Matrix switch options for modulation routing?

From circuitbending to homebrew stompboxes & synths, keep the DIY spirit alive!

Moderators: luketeaford, Joe., lisa, Kent

Post Reply
schmidtc
Common Wiggler
Posts: 198
Joined: Fri Jan 22, 2010 2:34 pm
Location: Boston

Matrix switch options for modulation routing?

Post by schmidtc » Thu Oct 20, 2011 10:33 am

I'm working on a synthi inspired monosynth, and it time to start thinking about modulation matrix switches. I can only find the pin ghielmetti and sealectro type overseas, and they seem really expensive even used. Plus, I think they use diode pins so I'd need to add some circuitry. Is there a led switch matrix available ala miniature monome?

User avatar
synthnut
Common Wiggler
Posts: 249
Joined: Tue Apr 27, 2010 7:47 am
Location: UK

Post by synthnut » Thu Oct 20, 2011 12:03 pm

Hi,
Well, there is this little beastie if you arereally serious about it:
http://www.hinton-instruments.co.uk/pap ... /index.htm
They aren't cheap, but there's nothing else that I know of that is available in a useable format already.
Of course Mr Hinton is a bit of a Synthi expert too, so it would have some pedigree!

TTFN,
Ben
www.bigbluewave.co.uk & Re:Synthesis

schmidtc
Common Wiggler
Posts: 198
Joined: Fri Jan 22, 2010 2:34 pm
Location: Boston

Post by schmidtc » Thu Oct 20, 2011 12:50 pm

This 4 x 4 from spark fun comes close but all the columns are tied to ground. It could be modified, but it's bigger than I'd like also.

http://www.sparkfun.com/datasheets/Comp ... eakout.pdf

User avatar
Cat-A-Tonic
Lobsters love Muff
Posts: 3734
Joined: Mon Feb 04, 2008 12:03 am
Location: Yokohama, Japan

Post by Cat-A-Tonic » Thu Oct 20, 2011 1:04 pm

4MS Bend Matrix seems like a good option.
I just got a PCB and front panel for this one.
Allows you to sequence your routings. :party:

Aside from that, Hinton Switchmix.
Fewer features, but more compact.

schmidtc
Common Wiggler
Posts: 198
Joined: Fri Jan 22, 2010 2:34 pm
Location: Boston

Post by schmidtc » Thu Oct 20, 2011 1:45 pm

Hey, that might work! My only fear is that if I build a synthi type clone with one of those I will confused the hell out of my self with so many options and features. Sequencing memorized modulation routing is a novel idea, but I'd have to think pretty carefully about how to trim the modulation sources down to only 4. Not sure if i could live with that.

The MT8816 datasheet is inspiring:

http://www.datasheetcatalog.org/datashe ... Xutxzt.pdf

User avatar
Graham Hinton
Super Deluxe Wiggler
Posts: 3293
Joined: Fri Jul 02, 2010 6:28 pm
Location: England
Contact:

Post by Graham Hinton » Sun Oct 23, 2011 5:53 am

schmidtc wrote:The MT8816 datasheet is inspiring:
It does not seem to matter how many times I explain it there is always someone who finds a crosspoint switch data sheet and thinks they can build a patch matrix out of one. You can't, you can only waste time and money trying.

A crosspoint switch is really a set of selectors made from FET switches and has no current limiting apart from the FET Ron and no voltage protection. Turning many on together just shorts everything together, you don't get mixing, summing or multiple output combinations. It is the same as a hardwire matrix with shorting pins, which will not behave like an EMS matrix either, except that a shorting pin may damage outputs shorted together rather than itself.

Apart from a few failed product attempts, the only working product I know of that is based on a crosspoint switch is the SSL X-Patch and that is expensive for what it is. They have hidden its shortcomings with complex software control that only allows "chains" of connections to prevent multiple shorts.

