The cello doesn’t have discrete notes like a piano. Instead, the cellist must find the note somewhere along the continuous spectrum of notes found on each string. I’m a beginner so finding those notes is difficult. So when practicing scales and songs, it’s really helpful to have a piano on hand to help me find the note that I’m targeting.
I don’t have space for a piano. Or even a keyboard.
What I do have is an AKAI MPK mini, a MIDI controller with 24 small piano keys. Unfortunately, it’s not a synth so it won’t play any sounds without plugging it into a computer.
Like a RaspberryPi.
Or a laptop or a phone. I’ve tried both of those but what I’d really like is to be able to sit down and start practicing without having to load up software and plug cables in. With a RaspberryPi all I’d have to do is power it up and it’s good to go.
It would be almost as good as just buying a keyboard with a synth. (-:
But then I wouldn’t learn anything.
So what’s the plan?
Get my old RaspberryPi booting and connected to the internet. Connect to the Pi via SSH. Install some sort of synth software. Get it to load up said software automagically on boot.
Flashing RaspberryPi OS to an SD Card #
Find the name of the SD card device:
ls -la /dev/sd*
or
sudo fdisk -l
Flash the image to SD (via https://askubuntu.com/a/227889)
sudo dd if=/home/username/Downloads/2012-10-28-wheezy-raspbian.img of=/dev/sdc status=progress bs=4M && sync
Finding the IP address of the RaspberryPi #
First get the subnet mask
ip -o -f inet addr show
Then use it in this command to scan the network:
nmap -sP 192.168.0.255/24
Oh, wait, looks like I can just do:
ping raspberrypi.local
Speaker Test #
speaker-test -t wav -c 2
List audio devices for playback:
aplay -l
My output had “card 0”, “card 1”, and “card 3”… just need to remember the number.
Test that device using:
speaker-test -t wav -c 2 --device hw:3
…where “3” is the card number above. (Where did “card 2” go?)
Install ntp #
The script synchronizes the system clock but I was getting command not found. Fixed with:
sudo apt install ntp
Mount point not mounted or bad option #
The offending line from synth.sh:
sudo mount -o remount,r /
…
Okay, now I’m just stuck in an endless loop printing “Starting fluidsynth server…”
The command from the script:
screen -dmS FluidSynth0 bash -c "sudo nice -n -20 fluidsynth -i -s -g 0.6 -a alsa -o audio.alsa.device=hw:3 -c 1 -z 1 -o synth.cpu-cores=4 -o synth.polyphony=128 /usr/share/sounds/sf2/FluidR3_GM.sf2"
Trying:
fluidsynth -i -s -g 0.6 -a alsa -o audio.alsa.device=hw:3 -c 1 -z 1 -o synth.cpu-cores=4 -o synth.polyphony=128 /usr/share/sounds/sf2/FluidR3_GM.sf2
Tried c 2 and z 64. That helped with some errors.
fluidsynth -i -s -g 0.6 -a alsa -o audio.alsa.device=hw:3 -c 2 -z 64 -o synth.cpu-cores=4 -o synth.polyphony=128 /usr/share/sounds/sf2/FluidR3_GM.sf2
Screen #
Huh. Screen lets you multiplex terminals. The script uses it to launch fluidsynth.
screen
Run a long-running process then Ctrl + A then ’d’ to detach.
Then I can do
screen -list
to see what’s going on and
screen -r 12345
to reattach the screen. Cool.
Ctrl + A then ’d’ can be used to detach again. Ctrl + A then ‘i’ to get information about the current screen.
The script uses the options -dmS. -d and -m together start a screen in detached mode so it doesn’t switch to that screen when it is created. It just loads it in the background. The -S just specifies the session name, in this case FluidSynth0.
Nice #
Okay, now what does nice do? Ah. Process priority. Negative values are more selfish and will take higher priority. -20 is the least nice it can be and it requires two hyphens to specify a negative number. nice --20. Gosh that’s gotta cause confusion. But that must be how to get rid of the thread priority error from fluidsynth.
The command uses -20 so at first I thought that’s positive 20 but then there’s an -n which means it’s adjusting the niceness by a value, and I think that value in this case is -20 so. I think it’s being demanding and not nice.
So running it with nice looks like this:
sudo nice -n -20 fluidsynth -i -s -g 0.6 -a alsa -o audio.alsa.device=hw:3 -c 2 -z 64 -o synth.cpu-cores=4 -o synth.polyphony=128 /usr/share/sounds/sf2/FluidR3_GM.sf2
To check the niceness of a process you have to:
ps ax -o pid,ni,cmd
And yeah, fluidsynth has a niceness of -20 in this case so quite greedy and rather not nice.
With both screen and nice:
screen -dmS FluidSynth0 bash -c "sudo nice -n -20 fluidsynth -i -s -g 0.6 -a alsa -o audio.alsa.device=hw:3 -c 2 -z 64 -o synth.cpu-cores=4 -o synth.polyphony=128 /usr/share/sounds/sf2/FluidR3_GM.sf2"
Connect MIDI input to Fluidsynth #
List the MIDI output devices:
aconnect -o
pi@raspberrypi:~ $ aconnect -o
client 14: 'Midi Through' [type=kernel]
0 'Midi Through Port-0'
client 24: 'MPK Mini Mk II' [type=kernel,card=2]
0 'MPK Mini Mk II MIDI 1'
client 128: 'FLUID Synth (9465)' [type=user,pid=9465]
0 'Synth input port (9465:0)'
I want to connect the MPK Mini to FLUID Synth so:
aconnect 24:0 128:0
List MIDI output devices and include each of their connections to verify:
aconnect -lo
It works! #
Yay! I tried with the RaspberryPi’s built-in sound card and I was getting some artifacts in the sound. Switched to a USB sound card and it went away.
Now I need to figure out how to get it to work using the script.
Also, how do I change the instrument? I would like something that can drone, preferably even after I’ve stopped pressing the key, at least for a few seconds. That sounds like a custom instrument.
…
Made a couple adjustments to the script and it works, too.
Shutting Down Fluidsynth #
Just for future reference, the easiest way I found to shut down the synth daemon is to reattach its screen and terminate the process.
screen -list
screen -r 1234
<Ctrl + C>
Useful when troubleshooting. Easier than ps and kill. Arguably. (-:
References #
https://www.koopinstruments.com/information/raspberry-pi-midi-synthesizer https://jereme.me/post/raspberry-pi-midi-sound-module/#software