[Chromako]

I don't know, but since the performance of dynamic drivers is now hard to drastically improve, why not try to enhance with some active technology..? I would like to see an engineers perspective on the reasons why or why not to do this. 

If it sounded better, I see no reason why some head-fiers wouldn't be interested of such concept.

But anyway, Sophisticated, granular EQ requires complex circuits that are quite bulky. Anything else, like counteracting square wave overshoot, slow impulse reaction, harmonic distortion, etc, will require DSP. To do that, and to make it into a one puece package, we would have to do this:

Digital file>desktop DAC>desktop amp>special-custom-analogue to digital converter in the headphone>DSP> special-custom-DAC> headphone driver. That sucks, and all that conversion would sound terrible.

If we applied the corrective DSP on the computer / DAC end, well, that'd be interesting, but I don't see it happening in a way that can make a Grado sound like Sennheiser HD800. There is only so much that can be done to counteract physics.

 

[Chromako]

[quote name="xnor" url="/t/682628/is-80-of-hi-fi-just-eq/0_30#post_9858970"
But to be honest, even if it's a perfect combination, which long time members on head-fi would buy such a thing? Buying a headphone and plugging it straight into the USB port on your computer ... isn't that way too easy? No DACs/amps to fiddle around with, no interconnects to match to such components (by magical criteria), ...
[/quote]


It's done. Audio Technica ATH-D1000. Sadly, they are quite rare nowadays, and I'm not selling mine

And they sound awesome.

 

[xnor]

I know there are many even really cheap (made in China) headphones with just a USB plug. Some are headsets and therefore also include an ADC.
 
I just don't think that there is any DSP going on*. Is the ATH-D1000 different in that respect? (Not talking about upsampling which can be thought of part of the D/A conversion process.)
 
 
*) well there's active noise cancellation, but I'm thinking of EQ

 

[xnor]

If you check out InnerFidelity, you'll see that there are many other parameters than frequency response (which can be corrected with EQ 80% of the way, as you said). Frequency response is, when making AB comparisons the most obvious difference.

There is also Impulse speed, harmonic distortion, and resonance, to start. None of those can be corrected with EQ.

Impulse response is the same as frequency response, just visualized in a different domain (time vs. frequency). The only time the impulse response is a nice addition is with multi-driver headphones with crossovers that delay different parts of the signal differently.
Electrical impedance and phase does not change how the headphone sounds, nor does isolation.
Square waves, as has been shown in other threads, are difficult to "interpret" because inaudible phase shifts can change their shape significantly. Also, if you have the impulse response you can also calculate how the square waves would look like (without nonlinear distortion).
And for non-linear distortion there is the THD plot, which is probably the second most important measurement.
 
 

And here are some cool plots that talk about resonance. (That's why the HD800's look so weird and asymmetrical and are made of plastic. It's not plastic because it's cheap.) None of this can be corrected with EQ.

Well, I disagree here as well. I doubt plastic has better acoustic properties than other materials, and even if it had they wouldn't have to put the spray painted plastic on the outside.
 
The HD280, for example, is made of plastic as well and looks sturdier but has serious resonance problems due to the plastic. I guess the enclosed volume and isolation are significantly more important.

 

[Joe Bloggs]

anyway, Sophisticated, granular EQ requires complex circuits that are quite bulky. Anything else, like counteracting square wave overshoot, slow impulse reaction, harmonic distortion, etc, will require DSP. To do that, and to make it into a one puece package, we would have to do this:

Digital file>desktop DAC>desktop amp>special-custom-analogue to digital converter in the headphone>DSP> special-custom-DAC> headphone driver. That sucks, and all that conversion would sound terrible.

Actually the signal can take several DAC ADC roundtrips while sounding transparent. Audiophiles *think* such a roundtrip will sound terrible, but that's a different matter

 

[Chromako]

  Impulse response is the same as frequency response, just visualized in a different domain (time vs. frequency). The only time the impulse response is a nice addition is with multi-driver headphones with crossovers that delay different parts of the signal differently.
Electrical impedance and phase does not change how the headphone sounds, nor does isolation.
Square waves, as has been shown in other threads, are difficult to "interpret" because inaudible phase shifts can change their shape significantly. Also, if you have the impulse response you can also calculate how the square waves would look like (without nonlinear distortion).
And for non-linear distortion there is the THD plot, which is probably the second most important measurement.
 
 
Well, I disagree here as well. I doubt plastic has better acoustic properties than other materials, and even if it had they wouldn't have to put the spray painted plastic on the outside.
 
The HD280, for example, is made of plastic as well and looks sturdier but has serious resonance problems due to the plastic. I guess the enclosed volume and isolation are significantly more important.

A couple things: 
 
Yes, impulse response is similar to frequency response, in a way. It gives us a way to tell what the highest reproducible frequency could be. However, it cannot be faked/changed based on EQ or any other processing. And not only do we care about the impulse speed, but also the size of the oscillation that happens afterwards (distortion). 
 
And the plastic case in the HD800, the plastic actually does have better acoustic qualities -when you design for it. Not all plastic is equal- nylon, teflon, PVC, PLLA, PA66, etc all have different acoustic qualities (and yes, I actually used to work engineering new types of plastics, though you wouldn't want to use the stuff I developed in your headphones as it's both carcinogenic water soluble). What's in the HD800's is a rather expensive type called Leona, made by Asahi Kasei, and much of  that is actually filled with glass fibre (again, fine tuning the acoustics).  It's actually very expensive to engineer headphones in this way. 
 
