11-03-2019, 10:45 PM
Very cool, thanks for sharing.
At the moment, the tone generator in fact needs to be reused like you did. The problem is that phyphox currently only supports playing back a sound wave that has been calculated by the analysis modules of the experiments. This is slow, requires pre-calculation (an analysis run) before playback and has plenty of problems if the frequency does not perfectly match into the playback period. We are planing on adding a parametric sound generator output, so you just need to send the parameters of the tone to an output and phyphox will process everything in the background, which should be faster and independent of a buffer period.
Because of the current limitations, there is one thing that you could change which might help with your problems:
At the moment, you have set the "sleep" attribute of your analysis block (found on the main tab) to zero, which means that in theory a new audio snipped will be calculated as fast as possible and that the analysis is not waiting for the audio buffer to be actually filled. Therefore, depending on the phone (this can differ heavily between different systems due to different minimum recording buffers and processing speed), you might have one or multiple of the following problems:
- Your analyzed recording might contain multiple frequencies
- Your frequencies might change extremely fast with crackling sounds since there is no control over phase changes or gaps between playbacks.
- The analysis process might run a few times with an empty buffer before running with a rather well-filled buffer.
So, setting sleep to a value of 0.2 "might" solve these problems and your experiment should run almost identically, with the only difference that you get "only" about 5 values per second, which is probably plenty (not sure about your requirements, though).
At the moment, the tone generator in fact needs to be reused like you did. The problem is that phyphox currently only supports playing back a sound wave that has been calculated by the analysis modules of the experiments. This is slow, requires pre-calculation (an analysis run) before playback and has plenty of problems if the frequency does not perfectly match into the playback period. We are planing on adding a parametric sound generator output, so you just need to send the parameters of the tone to an output and phyphox will process everything in the background, which should be faster and independent of a buffer period.
Because of the current limitations, there is one thing that you could change which might help with your problems:
At the moment, you have set the "sleep" attribute of your analysis block (found on the main tab) to zero, which means that in theory a new audio snipped will be calculated as fast as possible and that the analysis is not waiting for the audio buffer to be actually filled. Therefore, depending on the phone (this can differ heavily between different systems due to different minimum recording buffers and processing speed), you might have one or multiple of the following problems:
- Your analyzed recording might contain multiple frequencies
- Your frequencies might change extremely fast with crackling sounds since there is no control over phase changes or gaps between playbacks.
- The analysis process might run a few times with an empty buffer before running with a rather well-filled buffer.
So, setting sleep to a value of 0.2 "might" solve these problems and your experiment should run almost identically, with the only difference that you get "only" about 5 values per second, which is probably plenty (not sure about your requirements, though).