[fjrabon]

TL;DR: If you see a major difference in THD vs frequency at 90dB and 100dB, it predicts that the headphone will not handle EQ well at those frequencies where the difference is large.  If the square wave response quicly falls off, it predicts EQ adjustments will not produce as noticeable of a change.
 
I'm kind of making this thread as a future reference but also for any discussion.  What I do not intend here is a debate on the merits or failings of EQ in general.  What this thread is about is how to know if a headphone will "take" EQ well.  By that I simply mean whether or not if you EQ for more bass, whether or not the headphone reacts well, and provides additional, clean tight bass, or if it descends into a distorted, muddled mess.
 
 
I've owned a lot of headphones over the years.  Some of which could take EQ well (HiFiMan HE400i for example) and some of which could not (Audio Technica AD700).  At first I just kinda thought it was a crapshoot, or voodoo, or mysterious (amp synergy) or something like that.  But after doing some thinking, and observations of measurements of headphones known for taking EQ well, and those that have reputations for not being able to take EQ well, here is what I came up with.
 
There are doubtless more factors to it than this, but I chose these two factors because they are easily found and because I find them highly predictive, if not outright decisive on whether or not a headphone can take EQ.  
 
(side note, always leave the frequencies you want more of alone, and instead EQ the other frequencies down.  You should generally avoid “boosting” anything with EQ, instead lessening the other frequencies and then using your amp to provide the needed boost)
 
I thank Tyll for the measurements, taken from innerfidelity.
 
First, let's look at a headphone that, in my experiences, is very good  at “taking” EQ, the HE400i.
 
Let’s look at what I think is the single most important graph to understanding if a headphone will take EQ: a THD vs. frequency graph.  What this graph shows is how much distortion a headphone puts off at 90 dB and 100 dB for any given frequency.  For this, the headphone’s frequency response doesn’t matter, because it simply sets the volume at all tones to a constant 90 (and 100) dB.

Here we see that the HE400i performs very well across the board minus a couple of very isolated spikes in THD.  Which is great, but not really our purpose here.  What I want to point out is how the THD is just as low in the bass region as anywhere else AND that increased volume doesn’t increase THD.  We see virtually no issue with increased THD when we add 10dB in the bass region (or anywhere else.  This simply shows that additional volume at any given frequency (ie EQ) won’t result in additional distortion.  This is shown by the red and blue lines staying very close to the yellow and green lines throughout.
 
Now let’s contrast this with a headphone known (or notorious, depending on your view) for not only bass roll-off, but for bass that gets ugly in a hurry if you try to EQ it in, the AKG K701.  Now I want to say, to ward off any potential wars with K701 owners, that the K701 is a great headphone.  Gorgeous treble detail and incredible soundstage.  But it isn’t good at bass, and it certainly doesn’t respond well to EQing in a bass boost.

 
 
Here we see, first off, that there is substantially more distortion in the bass regions at both volume levels, as the K701 really struggles to get to 90 dB in the first place without starting to get low level clipping.  This won’t sound like what you might think of as distortion, but will rather sound like muddy, uncontrolled bass.  And this is what I experience when I try to EQ the K701’s bass to be flat.  Secondly, we see that you get substantially more THD at the 100dB mark all the way through 1000 Hz.  We see this because the yellow and green lines are well above the red and blue lines over a large part of the graph.  Meaning not only is the K701 resistant to EQ in the bass, but also doesn’t do particularly well in the midrange either with EQ.
 
Next, what I’ve found substantially predictive of ability to EQ, yet maintain a crisp response is how a headphone does in the 30Hz square wave.  A headphone that can maintain a strong square wave response can typically hold the additional power you are giving it at a given frequency.  A headphone that sharply falls off in the square wave response will struggle to do much with any additional power you’re sending it’s way in a particular frequency.  

Here we see the HE400i Maintains a strong, linear decline in the square wave test.  

The K701 on the other hand gives you a quick burst, and then rapidly falls off.  Now for some, this fools them into thinking that the K701’s bass is “fast” and not “bloated” because it will make bass transients seem faster.  But that’s simply because it’s cutting bass transients off before they even finish.  For our purposes here, what this also unfortunately means is that the K701 is going to make almost all bass sound at roughly the same level, because the magnitude of the bass falls off so quickly.  Almost all lower frequencies will be experienced as roughly the same “overly polite” low volume, even if you try to EQ them up.  This compounds the THD issue we saw in the THD plot above, because it means not only does the K701 distort when you add bass, but that it tends to quickly roll off the bass you do add.
 
