[SilverEars]

Why does headphones FR graph not look neutral like a straight line like output of devices?  Why is the high end have lots of peaks?
 
What if it was a complete straight line?  Is that neutral?

 

[higbvuyb]

  Why does headphones FR graph not look neutral like a straight line like output of devices?  Why is the high end have lots of peaks?
 
What if it was a complete straight line?  Is that neutral?

Headphone FR graphs aren't a straight line, because they're measured at the eardrum level.
 
When sound from, say, a real violin hits your ear, the ear amplifies and reduces different frequencies, so when the sound finally reaches your eardrum, the frequencies aren't totally flat.
 
The result ends up something like this: http://en.goldenears.net/en/files/attach/images/252/464/554421849cb29cdf3175f5a0f414c281.jpg
However, the graph is different for every person, and also depends on the direction the sound comes from.
 
 
The high end has lots of peaks because it's hard to design good headphones.

 

[proton007]

  Why does headphones FR graph not look neutral like a straight line like output of devices?  Why is the high end have lots of peaks?
 

 
Headphone FR graph represents the response of the transducer. Most transducers can't reproduce the whole spectrum of frequencies equally well (physical limitations). For instance, lower frequencies need bigger transducer size (woofers and subwoofers are big), and high frequencies need a smaller size (tweeters are small), but headphones try to balance it all with a single transducer, so an 'unbalanced' FR comes as a tradeoff.
 
Secondly, a speaker/headphone is an electrical system that produces physical oscillations, and based on the frequency of the input signal, the speaker/headphones' physical and electrical characteristics vary and affect each other.
 
Regarding the high end peaks and valleys, its because measuring the FR simulates the human ear, and High Frequency sounds tend to have a shorter wavelength and can be easily affected by interference due to reflection in the ear canal.
Also, some headphone makers 'tune' the high frequencies to sound more pleasant.
 

  What if it was a complete straight line?  Is that neutral?

 
 
From a headphone's perspective, yes.
From a listener's perspective, there might be slight adjustment needed, because of the Equal Loudness Contour.
Basically, what this means is that in order for us to perceive two different frequencies as equally loud, their intensity need not be equal.
However, a 'flat' headphone will respond well to an EQ, and is a better candidate for sounding completely neutral.

 

[SP Wild]

Very wrong answer.  A flatline will make a headphone sound bright at reference db.  For some it may not.  An articulate headphone will respond best with EQ - regardless of flatness.

 

[bigshot]

The closer to the goal response the better. When you try to push something farther than it wants to go, you start running into trouble. Especially at low frequencies.

 

[SP Wild]

Yes...I found this with the K701.

 

[SilverEars]

So this means FR respons graph is meaningless that people bring up from time to time?  The Etymotic graph showing how neutral their phones are not really reality?  They should have a perceived graph instead, correct?

 

[SP Wild]

Etymotic was using an older model, Diffuse Field compensation.  At the time it was believed to be a good one, over-time many people reported how not-neutral it sounded.  Dr. Olive at Harman is finalising research that shows differently.
 
If you are going by graphs from innerfidelity (you must pick one graphing system and only one) then the ideal curve would be a downward slope - not at all a straight line.  This goes very well with my subjective experiences.
 
You are fundamentally correct though.

 

[RazorJack]

  So this means FR respons graph is meaningless that people bring up from time to time? 

 
Basically uhm, yeah. You'll also notice how people often make their statements after having seen the frequency response, and then be all like "see, what I heard corresponds with the measurements!" 

 

[SP Wild]

   
Basically uhm, yeah. You'll also notice how people often make their statements after having seen the frequency response, and then be all like "see, what I heard corresponds with the measurements!" 

 
Bingo!
 
But there were a minority handful of people that insisted headphones that average a downward slope as sounding more accurate - upon deaf ears (literally?).  This was well before the new Harman research...funnily enough, the new research seems non-existent to a large number of elitists.

 

[higbvuyb]

  So this means FR respons graph is meaningless that people bring up from time to time?  The Etymotic graph showing how neutral their phones are not really reality?  They should have a perceived graph instead, correct?

