Midi to CV design -- Soliciting advice

[elkayem]

I am currently building the MFOS Ultimate and Ultimate Expander, and would like to add a Midi to CV controller to the box, so I can use my keyboard with it.

I am designing the Midi to CV out of an Arduino board, a few 12-bit DAC chips, op-amps, and other miscellaneous parts. The nice part is that I have quite a bit of flexibility on how I design it. However, I have no direct experience (yet) with analog synthesizers, so it is hard for me to judge what I will really need.

I was thinking I would include the following outputs:
* CV note output (top note, 1V/oct)
* Gate
* Trigger
* Pitch bend (using pitch wheel on my keyboard)
* Pedal

Are there any of the above that I should omit? Anything I haven't considered that I should add? As an alternative, I was also considering adding multiple CV note outputs so I could drive multiple VCOs, but I assume this means I would also want to add multple gates and/or triggers for each output.

Any tips? What features are key vs not useful? I'm sure there are plenty of others who have assembled similar things, so I'm also interested in any lessons learned.

[elkayem]

I should add... I also am thinking about velocity as another CV output, for a total of four CV outputs: note, pitch bend, pedal, and velocity. Too many?

[JovianPyx]

A dsPIC33F (Microchip) has multichannel 12 bit ADC built in. May be a better choice than Arduino. It can be programmed in C tho I like Assembly better. dsPIC is also a lot faster.

I would also consider something better than PWM for pitch CV output. Of all the CVs you want accurate, pitch CV is the one...

[elkayem]

Thanks for the tips, Jovian!

I've got a drawer full of Arduino Nanos. I love them and use them for everything. The dsPIC does look like a nice chip, I'll need to check it out.

I'm not using PWM for the analog control. Instead, I have two MCP4822s, which are dual-channel 12-bit DACs and communicate over SPI. I've written the code and breadboarded the circuit already, and can successfully control all four DAC outputs with my keyboard over MIDI. But we'll see if the DAC linearity is good enough for controlling my (under construction) synthesizer.

I think the point of no return is once I put together the panel with my selected outputs. I figure I can always rewrite the code, or put in better performance DACs or op amps, but the front panel will be more of a pain to switch around once I put it together. It's hard to know what I will really want until I have time to play with it.

[elmegil]

I would definitely add velocity.

[Grumble]

you could also add a clock output that gives a clock pulse of which the repetition time may be set by a potentiometer (and CV), or have it be controlled by a midi clock signal coming from a drum computer or similar device. Very usefull for controlling a sequencer or adsr.

[elkayem]

Nice idea, Grumble! My keyboard (Yamaha CP50) does a clock output over Midi that is tied to its metronome and drum machine. I can easily turn that into a pulse pattern.

I do notice that the Sound Lab Ultimate I am assembling has a Repeat Gate and Sample & Hold, which outputs a square wave and trigger with rate set by a potentiometer. I assume this is also what you were suggesting.

[Grumble]

[elkayem]

Update: I've completed my MIDI to CV converter, and installed it into my Sound Lab Ultimate & Expander (still under development).

The outputs look great in my oscilloscope, though I haven't yet been able to try it in my yet-to-be-completed synth. Man oh man, that beast takes a lot of soldering! Maybe I'll have the full thing completed by the end of the year.

For anyone interested in building or improving upon the design, I've made the Arduino code and schematics available here:

https://github.com/elkayem/midi2cv

Of course suggestions for improvements are always welcome!

[elmegil]

I would be inclined to suggest that you buffer the clock/trigger/gate.

You absolutely need protection resistors on the velocity and control and pitch bend, and probably would be best to buffer them as well. A 4822 is not a cheap little chip, and you're set to fry it when it's sending a high voltage and you patch it, when it will momentarily be shorted to ground.

Your LM324 has no output protection either; it's common to put at least a 330R in line with them, older circuits will typically use 1K. If you are concerned about drop in the note value, you can move the 330R inside the feedback loop (between pin 1 and the 7.7K resistor).

Speaking of the LM324, that's a quad op amp, and you're only using one? It would be trivial to buffer pitch bend, control, and velocity with the other three op amps. If you use another LM324 for clock, trigger & gate, remember that the unused op amp should be wired as a follower with a ground input (i.e. ground + input, connect - input to output). Letting them float will affect other op amps on the same die.

Where are you getting a 7.7K resistor?? Edit: ah, I see, you're combining a couple. Might be better to use a trimmer there, if you need some precision.

Edit 2: Ok, you're using bananas, so short to ground when patching shouldn't be an issue; however, it is still good practice to buffer and protect the outputs.

[elkayem]

Hi Elmegil,

Thanks for reviewing the schematic! I do appreciate the feedback.