Now, at this point you may be wondering “what about the other frequencies?  Why do you keep focusing so much on bass?”  Well, 1) when having these conversations, most of the time they come up when people are talking about wanting a “bass boost.”  Very rarely do I see people saying “you know, I would really love some more 6kHz in here.”  It does happen from time to time, but it’s less common.  2) Bass is much harder to EQ in, because bass frequencies require way more power, and are much harder for a small diameter driver in a headphone to reproduce well.  So, this is where you are more likely to see more problems.  Most of the time, if you simply want to cut off some 8kHz, you will be fine with almost any headphone.  3) the lessons from the THD chart still apply here.  As we saw with the K701, it wasn’t particularly good when EQing in anything all the way up to 1000Hz.
 
One final thing to consider when EQing in a bass boost is the requirements of your amp.  Again, how I recommend EQing in more bass, is to instead cut all the other frequencies out.  Now, if what you want is a bass boost, how do you accomplish a boost, when all you’ve done is cut?  You turn your amp up.  Think of it as the “bass” comes from your EQ adjustment and the “boost” comes from your amp.  Now, what you’ll have to remember is to add a 5dB bass boost, you’ll need your amp to have more than double the power you would need without the bass boost.  If you are an owner of the FiiO E10K, you may have noticed that if you enable the bass boost circuit that it clips above half volume when the bass boost is enabled.  That’s because the FiiO’s bass boost circuit is essentially like a rudimentary bass boost EQ adjustment, and then the volume knob is “recalibrated.”  The FiiO is fine until you start running out of its headroom, which is quickly gobbled up by the bass boost circuit. 
 
For a more detailed explanation of how to EQ, look HERE
 
Hope somebody finds this helpful.

 

[castleofargh]

the subject is indeed interesting. thank you for trying to give directions.
here is what I do, no idea if it's reliable but here I go:
I start obviously by looking at the signature of the headphone I might want. if the frequency response is far off of what I like I won't waste my time wondering if it can be EQed. my only exception to that is having too much bass or too much trebles. because cutting them down is easy on 100% of headphones, while compensating for a roll off is often doomed.
 
in practice I don't care too much for trebles, so I tend to go with headphones that have the signature I know I like(needs some experience with headphones and reading graphs from a given website), or the headphone that has the mids and trebles I want but too much bass.
 
then if I feel like I'll need to do more than cut the bass or reshape it to my liking(or a moderate boost), then I look at the distortions values, thinking how much I'll need to boost, and indeed if when raising the volume the distortions seem to rise abnormally, then it's not a great start for an EQ boost in that area. but TBH if it's still rather low even at 90db then I don't really mind.
what I really worry about a lot are resonance points. and for that I check if there is anything weird on the phase/impedance at the frequency I plan to EQ. often the distortions are enough to show that because it's likely that a huge resonance will also create a big mess in how the driver will move so distortions will tend to be high and already a warning.
 
overall, I feel like focusing on getting a headphone we will want to EQ mostly down and not up is the real winner headphone. ^_^
 
 
about your comment on square waves, while I believe you're showing the right thing, I'm sure you're giving the right reason(but it's not like I'm a square wave expert so maybe I'm the one who's wrong?).
to me you can make some kind of caricature in your mind where it's showing a frequency response in reverse. if the start of the square is clean and flat then the high freqs are good. if the right side of the square wave goes down, it suggest a roll off for the low freqs. in truth it's probably more indicative of phase shift, but as they often tend to move hand in hand(adding bass will create a phase shift), I find that my caricature works ok-ish in practice. except obviously that a ringing in the square wave should be seen as a ringing first. I'm only talking about rather regular changes in direction.
so in your example it just goes to show how the k701 rolls off in the sub way too much. so I agree with your conclusion that it's going to suck trying to get that back up with EQ, but IMO the frequency response was enough to know that.
 
 
 
about how to EQ, again I agree with the suggestion that we much avoid the risk of clipping, but why not just EQ how you like, and then on your picture, just lower the preamp by the value of the higher boost? ^_^ the result is the same and it's friendlier to use.

