CSOUND IN PD
You can embed Csound in PD via the external csoundapi~,2 which has been written by Victor Lazzarini. This external is part of the Csound distribution.
On Ubuntu Linux, you can install the csoundapi~ via the Synaptic package manager. Just look for "csoundapi~" or "pd-csound", check "install", and your system will install the library at the appropriate location. If you build Csound from sources, you should also be able to get the csoundapi~ via the scons option buildPDClass=1. It will be put as csoundapi~.pd_linux in /usr/lib/pd/extra, so that PD should be able to find it. If not, add it to PD's search path (File->Path...).
On Mac OSX, you find the csoundapi~ in the following path:
Put this file in a folder which is in PD's search path. For PD-extended, it's by default ~/Library/Pd. But you can put it anywhere. Just make sure that the location is specified in PD's Preferences > Path... menu.
On Windows, while installing Csound, open up the "Front ends" component in the Installer box and make sure the item "csoundapi~" is checked:
After having finished the installation, you will find csoundapi~.dll in the csound/bin folder. Copy this file into the pd/extra folder, or in any other location in PD's search path.
When you have installed the "csoundapi~" extension on any platform, and included the file in PD's search path if necessary, you should be able to call the csoundapi~ object in PD. Just open a PD window, put a new object, and type in "csoundapi~":
<CsoundSynthesizer> <CsOptions> </CsOptions> <CsInstruments> ;Example by Joachim Heintz sr = 44100 nchnls = 2 0dbfs = 1 ksmps = 8 giSine ftgen 0, 0, 2^10, 10, 1 instr 1 kFreq invalue "freq" kAmp invalue "amp" aSin oscili kAmp, kFreq, giSine outs aSin, aSin endin </CsInstruments> <CsScore> i 1 0 10000 </CsScore> </CsoundSynthesizer>
Save this file under the name "control.csd". Save a PD window in the same folder and create the following patch:
Note that for invalue channels, you first must register these channels by a "set" message.
As you see, the first two outlets of the csoundapi~ object are the signal outlets for the audio channels 1 and 2. The third outlet is an outlet for control data (not used here, see below). The rightmost outlet sends a bang when the score has been finished.
Audio streams from PD can be received in Csound via the inch opcode. As many input channels there are, as many audio inlets are created in the csoundapi~ object. The following CSD uses two audio inputs:
<CsoundSynthesizer> <CsOptions> </CsOptions> <CsInstruments> ;Example by Joachim Heintz sr = 44100 0dbfs = 1 ksmps = 8 nchnls = 2 instr 1 aL inch 1 aR inch 2 kcfL randomi 100, 1000, 1; center frequency kcfR randomi 100, 1000, 1; for band pass filter aFiltL butterbp aL, kcfL, kcfL/10 aoutL balance aFiltL, aL aFiltR butterbp aR, kcfR, kcfR/10 aoutR balance aFiltR, aR outch 1, aoutL outch 2, aoutR endin </CsInstruments> <CsScore> i 1 0 10000 </CsScore> </CsoundSynthesizer>
The corresponding PD patch is extremely simple:
The csoundapi~ object receives MIDI data via the keyword "midi". Csound is able to trigger instrument instances in receiving a "note on" message, and turning them off in receiving a "note off" message (or a note-on message with velocity=0). So this is a very simple way to build a synthesizer with arbitrary polyphonic output:
This is the corresponding midi.csd. It must contain the options -+rtmidi=null -M0 in the <CsOptions> tag. It's an FM synth which changes the modulation index according to the verlocity: the more you press a key, the higher the index, and the more partials you get. The ratio is calculated randomly between two limits which can be adjusted.
