Frequency Response Graphs

[KappaKing]

How influential should frequency response graphs be when I'm deciding between headphones?

[Peter Vaughan Truslow]

they show you how the headphones represent different parts of the music. if you like bassy headphones, go for something that doesn't taper off much in the bass region. you generally want something that is smooth and somewhat flat in the midrange

for me, I found it relatively influential, but I also took user account into effect when I bought my M50, and I absolutely love them, and how they minimally color the music.

it doesn't really tell the whole story though. definitely try them out if you can, but remember that when switching headphones, there is about a 1 week period where it will sound weird because you're not used to the sound signature of the headphones.

[creatip123]

How influential should frequency response graphs be when I'm deciding between headphones?

 

The correct answer should be: minimal.

 

FR graphs are used to give a general idea of how the sound is like, and give no information on the actual sound quality. Most of the use is when you're looking for, or avoiding a particular dominant frequency range, i.e. looking for more mid frequencies, or avoiding too much high frequencies.

 

Think about it this way: you're deciding whether or not to buy a particular dish at a diner, and you can read the information on the menu that it got shrimps, corn, vegetables, broccoli, and whatnot. It's useful if you're looking for shrimp dishes, or avoiding broccoli, but say nothing about how tasty it is.

[creatip123]

it doesn't really tell the whole story though. definitely try them out if you can, but remember that when switching headphones, there is about a 1 week period where it will sound weird because you're not used to the sound signature of the headphones.

 

Agreed, but it could work both ways. For 1 week period, there's the wow factor, and after a week, it's just meh....

[Peter Vaughan Truslow]

Agreed, but it could work both ways. For 1 week period, there's the wow factor, and after a week, it's just meh....

for me it just sounds off, as I love the sound signature of my M50s so much, and use them so frequently, that anything else sounds wrong.

I'm not saying they're the best headphone on the earth, not even saying they're the best for the money(although I'd say they might be), but I am saying it's what I'm used to.

edit: nice logo. I love my AD700 too, but for different reasons. fantastic for gaming.

[Guest]

Frequency response graphs are highly informative - IF you know how to read them. They tell a great deal about how a headphone sounds, but the hard part is interpreting the graph and how it translates to the sound of that headphone on your own head.

 

Think about it this way: you're deciding whether or not to buy a particular dish at a diner, and you can read the information on the menu that it got shrimps, corn, vegetables, broccoli, and whatnot. It's useful if you're looking for shrimp dishes, or avoiding broccoli, but say nothing about how tasty it is.

 

This doesn't make any sense. "Contains nuts" is not analogous to a direct measurement of the behavior characterstic of a headphone anymore than it is to what food actually tastes like. A more close analogy would be the breakdown of chemical compounds in a beer and interpreting how those affect taste.

 

Like I said above, a FR graph contains a wealth of information, the only hard thing is extracting it and translating it into meaninful information.

[creatip123]

for me it just sounds off, as I love the sound signature of my M50s so much, and use them so frequently, that anything else sounds wrong.

I'm not saying they're the best headphone on the earth, not even saying they're the best for the money(although I'd say they might be), but I am saying it's what I'm used to.

edit: nice logo. I love my AD700 too, but for different reasons. fantastic for gaming.

 

It's like perfume, where using just one brand could get you a numb nose to that brand, or to the other brands. A friend that work in an audio store once told me that the best is to rotate using headphones, as to keep your ears as objective as possible. For example, too much of a bassy headphone, and the ears could go numb on basses. When trying on other headphones that's actually bassy, it could feel lacking in bass.

 

The AD700 is collecting dust now. I use HE-400 for everything now

 

Frequency response graphs are highly informative - IF you know how to read them. They tell a great deal about how a headphone sounds, but the hard part is interpreting the graph and how it translates to the sound of that headphone on your own head.

