Time Domain in measurements.

[DNCAgain]

While watching an interview with an audio measurement blogger, I noticed that some people believe it's not objective that he doesn't measure impulse response and phase distortion.

I've read a few posts and learned that after a Fourier transform, the impulse response is virtually equivalent to the frequency response, and phase distortion/group delay doesn't seem to have a significant impact. (Unfortunately, I don't fully understand why, as I'm unable to access Amir's YouTube videos.)

What is the relationship between impulse response and phase distortion/group delay? For what reasons are they not discussed frequently?

 

[amirm]

For what type of product?

In general, time and phase are not important parameters. Audiophiles have latched on these terms as to then use them to complain about measurements. Your room and headphones massively distort both of these factors. And at any rate, much of our understanding of hearing is in frequency domain, not time.

In my measurements of headphones and speakers I do show group delay and step response. But they rarely show anything useful.

What some people do is that they subject audio electronics to type of time domain signal that does not exist in real life. A step signal by definition has infinite bandwidth. A 44.1 kHz music file has a bandwidth of only 22.05 kHz. No way would that impulse look like an impulse if you limited its bandwidth that low.

 

[DNCAgain]

For what type of product?

In general, time and phase are not important parameters. Audiophiles have latched on these terms as to then use them to complain about measurements. Your room and headphones massively distort both of these factors. And at any rate, much of our understanding of hearing is in frequency domain, not time.

In my measurements of headphones and speakers I do show group delay and step response. But they rarely show anything useful.

What some people do is that they subject audio electronics to type of time domain signal that does not exist in real life. A step signal by definition has infinite bandwidth. A 44.1 kHz music file has a bandwidth of only 22.05 kHz. No way would that impulse look like an impulse if you limited its bandwidth that low.

IEMs Headphones dac and amplifiers

 

[DNCAgain]

IEMs Headphones dac and amplifiers

They also send me a document to prove that it's a important data
I can't understand it due to my poor English

 

[amirm]

There is zero reason to worry about time domain with DACs and amps. IEMs as well for that matter.

 

[TimVG]

For speakers: neutral on- and off-axis magnitude is where the biggest audible gains are.

That said, there is indeed credible work that claims excess group delay in the lower frequencies can be audible with the right program material. This could be one reason why loudspeakers with large drivers and no crossover in the bass range are often perceived as more "punchy" or tighter compared to systems where the low end passes through a crossover network, the added phase shift from the crossover introduces group delay that smears transients enough to be perceived on certain content, even if it doesn't show up obviously in the frequency response.

This is not my research, and I'll keep my thoughts about it to myself for the time being, but for the sake of discussion I am mentioning it.

 

[DNCAgain]

What about speaker s?

There is zero reason to worry about time domain with DACs and amps. IEMs as well for that matter.

 

[amirm]

They also send me a document to prove that it's a important data
I can't understand it due to my poor English

That is a book, not a paper.

Here is the issue: there is no study indicating preference for speakers being based on time domain. Everything we know points to proper frequency response, on and off axis. If someone wants to make the phase linear, go ahead. But it better not come at the expense of frequency domain. Sadly, often it does.

 

[DNCAgain]

Have researchers specified how much is considered "excess"?

, there is indeed credible work that claims excess group delay in the lower frequencies can be audible with the right program material.

 

[DNCAgain]

That is a book, not a paper.

Here is the issue: there is no study indicating preference for speakers being based on time domain. Everything we know points to proper frequency response, on and off axis. If someone wants to make the phase linear, go ahead. But it better not come at the expense of frequency domain. Sadly, often it does.

Thanks for that!

 

[voodooless]

They also send me a document to prove that it's a important data
I can't understand it due to my poor English

My AI summarized it as this:

an enthusiast’s cherry-picked brief for active DSP linearization, with some real references mixed into a lot of overclaiming and square-wave fetishism.

What is the relationship between impulse response and phase distortion/group delay? For what reasons are they not discussed frequently?

It’s just math.. what is there to discuss? You can convert from one to the other using a Fourier transform.

 

[TimVG]

Have researchers specified how much is considered "excess"?

Here and here you can find the two studies

 

[DNCAgain]

Here and here you can find the two studies

Thanks for sharing!

