Latency and FFT parameters


I am trying to understand the latency of various flucoma analysis tools. I have a hard time understanding the dependency it has in relation with the FFT parameters. I am using the MAX version.

For example, with fluid.onsetslice~, the latency is equal to the Hop size specified in the fftsettings. But with fluid.sinefeature~ or with fluid.chroma~ it seems that the Hop size has no influence on the latency. The latency there is related to the FFT size.

I do not know the exact computations done in each of these tools but I was expecting the latency of FFT related tools to be governed by the Hop size more than by the FFT size.

I may miss something rather obvious… but any help in understanding this latency issue would be very welcome.

Thanks a lot in advance.

I can’t speak to massive detail, but I believe it comes down to the algorithm at play.

fluid.onsetslice~ only requires an FFT window amount of time to compute, and since it’s doing it at a rate of the hop, your latency is then a hop.

Something like fluid.hpss~ does different/more stuff, so the latency there is a function of the FFT window and the harmfiltersize.

Not checked across all the objects in detail, but most are hop, or window+hop, but a few others also take other parameters into account, depending on what is required.

At least now all the objects explicitly report their latency as a value. In earlier versions you had to calculate it yourself based on all the fiddly-ness above!

Thank you very much for the feedback. I agree it is good to clearly know the latency value!

1 Like