 

 

This doesn't make any sense. "Contains nuts" is not analogous to a direct measurement of the behavior characterstic of a headphone anymore than it is to what food actually tastes like. A more close analogy would be the breakdown of chemical compounds in a beer and interpreting how those affect taste.

 

Like I said above, a FR graph contains a wealth of information, the only hard thing is extracting it and translating it into meaninful information.

 

It's in the context of 'avoiding something' or 'looking for something'. Originally wanted to go with that, e.g. 2 tbs of salt, 3 tbs of corn starch, etc, but yeah...

[xtroria]

It gives you somewhat a general idea of what is the sound signature of the gear. Ofc, something like soundstage or other characteristics can't be described with FR graph but at least you know what you're expecting from it.

 

Go here and make a graph between Sennheiser HD558 and HD600. You'll find them to have a very similiar graph from sub-bass all the way to 10Khz, but despite that, I still can hear their difference in sound. Do they have similiar sound signature? Yes. Is there a pretty noticeable improvement despite the similiar FR chart? Yes.

[Guest]

It gives you somewhat a general idea of what is the sound signature of the gear. Ofc, something like soundstage or other characteristics can't be described with FR graph but at least you know what you're expecting from it.

 

Go here and make a graph between Sennheiser HD558 and HD600. You'll find them to have a very similiar graph from sub-bass all the way to 10Khz, but despite that, I still can hear their difference in sound. Do they have similiar sound signature? Yes. Is there a pretty noticeable improvement despite the similiar FR chart? Yes.

 

What are the differences?

[mr moose]

A 3Khz 80dB wave will look the same on an FR graph regardless if it is a saw tooth wave, a square wave or a sine wave.    Interfering frequencies can pop up on graphs but there is no guarantee it'll sound the same across different headsets even if the graph looks the same.  

 

Basically it will not tell you how good a headphone is, just what type of response you can expect I.E more bass or less bass.

[Guest]

A 3Khz 80dB wave will look the same on an FR graph regardless if it is a saw tooth wave, a square wave or a sine wave.    Interfering frequencies can pop up on graphs but there is no guarantee it'll sound the same across different headsets even if the graph looks the same. 

 

What? That doesn't even make sense. A square wave is the sum of the fundamental frequency and all odd harmonics up to infinity. If you graphed it it would appear as discrete sines; hardly the same.

[mr moose]

What? That doesn't even make sense. A square wave is the sum of the fundamental frequency and all odd harmonics up to infinity. If you graphed it it would appear as discrete sines; hardly the same.

 

A squarewave is a wave in it's own right, it does not have to be the sum of anything.  

[Guest]

A squarewave is a wave in it's own right, it does not have to be the sum of anything.  

 

I'm not saying it has to be, it is. A square wave necessarily contains higher frequency content. If you send a square wave through a transducer, it will oscillate at the fundamental frequency and all the odd harmonics up until the physical bandwidth limit of the system.

 

 

Saying that a square wave looks like a sine wave on a frequency response chart is nonsensical. Yes, a 20hz square wave has the same fundamental frequency as a 20hz sine wave; but the square wave necessarily contains harmonics that the sine wave does not, being a pure tone. That's what makes it a square wave.

 

 

[mr moose]

I'm not saying it has to be, it is. A square wave necessarily contains higher frequency content. If you send a square wave through a transducer, it will oscillate at the fundamental frequency and all the odd harmonics up until the physical bandwidth limit of the system.

 

 

Saying that a square wave looks like a sine wave on a frequency response chart is nonsensical. Yes, a 20hz square wave has the same fundamental frequency as a 20hz sine wave; but the square wave necessarily contains harmonics that the sine wave does not, being a pure tone. That's what makes it a square wave.

I never said a square wave would look like a sine wave.  My point is an FR graph does not show the wave shape at all,  it purely measures the amplitude of any given frequency.  Not the way in which that amplitude is reached.  For intents and purposes a square wave at 3Khz will measure the same on an FR graph as a sine wave or saw tooth wave.