 

[DNCAgain]

Wow! It seems I shouldn't blindly believe in everything presented in the form of research.

My AI summarized it as this:


It’s just math.. what is there to discuss? You can convert from one to the other using a Fourier transform.

 

[DNCAgain]

Wow! It seems I shouldn't blindly believe in everything presented in the form of research.

But to what extent and in what ways is square-wave measurement actually useful?

 

[voodooless]

Wow! It seems I shouldn't blindly believe in everything presented in the form of research.

It's not research. It's just masquerading as such.

 

[MaxwellsEq]

But to what extent and in what ways is square-wave measurement actually useful?

It's not very useful. Every sound source, carrier or receiver on the planet is band-limited, especially your ears.

Since a pure square wave or step response requires infinite bandwidth, every method of conveying it as sound will "mess up" the squareness.

There are many mathematically correct "signals" which are simply not reproducible. Their fundamental value is testing edge cases - e.g. something not working properly.

 

[Maxim-ICOM]

In general, time and phase are not important parameters. Audiophiles have latched on these terms as to then use them to complain about measurements. Your room and headphones massively distort both of these factors. And at any rate, much of our understanding of hearing is in frequency domain, not time.

It is too simple view on what we have in real world. Brain + ears are very sensitive to amplitude and phase.

Imagine that you have some transducer 6m from you on left and play some tones. Ears + brain will detect it's direction with phase shift and amplitude.
Amplitude difference for L amnd R ear is by very small (0.2db) and many claim that that small difference in to small to hear. There will be also phase shift.
Phase + amplitude + reflected signal amplitude and phase will allow the brain to locate objects.

Another known fact is that the first signal what comes to brain is accepted as "main" and will be louder then what comes with delay.
Speakers are non-linear devices as with frequency response too with phase and if phase have some shift in some frequencies in one speaker but differs in other while both have same frequency response the brain will think that speakers have different "character" - one may sound brighter, another more bassy.

Sadly measurement software can only show us GD, Phase, Impulse, FR, STEP, but it is a kind of same things presented in a different 2D graphs and while we look on it it is hard to decide which one is better if they are very close to each other. There are some things what we need to see in speakers measurements.

There would be good presentation if "time parameters" if FR+Time will be presented as "calibrated to brain+ears", but we cannot measure brain and ears, as we do not know what to measure and each compo is unique.

 

[MAB]

They also send me a document to prove that it's a important data
I can't understand it due to my poor English

Even in native English, it's not useful. A collection of personal anecdotes from the author, completely unsupported. And some blatantly false assertions and incorrect statements of fact. With the usual word-salad of random audiophile nonsense. It uses technical terms in a meaningless way. Some of the sections are empty of words: 4. Depth and 5. Resolution, and I am thankful for that since the rest of the sections are also empty of meaning.

It was funny to see the author suggest (or one of the blogs he cited suggested) one piece of DSP was incapable of correcting for Phase and Group Delay, while another was not. Odd since anybody can do a correction with basic DSP, like demonstrated here:

First speaker to reproduce a 60hz sq wave

ASR is a tough crowd to impress. People would have questions, "shut up and take my money!" would be the last statement on the last page if it would ever happen. Perhaps doing some market research was an honest intention but clickbaits aren't popular over here. It has been more than 24 hours...

www.audiosciencereview.com

The above square-wave demonstration is not unique to high end processors, and is a bit of a parlor-trick rather than some fundamental discovery about speakers and phase.

 

[amirm]

It is too simple view on what we have in real world. Brain + ears are very sensitive to amplitude and phase.

Only in their differntial.

Imagine that you have some transducer 6m from you on left and play some tones. Ears + brain will detect it's direction with phase shift and amplitude.
Amplitude difference for L amnd R ear is by very small (0.2db) and many claim that that small difference in to small to hear. There will be also phase shift.
Phase + amplitude + reflected signal amplitude and phase will allow the brain to locate objects.

We are not talking about differences between channels. Those amplitude and timing differences indeed change localization. Such information is massively in the content as opposed to being in the equipment.

The conversation here is about both channels being identical, yet timing/phase mattering. Such differences have been shown with headphones and anechoic chambers but in real rooms, it is a tiny factor if that.