DSP sampling frequency vs. real sampling frequency
Jun 22 2010 | 2:42 pm
Hello,
I would like to understand something with the loading of a buffer~ object via peek~.
Here is my patch :
If you hit space bar and move the number box until 512 values are displayed, and then press the 'c' key you will load the buffer with this values.
I made it via a coll object and a counter+metro because my values are coming (A1x) with a non constant sampling rate ... (around 200Hz) and I thought that I could load the buffer @200Hz after the coll object.
This was a bad idea. I forgot that Max will load the buffer with the sampling rate in DSP status.
So my question is : If I receive 512 values @ 200Hz from my sensor (so it is around a 2.5 sec buffer), how can I make Max load the buffer with the effective values at the effective sampling rate instead of loading it at 48000 KHz in my case ??
I would like to get the FFT of this buffer, so I would like it to be sampled at the same frequency than my incoming values, and not the frequency of the DSP status. Am I wrong ??
I am quite lost on this issue,
Thanks a lot,
Have a nice day,
Mathieu
- Jun 22 2010 | 2:54 pmYou can send buffer~ an [sr 200 ] message to set its sample rate to 200Hz. I'm not totally sure if that will help with what you're asking.
- Jun 23 2010 | 9:32 amThanks it is better like that, I can get 512 values @ 200Hz in the buffer.Now it is really complicated to get the sonogram (spectrogram) of my buffer, I start thinking that Max is really good to get frequency content of sounds / music files but maybe I should use Matlab to get the frequency content of continuous data stream from accelerometers sensors ?Maybe someone have a solution to get the time-frequency content of a 512 values buffer @ 200Hz (when in DSP status I have fs=48000KHz, I/0 vector size =512 and signal vector size = 64).Here is my actual patch, I have to deal with frequency content after the peek~ object, playing the buffer into a FFT process :We can see that the buffer is well loaded, with 512 values @ 200 Hz. Could someone give me some directions at this point ? (Record a classical FFT and display it into a LCD or using sonogram to have real time frequency content, I don't know how to start and how to deal with FS, I/O and signal vector size).Thanks a lot,Have a nice day,Mathieu
- Jun 23 2010 | 12:21 pmIn spectroscope~ object reference we can read : " In left inlet: The input signal is analyzed and its spectrum is displayed ". Is there a way to know how the signal is analyzed ?If you try to make the following patch work, you will see that the sonogram displays something linked to the buffer content :But it is hard to say what is displayed and how.Any ideas ?Thanks a lot (again ^^),Mathieu
- Jun 25 2010 | 12:43 amForgive me if this shows a complete ignorance of what you're trying to do but could you not simply represent it as an audio rate signal interpolating between the missing samples?rampsmooth~ or line~ comes to mind but there are better ways surely. If its just for display I would think this would be accurate enough.
