Friday, 31 August 2012

Serial Port MIDI on the Raspberry Pi

I noticed a project at Raspberry Pi was having to use a microcontroller to buffer 38400 Baud serial port data down to the Midi standard of 31250. As part of an ongoing larger project I solved this problem by changing init_uart_clock = 2441406 in /boot/config.txt on the raspberry Pi.

My proto-type MIDI hardware is the usual 6N138 opto isolator for the MIDI Input and a 7414 for the output shown below. The output from the6N138 is 3.3v and the output from the 7414 is at 5V. This was chosen because although the MIDI specification usually talks about the signal as a current loop, there are a few devices that are MIDI powered and require 5v. MIDI powered devices work by doing away with the opto-isolater and the safety it provides.

There are a couple of changes that are needed as by default, Raspbian "wheezy" Linux distro, sets it up a serial port tty. Using a consol make these changes can be made as follows.
sudo nano/boot/cmdline.txt
Edit the line

dwc_otg.lpm_enable=0 console=ttyAMA0,115200 kgdboc=ttyAMA0,115200 console=tty1 root=/dev/mmcblk0p2 rootfstype=ext4  elevator=deadline rootwait
dwc_otg.lpm_enable=0 console=tty1 root=/dev/mmcblk0p2 rootfstype=ext4 elevator=deadline rootwait

exit saving the changes

Next by changing init_uart_clock=3000000 to init_uart_clock=2441406, all baud rates will be reduced by about 81%. This is not an ideal solution as if you wanted to use the serial port for other projects all the standard baud rate timings would be out by 19%. However, from a MIDI point of view you can simply open /dev/ttyAMA0 at 38400 baud and you are up and running.
Again using a consol.
sudo nano /boot/config.txt
Add these lines to the end of the file
# change uart clock to 2441406 for midi 31250 baud rate
save and exit, and reboot the Pi.

The Raspberry Pi can now talk MIDI by opening /dev/ttyAMA0 at 38400 baud or connected to ALSA using ttymidi driving FluidSynth.


  1. Hey, I'm looking to find the best (as far as reliability/timing) midi interface for my RPi, I'll be running a few signals in and a lot of signals out. What's the benefit of using this over a USB MIDI interface?

  2. I was getting between 2.26 - 3.94 ms latency using FluidSynth at the time when the RPi was having problems with a USB overhead of 20% CPU.

    With the MIDI shield, it's just tidier, with one one the USB ports containing a WiFi Dongle and the other free for a sound card/USB flash etc. Which in the near future will be used as a midi controller.

  3. [this is still me I'm just signed in elsewhere]
    Not sure that I quite understand, So you still intend to use USB for midi? And you were using a soft synth within the rPi, not sending midi out?

  4. I will not be using USB MIDI for no other reason than less cables.

    I only used the soft synth for testing the hardware. Ultimately it will be a MIDI controller.

    Leaving latency and jitter aside, the MIDI Baud rate of 31250 bps format is 1 start bit, 8 data, no parity, 1 stop bit. This gives us 3125 bytes of data or ~1000 MIDI messages a second in and ~1000 out (for the RPi).

    This is not a lot for the RPi to handle so long as the RPi is not busy running to many apps.

    The ALSA MIDI latency test will probably be of help to you.

    Which you can also use by connecting just RPi midi in and out together.

  5. this is exactly what I wanted thank you. and thank you for covering the things relating to midi serial baud rate. (it's an important concideration.) I am using my raspberry pi as a midi interface with pure data running in -nogui mode this is the next part of the build for me.

  6. hi silicon stuff, i have tried to use my raspberry for midi applications on its uart as you described, but the clockrate on the Tx signal remains 38400 baud, though the vcgencmd measure_clock uart wil prompt: 224100. What image did you use for this midi-project? There should be a Kernel to support this uart-change, isn't it?

  7. Hi Erik.

    Sorry, I havn't access to my R-Pi at the moment but I would have been using the current distro in August 2012.

    Did config.txt save? as I seem to recall this sometimes being an issue?

    I will recheck next week and post any updates.

    1. thaks for the reply....finally i got it working: apart from the additions in config.txt regarding the new clockrate, the following has to be added to the existing adapted line in cmdline.txt: bcm2708.uart_clock=3000000. So no it's working fine...i had allready made an circuitboard with a PIC controller on it to translate the 38400 to 31250....but that's for other purpuses now..once againg thanks and regards..erik

  8. Thanks for this. I managed to get a simple MIDI in/out interface working with these instructions. I even managed to "recycle" an opto isolator chip from an old modem circuit board that I had, which saved me a couple of pounds. One word of warning though - some opto isolators don't work fast enough for Midi, so check that the on/off time is at most about 15 microseconds.

    I also wrote a simple Midi message monitor in Python, which I'll upload to in a moment.


  9. Very good work, looking for just such a solution for a project of mine.

  10. Hello. Nice work. Thanks for sharing it. There are two capacitors on the diagram, but i haven't seen them on the photo. Did i miss something ?