“Human ears are better than measurements!” / “Humans can hear things we can’t even measure!”

[Keith_W]

Before @amirm's concert ...

After @amirm's concert.

 

[mpw]

You’re “taking the piss” here, right?

Do you understand what I am saying ?

 

[Waveform Fidelity]

You’re “taking the piss” here, right?

Well I think he could be alluding to how some people listen to music from their phone speakers and they can recognize the favorite music and enjoy it - because the brain is filling in the gaps... rather large ones!!!

 

[mpw]

I don't really know what ahofer is referring to but he could say exactly what he means. If we talking about 'gaps we are talking about very small differences between sounds that the human brain can make whole, like a piano chord that the human can clearly hear but maybe not the exactness/lackof a measurement but its not important to the listener because the whole is the music.

 

[MarkS]

during a huge speaker search some years ago in which I listened to tons of different speakers, when I played some piano track on one loudspeaker I was taken aback. More than any other loudspeaker I had the sensation of actually sitting in front of piano keys striking a large solid piano.

So which speaker was it??

But also, it would be a rather unusual recording that is supposed to sound like what the pianist hears, rather than what an audience member would hear at a notably larger distance ...

 

[danadam]

But also, it would be a rather unusual recording that is supposed to sound like what the pianist hears

Unusual? I have the impression that this happens more often than not, i.e., lower notes more on one side and higher notes more on the other. There is even thread here about this "30ft piano" effect

The Truth about many "Audiophile" Piano Recordings

I listen to a lot of classical and jazz in general, so when I establish if something sounds "well", piano features very high on the things I'd listen to. As does percussion and human voice - and an entire classical ensemble, too (I do avoid the cannon shot in 1812, though, I think that's the...

www.audiosciencereview.com

 

[solderdude]

I don't really know what ahofer is referring to but he could say exactly what he means. If we talking about 'gaps we are talking about very small differences between sounds that the human brain can make whole, like a piano chord that the human can clearly hear but maybe not the exactness/lackof a measurement but its not important to the listener because the whole is the music.

Well we can measure a single note being played over its entire time period far beyond audible limits.

Measurements can't say anything about you, me or anyone else liking what they hear. At least usually when measurements suggest high sound quality and good recordings are used the 'sound quality' is usually deemed good so there certainly is good correlation in general.

Analyzing a complete piece of music is nigh on impossible to do with measurements nor reporting what was measured. Yet the human brain can do so much better and you can even tell if you like the music and the sound quality.

That’s exactly what brains and hearing is far superior.
Hearing is not superior where it comes to being a reliable and accurate analyzer but can be used that way ( but limited) when as much as possible bias has been removed.

 

[uliwer]

Well, I think sometimes it would help musicians if they applied just a little bit of measurement to their work. One string player may think he has perfect pitch when he is way above 440hz, the other may think so as well when he is way below. So both are off, even though they hear and „feel“ they are right. With measurement they might find a path to correct their errors. When a composition makes a clear distinction between piano, mezzoforte and forte, there has to be a difference of more than 1db between the different parts - very often it is not. The musician may have „felt“ that he played louder though in measured reality he failed to do so.

Same for recordings. Just one example: Handel‘s Suites for harpsichord played by Scott Ross were recently re-released in „Hires“. The whole recording is drowned in enormous low bass frequencies which sound as if a 16’ container was thumped on ground by a crane. The strings of an harpsichord can never produce such results, it seems as if there was a disturbing reverberation from the floor and the large room caused by the instrument’s mechanics. Obviously nobody noticed this during the recording, perhaps as they were just listening for the music. If only someone had looked into the measured spectrum of this recording he would have found out that they had a problem.

 

[Dogill]

If someone claims a violin is a Stradivarius what measurements can you use to show it's not ? I assume you are in possession of a recording of a real one.

 

[solderdude]

If someone claims a violin is a Stradivarius what measurements can you use to show it's not ? I assume you are in possession of a recording of a real one.

Perhaps look closely at and in the violin. The sound of a violin depends on so many factors including how notes are played.
I also don't think your average music lover is able to tell anyway. Even knowledgeable people will have a hard time doing this blind if it is even possible at all.

