cheap 4-channel videoplayer

for the dance piece ich(a) by zufit simon i constructed a system with four raspberry pi mini-computers and buttons to trigger playback of four video streams. as the videos didn't need to run in exact frame-by-frame sync, this was a very cheap way to get four channel high-quality video playback. total cost was about (rpi 28*4)+(sdcard 6*4)+(5v power 1*7) ≈ 141 euro. i chose the model A of the raspberry pi to keep the cost and power consumption down. the four computers share a 5v power supply of 2 amps and are powered over the gpio pins. video cables run 50 meters down to the stage and in to separate flat screen monitors. the monitors are built in to boxes that can be piled up or rolled around independently.

the videos are stored on the 4GB sd cards that also holds the linux operating system. i converted the videos from dvd to mp4 using ffmpeg with the following settings...

ffmpeg -i concat:"/Volumes/MONITOR01_may2012_DVD/VIDEO_TS/VTS_01_1.VOB|/Volumes/MONITOR01_may2012_DVD/VIDEO_TS/VTS_01_2.VOB" -an -vcodec libx264 -profile:v high -preset fast -crf 18 -b-pyramid none -f mp4 MONITOR01_may2012.mp4

that'll take two chapters and convert to a single mp4 and skip the sound track (-an flag).

the python program running on each computer is here below. it plays a video to the end and waits for a button trigger. if a button is pressed before the video is finished, it'll stop and jump to the next video - all in a cyclic fashion.
#for a raspberry pi running raspbian
#this script will cycle through videos in sequence when a GPIO pin is grounded

#pinSwi (pulled up internally) - gnd this pin to switch to the next video
#pinOff (pulled up internally) - gnd this to shut down the system

videos= ['/home/pi/ICHA1.mp4', '/home/pi/MONITOR01_may2012.mp4', '/home/pi/BLACK.mp4', '/home/pi/FLESH.mp4', '/home/pi/TESTBILDER.mp4']
delays= [0, 0, 0, 0, 0] #extra start delay time in seconds - one value for each video
pinSwi= 23
pinOff= 24

import pexpect
from time import sleep
import RPi.GPIO as GPIO
import os
GPIO.setup(pinSwi, GPIO.IN, pull_up_down= GPIO.PUD_UP)
GPIO.setup(pinOff, GPIO.IN, pull_up_down= GPIO.PUD_UP)

def main():
        os.system("clear && tput civis")        #clear and hide cursor
        index= 0        #keeps track of which video to play
        while True:
                omx= pexpect.spawn('/usr/bin/omxplayer -rp '+videos[index])
                omx.expect('Video')     #play
                        if GPIO.input(pinOff)==False:
                                omx.send('q')   #quit
                                os.system("tput cnorm && sudo halt")
                omx.send('q')   #quit
                sleep(0.5)              #safety
                index= (index+1)%len(videos)

if __name__ == "__main__":

//--instructions for installing (you'll need a model B to prepare a sd-card, but then move it over to the model A raspberry)

//--prepare the rpi
* use Pi Filler to transfer 2013-05-25-wheezy-raspbian.img to the sdcard
* put the sdcard in rpi model B
* select 'Expand Filesystem' in and enable SSH under advanced in config menu
* select 'Finish' and reboot
* log in with pi/raspberry
* sudo apt-get update
* sudo apt-get upgrade
* sudo apt-get install python-pexpect avahi-daemon

//--copy files from osx
* open a terminal window on main computer
* cd to folder with videos
* edit the file and select which videos to use
* optionally add delaytimes if some videos should start later
* scp MONITOR01_may2012.mp4 ICHA1.mp4 BLACK.mp4 FLESH.mp4 TESTBILDER.mp4 pi@raspberrypi.local:/home/pi/

//--back to model B
* sudo pico /etc/rc.local
* add the following before the exit line: (sleep 1; python /home/pi/ & # autostart video player
* press ctrl+o to save and ctrl+x to exit
* sudo halt

//--start model A
* take out the sdcard from model B and put it in model A
* connect hdmi or composite video, gpio pins and apply power - the first video should start
* ground pin 23 to cycle through the videos
* ground pin 24 to turn off the computer

//--useful commands (connect keyboard to rpi model A, type pi/raspberry to log in)
sudo pkill omxplayer.bin     #might need to write this without the terminal being visible

ssh-keygen -R raspberrypi.local     #useful for resetting ssh/scp after changing sd cards

it's not pretty but it's working. some day i'll build it in to a real rackmount box.


under the hood changes 2

updating this blog to drupal 7. it is quite different from version 6 and things will be a bit chaotic for a while. sorry that some content here will be unavailable for a few days.

update 130607: fixed the layout and sound files should play again.

arduino livecoding

so here's some more in depth info on my performance at the live.code.festival / algorave in karlsruhe.