 

This might help:

 

[Guest]

I never said a square wave would look like a sine wave.  My point is an FR graph does not show the wave shape at all,  it purely measures the amplitude of any given frequency.  Not the way in which that amplitude is reached.  For intents and purposes a square wave at 3Khz will measure the same on an FR graph as a sine wave or saw tooth wave.

 

Gah.

 

My point is this doesn't make sense. A square wave exists in the time domain, not in the frequency domain. A frequency response plot is the latter. Which means, you can't measure a square wave on a FR graph. The notion is literally incoherent. In order to do so, the square wave must be decomposed, i.e. via Fourier transform, into it's component frequencies first: http://en.wikipedia.org/wiki/Frequency_domain

 

On top of that, square waves DO respond to non-flat frequency response, phase distortion, and so on: https://sites.google.com/site/zevaudio/turt/square-waves-and-phase/square-wave-frequency-response---minimum-phase

[mr moose]

Gah.

 

My point is this doesn't make sense. A square wave exists in the time domain, not in the frequency domain. A frequency response plot is the latter. Which means, you can't measure a square wave on a FR graph. The notion is literally incoherent. In order to do so, the square wave must be decomposed, i.e. via Fourier transform, into it's component frequencies first: http://en.wikipedia.org/wiki/Frequency_domain

 

On top of that, square waves DO respond to non-flat frequency response, phase distortion, and so on: https://sites.google.com/site/zevaudio/turt/square-waves-and-phase/square-wave-frequency-response---minimum-phase

 

I see the issue,  a square wave  also does not have to have a uniform duty cycle (change in frequency without a change the wave shape), but that is moot, the point I am making is that if a signal shifted through the various wave forms due to the transducer, it would not necessarily alter the fr graph, because it would still measure at the same amplitude and frequency. Thus the transducer can alter the sound without altering the frequency response. 

 

All this means is that you can't rely on the fr graph to tell you how well a transducer will recreate the wave shape, just the response characteristic at a given frequency.

[Guest]

I see the issue,  a square wave  also does not have to have a uniform duty cycle (change in frequency without a change the wave shape), but that is moot, the point I am making is that if a signal shifted through the various wave forms due to the transducer, it would not necessarily alter the fr graph, because it would still measure at the same amplitude and frequency. Thus the transducer can alter the sound without altering the frequency response. 

 

All this means is that you can't rely on the fr graph to tell you how well a transducer will recreate the wave shape, just the response characteristic at a given frequency.

 

Maybe I'm not understanding you then, because it is possible to predict how a transducer will reproduce/respond to a square wave, which will alter the sound. Is that not what you're talking about? Obviously a physical medium lacks infinite bandwidth necessary to produce a perfect square, if say it has rolled off highs on top of that (obvious on FR graph) then a square wave will also be distorted through that transducer.

[mr moose]

Maybe I'm not understanding you then, because it is possible to predict how a transducer will reproduce/respond to a square wave, which will alter the sound. Is that not what you're talking about? Obviously a physical medium lacks infinite bandwidth necessary to produce a perfect square, if say it has rolled off highs on top of that (obvious on FR graph) then a square wave will also be distorted through that transducer.

 

I think you might be taking the concept of a square wave too far for the concept I am trying to articulate.   As you know the music we listen to is not simply a single wave shape but a combination of saw, sine, square (distortion is a square wave) and many other weird nameless ones.  If the transducer alters the shape slightly from a saw to a sine because of physical characteristics or from a sine to a square because (I don't know, a sticky VC for arguments sake) The FR will not necessarily change but the sound will.  Obviously how obvious this is depends on many other things too.

 

I'll post some wave measurements later of a dodgy amp I have that altered the wave shape but not the FR.

