LEGO laser harp – part II

Este artigo é a parte 2 de 2 da série  LEGO Laser Harp

About 10 years ago I offered my wife a M-Audio USB MIDI Keyboard and installed Ubuntu Studio on a computer so she could play some piano at home. She was so amazed with the possibility to generate music sheet while playing that almost accepted the idea of using Linux… almost 🙂

I remember that at that time I used timidity++ as a software MIDI synthesizer, tuned ALSA (one of the several Linux sound systems, perhaps the most generally used) and the preemptive kernel to work almost flawlessly with the Creative Labs sound card. My wife didn’t enjoy the KDE experience, Firefox was OK for her but OpenOffice were terribly with the spreadsheets she used and finally, when our first kid was born, she attended some English lessons at Wall Street Institute and we found out that the online lessons required an odd combination of an old version on Java, ActiveX and IE… so she returned to Windows XP and never looked back.

10 years is a LOT of time in computer systems but ALSA is still around, even on ev3dev. So I installed timidity++ and tried to play a MIDI file… to find that an ALSA module that is not currently available in ev3dev kernel is required just for MIDI.

I googled for alternatives and found fluidsynth with an unexpected bonus: there is a quite interesting python library, mingus, that works with fluidsynth. So I installed it in my Ubuntu laptop and in a few minutes I was playing harp – amazing!

sudo apt-get install fluidsynthsudo easy_install mingus
python
>>> from mingus.midi import fluidsynth
>>> from mingus.containers.note import Note
>>> fluidsynth.init("/usr/share/sounds/sf2/FluidR3_GM.sf2", "alsa")
>>> fluidsynth.set_instrument(1, 46)
>>> fluidsynth.play_Note(Note("C-3"))

In the previous example I just import the fluidsynth and Note parts of the library, initialize fluidsynth to work with ALSA loading the soundfount that cames with it, choose harp (instrument number 46) and play C3.

Well and polyphony? The correct way is to use a NoteContainer

from mingus.containers import NoteContainer
fluidsynth.play_NoteContainer(NoteContainer(["B-3", "C-3", "F-3"]))

but the lazy way is… just play several notes in a fast sequence.

So, let’s do it in the ev3dev!

Oops, fluidsynth also needs an ALSA module not available in current ev3dev kernel.

I’m not a linux music expert. Not even a linux expert! So after some more googling I gave up and asked for help in ev3dev’ GitHub project. And once again David accepted to include ALSA MIDI suport in future kernels, so I’ll just wait a bit.

Oh, but I can’t wait…

And if I read the color sensors in ev3dev and play the music in my laptop?

ALSA, once again, suports something like client/server MIDI communication with “aseqnet” and “aconnect” commands and some people are already using it with Raspberry Pi!

Yeah, I should have guessed… “aconnect” requires an ALSA MIDI module that is not available in current ev3dev kernel.

OK, let’s use MQTT: configure my EV3 as a publisher and my Ubuntu laptop as a subscriber and just send some notes as messages.

On the EV3:

sudo apt-get install mosquitto
sudo easy_install paho-mqtt

The publisher script is “harp-mqtt-pub.py”:

#!/usr/bin/env python

from ev3dev.auto import *
from time import sleep
import paho.mqtt.client as mqtt

DELAY = 0.01

# should have an auto-calibrate function
AMB_THRESHOLD = 9

sensor1 = ColorSensor('in1:i2c80:mux1')
sensor1.mode = 'COL-AMBIENT'
sensor2 = ColorSensor('in1:i2c81:mux2')
sensor2.mode = 'COL-AMBIENT'
sensor3 = ColorSensor('in1:i2c82:mux3')
sensor3.mode = 'COL-AMBIENT'
sensor4 = ColorSensor('in2')
sensor4.mode = 'COL-AMBIENT'
sensor5 = ColorSensor('in3')
sensor5.mode = 'COL-AMBIENT'
sensor6 = ColorSensor('in4')
sensor6.mode = 'COL-AMBIENT'

# there is no sensor7 yet, I need another MUX

s1 = 0
s2 = 0
s3 = 0
s4 = 0
s5 = 0
s6 = 0
s7 = 0

client = mqtt.Client()
client.connect("localhost",1883,60)

print 'Running...'