being fascinated since long by the sound of serial transmission, i got into trying to make music out of it in some way. by trial-and-error i figured out that if i connect a small speaker to the tx line of an arduino, i could upload programs that send serial data and listen to the sound of it.
it is all very basic: if i make the arduino program send data with delays in between, it play click rhythms. and programs with faster streams of data play tones. more elaborate combinations of delays and patterns of data produce chords, melodies and a variety of noises. so it works like some sort of one-bit music system that is nice and challenging to play with.

the programs i [live]code can look like this...

byte cnt= 0;
void setup() {
void loop() {

and the resulting sound is this... (raw and unfiltered)

and a more elaborate program...

byte cnt= 0;
void setup() {
void loop() {
  for(int i= 0; i<100; i++) {
  for(int i= 0; i<200; i++) {
  for(int i= 0; i<100; i++) {
  for(int i= 0; i<100; i++) {
  for(int i= 0; i<100; i++) {
  if(random(2)==0) {

sounds like this...

and of course the sound of the uploading (verification really) is great in it self. it typically sounds like this... (raw and unfiltered)

i think the uploading sound changes subtly depending on program length and i also guess it will change with different arduino bootloaders and whatever baudrate they are using.

and you also have a bit of control over the timbre of the sounds. certain 8bit numbers are more square-wave like than others e.g. 170 (0b10101010), and 85 (0b01010101) sound more 'clean' and 15 (0b00001111) and 240 (0b11110000) also have a more distinct pitch.
different baudrates have a huge effect on the sound - mainly working as frequency transposition.

but the real fun starts when one connects five arduinos to a mixer and start playing with volumes, panning and filters. by having five arduinos connected to an usb hub while running five copies of the arduino ide software, i can write little programs on the fly that will address the different boards and play different sounds on the tx lines. (listening to the rx line also works but then the upload process fails. it'll require extra circuitry to tap into this data without disrupting the uploading).

the reason i used five arduinos is because that's all i could connect to my laptop (2x usb) with my 4-port usb hub. that in combination with the limitation of computer screen space. it is hard to have more than five arduino ide programs open and visible at the same time.

anyway, as the voltage of the standard arduino is 5v and really a bit too much for audio equipment, i bring this down a bit with a simple voltage divider. i'm using a 10k and a 1k resistor.

here are some pictures of the setup. i'm using the arduino clone red board from sparkfun.

the complete setup (without mixer and laptop)...

one issue with the setup is that one can't trust the arduino ide to remember which serial port it was connected to. so every time i start the program i need to double check that the five arduino ide programs are set to the right arduino board. and as i like to know which board is connected to which mixer channel, i also need to check that and possibly reconnect the sound cables.

live at the live.code.festival in karlsruhe (algorave night 20apr, 2013). five arduino boards all with their serial port (tx line) connected to a mixer (with simple protective circuitry in between). so all sound are generated from what the arduino boards are programmed to transmit serially. note that the sound is heavily distorted. sorry.


since the category 'visuals' is underrepresented in this blog and i don't like to embed video in my standard [html] pages, i thought i'd include this old piece here. this is the shorter abridged version of the full piece. the quality isn't the best - it's pixelated and stuttering. one day i should re-render it in 60fps at a higher resolution. it looks a lot better when running in realtime from a computer.

Ström by Mattias Petersson (music) and Fredrik Olofsson (video) is, in its full version, a 45 minute minimalistic piece for five loudspeakers, live-electronics and live-video, based on an open-minded, artistic approach towards electricity. The piece is an attempt to transfer electric currents via sound to the audience. The five speakers in the surround system struggles to take over the sonic stream like electro-magnets. Sine waves and noise rotates with breakneck speeds around the listeners, tries to charge them with static electricity and, as an ultimate goal, even make them levitate. The video part is in direct connection with the sound and is generated out of five discrete lines – one for each channel in the surround system. The lines are treated in different ways, and as the high voltage builds up in the music they look more and more like electric wires, inflicting each other with violent discharges and eruptions. This version was made for a promotional DVD release on Swedish sound art.

also see here

traer physics library for supercollider

a while ago i started porting the java/processing library TRAER.PHYSICS 3.0 by Jeffrey Traer Bernstein to supercollider. it's a simple and elegant particle system and a physics engine all in one. there are already ports to actionscript3, javascript and c++ (cinder), but i haven't seen anyone working with it in sc yet. so i had a go - both to learn more and to have an alternative to my own physics library quark redUniverse.

it is now finished and released as a quark. this is the initial version and there might still be bugs. i _did see sc crash once in a strange way after spawning lots of particles, so watch out for memory leaks.
to install it run the following code and recompile sc...


i also wrote a few simple examples to go along with the helpfiles. here's a screenshot of one...


a tiny circuit i designed and built in five copies for dancer Raffaella Galdi. with the help of a small magnetic sensor this circuit makes it possible to start and stop sound coming from a mp3 player. because the five circuit boards, speakers and mp3players are mounted inside pointy hats, the electronics had to be light and draw very little current from the battery.
for the sound volume control i used a vactrol (ldr+led) and the timing and fade in/out logic are encoded in the firmware of a little microcontroller (ATtiny45). to save battery, the tiny45 is put to sleep and is only active when the magnetic reed sensor is triggered. i used the great JeeLib.h for controlling the sleep cycles of the microcontroller.

schematics, firmware and partslist attached below.

