Matrix mixer expanded edition

OK. I’m finalizing the design for the expanded edition.

Current features:

Each of the 3 columns has:
3 normal channels
1 inverted channel
switchable gain setting for the normal channels and the inverted channels

6 inputs jacks, 3 output jacks
15 potmeters total
HP: unclear as for now. bigger than 14HP

There are 3 potmeters that control overall gain (currently only the normal channels, not the inverted)
I might change this and add a attenuverted channel instead of the inverted one. I think I can then set the overall gain with one pot for all these channels
I might be tempted to make all channel’s attenuverters. but this may not be ideal for zeroing out certain channels.

Are there features you want added?

There was some talk about chaining. please give a detailed setup idea.


I love that each channel will have its own inverted input!

is there any kind of normalization between the inverted channels? So if you plugged a key into inverted R input would it cascade across to the others?

I’m not really worried one way or another about attenuverters.

The overall gain pots seems like a nice added bonus

RGB pre patching?

slightly different take on chaining

it could be really interesting to have a system for pre patching RGB signal chains between modules behind the rack.
So if you have a matrix mixer module and an output module with this pre patching setup you would have a cable behind the front panels connecting them together. So if you patched a RGB video from visual cortex into the matrix mixer it would automatically go to the output module even with nothing patched on the front panel. However if you patched with the matrix outs this normalization would be broken.

How I'm using chaining

RGB out from Memory Palace > Polar fringe > your matrix mixer > visual cortex channel

this allows me to get the positive and negative keys from the PF while also passing through the original unprocessed signal. I do some coloring with your mixer and then send to the cortex for compositing vs another channel

it is helpful to reduce the amount of multing that has to happen in a system.
I would have to mult the RGB signal off before sending it around otherwise.

so for a matrix mixer it would have a set of pass thru RGB outputs

I 100% get it for the polar fringe. I could see it being nice for a matrix mixer but not essential.


Ooo! exciting news…

Martijn, do you have a rough sketch of the panel? It helps me to wrap my head around the options. My initial reaction is that I’d prefer attenuverter channels to the overall, post-mix RGB gain.


yes, that is possible. good that you mention it :slight_smile:

I don’t know how to achieve this, as my output module does not have inputs on the back. does a VC has this?

So you’d want 2 sets of outputs , 6 instead of 3?
I can buffer these, or give them a resistor at least.

There is more space on the panel, so this could be possible

something like this. note that this design is very preliminary.
I’m still deciding on the type of switches (low height slide switches would be preferrable)
and the pots / jacks / switches location, width of the panel etc.

1 Like

Oh yeah, I also added a 1v reference input to the jacks
to do matte color mixing.
The previous Matrix mixer design was capable of doing this, but a 1v ref voltage will make it more precise. I have to think about the normalisation of this feature.

option 1: jack 1 copies the 1v to the next input. when a jack is inserted, the 1v is removed from the whole chain

option 2: each jack has a jumper to set it to 1v or previous input normalisation.


the visual cortex expander kind of does this but it outputs RGB on the back (it has solder points there isn’t a header installed).
the difference is the expander doesn’t work if it isn’t plugged in to the cortex.
but if instead of a cortex expander those RGB signals got sent to the matrix mixer you could patch with “nothing” plugged into the front panel.

the idea just popped into my head while I was reading this thread
it could make for an extremely tidy system. I know with the expander you have to have it sit right next to the visual cortex or else there is some potential noise that creeps in so maybe this is just a pipe dream.

I’m mostly just thinking out loud with this idea :slight_smile:

with chaining modules if it had 6 outs that could be neat
“channel 1” would be the regular output that is processed through the module
“channel 2” would be an unprocessed copy of whatever comes in at the input

this could be useful for that wonderful parallel processing that really makes video work sing!

1 Like

With chaining I think of connecting several modules together.
If I add 3 outputs that copy the input signal, I have to rethink the normalisation option…

What is the user case for this?
Is it useful to set the whole 3 channel mix to -1v?
I see this mixer as a possible output module, so a voltage of 1 to 0 to -1 is not really usable.
Normally one would only use 0-1v , right? (or maybe a bit higher to get some clipping / softkeying)

1 Like

let me rephrase that:

“Rather than a post-mix RGB gain, I’d rather have attenuverters on the 6 inputs”

I hope that’s clear?

In other words, for me the downstream RGB gain pots are redundant, and less useful than attenuverters. The downstream post-mix RGB gain pots remind me of the LZX Color Encoder and the inverted channel reminds me of the first edition of the Video Blending Matrix. With attenuverters for each input, you don’t really need either the invert channels nor the post-mix gain.

ok. but if you have attenuverters on all channels, you do not need the ‘inverted channels’
so 3x3 matrix with attenuverters = 9 attenuverters

I have to think on this.

I rather like the idea to have downstream gain , but maybe that’s just me :slight_smile:
It would act as a macro control for color mixing / feedback settings



Hi Martijn, any idea when the pcb’s will be ready?

not yet, as I am still figuring out the feature set.
but I expect somewhere next month


ok. but if you have attenuverters on all channels, you do not need the ‘inverted channels’

correct, all channels could be attenuverted, making the dedicated invert channels unnecessary

downstream gain

personally, for my work flow I would balance all the input channel levels with the 9 individual attenuverters. in your example, the feedback input levels can already be controlled if every channel has an attenuverter.

for downstream color tweaking, I usually prefer ‘proc amp’ style controls (brightness/contrast/hue/gain) or something much more sophisticated like R/G/B and Gain for separate luminance bands (as with the Videotek SDC-101) or even software based color correction tools (Look-Up Tables in VDMX, Red Giant filters, etc.)

However, I love the idea of one downstream master attenuverter knob to affect all 3 channels, in order to fade to black or white like the Roland V4


I did a simulation in Falstad. looks like this works.
the only thing I am not sure about is the effectiveness of the global gain pot, as it might be difficult to set it to zero (because of the both positive and negative input) - should this be an attenuverter too?)
I have to test the gain switches, which are now located before the non-inverting amplifier.

and should I add some sort of clipping? max voltage is now +9v if all pots are up and the switch is set to 3x amp

Might i suggest some sort of “soft” clipping past ±1V?
This would make no difference as long as you stay inside the limits but would be an interesting effect you can access by intentionally going outside the limits (e.g. by turning on 3x on an already full-scale signal)

thanks for your reply. yes , you might suggest! I’m learning a lot while designing this :slight_smile:
what kinds of soft clipping are there?

The 3x gain has cool effects, but more is not useful

1 Like

I think most approaches have some sort of amp with a resistor in parallel with antiparallel diodes in the feedback network. I can’t find a good example right here/now, but i can try to sketch something after work


hard clipping:

1 Like

Yesterday turned out busier than planned, but something along these lines should work:

The general idea is that at low output voltages R2 dominates the response giving a mostly linear output, but as the feedback current reaches into the milliamp range the diodes will increasingly absorb a larger portion.


This circuit is inflexible in clamp level since it is a property of the diodes, so you’d have to have another amp at the output to adapt it.

Alternately, i think you can get it to have a decent response clipping at around 1V by using two sets of antiparallel diodes in series instead of just one.

The ratio of R2 to R1 sets the gain in the linear segment, and their absolute sizes should change the “softness” of the clip

Edit: My brain is broken today - but on second thought i’m pretty sure the absolute sizes just shifts the cut-off rather than affect its shape.


I’ll test it in the simulation.
So if I understand correctly: this should be placed as 2nd stage in the matrix mixer? (after the summing resistors).
or not?

1 Like