<CsOptions> -+rtmidi=null -M0 </CsOptions> <CsoundSynthesizer> <CsInstruments> ;Example by Joachim Heintz sr = 44100 ksmps = 8 nchnls = 2 0dbfs = 1 giSine ftgen 0, 0, 2^10, 10, 1 instr 1 iFreq cpsmidi ;gets frequency of a pressed key iAmp ampmidi 8;gets amplitude and scales 0-8 iRatio random .9, 1.1; ratio randomly between 0.9 and 1.1 aTone foscili .1, iFreq, 1, iRatio/5, iAmp+1, giSine; fm aEnv linenr aTone, 0, .01, .01; avoiding clicks at the end of a note outs aEnv, aEnv endin </CsInstruments> <CsScore> f 0 36000; play for 10 hours e </CsScore> </CsoundSynthesizer>
Score events can be sent from PD to Csound by a message with the keyword event. You can send any kind of score events, like instrument calls or function table statements. The following example triggers Csound's instrument 1 whenever you press the message box on the top. Different sounds can be selected by sending f events (building/replacing a function table) to Csound.
<CsoundSynthesizer> <CsOptions> </CsOptions> <CsInstruments> ;Example by Joachim Heintz sr = 44100 ksmps = 8 nchnls = 2 0dbfs = 1 seed 0; each time different seed giSine ftgen 1, 0, 2^10, 10, 1; function table 1 instr 1 iDur random 0.5, 3 p3 = iDur iFreq1 random 400, 1200 iFreq2 random 400, 1200 idB random -18, -6 kFreq linseg iFreq1, iDur, iFreq2 kEnv transeg ampdb(idB), p3, -10, 0 aTone oscili kEnv, kFreq, 1 outs aTone, aTone endin </CsInstruments> <CsScore> f 0 36000; play for 10 hours e </CsScore> </CsoundSynthesizer>
If you want Csound to give any sort of control data to PD, you can use the opcodes outvalue or chnset. You will receive this data at the second outlet from the right of the csoundapi~ object. The data are sent as a list with two elements. The name of the control channel is the first element, and the value is the second element. You can get the values by a route object or by a send/receive chain. This is a simple example:
<CsoundSynthesizer> <CsOptions> </CsOptions> <CsInstruments> ;Example by Joachim Heintz sr = 44100 nchnls = 2 0dbfs = 1 ksmps = 8 instr 1 ktim times kphas phasor 1 outvalue "time", ktim outvalue "phas", kphas*127 endin </CsInstruments> <CsScore> i 1 0 30 </CsScore> </CsoundSynthesizer>
SEND/RECEIVE BUFFERS FROM PD TO CSOUND AND BACK
A PD array can be sent directly to Csound, and a Csound function table to PD. The message tabset [tabset array-name ftable-number] copies a PD array into a Csound function table. The message tabget [tabget array-name ftable-number] copies a Csound function table into a PD array. The example below should explain everything. Just choose another soundfile instead of "stimme.wav".
<CsoundSynthesizer> <CsOptions> -odac </CsOptions> <CsInstruments> sr = 44100 ksmps = 8 nchnls = 1 0dbfs = 1 giCopy ftgen 1, 0, -88200, 2, 0 ;"empty" table giFox ftgen 2, 0, 0, 1, "fox.wav", 0, 0, 1 opcode BufPlay1, a, ipop ifn, ispeed, iskip, ivol xin icps = ispeed / (ftlen(ifn) / sr) iphs = iskip / (ftlen(ifn) / sr) asig poscil3 ivol, icps, ifn, iphs xout asig endop instr 1 itable = p4 aout BufPlay1 itable out aout endin </CsInstruments> <CsScore> f 0 99999 </CsScore> </CsoundSynthesizer> ;example by joachim heintz
Make sure that the Csound vector size given by the ksmps value, is not larger than the internal PD vector size. It should be a power of 2. I'd recommend to start with ksmps=8. If there are performance problems, try to increase this value to 16, 32, or 64.
The csoundapi~ object runs by default if you turn on audio in PD. You can stop it by sending a "run 0" message, and start it again with a "run 1" message.
You can recompile the .csd file of a csoundapi~ object by sending a "reset" message.
By default, you see all the messages of Csound in the PD window. If you don't want to see them, send a "message 0" message. "message 1" prints the output again.
If you want to open a new .csd file in the csoundapi~ object, send the message "open", followed by the path of the .csd file you want to load.
A "rewind" message rewinds the score without recompilation. The message "offset", followed by a number, offsets the score playback by an amount of seconds.