There are internal buffers in the DAC (AD5668) http://www.analog.com/media/en/technical-documentation/data-sheets/AD5628_5648_5668.pdf and then there are external 220ohm resistors in series with each analog out. Full schematics https://code.soundsoftware.ac.uk/attachments/download/1495/beaglert_cape_rev_A3.zip (the current board layout is slightly different, but electronically it is the same).

The RD-0512D seems a nice piece of kit, but don't you think the 150mV ripple is a bit too large? That is about -38dB, while an LM7812, for comparison, would normally come with about -70dB ...

So I managed to interface an VCO to the Bela, and as @Hypasus said, it is amazing ! I'm using Supercollider to get the audio in, and everything sounds and works perfectly. I'll put it on hold for now though, I'm waiting for some components I didn't have so I can secure the inputs, then I'll start experimenting 😃

I'm following your discussion with great interest. My signal is still a bit noisy. Can I blame it only on the "breadbording"/low grade components effect or a a better 5V source would improve it (as I'm dividing the 5V output of the Bela to get my 2.5 reference) ? Or both ?

Good news!
Difficult to say what the source of your noise is, without more details on your setup and on your noise.
Surely using Bela's 5V to passively obtain a 2.5V reference is not a good idea...

What is your current setup?
and
What is your noise like?

It is basically the input of this one:

http://interface.khm.de/wp-content/uploads/2008/11/arduinoaudiofig1.gif

Except that now I'm getting my 2.5 from dividing the output of a 5V external power supply. And it is not a 4.7n but 100n, I don't have any other values at the moment...

It's not noise, it's more like interferences, discrete little spikes. Not continuous noise. I blame the dirty way of wiring all of it together, but I'm (surely) maybe wrong. I could record a gif of the scope if you want.

100n should be fine, as long as it is not electrolytic ( I doubt they manufacture electrolytics in that range).
A pic of the scope could help, also some sort of frequency-domain representation would help. Ideally, you'll want to scope your 5V reference at the same time as your signal, to see if the noise comes from the 5V.

Let me see if I get this right: you connected one of the analog outputs from Bela to the "audio input" on the schematic and then connect "analog input 0" from the schematic to your VCO's CV input. Correct?
How are your grounds connected?

What is your external power supply like? Lots of 5V PSUs are switching and not necessarily well filtered, depending on the application they were designed for, but I would expect more like a 5-6kHz periodic signal from them rather than discrete spikes.

The other way around. The VCO output is the Audio Input, the Analog Input 0 is on the Bela. I get a wave of max 2.5Vpp centred around 2.5V, approximately.

So this is, in blue, the signal from the VCO with 2.5 from the PSU:

http://www.balladeenruaba.fr/blog/wp-content/uploads/2016/07/scope1.png

In red is just a phasor I made for reference.

Still in blue, the PSU with the signal unplugged:

http://www.balladeenruaba.fr/blog/wp-content/uploads/2016/07/scope2.png

Any advice for a better PSU ? I'll start looking on my own too 🙂

It looks like troubles with the PSU. Just to confirm, why don't you make one of the analog outputs output a constant voltage and connect it to your input, just to make sure you do not see any of those spikes?
I also like to "listen to the PSU": connect the PSU to an AC-coupled audio input (maybe put a few hundred ohms in series, just in case) and listen to it. The noise you hear there will leak into your signal.

You do not really need a PSU for this application, a voltage reference is all it takes. A few silicon diodes in series or a 4V zener would probably give you a stable enough reference (remember the analog input range is 0-4.096V), alternatively, look for a proper voltage reference IC. Also an LM7805 would probably be good enough for this application, but voltage references are less noisy and you do not need much current, so they are better suited. Note that stabilizers and references have up to 2V of dropout, so you need either a PSU with higher voltage or pick a low-dropout device. You can also try if you are lucky using one of the voltages available on the BBB, like the 3.3V on P9-3 and P9-4 (you will have to change your upper resistor in the divider to match the new voltage).

I have no recommendation about commercially available power supplies... but prices can be quickly become esoteric if you google for "audio power supply", at which point you are probably better off with a transformer, bridge rectifier, cap and LM78xx ...

Allright ! I already had some zeners shipped, I'll read more about how to get a proper voltage reference. I'll look into the LM7805 too.
Thanks a lot for your time and advice, I learned a lot from them. It seems a lot easier now to understand how to achieve this particular need to make Bela and the Euroworld talk to each other.

One thing I forgot to mention is that this circuit is far from ideal for serving as a general purpose interface between Bela and a synth (either way), because of many reasons, including that it does not provide any protection to the Bela inputs and it removes the DC component of the original signal, which means that LFO, gate, envelope signals would be badly distorted by passing through it.

Yes, I used it only to get audio from a VCO with a known voltage output. It was really for me to understand how to wire, divide, etc. It serves absolutely no other purpose than that.
I'll get to work to have a proper way to get anything from the euroworld into the Bela once I get components to protect inputs and be able to switch to DC signals.

5 days later

The idea of listening to the power supply noise is a nice one. Maybe a piece should be done around the concept.
Thank you for the schematics (I saw the update in the other thread), very useful.
I still designing to understand the power consumption of the whole thing. I am considering now to go with an external 12V/2-3A since I am adding an LCD running also from a 5v source.
Do you have an opinion about demultiplexing one of the analog inputs with something like the 74HC4051PW to add extra pots. Or is it better to add an extra mcu. I would like to reserve all the possible cpu for audio.

