DaisySP DSP Library for the Teensy 4.x - now compatible with Raspberry Pi Pico 2
Updated Jan 2026 for Pico 2. Works very well on Pico 2 using just the DaisySP library - no Teensy audio framework needed. See the 2HPico-Sketches archive for examples:
https://github.com/rheslip/2HPico-Sketches
All the examples use both cores on the Pico for efficiency with most of the DSP running on the second core.
The Teensy Audio framework has been ported to the Pico 2 as well https://github.com/ghostintranslation/pico-audio
This is a port of Electrosmith's DaisySP signal processing library
https://github.com/electro-smith/DaisySP
to the PJRC Teensy 4.x for Arduino Teensy and Arduino Pico. This library allows you to use the features of the Teensy Audio library and DaisySP together, or just DaisySP by itself. Teensy Audio is fast and memory efficient but has a limited selection of DSP functions. DaisySP is a bit of a CPU and memory hog but it has some sophisticated audio processing functions. The Teensy 4.x is VERY fast and you can do a lot of signal processing with it. The Raspberry Pi Pico 2 is quite a bit slower than the Teensy 4 but it still capable of fairly complex digital signal processing.
DaisySP consists mostly of code collected from other projects - Csound, Soundpipe, and Mutable Instruments eurorack modules. DaisySP is quite similar to Soundpipe but much better written and documented. I ported Soundpipe first and then realized DaisySP is much better.
DaisySP uses floating point for all DSP operations and as such will run slowly on the Teensy 3.x - this has not been tested. On a Teensy 4.x most DaisySP functions (oscillators, envelope generators etc) consume roughly 1% of the CPU so you could create a polyphonic synth with 10 oscillators, 10 envelope generators and 10 filters and still have lots of CPU left. The physical modelling functions use quite a lot of CPU. The sine oscillator uses up more CPU since its implemented as a trig function. DaisySP has antialiased polyblep oscillators which are quite CPU efficient - much better than the simplistic and noisy waveform generators in the Teensy Audio library.
On the Pico 2 you will typically process audio samples in a tight loop - the examples linked above use the second core of the Pico 2 for this, leaving the first core available for handling physical devices and user interfaces. Take a look at ElectroSmith's DaisySP examples as well to see how the signal flow works.
On the Teensy we use the Teensy Audio framework which uses a somewhat different approach - 128 samples are processed as a block in a background interrupt process. Teensy Audio also uses fixed point math and the ARM Neon DSP instructions for speed. Since DaisySP uses floating point, samples must be converted from integer to float and back for each block of samples when using the DaisySP library functions.
This implementation adds a DaisySP Teensy Audio synth object and a DaisySP Teensy Audio effect object to the Teensy Audio library. The DaisySP synth object has no inputs and one output ie its a generator of audio samples. The DaisySP effect object has one input and one output. The implementation currently supports only one instance of a DaisySP object - more may be possible but I'm not good enough with C++ to figure it out. The simplest setup is a DaisySP synth object to the Teensy Audio Shield object which is set up like this:
AudioSynthDaisySP Mysynth; // create the daisysp synth audio object
AudioOutputI2S out; // audio shield object
AudioControlSGTL5000 audioShield; // control channel
AudioConnection patchCord1(Mysynth,0,out,0); // patch mono synth to right and left channels
AudioConnection patchCord2(Mysynth,0,out,1);
A DaisySP effect object would be set up like this:
AudioEffectDaisySP Myeffect; // create the daisysp effect audio object
AudioInputI2S audioInput; // audio shield: mic or line-in
AudioOutputI2S out; // audio shield object
AudioControlSGTL5000 audioShield; // control channel
AudioConnection patchCord1(audioInput,0,Myeffect,0); // mono input
AudioConnection patchCord2(Myeffect,0,out,0); // patch mono synth to right and left channels
AudioConnection patchCord3(Myeffect,0,out,1);
Teensy Audio processes 128 16 bit integer samples at a time and uses a pool of sample buffers which are passed between audio objects. This is memory and CPU efficient but can make coding audio objects quite complicated because of the sample buffer management. Every audio object has a callback function which processes blocks of samples approximately every 2.3 ms @ 44.1khz sample rate.
In contrast, DaisySP processes one sample at a time using floating point and each function allocates its memory statically. Simple, but uses a lot of memory for things like reverbs and delays and its pretty CPU intensive.
To use DaisySP with Teensy Audio we process 128 samples at a time, convert the floating point results to integer and pass them to the next Teensy audio object in the patch. The DaisySP audio object has a callback function called AudioSynthDaisySP::update (or AudioEffectDaisySP::update) which does this. You must have this function in your sketch and this is where you call DaisySP library functions. Look at the example sketches to see how this works.
Installing the library:
Copy the contents of the DaisySP folder to your arduino/library folder
Copy the file Teensy/Audio/synth_daisysp.h and Teensy/Audio/effect_daisysp.h (the DaisySP audio objects) to your Teensy audio library - usually this will be your_Arduino_installation_directory/hardware/teensy/avr/libraries/audio.
Teensy/Audio/Audio.h has the lines #include synth_daisysp.h and #include effect_daisysp.h added so you can replace your Teensy audio library Audio.h with this file. Its probably better edit your existing Audio.h file - I can't guarantee I will be tracking future changes to the Teensy Audio library.
I decided to structure the library so you have to manually include the DaisySP *.cpp files you are using in your sketch vs compiling the whole library into the sketch. This is a bit of a pain but including the whole library uses almost 500k of program memory and close to 500k of RAM which leaves very little RAM for the rest of your code. There is currently no provision for using the optional PSRAM on the Teensy 4.1.
I have not tested the library extensively. It should be fairly simple to add a DaisySP audio effect object that has two outputs for the DaisySP functions that generate stereo audio. I have not had a need for this yet so its not in the library.
There are compile problems with the resonator objects. I think this an issue with derived classes but I have not been able to fix it as yet.
Originally tested with Arduino 1.85 and Teensyduino 1.53. Latest updates tested with Pico 2 and Arduino 2.2.7.
Rich Heslip rheslip@hotmail.com