Kosmo Teensy Module

CD4504 Logic level converter… one package to replace all 4 of those 2n7002s and it’s made for that job.

HOWEVER… according to your ADC’s datasheet it can run as low as 2.7V VCC… so just run it on 3.3V and you won’t need level conversion :slight_smile:

EDIT: Also you might consider consolidating some of these TL072s in the filters down to TL074s. Less chips, same function.

EDIT#2: No 10u caps on the power input? I’ve not seen that particular configuration before, maybe yours is better?

1 Like

good progress :slight_smile: …f

It’s the “slighly fancier polarity protection” mentioned here. Basically series diodes, but using a low-resistance MOSFET to get a negligible voltage drop.

1 Like

@BratAttack Thanks for looking at the schematics and for the feedback!

Indeed, you are right that this would simplify the design. One of the goals was to make it possible to order everything in one go from Tayda. But there are four things that would be better to get from somewhere else:

  • the voltage ref and logic level converter (as you suggest)
  • the buck converter
  • a vertical USB-B port

Any thoughts on requiring some parts to be ordered somewhere else, as long as they are fairly standard?

It is true that the chip runs at that voltage and I have tested it like that, but the catch is that the sampling rates scales with the supply voltage. This means, the chip only runs at 100k samples if it is supplied with 5V!

I don’t understand this. The CV inputs are grouped in pairs, and each pair uses one TL074. There is only one TL072 used to buffer the voltage references.

There is one cap on the Teensy and one on the ADC. The op-amps might like some caps as well, perhaps, but I’m not sure if they are really needed. Any thought?

I’m not sure whether @BratAttack was referring to that, but yes, the module has some “fancy” protection to avoid the destruction of expensive hardware while being drunk etc. :slight_smile: I designed this using MOSFETs because the power loss and heat generation with series diodes is unjustifiable to me :wink: And having resistors or diodes that will burn out in case of an issues is unacceptable practice, even though often practised (…), because normal users would be left with a non-functional module and perhaps power supply. Using a dedicated controller is a bit overkill here though, especially one would need two for the two power rails.

1 Like

Snap! good to know!

Cool. I must have misread.

Usually we put 10u caps on the power input to deal with line capacitance and 100n on chips for 60hz noise. How necessary it really is? IDK because I just do it, but yeah i usually see people decouple their op-amp supplies too :slight_smile:

I definitely want to read-up and test that power input protection circuit though, looks shiny!

I will it add some caps for good style then :wink:

The power input protection has btw three safe guards: over-current (fuses) that protect from short-circuits (e.g., wrong soldering), reverse polarity (the MOSFETS) and over-voltage (the 13V Zener diodes). The last part is probably over-kill, if we assume that the module will ever only be exposed to +/-12V. Also, the positive rail p-channel MOSFET is a standard configuration. For the negative rail, I have turned everything around (n-channel, doide). I have simulated but not tested this part of the circuit. So, try it first before you use it for protecting anything :slight_smile:

Finally, I would like to still have some thoughts on having parts in the module that are not available on Tayda. Would that pose an issue for anyone interested in this module?

1 Like

Some more progress on the design: I have drafted the front panel to have an idea where the components have to go. Lets see whether I can keep up my artistic vision. And yes, I plan to make it yellow with black silkscreen like on the first picture. The module is 15cm wide. Enjoy!


This is a surprisingly complex project! The PCB design has taken me a while because I had to learn many things from scratch and fix some mistakes in the design. Here are some rendered images of the final design.

Front panel

The complete module (unfortunately without silkscreen and colour)

The back of the module to give you an idea of the complexity.

I went ahead and ordered the three PCBs (front panel, mainboard and the board with the jack sockets). When I have some time, I will also prepare the BOM and order the components that I am missing (including some tips for SMD soldering). Next week, I’m organising a conference and take some exams but the week after I should have some time again.

And then I hope that everything works… Fingers crossed! If anyone is interested in getting some of the remaining PCBs, under the provision that the design works, please let me know. I will have 4 spare modules. I can also offer to already place the SMD components. Just send me a PM.



Ok, I have learned in the design process quite a lot. The channel of Robert Feranec on YT was very, as were the KiCAD videos by Phil’s lab. On the topic of decoupling capacitors, Robert also has some videos. One of the eye-openers on this topic was, however, this talk by Eric Bogatin:

Indeed, the decoupling capacitors are not just “for good style” as I dismissed them, but rather crucial to decouple line inductance. Should really do some reading on electromagnetism!

Bypass capacitors, as they are more properly called (decoupling caps are series caps to remove DC bias), are good style but not usually particularly necessary for audio electronics. I always include them but Sam frequently omits them and gets away with it. For RF/digital electronics it’s another story, though even there, you evidently can survive with fewer bypass caps than are usually used (though best not to assume that):

With Arduino (or Teensy, I suppose) I tend to assume the on-board bypass caps are sufficient and I don’t need one on the board power pin.


Amazing work! :grinning:
A couple of questions - how do you make the assembled rendering? Is that in FreeCAD?
Also did you make the jack 3d models yourself or download one? I looked and couldn’t find one, and decided it wasn’t worth the time and effort to design one myself.


As I understand it, decoupling and bypass capacitors are both to ground but the former are close to the load and the latter are close to the source. Eric Bogatin also uses decoupling capacitors that way. The main issue that they solve is that sharp edges with high current tend to cause voltage drops (see the demonstration in the video where Bogatin uses a signal with 50ns rise time and 0.4A current flowing). In the case of audio, this does indeed not happen often. We may have sharp edges, but the current consumed by op-amps are very low. But then I’d like to be on the safe side, as the TL07x datasheets recommends them :slight_smile: You are right that decoupling capacitors should not be needed for complete μController boards. I also have an ADC and a multiplexer in the design, which are certainly recommended to have bypass capacitors.

About the EEVblog video: I have seen that, and I suppose that the PCB itself has certainly some capacitance and there are also other factors that play a role. Again, the load on a signal is a key factor. As you say: it is best not to assume that you can get away without.

It does, but that’s a handful of picofarads, so several orders of magnitude smaller than your typical bypass capacitor. Not that it matters that much; you’re building imprecise analog synths, after all.

1 Like

Thanks! That is all in KiCAD. It is a bit limited but for checking the mechanical design it should be sufficient. In this case, I have three separate projects and I export two of them to .step files, which I assign as second 3D model in the third project to some footprints of logos. Then you have to manually rotate and move these models around, until they fit. See here:

Yes, I have downloaded 3D models of the jacks from 3D Content Central (drop me a PM if I should send it to you) but there are also other places for 3D models: SnapEDA, Octoparts, GrabCad. Just watch out that you get the right measures. Comparing with datasheets takes a bit of time. There is also a nice repository that contains plenty of potentiometers.

1 Like

Out of curiosity, I have run some calculations and indeed, even with very close layer etc. you need a huge area for capacitance in the nF range! So then that will not help, indeed :slight_smile: Thanks for pointing this out!

Excellent, I didn’t know about 3D Content Central. They ask way too many questions when you sign up and they run the risk that someone might possibly give less than truthful responses but I wouldn’t know about that. And that Haillant repo looks like a winner too.

I gave them some fake data and a throw-away email address, which you have to confirm though. They did not complain about it.



@moscione Thanks so much for this! I looked on Grabcad and Snapeda but gave up. Never heard of 3d Content Central, but now they have my name(Donald Trump), address(55 Klumpeffery Strand, Bilboffer) and a throwaway email address - it seems promising. The video is also really helpful as all my Kicad skills are self taught and I would never have figured this out on my own.

1 Like