Head-Fi.org › Forums › Equipment Forums › Sound Science › Equalizing headphones: What equipment do I need?
New Posts  All Forums:Forum Nav:

Equalizing headphones: What equipment do I need?

post #1 of 52
Thread Starter 

So I was thinking about trying to equalize my headphones with a measurement microphone, but I have a few questions:

 

Is that all the equipment I need other than some sort of device to seal the microphone within the headphone cups?

 

And are <$100 measurement microphones adequate? I noticed there are some that go for over $600 but they claim to measure flat well past human hearing ranges.

A couple I was looking at: http://www.amazon.com/DBX-Driverack-RTA-Condenser-Microphone/dp/B0002DVCAM/ref=sr_1_3?ie=UTF8&qid=1354573348&sr=8-3&keywords=measurement+microphone

 

and http://www.amazon.com/Nady-Reference-Measurement-Condenser-Microphone/dp/B00095MG6M/ref=sr_1_4?ie=UTF8&qid=1354573348&sr=8-4&keywords=measurement+microphone

 

Also, these things come fully calibrated right? Not sure how I would go about calibrating a device that I'm using for calibration in the first place.

post #2 of 52

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.

post #3 of 52

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.

post #4 of 52
Thread Starter 
Quote:
Originally Posted by jaddie View Post

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.

The one thing that I have trouble with that is that I'm guessing on the +/- dB parts of the EQ. I don't really get how to interpret it, it says that some parts go as much as -16dB but that would obviously sound ridiculous if I had an EQ bumped up that much.

 

The headphones I have don't really rely on a seal or anything though, would a dummy head still be completely necessary?

post #5 of 52

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

post #6 of 52
Thread Starter 

What's the alternative? By ear?

 

My ears aren't exactly trained, and I can't stand listening to sine sweeps. Especially since the main part in my headphone that needs equalization is the most sensitive part of the FR to human hearing.

 

I feel like I could probably do it via pink noise(at least for making them roughly neutral) since it sounds so drastically different with any equalization, but I would first need to hear some neutral pink noise for reference...


Edited by chewy4 - 12/4/12 at 2:21pm
post #7 of 52
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.
post #8 of 52
it is my opinion that you should never eq a headphone.
post #9 of 52
Thread Starter 
Quote:
Originally Posted by bigshot View Post

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.

Is there a science behind translating the dBr of the measurements to the dB levels of the EQ though? It's easy to get all the peaks and such that need changed, but I'm pretty much guessing on how much.


Here's what I've done so far:

 

 

 

Note the scaling on the dB levels, headroom measurements go down to ~-14 dB while I only push mine up about 3 or so(using what was about -2 or -3 as neutral).

 

Not enough? Too extreme? It actually does sound pretty good so far.

post #10 of 52
Quote:
Originally Posted by chewy4 View Post

Is there a science behind translating the dBr of the measurements to the dB levels of the EQ though? It's easy to get all the peaks and such that need changed, but I'm pretty much guessing on how much.


Here's what I've done so far:

 

 

 

Note the scaling on the dB levels, headroom measurements go down to ~-14 dB while I only push mine up about 3 or so(using what was about -2 or -3 as neutral).

 

Not enough? Too extreme? It actually does sound pretty good so far.

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.  

post #11 of 52

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.

post #12 of 52
Quote:
Originally Posted by chewy4 View Post

The one thing that I have trouble with that is that I'm guessing on the +/- dB parts of the EQ. I don't really get how to interpret it, it says that some parts go as much as -16dB but that would obviously sound ridiculous if I had an EQ bumped up that much.

 

The headphones I have don't really rely on a seal or anything though, would a dummy head still be completely necessary?

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.

post #13 of 52
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.
post #14 of 52
Thread Starter 
Quote:
Originally Posted by jaddie View Post

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.  

Huh? Isn't that exactly what I'm trying to do here?

 

 

Quote:
Originally Posted by jaddie View Post

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.

That's definitely an interesting thing to try and do, but I'm not trying to replicate a high-end headphone - without any alteration mine is said to sound similar to some high-end headphones already - I'm just trying to flatten out mine's FR a bit. I actually am satisfied with the way my headphones sound, but I know they're a bit colored  and I'd like to play around with a more flat signature.

 

As I'm not willing to go all out with the dummy head and all that, really all I can shoot for is more flat. Once I throw in different, more complicated curves to try and accomplish that just increases the chance of error IMO(although I guess that depends on the curve). But as you can see, I didn't really mess around with certain things like the high-treble roll-off. I just sorta let that be since measurements are usually out of whack in that range, and I figure the rolloff is there for a reason.

 

And there's a lot of varience in high-end headphones. As you said, there's plenty of extremes. There are some like the HD800 that put a treble shelf on all the high frequencies so people listen and think it has a ton of detail due to the fact that those frequencies that are boosted are used in the noise floor more than any other instruments, so they hear noise floors that they never noticed before. Others like the LCD-2 are just flat with a rolled off treble. I don't think flat is too far off for an ideal headphone sound, but I'll certainly look into that more. Thanks for the advice.

 

The EQ I am using is parametric by the way. The curve shown is not a control, it is the resulting curve of the controls that I have set. All the dots are the bands, and they're peak bands(I've got my choice of a plethora of different kinds).

 

 

Quote:
Originally Posted by TMRaven View Post

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.

I'm gonna look more into how to correctly tune using tones and sine sweeps and such, although there's only so much of that I can stand. Luckily I don't have to worry about extreme lows as they're already flat, and I don't care about extreme highs as my hearing is cut off somewhere about 17.5k.

 

Pink noise though... like I said before I'd have to know what neutral pink noise sounds like.


Edited by chewy4 - 12/5/12 at 7:49am
post #15 of 52

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 redface.gif But you can pm me if you're interested and I can try to walk you through it smily_headphones1.gif
 

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
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Equalizing headphones: What equipment do I need?