RIP Prof. Angelo Farina — ESS Pioneer

[thend]

Hi everyone,

I’m sad to share that Prof. Angelo Farina, a major figure in acoustics and audio-measurements, passed away on 22 March 2025.

One of his biggest contributions was the Exponential Sine Sweep (ESS) method: a powerful way to measure impulse responses and separate linear response from distortion. His work on ESS has become a standard tool in audio measurement.

Passing away of Prof. Angelo Farina - EAA Fenestra - European Acoustics Association

About the European Acoustic Association, its national societies, board and executive council.

euracoustics.org

 

[myHRTF]

This is sad news!
Prof. Farina contributed a lot to the field of audio, and he will be remembered for that by the audio community!
In fact, I still remember hacking the content of his AES paper on the ESS method into MATLAB more than 15 years ago. Grazie mille, professore!
My thoughts are with his family.

 

[SSS]

Did not know of it. Sad. Gone too early. RIP

 

[andyc56]

That is very sad

At the bottom of the REW homepage, in the "Reference Material" section, John Mulcahy lists one of Farina's many papers in the reference below. This is for the exponential sine sweep method used in REW:

Angelo Farina, "Simultaneous measurement of impulse response and distortion with a swept-sine technique", 108th AES Convention, February 2000. Available as entry number 134 in this list of papers

I tried going to that page, but it timed out. In case the site is taken down, which I hope it is not, here's a link to the Wayback Machine version of the page, from which you can download many other interesting papers in addition to that one.

 

[Curvature]

Very sad. He was an important person in the science.

 

[amirm]

Oh no! The world of acoustic measurements has benefited so much from his discovery of log sweeps. If you watch his video on it, you see that he used very primitive audio workstation software and forced it to perform this test.

As an interesting aside, even though his class was in Italy, he forced his students to speak English and taught the same! He wanted to make sure that his students could read the body of research out there in English. I watched his entire semester lecture some 15 years ago and it was so wonderful the simple way he explained things. Nothing stuffy like typical college class.

May he rest in peace and memory of him be a blessing for his family and friends.

 

[Waveform Fidelity]

Indeed it is very sad that he is no longer with us. I found an old email exchange I had with him in 2005 on the general topic of analysing the impulse response and the motivation for flat linear phase response.

On the FFT topic, he made more aware that the FR obtained from the IR doesnt not reveal the all import transient response of the loudspeaker and he suggested instead to use the Gabor transform. (SFST) I found this thougt provoking since few if any of use analyses the raw impuse response, but only the steady state FR.

Here is couple of a snip of one conversation

Angelo Farina <[email protected]>
wrote:

If You use Fourier for passing from the impulse response to the frequency response function (FRF), You are inherently forcing yourself in the strict boundaries required by Fourier theorem: stationary signal, no transients! Although it is true that the FRF still contains the complete characterization of the system's transfer function (provided that it is a Linear, Time-Invariant system) this information is presented in a way which is misleading in terms of representation of the human hearing mechanism. Our ear does not use FFTs, it uses a parallal filter bank, with a time integration constant variable with frequency. A sort of family of wavelets! So, when the signal is not stationary (as with music), a single, very long FFT does not exploit the information which is contained in the impulse response, as packs all the "TAIL" in the phase information, which is ambiguous... OK, You can unwrap the phase getting a better picture, but no one looking at the magnitude/phase plot can tell You how long is the reverberant tail at a given frequency...

We were also discussing the phase of the sound field and here is a bit mor of that conversation regarding the Hilbert transform to examine the imaginary part:

The Hilbert transform is a time domain representation of the impulse response of a system, which represents a sort of imaginary waveform. The pair constituted by the real impulse response and its imaginary counterpart is called the AnalyticalSignal. You can get an "instantaneous phase" information by the ratio of imaginary and real part, for any sample of this time-domain representation.

This "instantaneous phase" has absolutely no trivial relationship with the phase information of the FRF, in frequency domain. These are two very different physical concepts, which are sons of two different facts: the frequency-domain phase is a representation of thedelay which occurs during the propagation (and hence, of the duration of the tail, although in a not trivial way). Instead, the "intataneous phase" obtained by the Hilbert transform is a measure of the "reactive" part of the sound filed, that isthe energy oscilating without propagation. However, this is only a rough estimate of the reactivity: the modern energetic analysis of sound fields is based on the Schiffer/Stanzial theory, which provides a much more correct analysis of the reactivity of the sound field, and can even give You the directional information about the spatial distribution of propagating and oscillating energy. I can provide You with some of their papers, if You want to study the topic, although I think it is irrileveant for Your goals. However, any serious acoustical laboratory is nowadays equipped with a Sound Intensity probe (usually mono-directional), or even of a full 3D probe (the so-called Soundfield microphone), so noone nowadays relies anymore on the obsolete Hilbert transform for "guessing" theimaginary part of the impulse response, as now we can directly MEASURE it!