 

[voodooless]

Million-dollar Strads fall to modern violins in blind ‘sound check'​

Science | AAAS

www.science.org

 

[voodooless]

Million-dollar Strads fall to modern violins in blind ‘sound check'​

Science | AAAS

www.science.org

Now, if one were to do the same experiment, but with a recorded version of the same violins, I have no doubt that the preferences would be the same. So, the recording would not influence the outcome, even though it would not capture exactly how we would hear things in a live environment.

 

[DenonMate]

If you can raed waht is wetritn here, yuor biarn has done the mian wrok. The oigrnial maetiral is jlbemud leetrts. The ctupomer wlil crautpe tihs jelmbud lrteets bit-pcfretly. Nhoitng mroe.

 

[ahofer]

Do you understand what I am saying ?

Perhaps not. It doesn’t seem relevant to measurements and fidelity, so I’m not sure what the point is. Usually raising those sorts of points in the context of a measurement debate is intended to sow confusion about the objective aspects of audio reproduction, but perhaps I misjudged.

 

[ahofer]

If someone claims a violin is a Stradivarius what measurements can you use to show it's not ? I assume you are in possession of a recording of a real one.

We aren’t talking about verifying provenance here, we are talking about whether the sound waves produced by the two instruments are audibly indistinguishable.

There are apparently some inaudible but measurable tells with the sound, appearance, and chemistry (via tiny finish samples) of older instruments. This whole digression seems like a red herring.

Audio reproduction first must reproduce sound waves as accurately as possible, at least to the point of human inaudibility. The content of those sound waves is a matter of source material, not reproduction. Similarly, the brain’s ability to de-garble audibly distorted sound waves is not really relevant to the fact that they are audibly distorted.

 

[uliwer]

For measurements you need sufficient instruments which were made for the purpose. Instruments which measure the (recorded) sound of a violin are certainly not sufficient to proof or disproof the violin‘s origin. You need instruments which can analyse the structure of the wood the thing was made of, how old is this wood, does it come from the region of Cremona, does it fit into the pattern of known violins by Stradivari etc. Though you certainly need measurements, „feeling“ don‘t help.

 

[rjp]

If someone claims a violin is a Stradivarius what measurements can you use to show it's not ? I assume you are in possession of a recording of a real one.

Interesting question for a science project! First of all, based on the question as worded, I take it to mean the test need not have anything to do with how it sounds to humans, but can instead be based entirely on measurements. With just one Stradivarius sample I do not think this is possible, but if you have access to many true strads and non-strads, you could probably train a neural network encoder to identify the strad class as follows.

Digitally sample the violins and use a common neural network encoder to generate a feature vector representation of the sound using a fixed number of bits (start with, say, 128). Train this network using as many true Stradivarius samples as you can get using either an auto-encoder or triplet loss function topology. Over time, the network will learn how to generate feature vectors that best separate the class of Stradivarius from others.

Once trained, this network should be able to validate any unknown violin in a single sitting.

Of course this same approach can be used to identify amplifiers or any audio equipment, which I assume is the reason for the analogy.

There are always tiny details in signals produced (fingerprints so to speak). In fact, it is often possible to distinguish two different serial numbers of the same amplifier model this way. This technique is called single emitter identification, and yes, it works.

 

[Waveform Fidelity]

There was also a debate on the LINFIR thread about these violins which you might find of value. We just agreed to stop debating in the end.

 

[voodooless]

For measurements you need sufficient instruments which were made for the purpose. Instruments which measure the (recorded) sound of a violin are certainly not sufficient to proof or disproof the violin‘s origin. You need instruments which can analyse the structure of the wood the thing was made of, how old is this wood, does it come from the region of Cremona, does it fit into the pattern of known violins by Stradivari etc. Though you certainly need measurements, „feeling“ don‘t help.

Listeners had the same preference for modern violins as the players, so no, you don't need any of this; you only need the sound. And as @rjp mentions, you can likely even train a model on it to distinguish a Stradivari from a plain violin with high confidence and accuracy.

 

[MattHooper]

So which speaker was it??

This one.

(Didn’t want to go down the road of defending the speaker again, but the relationship of what Amir was describing to what I experienced in what drew me to that speaker seemed relevant)