• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Headphones transients

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
Can anyone please help with what conclusions you would read from these impulse response measurements.

Not practical conclusion (like no difference in sound quality) but moreso technical / theoretical conclusions. So I can learn.

My initial question is about the higher peak for the MrSpeakers AEON. I assume the input signal is the same for both. So what does this higher peak actually mean?

Also, the output appears 'cleaner' for the MrSpeakers.

MrSpeakers AEON:
1588774367322.png




Focal Elear:

1588774382997.png
 

617

Major Contributor
Forum Donor
Joined
Mar 18, 2019
Messages
2,403
Likes
5,296
Location
Somerville, MA
The impulse response is simply another way of showing the exact same data in the frequency response. Not particularly useful, especially for headphones.
 
OP
M

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
The impulse response is simply another way of showing the exact same data in the frequency response. Not particularly useful, especially for headphones.

What does the higher peak for the AEON above mean?
 

Fluffy

Addicted to Fun and Learning
Joined
Sep 14, 2019
Messages
856
Likes
1,423
I tried for a long time to connect those impulse responses from Innerfidelity with any sort of audible effect, but I didn't really find anything useful. The frequency response is the most useful measurement in assessing headphones, followed by the THD vs frequency (only if there are severe irregularities), and sometimes CSD plot (again, if there is something that really stands out). All of the square waves, impulse responses, etc doesn't give any information that doesn't already exists in the FR graph. The phase and impedance plot are sometimes useful, but it's effects on sound is much more obvious in the damping factor comparison plots at diyaudioheaven.
 
OP
M

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
mpulse responses, etc doesn't give any information that doesn't already exists in the FR graph.

Noted, but if the peak of an impulse response is double that of another headphones (an example), how does that translate to frequency response?
 

Fluffy

Addicted to Fun and Learning
Joined
Sep 14, 2019
Messages
856
Likes
1,423
Actually I'm not sure. He never really explained the impulse response part of the measurements. Let me correct that – the FR information exists in the square eave graphs. The impulse response should give information similar to CSD plot, but it's not as helpful. The thing is, an impulse doesn't imitate how an actual musical soundwave behaves, so it's hard drawing conclusions based on that. a CSD plot show you the decay vs spectrum, and that is more useful to predict what the headphone will do when playing music.

I don't know if the relative peak height means something. the units of the vertical axis are not clear, and it's unknown if these plots are somehow normalized across different models. Based on the fact that headphones with vastly different sensitivities get to about the same number there, they are probably normalized so you can't draw conclusions from the height alone.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
It's almost impossible to interpret. Tyll was actually wrong on that but not necessarily totally wrong in square wave interpretation.

Real way to interpret is overlaying with a target impulse response. As the target frequency response is not flat, target impulse response is not the delta either.

Better way to see the information is to use frequency response and phase response. And the phase response needs to be excess phase.
 

617

Major Contributor
Forum Donor
Joined
Mar 18, 2019
Messages
2,403
Likes
5,296
Location
Somerville, MA
No transducer with limited bandwidth can reproduce a square wave. Square wave is even less useful than impulse response.

The truth is that even frequency response measurements of headphones are almost meaningless due to measurement variation. Removing and reseating the headphones on the measurement platform produces big variations in frequency response.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
The problem is Tyll's measurements are not so valuable in this regard as there is no reference given (target) and the fake ear canal/Pinna resonances are also shown and sometimes more prominent than the actual ringing.
One should realize 'speed' and 'transients' are in the 6 to 7kHz range not in the 20kHz and up range.

For this reason I use a wider transient that lies in the 5kHz range and no pinna/ear canal and a scope + real squarewave generator so you can actually see the drivers response. This says a bit more. Though the steepness of the rising edge is limited by that of the microphone.

I happen to believe the squarewave and 'impulse' response is valuable as differences ca seen more clearly on a linear scale than a dB scale.
No... headphones can't reproduce a 'perfect' squarewave. That is impossible but also not the goal. The goal is to see how long the ringing is after an initial 'jump' in position.

Below the HD650's 'needle' response.
Notice the time delay which is caused by the driver-mic distance and speed of sound.
100us needle HD650 no foam, new pads.png


and 440Hz squarewave response.

1mm felt 440.png


green is the applied signal AND target level. Red is the actual measured waveform.
 
Last edited:
OP
M

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
The goal is to see how long the ringing is after an initial 'jump' in position.

Thanks. And in the case of your impulse response measurements (not Tyll's), what does a higher peak mean?

Let’s say one reaches the target peak and another headphones reaches half it’s target peak?

Assume both have clean/equal ringing after the peak.
 
Last edited:

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
The problem is Tyll's measurements are not so valuable in this regard as there is no reference given (target) and the fake ear canal/Pinna resonances are also shown and sometimes more prominent than the actual ringing.
One should realize 'speed' and 'transients' are in the 6 to 7kHz range not in the 20kHz and up range.

For this reason I use a wider transient that lies in the 5kHz range and no pinna/ear canal and a scope + real squarewave generator so you can actually see the drivers response. This says a bit more. Though the steepness of the rising edge is limited by that of the microphone.

I happen to believe the squarewave and 'impulse' response is valuable as differences ca seen more clearly on a linear scale than a dB scale.
No... headphones can't reproduce a 'perfect' squarewave. That is impossible but also not the goal. The goal is to see how long the ringing is after an initial 'jump' in position.

