[markanini]

For the sake of addressing frequency response differences due to personal positioning, I have spent some time figuring out manual EQ based on available averaged measurements, such as those by oratory1990. The strategies detailed below avoid unwanted harshness or dullness. I will explain how these unwanted effects occur despite being correct on paper and will provide evidence for why this happens and why my method avoids such issues. In other words, this guide shows how to EQ headphones without sounding like a mess and let your ears be the final judge.

This is my EQ curve for Creative Aurvana Live! using oratory1990s measurement.

Filter 1: ON PK Fc 32,0 Hz Gain 2,0 dB Q 1,41 Filter 2: ON PK Fc 125,0 Hz Gain -3,0 dB Q 1,41 Filter 3: ON PK Fc 250,0 Hz Gain -2,0 dB Q 1,41 Filter 4: ON PK Fc 500,0 Hz Gain 2,0 dB Q 1,41 Filter 5: ON PK Fc 1000,0 Hz Gain -3 dB Q 1,41

At the start I loaded the Harman target and headphone measurement curve into Room EQ wizard. Following that I fill in the octave bands into the EQ Filters window.

Next I start iterating "optimize gains" at various target levels, each time a new filter set is generated I input the values into EqualizerAPO and listen to a familiar song. Effectively the target level control is moving up and down what the correction routine considers to be zero level. After repeating this enough times I become confident on which target level sounds best. At this point I have reached a set of filters that can be used as-is or tweaked further by ear outside of REW. I rounded to the closest 1 dB because unit variation is still a larger factor so I'd rather gain fast input of filter values.

In this case I'm leaving the 2kHz band off, because raising it also raises the adjacent 3kHz peak. That would sound worse than leaving the dip around 2200Hz uncorrected. In part 2 I will use a 2kHz filter on a different headphone and you will see why.

Bands 4000hz 8000Hz and 16000Hz are left off always as correcting those leads to variations of dull and harsh.
This us why iterating the target level is necessary for proper integration above 2000Hz.

[More in part 2]

 

[markanini]

This is my EQ curve for AKG K612 using oratory1990s measurement:

Filter 1: ON PK Fc 32,0 Hz Gain 7 dB Q 1,41 Filter 2: ON PK Fc 64,0 Hz Gain 6 dB Q 1,41 Filter 3: ON PK Fc 125,0 Hz Gain 1 dB Q 1,41 Filter 4: ON PK Fc 250,0 Hz Gain -1 dB Q 1,41 Filter 5: ON PK Fc 500,0 Hz Gain 1 dB Q 1,41 Filter 6: ON PK Fc 1000,0 Hz Gain 2 dB Q 1,41 Filter 7: ON PK Fc 2000,0 Hz Gain -2 dB Q 1,41

As with the previous headphone, I experiment with target level values and review the generated filter set each time on a familiar song.
Unlike the previous headphone the stock response at 2kHz overshoots the target, and there's no peak to the right of it that I want to avoid raising. That's why I'm using the 2kHz filter to lower the band by 2dB this time.


Objection section:
"There's no evidence that EQ above a certain frequency is inherently bad."
At the same time you would be treading on thin ice because above a certain frequency the effects of personal positioning and unit variation cause differences on a larger scale. Because of that you can't be certain that EQ above a certain frequency is improving the sound or making it worse. You can be more certain of your subjective impressions in such a case.

"Octave-band EQ is too imprecise"
Heaphone manufacturing is more imprecise. Personal fit and positioning is more imprecise. You could use several parametric filters to correct peaks and dips though the whole frequency range and it will look great on paper. But it won't fix the lack of precision due to varying personal positioning. Correcting for such a moving target makes little sense. You end up with erroneous corrections to sharp nulls and peaks which can shift by thousands of Hz in the treble range.

"oratory1990 EQs his headphones at higher frequency ranges. Are you calling him stupid?"
His approach make sense because he's measuring the same unit that he's EQing and evaluating the sound throughout the process. Under different circumstances it makes less sense.

 

[markanini]

To provide concrete evidence on why EQ-ing above 2kHz is questionable take a look a this graph, from Tyll Hertsens:

You can see that the effects of varying positioning largely affect frequencies above 2Khz.

This leads to the conclusion is that there are two rational solutions for this type of corrective EQ:
1. Adjust higher frequencies by ear.
2. Leave higher frequencies un-EQ'd and focus on lower range EQ, and the intersection with higher frequency ranges, by ear. This is my chosen solution.

This concludes my write up. If you are curious about comparing this method to Oratory1990's presets or AutoEQ on your headphones let me know.

 

[bigshot]

I'd suggest using a range of different kinds of well recorded music to tune by ear, not just one recording you're familiar with. If you go with just one, you'll end up with a perfect curve for that song and other stuff won't sound as good. Listen to a bunch of recordings and average them.

