
> >> 2. later add a panel with 18 switches. Plug the hole for that old
> >> switch and pretend it's not there, or put a nice blinking diode there,
> >> or something.
> >> 2b. add a panel with 18 switches. Replace the original switch with a
> >> three-pole and make it AND/OR with the other switches (can be done
> >> without gates, just by routing the voltages the right way). Up means
> >> "all outputs have gain compensation", center means "each output has
> >> gain comp according to its switch", bottom means "no output has gain
> >> compensation".
> >> 3. Start out with just the 18 jack outputs and add another panel later
> >> with 18 jacks with gain compensation
> >> 4. Start out with just the 18 jack outputs and add another panel later
> >> with 4 jacks for gain-compensation of different kinds of outputs
> >>
> >> ...
> >>
> >>
> >> Cheers,
> >> D.
> >>
> >> On 19/10/2010, David G. Dixon <dixon at interchange.ubc.ca> wrote:
> >> >> It's enough of an important function to have it. With complicated
> >> >> multimode filters it's important to be able to run multiple of them
> in
> >> >> parallel without having any funny issues with the levels etc; for
> >> >> example you might want to rely on this when doing an all-pass filter
> >> >> out of multiple such filter units and then changing their levels
> >> >> somehow so that they animate in the music.
> >> >
> >> > If it's just a question of gain, a post-module mixer would do it.
>  Also,
> >> I
> >> > can pull as many outputs from each stage as I want, because they're
> all
> >> > buffered separately.  Ditto with the 4P LPF outputs.  Hence, one
> could
> >> use
> >> > as many of these matrix filter interfaces as one wanted, all driven
> from
> >> the
> >> > same lowpass filter.
> >> >
> >> >> On the other hand it's also important to be able to switch or
> >> >> crossfade between the different filter modes without gain issues
> which
> >> >> can be very unmusical.
> >> >
> >> > All possible filter modes will be available simultaneously, so
> >> > crossfading/switching could all be done downstream in a separate
> >> > mixer/scanner.
> >> >
> >> >> It's just one switch on the front panel. And it's 1u already..
> >> >
> >> > 18 jacks spaced 7/8" apart on a 1U MOTM panel take up all the
> available
> >> > space (not counting 1/2" on either end for mounting).
> >> >
> >> >> If you don't want to do it, at least put a location on the PCB for
> that
> >> >> switch.
> >> >
> >> > Hmmm.  Maybe.
> >> >
> >> >
> >