    Hypasus do you mean multiplexing the inputs ? The multiplexer capelet does just that. Schematics are here https://github.com/BelaPlatform/bela-hardware/tree/master/capelets/multiplexer/mux_capelet_A2 .

    I have no idea what current is required by your LCD, but 3A seems a bit overkill for this application... Rather, unless you are using an LDO regulator, I would go for 15V AC transformer if you plan to use ±12V DC supply rails. I used to have troubles with 12V AC to be used for 12V DC supply rails because of fluctuations in the power grid in my area. Power outlets are normally rated ±10%.. When I had this issue, my synth would be unstable at random times of the day, or depending on the location, depending to what the mains voltage was in the area at the time.

    Using two 1n4007 in a bridge rectifier configuration, you get a combined voltage drop of about 1.4V, so you get an unregulated DC voltage of
    (12V - 1.4V)*sqrt(2) = 15 V

    But if your mains voltage is 10% below the nominal value, you effectively get 13.5V. Considering the LM78xx/LM79xx series has a dropout of about 2V, an LM7x12 would not be capable of producing a regulated 12V output from an unregulated 13.5V input. 15V AC gives you a better safety margin on this. Again, your power requirements should not be huge and you should be able to comfortably fit in the 500mA provided by standard LM78xx series. Otherwise, I believe there must be a high-current version of these regulators

      6 days later

      Yes, the multiplexer of course by 64 inputs is a bit too much for what I had on mind (4 extra pots and 4 encoders).
      Well, I will keep playing maybe the 8 analog inputs will be enough for this project.
      For the power supply, the real issue is sourcing the Beaglebone rated at 2A and the opamps for the synth level shifting: 500mA will be enough. I am experimenting with a small oled display, so that shouldn't be a big extra load.
      That's why I initially thought to start with a strong 5V power supplyy and the boost, invert and regulate the 12/-12v
      I am expecting the DC-regulators to make some measurements.
      In between I put in a strip board a test circuit for 1v/octave input a trigger input, one unipolar and two bipolar inputs for LFOs, with another technic: generating a negative offset with a MAX1044 from the 3.3v supplied​ by the beagle, seems to work.
      https://drive.google.com/file/d/0BxadeX5mevFULVpfRjFtZDgyT28/view?usp=sharing
      The dac outputs are unipolar, will be enough to use a decoupling capacitor for bipolar signals? I tried with a 100uF and got a bit distortion.

      Decoupling capacitors are useful for generating audio but not for CVs where you want response down to DC.

      One way to get a bipolar output from a unipolar DAC is to use an op-amp mixer circuit. This is an inverting amplifier configuration with one feedback resistor and two input resistors. One of the inputs goes to the DAC, the other goes to a low-noise, regulated negative voltage, say -5V. The op-amp sums the two signals together at the output. Try, say, a 20k feedback resistor, 20k to -5V, and 10k to the DAC. This will turn the 0-5V range of the DAC into +5V to -5V (inverted), i.e. 0V from the DAC = +5V out, 2.5V from the DAC = 0V, 5V from the DAC = -5V out.

      6 days later

      Strangely enough,the problem I have with "spikes" in the signal is not bound to the source. I tried a simple connection, taking the 3.3 from the Bela, into a 10k pot, into the analog 0.
      The reading is dead flat, until I move the pot. Then some spikes emerges, until I stop making the signal varying. then dead flat again.
      What can be the issue here ?

      It sounds like a scratchy pot to me. have you tried with different pots?
      Can you post a still of the scope? If you set it to single trigger mode you should be able to capture your noise.

      a month later

      Sorry for the late answer, but I ran in this problem again, not having interfacing anything lately. The spikes are not occurring when I move the pot, but at random. It affects everything. I linked a video of it, the blue line is the sin output from the examples project, the red line is a simple 10k pot into Analog 0 wired to the 3.3v from the Bela. I simply multiply the reading from AnalogIn 0 with the output from the sine wave.

      As you can see, the blue waveform is affected without me touching the pot. I tried a lot of them too. And also to change basic settings like sample rate, but it's always like this.
      http://ovh.to/iw9ZH5g
      Soundwise, you can hear pops now and then when it's strong enough, but you can hear it when modulating, ie using a pot to FM an oscillator in SC can get quite messy.

      So I tried with just the simple sinetone project, and added a scope, and there are tiny spikes too:
      http://ovh.to/Mkwp775

      Is it me not understanding something ?

      Interesting.
      Where are you taking the ground signal from the board?
      Is there anything else connected to the board?
      Is your computer powered by battery or power supply?

      Can you try instead of using the 3.3V source to use one of the analog outputs as your voltage reference? For instance, add this at the bottom of your render() function:

      analogWrite(context, 0, 0, 0.66)

      This should give you 3.3V out of analog 0.

      For the recording of the two clips, ground was the pin right under the 3.3v supply. Nothing else was connected but 3.3, grd and analo in 0. I tried with my laptop wired or only on battery, and without the battery and only wired, same issue. I used to power it to get the amplifiers, but even without PSU and input/output (so nothing but the patch running), the sine get spikes.

      I tried with the 3.3 from analogWrite, same issue. Also I have to write it every frame, only 0 makes it goes back to 0 every other frames. Here with the audio input on, a tiny one 😉

      http://ovh.to/Y5o4NhF