skrasms

I am trying to get more into non-linear feedback systems and look at what can be done in real-time inside Max. I have been avoiding Java (not another runtime!), but may have no choice.

For example, taking the terribly low-fidelity non-linear difference equation...

y[n] = y[n-1] + T*((x[n]>=y[n-1])*2 – 1)

...where each sample depends on the input and the last sample of the output. Obviously care has to be taken with the result to get rid of DC offset and keep the amplitude range limited, but I am just focusing on implementing similar sample-level math.

I tried working with poly~ so I could change the vector size to 1 (or at least very low relative to the rest of the patch), but it makes everything very unstable and kills my CPU. Sometimes the DSP in Max stops turning on or drops out.

What other methods are there? LUA? A different environment? Should I just cave and learn about making mxj objects?

I am trying to get more into non-linear feedback systems and look at what can be done in real-time inside Max. I have been avoiding Java (not another runtime!), but may have no choice.  

x[n] = current input
y[n] = current output
y[n-1] = last output
T = Sample Period (1/Fs)

...where each sample depends on the input and the last sample of the output. Obviously care has to be taken with the result to get rid of DC offset and keep the amplitude range limited, but I am just focusing on implementing similar sample-level math. 

I tried working with poly~ so I could change the vector size to 1 (or at least very low relative to the rest of the patch), but  it makes everything very unstable and kills my CPU. Sometimes the DSP in Max stops turning on or drops out. 

What other methods are there? LUA? A different environment? Should I just cave and learn about making mxj objects?


single-sample-feedback-methods

Java allows you to program such kind of things pretty easily without 

having to take care of complicated compiler issue. AFAIR, there should 

be an example of filtering somewhere in the Max folder.

Feel free to post any question on the java forum.

> You may even, after the newer SDK is released, look into writing 

> your own MSP externals in C(will require learning C a bit but at 

> least this is easier than writing software at the operating-system 

> level). This would probably be the best way of manipulating at the 

> sample level because with the "perform" function of your MSP object, 

> you can establish control within a signal vector without having to 

> worry about changing vector sizes to 1 sample within the DSP-status 

> Universal Binary SDK for Max4.6.3 is still here:

https://cycling74.com/twiki/bin/view/ProductDocumentation/MaxUBSDK

> The newer SDK soon to come out will allow you to add features 

> particular to Max5(or so I hear from the dev-list) but externals 

> written according to the older UB-SDK can still work pretty well 

Java allows you to program such kind of things pretty easily without  
having to take care of complicated compiler issue. AFAIR, there should  
be an example of filtering somewhere in the Max folder.

>
> You may even, after the newer SDK is released, look into writing  
> your own MSP externals in C(will require learning C a bit but at  
> least this is easier than writing software at the operating-system  
> level). This would probably be the best way of manipulating at the  
> sample level because with the "perform" function of your MSP object,  
> you can establish control within a signal vector without having to  
> worry about changing vector sizes to 1 sample within the DSP-status  
> window.
>
> Universal Binary SDK for Max4.6.3 is still here:
> https://cycling74.com/twiki/bin/view/ProductDocumentation/MaxUBSDK
>
> The newer SDK soon to come out will allow you to add features  
> particular to Max5(or so I hear from the dev-list) but externals  
> written according to the older UB-SDK can still work pretty well  
> with Max5. Best of luck!


Single-Sample Feedback Methods