On Thu, 2018-12-13 at 13:21 +0100, Peter P. wrote:

• Roman Haefeli reduzent@gmail.com [2018-12-13 11:58]:

...

Hey all

I once read about a simple and robust way to perform latency measurements with an audio signal.

Explained in a few words, the test signal consists of a sweeping sine tone. The return signal ring-modulates the source signal and the resulting signal consists of two frequencies, the sum ( f_src+f_ret ) and difference ( f_src - f_ret), while the lower frequency is proportional to the latency and can be detected quite easily with with e.g. [sigmund~].

I have troubles finding the name of this algorithm and don't remember the original source. I would like to read more about it and correctly attribute the original author / inventor.

See jack_delay on https://kokkinizita.linuxaudio.org/linuxaudio/

That's a different algorithm and probably much more precise. The one I'm looking for is simpler (and I'd probably know how to implement it in Pd).

Anyway, thanks for the input. It's certainly an interesting approach, too.

Roman

On Thu, 2018-12-13 at 15:09 +0100, Peter P. wrote:

• Roman Haefeli reduzent@gmail.com [2018-12-13 13:52]:

...

On Thu, 2018-12-13 at 13:21 +0100, Peter P. wrote:

...

• Roman Haefeli reduzent@gmail.com [2018-12-13 11:58]:

...

Hey all

I once read about a simple and robust way to perform latency measurements with an audio signal.

Explained in a few words, the test signal consists of a sweeping sine tone. The return signal ring-modulates the source signal and the resulting signal consists of two frequencies, the sum ( f_src+f_ret ) and difference ( f_src - f_ret), while the lower frequency is proportional to the latency and can be detected quite easily with with e.g. [sigmund~].

I have troubles finding the name of this algorithm and don't remember the original source. I would like to read more about it and correctly attribute the original author / inventor.

See jack_delay on https://kokkinizita.linuxaudio.org/linuxaudio/

That's a different algorithm and probably much more precise. The one I'm looking for is simpler (and I'd probably know how to implement it in Pd).

What about 7.stuff/tools/latency.pd?

Cool, the collection of algorithms is growing :-)

This is again a different approach. The test signal is a pulse and the patch measures time between sending and receiving the click. The cool thing is it's done in a way that doesn't need any "magic" objects like [sigmund].

However, it's still not the one I was describing earlier.

Roman