while True:
    key_touched = False
    s1 = sensor1.value(0)
    s2 = sensor2.value(0)
    s3 = sensor3.value(0)
    s4 = sensor4.value(0)
    s5 = sensor5.value(0)
    s6 = sensor6.value(0)
#    s7 = sensor7.value(0)

    if s1 < AMB_THRESHOLD:
        client.publish("topic/Harp", "C-3")
        key_touched=True
    if s2 < AMB_THRESHOLD:
        client.publish("topic/Harp", "D-3")
        key_touched=True
    if s3 < AMB_THRESHOLD:
        client.publish("topic/Harp", "E-3")
        key_touched=True
    if s4 < AMB_THRESHOLD:
        client.publish("topic/Harp", "F-3")
        key_touched=True
    if s5 < AMB_THRESHOLD:
        client.publish("topic/Harp", "G-3")
        key_touched=True
    if s6 < AMB_THRESHOLD:
        client.publish("topic/Harp", "A-3")
        key_touched=True
#    if s7 < AMB_THRESHOLD:
#        client.publish("topic/Harp", "B-3")
#        key_touched=True

    if key_touched == True:
        sleep(DELAY)

On the Ubuntu laptop side:

sudo easy_install paho-mqtt

The subscriber script is “harp-mqtt-sub.py”

#!/usr/bin/env python

import paho.mqtt.client as mqtt
from mingus.midi import fluidsynth
from mingus.containers.note import Note

EV3_IP = "192.168.43.35"

SOUNDFONT = 'Concert_Harp.sf2'
INSTRUMENT = 46 # Harp

NOTES = ['C-3','D-3','E-3','F-3','G-3','A-3','B-3']

def on_connect(client, userdata, flags, rc):
    print("Connected with result code "+str(rc))
    client.subscribe("topic/Harp")

def on_message(client, userdata, msg):
    global i
    if (msg.payload in NOTES):
        print msg.payload
        fluidsynth.play_Note(Note(msg.payload))
    
client = mqtt.Client()
client.connect(EV3_IP,1883,60)

client.on_connect = on_connect
client.on_message = on_message

fluidsynth.init(SOUNDFONT, "alsa")
fluidsynth.set_instrument(1, INSTRUMENT)

client.loop_forever()

And guess what? It works!!! I just love linux and open source!

I will keep waiting for David Lechner to include ALSA MIDI support in ev3dev’ kernel. I’m not so sure if there is enough horsepower in the EV3 to load a soundfont and play it with acceptable latency but if I can at least use the MIDI client/server functionality I can drop MQTT.

An interesting possibility that this client/server design allows is to scale my harp easily: with just a second EV3 (2 MUX each) I can make a 13-string harp with almost no modification on my code.

LEGO laser harp – part I

Este artigo é a parte 1 de 2 da série  LEGO Laser Harp

This is an idea I’ve been postponing for several months but the time has finally come: an laser harp.

After tinkering with lasers, fog, sound, color sensors and python I found myself wondering how to give a proper use to all that. Then I remembered Jean-Michel Jarre and how his laser harp made such a big impression on me at late 80’s when I finally decided “hey, i wanna study Electronics!”

For a first version, let’s call it “a proof of concept”, I just want a simple 7-string harp that can play the basic 7 notes. Polyphony would be great but I doubt that the EV3 sound capabilities allow that (and I cannot afford the brute force solution of using 7 EV3 so that each one plays only a single note).

So in the last months I’ve been buying EV3 color sensors and I finally have 7. Since the EV3 only has 4 input ports I need some kind of sensor multiplexer but thanks to mindsensors.com I already have one EV3SensorMux (and a second one is on the way, from an european distributor – portuguese customs DO stink!)

With 2 MUX it’s possible to connect up to 8 sensors to the EV3. Since I just want 7 “strings” I am considering using an 8th sensor to control the amplitude of the notes. I’ll try an ultrasonic sensor but I’m not sure if it has enough “wideness” to cover the whole harp, let’s see.

So of course I’ll be using ev3dev and python.

Using the EV3SensorMux is easy: just plug it to an input port and ev3dev immediately recognizes it:

lego-port port8: Registered 'in1:i2c80:mux1' on '3-0050'.
lego-port port8: Added new device 'in1:i2c80:mux1:lego-ev3-color'
lego-sensor sensor0: Registered 'ms-ev3-smux' on 'in1:i2c80'.
lego-port port9: Registered 'in1:i2c81:mux2' on '3-0051'.
lego-port port9: Added new device 'in1:i2c81:mux2:lego-ev3-color'
lego-sensor sensor1: Registered 'ms-ev3-smux' on 'in1:i2c81'.
lego-port port10: Registered 'in1:i2c82:mux3' on '3-0052'.
lego-port port10: Added new device 'in1:i2c82:mux3:lego-ev3-color'
lego-sensor sensor2: Registered 'ms-ev3-smux' on 'in1:i2c82'.
lego-sensor sensor3: Registered 'lego-ev3-color' on 'in1:i2c80:mux1'.
lego-sensor sensor4: Registered 'lego-ev3-color' on 'in1:i2c81:mux2'.
lego-sensor sensor5: Registered 'lego-ev3-color' on 'in1:i2c82:mux3'.