 

[fjrabon]

about your comment on square waves, while I believe you're showing the right thing, I'm sure you're giving the right reason(but it's not like I'm a square wave expert so maybe I'm the one who's wrong?). to me you can make some kind of caricature in your mind where it's showing a frequency response in reverse. if the start of the square is clean and flat then the high freqs are good. if the right side of the square wave goes down, it suggest a roll off for the low freqs. in truth it's probably more indicative of phase shift, but as they often tend to move hand in hand(adding bass will create a phase shift), I find that my caricature works ok-ish in practice. except obviously that a ringing in the square wave should be seen as a ringing first. I'm only talking about rather regular changes in direction.
so in your example it just goes to show how the k701 rolls off in the sub way too much. so I agree with your conclusion that it's going to suck trying to get that back up with EQ, but IMO the frequency response was enough to know that.
 

I think this is sort of right in some ways.  I didn't really want to go into a long explanation of square waves in the original, as I really just wanted to talk about what to look for when EQing.  The 30Hz wave will tell you how it transitions from bass to treble, and how the headphone handles additional power.  A super quick roll off in the square wave doesn't really indicate anything about treble, but it does say that the bass is probably out of phase, which even beyond just the frequency response, means it's going to be resistant to adding more bass via any method.  The square wave will not just tell you that bass is weak, but that with more volume it becomes even weaker in comparison to other frequencies.  For instance, the HE6 isn't know for being a particularly bassy headphone, but at higher volumes (and EQ) it holds its bass well.  The K701, on the other hand, as you turn it up, the bass plays an even smaller role compared to the mids and treble.  You analogy of a "flipped FR curve" sort of works, in that a roll off does indicate weak bass on a 30Hz square wave.  However, the left part doesn't really tell us anything about treble, strictly speaking, as on a 30Hz wave the treble information is too compressed to make much our of (hence why Tyll use a 300 Hz wave too, where you can actually see the treble portion).  From just a 30Hz wave, we might still have a very mid heavy or a very treble heavy headphone.  
 
But the point I wanted to make more than anything is that square waves can tell us a bit more than just the frequency response, because square waves can also tell us how the headphone responds to additional energy in terms of adding additional power.  That is, it gives us an idea of what a FR curve would look like if you drew it at many different volume levels, something that you rarely find on measurement sites.  That's important for EQing, because what you're doing when EQing is essentially applying different power levels to the headphone at different frequencies.  Some headphones will nearly perfectly reflect those changes (HE400i), some will resist those changes with all they have (K701).  

about how to EQ, again I agree with the suggestion that we much avoid the risk of clipping, but why not just EQ how you like, and then on your picture, just lower the preamp by the value of the higher boost? ^_^ the result is the same and it's friendlier to use.

Yeah, six of one, half dozen of another.  I just gave that as an example, and chose not to fiddle with the pre-amp, since some EQ programs (or physical units) don't have a pre-amp slider.  Just remember to guarantee no clipping you have to pull the pre-amp slider down the same amount as the largest increase in the EQ.  Sometimes that can be a bit imprecise (it's difficult to see exactly what point a given slider is at sometimes).  Whereas just pulling all the other sliders down is always guaranteed not to cause line level clipping.

 

[watchnerd]

 
 
One final thing to consider when EQing in a bass boost is the requirements of your amp.  Again, how I recommend EQing in more bass, is to instead cut all the other frequencies out.  Here I give a very simple version of an EQ adjustment to illustrate what I mean. 


Notice everything is "cut" rather than "pushed up."  This is because almost all modern music is mastered super compressed and very close to the line clipping level.  This means that if you EQ'd anything (and especially bass it will push the line leve into clipping, meaning your signal is already distorted before it even reaches your amp, let alone the headphone.  
 
 

 
Why didn't you just drop the preamp by -6-10 db and then boost the bass?
 
That's why it's there.  And it's a lot less messy.
 
Also, you really should be using PEQ.  Graphic EQ is crude.

 

[fjrabon]

   
Why didn't you just drop the preamp by -6-10 db and then boost the bass?
 
That's why it's there.  And it's a lot less messy.
 
Also, you really should be using PEQ.  Graphic EQ is crude.

because this was just an explanatory example and not all EQ solutions have pre-amp sliders, with this I could just take a screenshot that easily gets the point across.  
 
Also, I don't really understand what you mean by "a lot less messy."  Lowering it by 6-10 dB in the bass would be way more than needed and make the amp work harder than is really needed.  As a rule, you should lower the pre-amp slider as little as absolutely necessary, as any more than necessary you're sending a needlessly low line level signal.  6-10 would likely be almost 5dB lower than needed, meaning your amp is having to make up that 5dB, which for a lot of amps that means needing to push a couple hundred extra mW (or if you have a hard to drive headphone on the other end, potentially as much as a full extra watt).
 
