Electronic – How to stretch high frequency digital signals (e.g., MIDI) into the audio range

Start with, say, a guide to the protocol.

Given that you've decided to use an Arduino, you can just speak MIDI more or less directly. There's even a handy tutorial on the Arduino website for MIDI out! MIDI input is slightly more electrically complex as you're supposed to have an optoisolator in there to avoid ground loops.

It's almost entirely a software project, but if you're willing to pick up code from the internet should be reasonably straightforward.

Edit: a bit more googling has found someone who's actually done this, in 40xx CMOS rather than 74: http://www.pykett.org.uk/a_midi_pedalboard_encoder.htm

That's MIDI output rather than input. I still maintain there's no point doing it by halves: if you're going to do MIDI decode in hardware, do all the audio in hardware too, don't just put an Arduino in there as a PWM generator. And if you do have the Arduino, let it do the protocol bit and consider moving the audio side to analog.

1. The official circuit uses three 220 Ω resistors (two in the transmitter and one in the receiver) to limit the current to about 5 mA.

   Dropping one resistor will increase the current to about 8 mA. This is unlikely to damage the receiver, but it might increase the time the optocoupler needs to switch off, and thus foul up the timing of the UART signal (for really cheap and non-compliant receivers that are already on the edge of not working). Anyway, there's no reason to waste power.

2. The "A" is just a label used in the remark below:

   Gates "A" are IC or transistor.

   When the MIDI specification was written, typical microcontrollers (e.g., 8051) were not able to sink 5 mA of current, so it was necessary to add some buffer. Nowadays, you would be able to use the output pin of most microcontrollers directly, or just use a single non-inverting 74LVC1G07 gate, but thirty years ago, the most easy way to construct a buffer would be to use two gates of a hex inverter. (Whether a Schmitt-trigger input is needed depends on the input signal of the inverter.)

   What exact chip or buffer circuit you use does not really matter, as long as it is capable of 5 mA. (Many older MIDI devices indeed used transistors.)

3. A Thru circuit is just a normal output circuit connected to a normal input circuit.

4. The DSS306 are EMI filters. Some kind of filtering is likely to be necessary nowadays; this is one of the reasons that the MIDI schematic you mentioned is outdated and was replaced with the Electrical Specification Update (2014).

   If you are not going to sell the device, or if you are a cheap bastard, don't bother about EMI. Otherwise, use something that filters out high frequencies (MIDI is slow, so you don't need more bandwidth than a few MHz). In practice, many devices use ferrite beads, but RC filters or even common-mode chokes are also used sometimes.

5. Transistors can handle larger currents and voltages than a logic gate, but that does not matter for MIDI.

   A transistor is just cheaper …

6. … but if you're using a hex inverter, you might just as well use all of its gates.