Even better: by default all 3 mux ports are configured for the EV3 color sensor, just as I wanted!

NOTE: as of today (kernel version ‘4.4.17-14-ev3dev-ev3’) my EV3 autodetection only works when booting with a non-default configuration:

sudo nano /etc/default/flash-kernel

 LINUX_KERNEL_CMDLINE="console=ttyS1,115200"

sudo flash-kernel
sudo reboot

this was suggested to me by David Lechner in another issue, hope will be fixed soon.

To use the color sensors in python I just need to know their ports. With the MUX in port ‘in1’ and 6 color sensors connected, these are the ports to use:

in1:i2c80:mux1
in1:i2c80:mux2
in1:i2c80:mux3
in2
in3
in4

And to play a note in python I just need to know it’s frequency to use with Sound.tone() function, so:

C3 = [(130.81, TONE_LENGHT)] 
D3 = [(146.83, TONE_LENGHT)] 
E3 = [(164.81, TONE_LENGHT)] 
F3 = [(174.61, TONE_LENGHT)] 
G3 = [(196.00, TONE_LENGHT)] 
A3 = [(220.00, TONE_LENGHT)] 
B3 = [(246.94, TONE_LENGHT)]

And so this was the first script for my harp:

#!/usr/bin/env python

from ev3dev.auto import *

TONE_LENGHT = 150

C4 = [(261.64, TONE_LENGHT)]   #Do4
D4 = [(293.66, TONE_LENGHT)]   #Re4
E4 = [(329.63, TONE_LENGHT)]   #Mi4
F4 = [(349.23, TONE_LENGHT)]   #Fa4
G4 = [(392.00, TONE_LENGHT)]   #Sol4
A4 = [(440.00, TONE_LENGHT)]   #La4
B4 = [(493.88, TONE_LENGHT)]   #Si4

AMB_THRESHOLD = 9

sensor1 = ColorSensor('in1:i2c80:mux1')
sensor1.mode = 'COL-AMBIENT'
sensor2 = ColorSensor('in1:i2c81:mux2')
sensor2.mode = 'COL-AMBIENT'
sensor3 = ColorSensor('in1:i2c82:mux3')
sensor3.mode = 'COL-AMBIENT'
sensor4 = ColorSensor('in2')
sensor4.mode = 'COL-AMBIENT'
sensor5 = ColorSensor('in3')
sensor5.mode = 'COL-AMBIENT'
sensor6 = ColorSensor('in4')
sensor6.mode = 'COL-AMBIENT'

# there is no sensor7 yet, I need another MUX

s1 = 0
s2 = 0
s3 = 0
s4 = 0
s5 = 0
s6 = 0
s7 = 0

while True:
    s1 = sensor1.value(0)
    s2 = sensor2.value(0)
    s3 = sensor3.value(0)
    s4 = sensor4.value(0)
    s5 = sensor5.value(0)
    s6 = sensor6.value(0)
#    s7 = sensor7.value(0)
  
    if s1 < AMB_THRESHOLD:
        Sound.tone(C4).wait()
    if s2 < AMB_THRESHOLD:
        Sound.tone(D4).wait()
    if s3 < AMB_THRESHOLD:
        Sound.tone(E4).wait()
    if s4 < AMB_THRESHOLD:
        Sound.tone(F4).wait()
    if s5 < AMB_THRESHOLD:
        Sound.tone(G4).wait()
    if s6 < AMB_THRESHOLD:
        Sound.tone(A4).wait()
#    if s7 < AMB_THRESHOLD:
#        Sound.tone(B4).wait()

So whenever the light level over one of the color sensor drops bellow AMB_THRESHOLD a note will play for TONE_LENGHT milliseconds.

Unfortunately the sound is monophonic (just one note can be played at a time) and it doesn’t sound like an harp at all – it sounds more like my BASIC games on the ZX Spectrum in the 80’s.

So I tried Sound.play(Wave File) instead. Found some harp samples, converted them to .wav files at 44100 Hz and it really sounds much better… but the length of the samples I found is to big so the “artist” have to wait for the note to stop playing before moving the hand to another “string”. Not good and also not polyphonic.

Next post I’ll show a better approach for both quality and polyphony: MIDI.