Regarding buffering the clock/trigger/gate, my thinking was that the 300 ohm resistors would be sufficient to prevent excessive current draw. Each GPIO can source or sink 40 mA, and grounding one of these outputs with a 300 ohm resistor would draw 17 mA. (Thinking this over again, maybe I should use 1K resistors instead, in case I accidentally plug this into another output sitting at -12V!) Would there be any advantage to buffering these outputs rather than using a series resistor?

I purposely did not buffer the analog outputs of the DACs. According to the MCP4822 datasheet, the outputs are already buffered, and on p. 26 states that the MCP4822 "does not require any external output buffer
for most applications." I assume that means I am in the clear. Or are there other concerns?

I did think about putting an output resistor on the op-amp, but opted not to because I was concerned about the voltage drop. According to the LM324 datasheet (absolute maximum ratings), the op amp can have the output shorted to ground continuously, with max current limited to around 40 mA. However, your suggestion to put a 300 ohm resistor at the output of pin 1 before the 7.7K feedback is a great idea, and cheap insurance in case I have misunderstood the datasheet. I hadn't thought of that.

I agree using a quad op-amp was wasteful, and I hated doing that! I had a dozen of them on hand and didn't have any of the single op amps when I put this together, so I was just being lazy. I did not know that allowing the other op amps to float would affect the one I am using. Thank you for pointing that out! I will wire the unused ones as you suggest, unless I really should be buffering some of the other signals.

[JovianPyx]

Regarding buffers:

A buffer will protect the GPIO far better than a resistor. Your example of shorting to -12 through a resistor can still damage the GPIO. Using a buffer will prevent damage to the GPIO - however, such a short would probably blow the buffer - but at least the buffer is inexpensive compared to replacing an MPU.

Also note that "most applications" is not the same as "any or all applications". Applications that allow external user patching can always be problematic if you "accidentally" plug into something that is inappropriate. Buffers are inexpensive and protect your gear, I would strongly advise using them.

[elkayem]

Ok, I've listened to the voices of wisdom and implemented buffers on all four DAC outputs. This takes care of the floating op amps and protects the DACs. Jovian, you are correct that "most" applications is not "all", and I see no indication on the data sheet that these DACs can tolerate a short to ground. (Though no indication they can't, either.) I have also increased the 300 ohm resistors on the clock/gate/trigger outputs to 1K, which should be sufficient when tying it to anything in the +/-12V range. (I know that buffers were recommended here too, but I'm willing to wager a $3 Nano clone on that, and willing to be proved wrong.) The soldering took all of 15 minutes, so no point in dragging my feet! GitHub repository is now updated.

Thanks all for the tips! One of these days I'll post my completed MFOS Ultimate and Expander, if I can ever finish it. Meanwhile, I'll be lurking around this site, hoping to glean some more useful knowledge...

[gabbagabi]

well done, also your expander is very good looking!

cheers, bb

[Sheer Kahn]

thanks a lot for this, exactly what I've been looking for, I will hopefully get to build it next week

[Sheer Kahn]

The DACs turned up faster than I expected, so I got this breadboarded today, worked first time and tracking very nicely (when I put my scope on the right setting ) .

Thanks a lot for sharing your work!


A couple of questions I have (sorry this is my first time using Arduino)

The pitchbend goes +1v, so that's +/-600cents (right?), is it fairy easy to get this to go a bit higher, even +2v, so I can have +/-1200cents. I'm looking at the lines of code:

[elkayem]

Hi Sheer Kahn,

Apologies for the month delay, but I just saw your note this morning!

[elkayem]

[Sheer Kahn]

No worries man! thanks for the reply, yeah I got it working great, nice one.


I didn't find the problem with the lockups, but for some reason it's stopped doing it at the moment, so I'll keep you posted on that one.

[elkayem]

Sheer Kahn, I'm glad you got everything working! It has been an essential addition to my synth.

I reviewed the code, and have a good idea on why your device may have been locking up. I made the implicit assumption in the code that the MIDI note would always be between 21 (A0) and 108, which is the normal range of an 88 key keyboard. If it returned a note outside this range (e.g., below 21), the code would write to unallocated memory. Doh! In that case, anything might happen. Anyway, I've fixed it and pushed the fix.

[Sheer Kahn]

I am still using this, it's great. Thanks for the latest fix.

Is there an easy way to add MIDI clock out to it?

[Sheer Kahn]

Yeah, I figured out that it already has MIDI clock.

You wouldn't have thought I'd been using it all this time and never realised. for some reason I didn't connect it up first time round.


Anyway, thanks again for the great project

[elkayem]

[hububalli]

Just built one of these Midi2CV and it is great.

I an trying to make an adjustment to the pitche bend. I would like it to shift just one tone in each direction. I have been messing around with the code but can't figure it out properly.

Thanks,

[elkayem]