Package icon hats.zip29.15 KB

more sc twitter

more audio recordings of my twitter sctweets. see and this post.
normally you run these lines of code (140 characters) in supercollider and it will play you some kind of generative music or soundscape (also graphics in rare cases). here i've recorded a few for those who are too lazy to install sc.




this one is using the built in .fib (as in fibonacci) method to generate pitches (\degree) for the Pbind. the Pbind in it self is quite boring to listen to, so by playing it out on audio bus 8 and then making a small distortion+echo effect synth reading from bus 8, we get a much more interesting sound.




again using the built in fibonacci method to generate patterns. here every 8th bar the pattern is scrambled (Dshuf), and in every 8th bar period the first 2 bars are transposed by a strange trick running the melody pattern (i.e. the frequencies) through a CombN.




here is something that sounds a bit like a couple of trombones playing a riff in a reverberant room. the riff just goes on and on and is made from a pair of slowly changing LFPar oscillators, scaled, offset and rounded to the nearest 50Hz.




sounds a bit like punk rock in 6/8 time signature. the crispness comes from the > BrownNoise combo and the rhythms from the Pulse. overall melody is the slowly running LFPar oscillator stored in variable f. note that this tweet only works in sc version 3.5 and above.




an ever rising tone cluster with some clicks. this is built using a localin/localout feedback chain. there are plateaus where one thinks the maximum frequency is reached, but those are only temporary and after a while the tone starts to rise again.




here the cutoff frequency of a BRF (band reject filter) is modulated with a SinOsc. the cutoff varies between 1 and 99 Hz in the left channel, and 1 and 100 Hz in the right channel. the BRF goes wild and outputs totally crazy sounds when modulated in this matter - just like the BPF used to behave in old sc versions (3.3 and earlier).




a deep fat bass. it sounds as lovely in a big speaker system as it sounds poor in laptop speakers. the patch is mainly doing phase modulation on a SinOsc with tanh distortion.




this tweet sounds much like a field recording. the noise comes from an exploding BRF (band reject filter) that is wrapped in a Limiter so that it keeps in range. last a GVerb is adding a metallic quality reverb to the overall sound.




this was coded, believe it or not, within a 5min time limit and under water pistol threat (part of sc2012 keynote talk in london). again it's a BRF misbehaving run through a comb delay with short modulated delaytime (from 0 to 0.1).




a quite poor tweet. the rhythms are not so interesting and it also have the problem of running out and stopping after a few seconds. anyway, the principle is that a BRF is generating strange sounds that phase modulate a SinOsc, that in turn goes through a ringing filter (Ringz). i only wanted to record it so that when someone fixes the BRF in some upcoming supercollider version, i can go back and listen to how it could sound.



Pbind(\freq,Pseq("SUPERCOLLIDER".ascii,inf)*Pstutter(64,Pseq([3,4,5],inf))*[1,2.045],\dur,0.03,\amp,Pseq([0,0.1],inf)).play// #SuperCollider

super annoying little thing. it is using the values of the ascii characters in the string "SUPERCOLLIDER" which is [83, 85, 80, 69, 82, 67, 79, 76, 76, 73, 68, 69, 82]. this is played in sequence and transposed and detuned. maybe a candidate for the official supercollider theme song?




very intense sounding tweet.




an even more intense sounding tweet.




this patch is heavy on the cpu. it consists of a synth with nested GVerbs all with random settings for roomsize and reverberation time. synths play for 6 seconds and then fades out over 5 seconds as another synth, with different reverb settings start. the result is overlapping sounds and quite dense texture.



a=LFTri;play{,128,128),[3,4],,0,8,12),0,32,128)).sin)/4,1,1/6,}// #SuperCollider

a rhythmic tweet. gets a bit annoying after a while but there are some nice details in there.




phasing melody in left and right channels. every 7th note has a slightly different timbre (the 0!6++500 part) and every time one starts this tweet the melody changes (the Dshuf((0..7) part). the phasing is done with two Saw oscillators running at 9 and 9.01Hz. they are in turn used as triggers for the timbre and melody sequences (the two Demand ugens).


Subscribe to f0blog RSS