The way to judge how expensive a device like this is is to examine the number of routing functions it offers and the cost per connection. The X-Patch is expensive because it offers less functionality at 16x16 than a passive professional jackfield at 48+48 and it cannot even provide multing. SwitchMix is not expensive because nothing else offers the same functionality at any price. Divide the cost of the product by the number of interconnections to see the true cost per connection and remember "if it only costs money it is cheap".

We have provided several SwitchMix OEM versions for people to build onto their own panels and connectors from £400. There are several new matrices under development which will be announced soon.

Every patching matrix I have ever built for a synthesizer has True Summing which means that you can add, say, a sequencer, a keyboard, a pitchbend voltage and an LFO and they will stay in tune. Then add different combinations to other outputs and they will be in tune too and not interact as you change them. Standard EMS matrices don't do that.

User avatar
Electronic Battle
Wiggling with Experience
Posts: 359
Joined: Fri Mar 27, 2009 12:36 pm
Location: UK
Contact:

Post by Electronic Battle » Sun Oct 23, 2011 12:07 pm


User avatar
ersatzplanet
Synthwerks Design
Posts: 6577
Joined: Fri Mar 06, 2009 1:18 pm
Location: Seattle WA

Post by ersatzplanet » Sun Oct 23, 2011 12:21 pm

Graham Hinton wrote:Every patching matrix I have ever built for a synthesizer has True Summing which means that you can add, say, a sequencer, a keyboard, a pitchbend voltage and an LFO and they will stay in tune. Then add different combinations to other outputs and they will be in tune too and not interact as you change them. Standard EMS matrices don't do that.
This is the key to the whole thing. This is the difference that truly matters. The EMS structure added problems too. The inputs of modules having no attenuation - the outputs had the attenuation. The EMS keyboards had to be re-scaled every time you added a modulator or another VCO. You could easily reach a point where you could not scale it to work.
It is pretty easy to build a passive pin matrix out of 3.5mm jacks ($0.20-0.30 each at mouser) and it will be quite sturdy but as Graham points out it will be nothing more than a large passive multiple with all the problems that passive multiples have. The voltage droop on CV lines is of course the obvious one but the matrix add the danger of easily accidentally connecting outputs together too, which some modules will not like at all.

Also the trouble with those Monome style switch arrays is that they are almost always momentary switches. The toggling action is in software.
-James
-James

James Husted - Synthwerks, LLC - www.synthwerks.com - info@synthwerks.com - james@synthwerks.com
Synthwerks is a proud member of the Mostly Modular Trade Association (http://www.mostlymodular.com).
Always looking to trade for Doepfer P6 cases

User avatar
ersatzplanet
Synthwerks Design
Posts: 6577
Joined: Fri Mar 06, 2009 1:18 pm
Location: Seattle WA

Post by ersatzplanet » Sun Oct 23, 2011 12:41 pm

I have been chasing this holy grail ever since I let my Synthi's go decades ago. Unless you make your synth with no jacks at the modules themselves I see no advantage of using a matrix in a modular. In fact there is the added dis-advantage of having to remember what modules are patched into a row or column and writing it down on the matrix and not changing it. That basically is non-ergonomic for me. Using one as a patch bay for external Efx units and mixers is truly golden though. They can remain patched that way and the workflow time setting up is reduced immensely.
I don't know whether the original poster was designing his system from scratch and not having patchable modules but wiring them directly (and only) to the matrix, but that is the only way it really is worth it to me. I once went as far as designing a large front panel that would house all the separate modules I wanted to use in a matrix system. I was going to use off the shelf modules and just take off the jacks and pots not needed (don't need multiple CV in with a active matrix) and hard wire it to a matrix and mount them to one custom front panel. After figuring the hassle in making even the simplest active matrix I just gave up. The compromises that went into the EMS design gave that machine some of its original character but they are not compromises that anyone would want to make today. I used to be a big advocate of rolling your own matrix after seeing hoe cheaply you could get jacks nowadays, but after seeing the limitations I will be sticking with a bunch of cables, passive and active multiples, and panels with switches on them. If I could afford Graham's switchmix I would buy a few of them though. That is a quality piece of kit there. I would get the version without the jacks and wire it directly.
-James
-James