You can experiment and find the exact needed pre-amp adjustment you'd need in your worst case (ie most compressed music), sure.  I'm not sure that's actually less messy than just EQing everything down from the start, which will guarantee you don't get line level clipping.  To me it's six of one, a half dozen of another.
 
Anyway, again, it was just a visual example of how to avoid line level clipping, which is the biggest reason why a lot of people believe "EQ sounds like crap" because they actually end up clipping the line level signal before it even gets to their headphone.
 
and if you don't make larger than 2dB jumps between bands on graphic, there's not really a real difference between parametric and graphic.

 

[watchnerd]

 
because this was just an explanatory example and not all EQ solutions have pre-amp sliders, with this I could just take a screenshot that easily gets the point across.  
 
Also, I don't really understand what you mean by "a lot less messy."  Lowering it by 6-10 dB in the bass would be way more than needed and make the amp work harder than is really needed. 
 
and if you don't make larger than 2dB jumps between bands on graphic, there's not really a real difference between parametric and graphic.

 
There is a huge different between PEQ and graphic...the major one being you can define the Q, as well as shelves.

 

[fjrabon]

   
There is a huge different between PEQ and graphic...the major one being you can define the Q, as well as shelves.

I understand how PEQ works.  The differences aren't really relevant to the original article.  Yes, I realize it's more detailed and exacting.  It's also way more trouble than I wanted to bother with to create a simple image that was solely for the purpose of illustrating EQing down vs EQing up.

 

[Maconi]

I find it helpful to have a visual representation of what you're actually doing. Just turning the dB up/down could result in a lot of spikes (looking line a rough sine wave). A (good) Parametric EQ lets you see the final result like this (my current EQ on an LCD-XC).
 

 

 

[fjrabon]

  I find it helpful to have a visual representation of what you're actually doing. Just turning the dB up/down could result in a lot of spikes (looking line a rough sine wave). A (good) Parametric EQ lets you see the final result like this (my current EQ on an LCD-XC).
 

 

yeah, I should have linked one of the many other "how to EQ" posts, there's a lot of them that explain it much better than I.  I didn't want to get sidetracked about the ins and outs of the process of actually EQing, but simply stick to what to look for when figuring out if a headphone will respond well to EQ, which as far as I could tell hadn't been talked about much.  Most people I've talked to about it seem to think that all headphones respond equally well to EQ.  I honestly wish I had not even put in the EQ picture at this point, haha.

 

[watchnerd]

  yeah, I should have linked one of the many other "how to EQ" posts, there's a lot of them that explain it much better than I.  I didn't want to get sidetracked about the ins and outs of the process of actually EQing, but simply stick to what to look for when figuring out if a headphone will respond well to EQ, which as far as I could tell hadn't been talked about much.  Most people I've talked to about it seem to think that all headphones respond equally well to EQ.  I honestly wish I had not even put in the EQ picture at this point, haha.

 
How one does the EQ is directly relevant to how much EQ it can "take".
 
PEQ does a better job of conveying the combined load you're putting on the signal than GEQ.

 

[fjrabon]

   
How one does the EQ is directly relevant to how much EQ it can "take".
 
PEQ does a better job of conveying the combined load you're putting on the signal than GEQ.

 
Yes, and there are many different threads on how to EQ.  This isn't a thread on how to EQ.  Anyway, I edited the OP to make that more clear, and hopefully less distracting.  Again, the graphic was a simple illustration, not a tutorial on how to EQ, and I've since removed it since it has turned this into a discussion on how to EQ, which is the primary thing I was attempting to avoid.  
 
I never said to use one type of EQ or another, because, again, it isn't what Im talking about.  And the issue I am talking about isn't going to go away with parametric EQ.  The K701 is still going to distort like crazy if you try to add 10 dB of bass, no matter what type of program you use, because the K701 has very bad THD measurements in the bass region, especially as power rises.  It is extremely bad at power handling in that region.  Further, it has phase issues in the bass that means it isn't as responsive to an attempt at adding more bass, 10 dB in EQ adjustment won't actually give you 10dB more bass on the K701, more like 3-5dB.  Yes, PEQ is more fine tuned, but it won't make those facts go away.