[Guest]

I think you might be taking the concept of a square wave too far for the concept I am trying to articulate.   As you know the music we listen to is not simply a single wave shape but a combination of saw, sine, square (distortion is a square wave) and many other weird nameless ones.  If the transducer alters the shape slightly from a saw to a sine because of physical characteristics or from a sine to a square because (I don't know, a sticky VC for arguments sake) The FR will not necessarily change but the sound will.  Obviously how obvious this is depends on many other things too.

 

I'll post some wave measurements later of a dodgy amp I have that altered the wave shape but not the FR.

 

Oh. You're just talking about harmonic distortion, then. Things make a lot more sense now

 

You're right, no way to see harmonic distortion on a FR graph. However, harmonic distortion is generally a lot less audible than FR differences, since real music contains tons of harmonics already; the added frequency content has a high probability of being euphonious, which as we know is why so many people like vinyl and tube amps.

 

And I'd be interested to see those amp measurements.

[mr moose]

Oh. You're just talking about harmonic distortion, then. Things make a lot more sense now

 

You're right, no way to see harmonic distortion on a FR graph. However, harmonic distortion is generally a lot less audible than FR differences, since real music contains tons of harmonics already; the added frequency content has a high probability of being euphonious, which as we know is why so many people like vinyl and tube amps.

 

And I'd be interested to see those amp measurements.

 

I just spent the last hour looking through all my backup dvds,  These measurements I took were well over a decade ago now.  I couldn't find them but suffice to say I did find an image on google that was not dissimilar.

 

 

Imagine for the sake of discussion that each trace represents a transducer/amp combination and is the same amplitude. 

 

As you can can see the frequency of each trace is exactly the same,  Assuming the input wave was a sine wave (which they are for FR measurement) and one headphone put out trace A and the other put trace B an FR graph would look identical, even thought he wave shape is not. 

 

The measurement I took showed a step not unlike the one in the above measurement except that it occurred at 0v. (At first I assumed this to be a result of the driver) but on closer looking I believe it was actually sloppy circuitry in the class AB amp I was using. 

 

So where I was going in the very first response:  and FR graph can tell us a lot about headphones, but it can't tell us how accurately they reproduce the wave shape.  I guess I should have said especially if they are paired up to a dodgy amp.

[Guest]

So where I was going in the very first response:  and FR graph can tell us a lot about headphones, but it can't tell us how accurately they reproduce the wave shape.  I guess I should have said especially if they are paired up to a dodgy amp.

 

Cool. I get what you're talking about, now. Hadn't heard harmonic distortion described as such before, therefore my confusion.

 

[mr moose]

that's because technically it isn't.  Although that's exactly how it looks,  the signal I used was a single sine wave set at one frequency,  where as harmonic distortion requires more than one signal.

[Guest]

that's because technically it isn't.  Although that's exactly how it looks,  the signal I used was a single sine wave set at one frequency,  where as harmonic distortion requires more than one signal.

 

Come on, man; yes it is. Harmonic distortion is when a system generates additional frequency components in the output of a single input signal.

 

Two frequencies is intermodulation distortion.

 

[mr moose]

Come on, man; yes it is. Harmonic distortion is when a system generates additional frequency components in the output of a single input signal.

 

Two frequencies is intermodulation distortion.

 

yes you're right, sorry. ID

[Mr.McMister]

FR graphs are quite useful in determining how a headphone might sound, but one also needs to see impulse response, square wave responses, total harmonic distortion noise plots and best of all a waterfall plot which basically shows all the information of the previous plots combined except THD noise.

 

FR by itself can be misleading as stuff like distortion can affect our perception of loudness and by it's nature make things sound less clear, and impulse response can determine whether a headphone sounds fast, punchy and resolving vs, slow and muddy.

 

So should FR have a big influence on headphone purchasing decisions, while yes, but not without factoring in the other measurements. However, it can be difficult for the layman to acquire the knowledge necessary to interpret these measurements and colate them between headphones they've already heard and those they haven't. So try before you buy is an important method to use for most people.

 

Innerfidelity has a great set of article on interpretting response graphs for headphones found here.