James Husted - Synthwerks, LLC - www.synthwerks.com - info@synthwerks.com - james@synthwerks.com
Synthwerks is a proud member of the Mostly Modular Trade Association (http://www.mostlymodular.com).
Always looking to trade for Doepfer P6 cases

User avatar
Cat-A-Tonic
Lobsters love Muff
Posts: 3734
Joined: Mon Feb 04, 2008 12:03 am
Location: Yokohama, Japan

Post by Cat-A-Tonic » Sun Oct 23, 2011 6:49 pm

The primary advantage of having a routing matrix is for quick repatching live so that you can switch between very different sound pallets very quickly.

This would potentially allow for complex song structures and even separate pieces in a set-list
that would be awkward/impossible with a modular without a matrix.

User avatar
ersatzplanet
Synthwerks Design
Posts: 6577
Joined: Fri Mar 06, 2009 1:18 pm
Location: Seattle WA

Post by ersatzplanet » Sun Oct 23, 2011 7:52 pm

Cat-A-Tonic wrote:The primary advantage of having a routing matrix is for quick repatching live so that you can switch between very different sound pallets very quickly.

This would potentially allow for complex song structures and even separate pieces in a set-list
that would be awkward/impossible with a modular without a matrix.
I think that would be easier with a "Presto patch" arrangement. There are many edge-card PCB prototype boards out there and for a live patching situation where "presets" are desired, a presto-patch layout is easier. Get a zero-insertion edge connector and wire it to a ton of sockets. The PCB programming card could even have trimmers and performance switches built into it. It would be like patching a bunch of cords at once. Pretty cheap too.
-James

James Husted - Synthwerks, LLC - www.synthwerks.com - info@synthwerks.com - james@synthwerks.com
Synthwerks is a proud member of the Mostly Modular Trade Association (http://www.mostlymodular.com).
Always looking to trade for Doepfer P6 cases

User avatar
Orbless
Super Deluxe Wiggler
Posts: 1202
Joined: Wed Jan 12, 2011 2:57 pm
Location: San Francisco, CA

Post by Orbless » Sun Oct 23, 2011 11:21 pm

I think that would be easier with a "Presto patch" arrangement. There are many edge-card PCB prototype boards out there and for a live patching situation where "presets" are desired, a presto-patch layout is easier. Get a zero-insertion edge connector and wire it to a ton of sockets. The PCB programming card could even have trimmers and performance switches built into it. It would be like patching a bunch of cords at once. Pretty cheap too.
Can you go further with this idea? I am unfamiliar with what you are referring to.

User avatar
ersatzplanet
Synthwerks Design
Posts: 6577
Joined: Fri Mar 06, 2009 1:18 pm
Location: Seattle WA

Post by ersatzplanet » Mon Oct 24, 2011 12:26 am

Orbless wrote:Can you go further with this idea? I am unfamiliar with what you are referring to.
Basically the Presto-patch is a system used by the MkII versions of the AKS Synthesizers. The rows and columns of the patch matrix were brought to the pins of a 32 pin PCB socket. Then a little PCB with the connections that were originally on the pin matrix are wired on it and when it is placed in the socket it re-creates the pins in the matrix. They looked like this:

Image

The idea for a module version would be a PCB socket wired to jacks (or directly to the jacks in the modules to be "patched"). Then you can make the interconnections on a prototyping PCB and plug it in the socket to make the connections. You can cut down something like this:

Image

Actually for something like this any multi-pin connection solution will work to do something like this - connecting a bunch of wires a certain way quickly. The "presto patch" like this can be expanded by putting switches and pots on the PCB to add flexibility. The cost can be cheap enough. It is not as flexible as a full matrix but it works well as a patch recall system.
-James