Usually, FR graphs are 'compensated'. This means they don't show the actual FR, they show the difference between the FR and the target curve. So the 'ideal' is a straight line.*
It doesn't mean that they're 'meaningless'.
 
e.g. on InnerFidelity, the light grey lines show the FR measurements before compensation.
 
* The target curve is supposed to be perceived as 'neutral' but there are a lot of different target curves and no absolute consensus on which is the best one. Because preference is subjective, you'll have to find the right one for yourself.
 
Etymotic uses a modified DF which has a slight downwards tilt in the treble area (except for the ER4B, which is designed for binaural recordings). They added the downwards tilt because they thought it was too bright. 
 
As SP Wild mentioned, in the past few years, Olive, Welti, and McMullin have come up with another target based on the idea that real listening rooms aren't really DF-like. It has more of a downwards tilt. They've done a bit of research that indicates that people prefer this new target and think it sounds more accurate.

 

[higbvuyb]

   
Basically uhm, yeah. You'll also notice how people often make their statements after having seen the frequency response, and then be all like "see, what I heard corresponds with the measurements!" 

Expectation bias derived from objectivity tends towards fact, while expectation bias derived from myth is still wrong.

 

[castleofargh]

 

  So this means FR respons graph is meaningless that people bring up from time to time? 

 
Basically uhm, yeah. You'll also notice how people often make their statements after having seen the frequency response, and then be all like "see, what I heard corresponds with the measurements!" 

I know what you're talking about and in a way I've seen it loads of times for bad reasons and people telling the complete opposite of what they were telling the day before, just so that it would go with the newly found graph. but given that the measurement is usually pretty close to the truth when it's not actually defining it in the manufacturing process, couldn't we call the action of "someone adapting his perception to a graph", something like "learning"? like a kid slowly getting that orange is not red by being confronted repeatedly to both colors with clear names stuck on both.
I'd rather have a few guys learning to recognize a signature by bending to the graphs. so that later we can all talk about sound using the same basis, rather than having the usual patchwork of neutral signatures going from W4 to ER4 and each team never budging from its own self righteous sens of "neutrality".
I see the graph as one of the tools to set common values and helps talking about sound. in that effort the more people to lean toward what the graph says, the better. no? (given that you didn't draw the graph on painter, and that people actually make the effort to confront what they hear and the graph values).

 

[James-uk]

Usually, FR graphs are 'compensated'. This means they don't show the actual FR, they show the difference between the FR and the target curve. So the 'ideal' is a straight line.*
It doesn't mean that they're 'meaningless'.

e.g. on InnerFidelity, the light grey lines show the FR measurements before compensation.

* The target curve is supposed to be perceived as 'neutral' but there are a lot of different target curves and no absolute consensus on which is the best one. Because preference is subjective, you'll have to find the right one for yourself.

Etymotic uses a modified DF which has a slight downwards tilt in the treble area (except for the ER4B, which is designed for binaural recordings). They added the downwards tilt because they thought it was too bright. 

As SP Wild mentioned, in the past few years, Olive, Welti, and McMullin have come up with another target based on the idea that real listening rooms aren't really DF-like. It has more of a downwards tilt. They've done a bit of research that indicates that people prefer this new target and think it sounds more accurate.

So is a 'perfect ' compensated FR at inner fidelity a straight line? Based on the diffuse field raw data?

 

[higbvuyb]

So is a 'perfect ' compensated FR at inner fidelity a straight line? Based on the diffuse field raw data?

In theory, on InnerFidelity, if the red/blue lines are perfectly horizontal, that's a 'perfect' FR.
 
But this is assuming that their compensation curve is perfect (for you). It won't be, because everyone's ears are different. For example, to SP Wild, a horizontal line sounds unnatural and he prefers a slight downwards slope. You will have to work out what's right for you.
 
 
Also, Inner Fidelity doesn't use a DF as their target, they use something called "Independent of Direction" which is apparently somewhere between DF and Free Field.