Below the HD650's 'needle' response.
Notice the time delay which is caused by the driver-mic distance and speed of sound.View attachment 62155

and 440Hz squarewave response.

View attachment 62165

green is the applied signal AND target level. Red is the actual measured waveform.
The response from tyll needs to be compensated by minimum phase eq in order to be interpreted. And it's essentially frequency response. It doesn't tell anything more than magnitude and phase response.
You can eq the headphones digitally using minimum phase eq then you will see that headphones now produce almost perfect square wave
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
Thanks. And in the case of your impulse response measurements (not Tyll's), what does a higher peak mean?

Let’s say one reaches the target peak and another headphones reaches half it’s target peak?

Assume both have clean/equal ringing after the peak.

You mean something like this:

sqr.png


It means the headphone sounds darker.

When the peak is overshooting the headphone may sound sibilant or sharp.

You can EQ a headphone 'flat' (at least acc. to the measurement gear+compensation it measures 'ideal') but that won't address ringing and distortion and how it measures on another test rig and sounds like on real ears.
 
Last edited:

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,202
Likes
16,982
Location
Riverview FL
My initial question is about the higher peak for the MrSpeakers AEON. I assume the input signal is the same for both. So what does this higher peak actually mean?

All else equal, higher = louder in an impulse response where the peak has not been "normalized".

This is from 24 measurement sweeps (in-room, with speakers), with each sweep 2dB louder than the previous sweep, overlaid, and not "normalized"

1588836049195.png


When "normalized" they all look (about) the same:

1588836086278.png


In the case of different headphones, all is not equal, so, ???
 
Last edited:

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
What Ray is showing is that when one doesn't normalize (or doesn't have a reference) the plots are pointless.

It is a bit more complicated than that though. In my plots the A440 (very easy to reproduce) is used as a reference (set at the 100% mark Ray is talking about). All other frequencies are then relatively higher, lower or the same.
In the case of my plots the top of the green line = 100% and should be reached.
When the pulse (which is much higher in frequency) is peaking above the green line the treble is louder than the mids and thus brighter/clearer/sharper and maybe even sibilant. When it remains below the green horizontal reference (target) the headphone has subdued upper mids/lower treble and thus sounds 'darker'.

In the case of Tyll's measurements all bets are off as the reference isn't there and the pulse is too short to ensure it could be reached.
It is a meaningless plot and thus says nothing about frequency response, nothing about ringing (as the fake ear canal rings and is also in the plot).
A pointless measurement. Just like the frequency response plots from Tyll have no relation to how they are perceived as the wrong compensation/correction has been used when he started. Similar to the pulse plots he later found out more but because he did not want to redo all reports and articles left it as it is. You can compare frequency plots of Tyll's measurements to his other measurements though.
 
OP
M

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,636
Likes
2,809
What Ray is showing is that when one doesn't normalize (or doesn't have a reference) the plots are pointless.

It is a bit more complicated than that though. In my plots the A440 (very easy to reproduce) is used as a reference (set at the 100% mark Ray is talking about). All other frequencies are then relatively higher, lower or the same.
In the case of my plots the top of the green line = 100% and should be reached.
When the pulse (which is much higher in frequency) is peaking above the green line the treble is louder than the mids and thus brighter/clearer/sharper and maybe even sibilant. When it remains below the green horizontal reference (target) the headphone has subdued upper mids/lower treble and thus sounds 'darker'.

In the case of Tyll's measurements all bets are off as the reference isn't there and the pulse is too short to ensure it could be reached.
It is a meaningless plot and thus says nothing about frequency response, nothing about ringing (as the fake ear canal rings and is also in the plot).
A pointless measurement. Just like the frequency response plots from Tyll have no relation to how they are perceived as the wrong compensation/correction has been used when he started. Similar to the pulse plots he later found out more but because he did not want to redo all reports and articles left it as it is. You can compare frequency plots of Tyll's measurements to his other measurements though.

Thanks, crystal clear now.
 

noel_fs

Active Member
Joined
Dec 12, 2018
Messages
270
Likes
181
What Ray is showing is that when one doesn't normalize (or doesn't have a reference) the plots are pointless.

It is a bit more complicated than that though. In my plots the A440 (very easy to reproduce) is used as a reference (set at the 100% mark Ray is talking about). All other frequencies are then relatively higher, lower or the same.
In the case of my plots the top of the green line = 100% and should be reached.
When the pulse (which is much higher in frequency) is peaking above the green line the treble is louder than the mids and thus brighter/clearer/sharper and maybe even sibilant. When it remains below the green horizontal reference (target) the headphone has subdued upper mids/lower treble and thus sounds 'darker'.

In the case of Tyll's measurements all bets are off as the reference isn't there and the pulse is too short to ensure it could be reached.
It is a meaningless plot and thus says nothing about frequency response, nothing about ringing (as the fake ear canal rings and is also in the plot).
A pointless measurement. Just like the frequency response plots from Tyll have no relation to how they are perceived as the wrong compensation/correction has been used when he started. Similar to the pulse plots he later found out more but because he did not want to redo all reports and articles left it as it is. You can compare frequency plots of Tyll's measurements to his other measurements though.
Could you link a comparison of tylls measurements?
 
Top Bottom