James Husted - Synthwerks, LLC - www.synthwerks.com - info@synthwerks.com - james@synthwerks.com
Synthwerks is a proud member of the Mostly Modular Trade Association (http://www.mostlymodular.com).
Always looking to trade for Doepfer P6 cases

User avatar
RealDudes
Lives in a Dudeplex
Posts: 1705
Joined: Wed Feb 23, 2011 2:23 pm
Location: Los Angeles

Post by RealDudes » Mon Oct 24, 2011 12:36 am

I think I got my next project here
thanks james!
the most important track you'll ever make is the next
https://soundcloud.com/doom_asylum
https://soundcloud.com/wade_blazer

User avatar
synthnut
Common Wiggler
Posts: 249
Joined: Tue Apr 27, 2010 7:47 am
Location: UK

Post by synthnut » Mon Oct 24, 2011 4:13 am

Hi,
Don't forget the point that Graham made was that it's not just a matter of hard-wiring points together in a summing matrix, there needs to be the element that allows the summing! I know there is a discussion about the output impedance of modules allowing mult sto be used as mixers, but the same problem will be encountered here too. With the Synthi's matrix the resistor in the pin allowed summing of sorts, but that this was problematic for things like CV! It's not too dificult to acheive, but making one that is actually accurate is that much harder.
I suppose it depends on if you want the idiosyncrasies of the Synthi or a more predictable result that can be used with pitch CVs without inducing madening problems. This is perhaps one reason why there is so little melodic use of the Synthis!

TTFN,
Ben

www.bigbluewave.co.uk & Re:Synthesis

User avatar
Graham Hinton
Super Deluxe Wiggler
Posts: 3293
Joined: Fri Jul 02, 2010 6:28 pm
Location: England
Contact:

Post by Graham Hinton » Mon Oct 24, 2011 5:28 am

synthnut wrote: Don't forget the point that Graham made was that it's not just a matter of hard-wiring points together in a summing matrix, there needs to be the element that allows the summing! I know there is a discussion about the output impedance of modules allowing mult sto be used as mixers, but the same problem will be encountered here too.
The irony is that if a summing matrix is executed properly passive mult mixing is redundant.
With the Synthi's matrix the resistor in the pin allowed summing of sorts, but that this was problematic for things like CV! It's not too dificult to acheive, but making one that is actually accurate is that much harder.
The standard EMS matrices constitute a weighted average set by the ratio of the resistor pin to the input impedance of the module being patched to and which varies. On a VCS3/Synthi A the pins are 2k7 which is almost like passive mult mixing. On the larger Synthi 100 the pins were 10k and this resulted in very dull mixing of audio, after three or more pins were inserted the sound became indistinct and muddy and the crosstalk atrocious, it was virtually unusable. With averaging adding more instances of the same signal gives no gain, with different signals/CVs what you get is what you get, but not what you want.

The small 16x16 matrices almost get away with it and the user interface allows patching that could not be done easily otherwise. It is still better than all the other synthesizers that have attempted to use matrices with shorting pins, but falls a long way short of one with my buffering mods which does what you expected it to do in the first place.

Another important consideration is repeatablity. This requires very high precision resistors, not the 5% and 2% tolerances that EMS used. We hand grade 0.1% resistors which we can do as we buy them in bulk, but it is not feasible to do for a one off project.
I suppose it depends on if you want the idiosyncrasies of the Synthi or a more predictable result that can be used with pitch CVs without inducing madening problems. This is perhaps one reason why there is so little melodic use of the Synthis!
The people that used to do melodic work on them had them heavily modified. That was the attraction. Synthis were a low cost synthesizer that were very easily adapted at a fraction of the cost that ARPs and Moogs were going for at the time. Now collectors have put the prices up beyond reach and are afraid to make any changes in case it alters "The Sound"--whatever that is.

