Equalizing headphones: What equipment do I need?

To make a valid measurement of headphones you need more than just a way to seal in a measurement mic.  You need an ear and head.  Like this:

http://www.head-acoustics.de/eng/telecom_hms_II_3.htm

 

If you just measure with a mic and no ear, ear canal, and head, you won't get a response measurement that represents how the headphones perform on real ears, and the resulting EQ curve will be fairly useless.

Why not use the test data that already exists on headphone.com, for example? You might not get to see all the details, but all you need for basic EQ is there, just set up the inverse curve on your EQ...remembering you aren't shooting for flat in this case.

I haven't had any luck at all with microphone based equalization. At best it just gives you a good starting point.

If you don't want to do it by ear, then using a microphone is better than not EQing at all. Otherwise, just look at published response readings for your cans and guestimate a correction.

it is my opinion that you should never eq a headphone.

The above technique would tend to compensate for all response anomalies and force the response to "flat".  That's not what you want to do.  

I wouldn't go off headphone graphs.  They all use different compensations and can be inaccurate if you're trying to EQ to flat.  They're mostly good for comparing amongst other measurements using the same parameters.

 

By ear using tones and pink noise is a much better idea.  Just be careful that the extreme frequencies as you near 20hz and 20khz are more about energy/pressure and not about actual volume level.

Hmm... that sounds like the wrong problem to be having when you're about to embark on this kind of project.  

 

Yes, for accurate measurements you need a dummy head, with outer and inner ear components.  Otherwise, there's really no point.  You also need to study up on why headphones should NOT have flat response.  Until you have a good understanding in hand, this is probably not the project to undertake.  But there's already good data, so you shouldn't need to measure anything to achieve your goal.

 

If you are finding the need to EQ your headphones in the first place, that shows a lack of satisfaction for how they sound now, and a desire to improve them.  

 

Let's break the project down this way.  Assuming the response of your headphones is somehow not ideal.  There are two questions to be answered.

 

1. What is the response of your headphones?

2. What is the ideal response for any headphones?

3. What is the inverse EQ curve needed to compensate for your headphones variance from the ideal?

4. How can you apply that curve accurately?

 

Since your headphones always operate into the acoustic system of the ear, there is no point in measurements make without that acoustic system.  All headphones of all designs always operate into the acoustic system of the outer ear and inner ear.  No exceptions.  So any measurements made without that system won't reflect their true performance. To answer the first question, you need good measurement data.  If you can't take it yourself, you need to access that taken by someone else with the proper equipment.  Hence, my suggestion of the graphs from headphone.com.  You need to read them carefully, it's all there.

 

Answering the second question is much harder.  Unless you can find something published from scientific research, all we have is the general idea that the ideal response contains a bass boost to make up for the lack of large sound waves striking the entire body, and a treble roll-off to compensate for the close proximity of the drivers to the ear.  How much, though?  One way to get a feel for what others seem to prefer is to take the response data from several high-end, top rated headphones and average them.  You probably only need a data point every 1/3 octave, perhaps less, for a general trend.  What you'd have then is an average response that some very learned designers have all agreed on.  If you average enough good headphone data, that would make a fine target curve for some less than ideal headphones.  In actual fact, averaging wouldn't be the best way to combine those measurements, but you can at least do it on a spreadsheet.

 

Now you have the response of your headphones, and the averaged response of "great" headphones of the world. If you subtract yours from theirs you'll have the difference that you need to correct for with your equalizer.  Invert that curve and you'll have the target for your equalizer's response (not the total response, just that of the EQ needed to pull your headphones to the average target).  The equalizer you'll need will be a multi-band full parametric, as you'll need to adjust gain, frequency and Q to hit the curve well.  But the key is, you need a way to either plot or predict, exactly, the actual response of the equalizer.  That's why a graphic wouldn't work, the curve you see represented by the controls isn't close enough to reality.  Match the actual equalizer's response to the desired correction curve, and you'll be fairly close to the target curve you derived in question 2.

 

And...when you're done...you won't have headphones that sound exactly like the average high-end ones anyway because there are internal resonances that EQ will not address.  EQ deals with response, and to a lesser extent, phase, but can't deal with ringing or resonances that carry on in time.  The technique you'll use here is essentially time-blind.  But probably as close as practical.

Just to tag this on...looking at about a dozen headphone response pots, it would seem that you could do quite a bit of data smoothing, especially at the high end. Some of these things are radically different, and it's not a bad idea to throw out the extremes. It's a job for fuzzy clustering, but I don't see that happening easily in Excel.

Usually you don't want a flat line on a headphone frequency response graph because the ear affects the sound from headphones differently from a flat pair of loudspeakers.  Headroom graphs have a compensation curve built in so a flat line on that graph supposedly sounds good.  However

1. It's for the dummy head model, your ears may be different

2. The frequency graph has been smoothed over

 

There are two threads on by-ear equalizing

http://www.head-fi.org/t/413900/how-to-equalize-your-headphones-a-tutorial

http://www.head-fi.org/t/615417/how-to-equalize-your-headphones-advanced-tutorial-in-progress

 

The first thread is well organized but the thread starter is gone and there are some inaccuracies in the information presented.  I started the second thread but never had the energy to finish the writeup so...

 

Basically I wanted to add two contributions to the first thread

1. When equalizing by ear one should not aim to make all frequencies sound the same loudness, because one needs to compensate for the equal loudness curve of human hearing

2. One can compensate for this by adding a second equalizer in line to the first, reproducing the equal loudness contour; then one can just aim for the same loudness in all frequencies with the first equalizer and you should have a flat headphone response when you remove the second EQ.

 

I then try to go into details on how this is done and failed miserably at properly describing it But you can pm me if you're interested and I can try to walk you through it
 

Finally, you want to add a "gain only" EQ band to your Electri-Q preset and lower the whole curve below zero to prevent clipping.

Edited by Joe Bloggs - 12/5/12 at 10:20am