Overlap-Add vs Overlap-Save convolution methods

I have some impulse responses (all sample rates, so should be same as the file’s sample rate) that i want to use to EQ my headphones. Which convolution method out of Overlap-Add or Overlap-Save will produce a more accurate (to the source file/EQ impulse response) convolution and/or sound better? Will there be any difference at all in the output from the two different methods, or, given my use case, are they just two different ways of arriving at the same output? Thanks. @jussi_laako

While waiting for Jussi to reply, I’ll add that I moved from using convolution to HQPlayer’s parametric EQ, for headphones EQ.

Prefer it this way.

Just one line of text for each channel, like:


With HQPlayer you use just highest rate one and HQPlayer scales it to match other rates as needed.

Practically yes, default is overlap-add which is a good choice.

1 Like

Thanks for that information. Does HQplayer convolution carry out the convolution at an equivalent point in the signal path to Roon’s built in convolution? Would there be an advantage to using HQplayer convolution over Roon’s built in convolution engine?

I don’t know about Roon’s, but if you use HQPlayer, it is better to perform all DSP in HQPlayer.

Likely biggest differences and benefits are for DSD → DSD paths.