User avatar
ersatzplanet
Synthwerks Design
Posts: 6577
Joined: Fri Mar 06, 2009 1:18 pm
Location: Seattle WA

Post by ersatzplanet » Mon Oct 24, 2011 10:56 am

Graham Hinton wrote:Synthis were a low cost synthesizer that were very easily adapted at a fraction of the cost that ARPs and Moogs were going for at the time. Now collectors have put the prices up beyond reach and are afraid to make any changes in case it alters "The Sound"--whatever that is.
This is very true. I didn't pay more than $500 for any of the three I owned. That was 1975-80 $500 though which was worth more than it is now. They looked cool but if you were a performing musician the goal was to be able to do covers of the pop hits and the EMS was not for that without a LOT of work. I was doing classic EM and "experimental" music at the time and they were perfect. I did some sound effect work for radio ads with them too. So easy to carry a AKS to a gig. Really miss that.
-J
-James

James Husted - Synthwerks, LLC - www.synthwerks.com - info@synthwerks.com - james@synthwerks.com
Synthwerks is a proud member of the Mostly Modular Trade Association (http://www.mostlymodular.com).
Always looking to trade for Doepfer P6 cases

User avatar
Bergfotron
Common Wiggler
Posts: 159
Joined: Tue Aug 23, 2011 8:24 am
Location: Sweden

Post by Bergfotron » Mon Oct 24, 2011 11:20 am

synthnut wrote:Hi,
Don't forget the point that Graham made was that it's not just a matter of hard-wiring points together in a summing matrix, there needs to be the element that allows the summing! I know there is a discussion about the output impedance of modules allowing mult sto be used as mixers, but the same problem will be encountered here too. With the Synthi's matrix the resistor in the pin allowed summing of sorts, but that this was problematic for things like CV! It's not too dificult to acheive, but making one that is actually accurate is that much harder.
I suppose it depends on if you want the idiosyncrasies of the Synthi or a more predictable result that can be used with pitch CVs without inducing madening problems. This is perhaps one reason why there is so little melodic use of the Synthis!

TTFN,
Ben

www.bigbluewave.co.uk & Re:Synthesis
I just want to mention that the AMORE modules have all their inputs as summing nodes. So if you have a suitable resistor (100 k or higher) in each pin/plug, you could patch any modulation sources to each input.
I think, however, that you should leave the signals that should track 1 oct/volt out of the matrix.

User avatar
Graham Hinton
Super Deluxe Wiggler
Posts: 3293
Joined: Fri Jul 02, 2010 6:28 pm
Location: England
Contact:

Post by Graham Hinton » Tue Oct 25, 2011 4:37 am

Bergfotron wrote: I just want to mention that the AMORE modules have all their inputs as summing nodes. So if you have a suitable resistor (100 k or higher) in each pin/plug, you could patch any modulation sources to each input.
I think, however, that you should leave the signals that should track 1 oct/volt out of the matrix.
Why?
I had a passive matrix interfaced with summing node modules 40 years ago, the whole point was that pitch CVs could be handled correctly. It requires matching accuracy to better than 0.1% and that means in the module nodes as well as the pins. This is not difficult if you design the whole system, but there will be errors when "foreign" devices are introduced, which is why I now favour active matrices with full control of the drive and summing.

A common mistake that I see people making now is driving a matrix from 1k output modules or directly off an attenuating pot.

User avatar
Bergfotron
Common Wiggler
Posts: 159
Joined: Tue Aug 23, 2011 8:24 am
Location: Sweden

Post by Bergfotron » Tue Oct 25, 2011 9:53 am

Graham Hinton wrote:
Why?
Well, if you look at what really needs to track 1 oct/volts, it's usually oscillator pitch. Does it really make sense to route the note signal to the oscillators through a matrix? I would say no. Of course, you can introduce frequency modulation of different types through the matrix. But the note data? Why?