 
Now, to be fair, they really could have used a better silver paint... I have no idea why they didn't think that one through. 

 

[Joe Bloggs]

A couple things: 

Yes, impulse response is similar to frequency response, in a way. It gives us a way to tell what the highest reproducible frequency could be. However, it cannot be faked/changed based on EQ or any other processing. And not only do we care about the impulse speed, but also the size of the oscillation that happens afterwards (distortion). 

And the plastic case in the HD800, the plastic actuallydoes have better acoustic qualities (not all plastic is equal). It's a special type called Leona, made by Asahi Kasei, and much of it is actually hollow and filled with glass fibre.  It's actually very expensive to engineer headphones to use that. Now, to be fair, they really could have used a better silver paint... 

You lost me there. Any change in the sound, any change at all, including EQ, will affect the impulse response. One can't help BUT change the impulse response when one applies any kind of signal processing. In fact a whole class of pricessing (linear transformations, which includes EQ) can be represented solely by the way they affect the impulse response. I have a whole thread on xda-developers dedicated just to this (becoz of technical difficulties I'll link to it in a separate post)

 

[Joe Bloggs]

http://forum.xda-developers.com/showthread.php?t=2372750
The above mentioned xda-developers thread

 

[xnor]

Indeed applying an EQ filter to a signal is equivalent to convolving the signal with the impulse response of that filter.
 
The impulse response of a headphone is just the output of the headphone when you feed it an impulse. If you apply EQ to that impulse it has to change. Therefore the headphone's output (impulse response) is going to change.
To put it very simple: peaks and dips in the frequency domain are achieved by ringing in the time domain. For example, ringing due to a -5 dB cut at 500 Hz can be completely removed by a corresponding boost of +5 dB at 500 Hz, that is ignoring nonlinearities.

 

[proton007]

  Indeed applying an EQ filter to a signal is equivalent to convolving the signal with the impulse response of that filter.
 
The impulse response of a headphone is just the output of the headphone when you feed it an impulse. If you apply EQ to that impulse it has to change. Therefore the headphone's output (impulse response) is going to change.
To put it very simple: peaks and dips in the frequency domain are achieved by ringing in the time domain. For example, ringing due to a -5 dB cut at 500 Hz can be completely removed by a corresponding boost of +5 dB at 500 Hz, that is ignoring nonlinearities.

 
This gives me an idea.
 
Is there any software/hardware setup that measures the sound produced by a headphone with a particular input, analyzes it, and then EQs the input so that the output is within the desired criteria?
 
For instance, I can give it a desired output curve, place the headphone in the device or mount it on a dummy head, press 'calibrate' and voila!
 
Not sure if such a system exists.
 
That would also determine a good headphone from a bad one.  A good one would be tuned much easily, while a bad one may not calibrate because it would never respond well to EQ.

 

[Joe Bloggs]

This gives me an idea.

Is there any software/hardware setup that measures the sound produced by a headphone with a particular input, analyzes it, and then EQs the input so that the output is within the desired criteria?

For instance, I can give it a desired output curve, place the headphone in the device or mount it on a dummy head, press 'calibrate' and voila!

Not sure if such a system exists.

That would also determine a good headphone from a bad one.  A good one would be tuned much easily, while a bad one may not calibrate because it would never respond well to EQ.

Smyth Realiser? It is much more involved than setting an "output curve" though (as well it should be)

 

[proton007]

Smyth Realiser? It is much more involved than setting an "output curve" though (as well it should be)

 
Not really. The Smyth Realiser corrects for positional changes.
 
What I meant is much simpler in theory. Something that finds the correct EQ based on measuring the headphone's response. More like finds the optimal control solution to a system. I understand its not an easy problem because transducers tend to be non-linear systems.

 

[Joe Bloggs]

That's much simpler--it also doesn't work. I answered another one of your posts here, it's relevant to the discussion at hand:

http://www.head-fi.org/t/670056/skeptico-saloon-an-objectivist-joint/255#post_9876045

Simply put, there is no universal correct FR for headphones. You need to take measurements at the individual's ears, with the headphones on and then again when sat in front of a reference speaker system. Then you can begin to calculate a correction filter that will work. The Smyth Realiser is the only commercial system out there that does this. That it also does head tracking is just icing on the cake.

If you thought headphone acoustics were so well controlled compared to speakers (as detailed by your post in the other thread), why do you reckon we still have such a plethora of wildly different sounding "hi-fi" headphones all enjoying success in the marketplace? There is no universal correct FR for headphones.

Not really. The Smyth Realiser corrects for positional changes.

What I meant is much simpler in theory. Something that finds the correct EQ based on measuring the headphone's response. More like finds the optimal control solution to a system. I understand its not an easy problem because transducers tend to be non-linear systems.

 

[bigshot]

Have you auditioned speakers lately? Talk about wildly different sounds!

 

[Joe Bloggs]

Have you auditioned speakers lately? Talk about wildly different sounds!

That's because listening room acoustics are just as inconsistent as ears, again negating the incentive for consistent speaker sound design.  But my point in the other thread was it is easier to measure and correct for loudspeaker-room acoustics than to do the same for headphones.