User avatar
ersatzplanet
Synthwerks Design
Posts: 6577
Joined: Fri Mar 06, 2009 1:18 pm
Location: Seattle WA

Post by ersatzplanet » Tue Oct 25, 2011 11:15 am

Bergfotron wrote:
Graham Hinton wrote:
Why?
Well, if you look at what really needs to track 1 oct/volts, it's usually oscillator pitch. Does it really make sense to route the note signal to the oscillators through a matrix? I would say no. Of course, you can introduce frequency modulation of different types through the matrix. But the note data? Why?
You can also only match the resistors on a few columns dedicated for VCO inputs - no need to do the whole thing, but your right. The majority of pitch oriented patching would be from the source directly to the VCOs. All other modulation inputs can go through the matrix. You would just need a precision summer for the VCO if you want more than one source (sequencer with transposing keyboard comes to mind)
-James

James Husted - Synthwerks, LLC - www.synthwerks.com - info@synthwerks.com - james@synthwerks.com
Synthwerks is a proud member of the Mostly Modular Trade Association (http://www.mostlymodular.com).
Always looking to trade for Doepfer P6 cases

User avatar
Graham Hinton
Super Deluxe Wiggler
Posts: 3293
Joined: Fri Jul 02, 2010 6:28 pm
Location: England
Contact:

Post by Graham Hinton » Tue Oct 25, 2011 2:07 pm

Bergfotron wrote: Well, if you look at what really needs to track 1 oct/volts, it's usually oscillator pitch. Does it really make sense to route the note signal to the oscillators through a matrix?
Yes. I've been doing it for 40 years.
I would say no. Of course, you can introduce frequency modulation of different types through the matrix. But the note data? Why?
Because if your patching system is a matrix then that is the obvious and most desirable way to do it. Most modular VCOs now only have one straight 1V/oct input and the others go through pots and do not track so well. Certainly no analogue keyboard synths or semimodular voices that I can think of have more than one external pitch input. It is easier to precision sum and go into one calibrated input and preferable to see the various combinations available on the matrix.

If you haven't used matrices maybe you are not thinking in terms of how many pitch sources you may want to sum or how many destinations you may want to track and what you could do if you had more.

There is a difference between saying "I don't need a high performance sports car" and "nobody needs a high performance sports car".

User avatar
robotmakers
Veteran Wiggler
Posts: 658
Joined: Fri Apr 09, 2010 10:52 am
Location: New York, NY

Post by robotmakers » Tue Oct 25, 2011 2:43 pm

Just wanted to mention that the Hordijk Active Matrix is a solid 5U implementation of the matrix concept which I use for modulation routing. This thread has a nice picture of my mod matrix with labels and everything!

https://www.muffwiggler.com/forum/viewtopic.php?t=38605&highlight=matrix

Cheers,
Roger

schmidtc
Common Wiggler
Posts: 198
Joined: Fri Jan 22, 2010 2:34 pm
Location: Boston

Post by schmidtc » Fri Dec 09, 2011 1:31 pm

Thanks for all the suggestions, and sorry to revive this tread after so long. It's especially nice to have an expert opinion from Graham.

I decided 2.5mm jacks were the most economical way to go (if I consider my time effectively worthless :zombie:). This kind of limited my options, but even with a 14 x 15 matrix there's still more than enough to work with. I'm planning to use 10K .1% summing resisters. The V/oct path and much of the audio path will be normaled and won't go through the matrix at all. Even with <1K output impedance, and >100K input impedance, I suspect Graham is right and the drop of added modulations may start to bother me. If this becomes a problem I can simply add buffers (on both ends so not to invert the signal). The OP27 seems the best balance of speed, bandwidth and offset that I happen to have in the drawer. Grahams mod for the synthi is the same design.

http://www.hinton-instruments.co.uk/ems ... ended.html

The panel is about half done. Here's a picture of the routing:

Image

User avatar
delayed
Veteran Wiggler
Posts: 546
Joined: Wed Oct 07, 2009 10:14 pm
Location: Nashville TN

Post by delayed » Tue Apr 01, 2014 4:50 am

How did this turn out?

Post Reply

Return to “Music Tech DIY”