In my experience balancing the midrange and upper mids is the most important. In a lot of cases, a wide Q to sculpt an overall curve is all you need. Narrow Q is just for occasional spikes. Above 6kHz, just get it so it sounds right and isn't blasting ultrasonic frequencies into your ears. No need to fiddle a lot up there. And your headphones will have a finite limit to how far you can push the low bass. Be conservative there to avoid boominess on some recordings. With my headphones, I'm lucky. All I need is about a eighth octave boost of 1.5-2dB in one spot.

A lot of the objections to EQ come from people who don't EQ themselves, so take those comments with a grain of salt.

 

[markanini]

A lot of the objections to EQ come from people who don't EQ themselves, so take those comments with a grain of salt.

For sure. Or they tried it without being aware of it's limitations and ended up dismissing EQ altogether cause they werent happy with the sound. That's why I wanted to address some of the common ones because it does raise interesting questions.

 

[castleofargh]

With headphones I don't mind going up to 8 or 10kHz if I feel like it's needed. I rarely touch above that for a mix of reasons, some that you mentioned, but I don't worry too much. Not all headphones have wild FR changes with different placement. They still have some, and obviously shorter waves(higher freqs) can be affected more. But in practice I rapidly get a sort of automated routine when it comes to placing a given headphone on my head. It's pretty consistent after months of using it. Nothing like when I move a headphone around to take measurement on my miniDSP E.A.R.S.
So while you're absolutely correct that the accuracy of what we'll do in the upper range probably doesn't allow precise EQ, and that we should probably not rely on measurements for those areas because our own headphone probably doesn't measure the same and it was on some dummy head instead of our head, I don't see too much of an issue for some moderate EQ by ear in the upper mid and treble range. I still would advise against big spiky boosts to try and get some recessed area up. Because it probably won't work, and because it probably won't work, but also because once shifted in frequency from placement or because the pads slowly gave in over time, we can end up having created a big spike. Which is the worst thing we can have in a frequency response, and what could potentially be bad for the ear(we set the loudness for the overall response and end up with that spike being too loud, all the time and slowly damaging our air cells that resonate most with that freq).


Now with IEMs, the slightest change in insertion will shift some main resonances(initially near 3kHz, 5 etc) of the ear canal. One better be confident about the way he inserts his IEMs and that he does get the same final response in both ears(which probably never happens at high freqs TBH but sometimes the differences become huge because of different ear canals and insertions). That alone is probably enough to go easy on EQ in the upper range unless it's to tilt the entire curve with something wide like bigshot mentioned. That isn't much of a risk, so long as we confirm by ear that we do prefer it.

IMO the first work of EQing transducers is probably to be careful about what we purchase in the first place. Some stuff are easy to EQ, like too much bass, a certain overall shape in the low to mid, an overall bright headphone. Some aren't, like massive roll off, some frequency cancellation(most unnatural dips have a chance to be one), some very local and big spikes.
I have gotten the habit of looking for those in graphs before purchasing anything. That significantly changed the way I shop. Initially I would just go for what feels closest to what I like, and if there was a little area I didn't like, oh well, it's "almost there". Now I'd rather get something with perhaps way too much bass and treble for my taste, but with a fairly smooth curve otherwize, because I know that's going to be a clean EQ that could leave me with good extension. or I know what low end profile I enjoy, so if I imagine that it's going to be hard to get there based on what's measured, I might give up and look for something else. I know I could do it, but if I need 6 bands just to make the low freqs response I want, maybe that headphone is just not for me. It's a slightly different way to shop.

 

[bigshot]

Wouldn't the possible range of correction depend on the headphones? Most headphones aren't likely to be 10dB out of whack, but if you really wanted to make a curve like that, aren't there some cans that could do it (as long as you aren't clipping)?

I totally agree that one should try to get as close as they can to their ideal curve to start with though. I don't know why people don't do that. Maybe they like the brand or the color or something. My cans are nice because they don't require amping and they don't really require EQing either. Close enough for government work.

 

[castleofargh]

Wouldn't the possible range of correction depend on the headphones? Most headphones aren't likely to be 10dB out of whack, but if you really wanted to make a curve like that, aren't there some cans that could do it (as long as you aren't clipping)?

The 2 main problems we can encounter, I think, are one about the headphone and one about the EQ itself:
1/ If some area ends up recessed for mechanical reasons for example, it's possible that trying to get that frequency loud(with EQ or by turning up the volume knob) will bring more distortions than amplitude boost to the headphone. In some cases, any extra boost might add a lot of distortion and rapidly become audible. Same thing the other way around though, lowering a given frequency that was struggling to output a certain amplitude at your listening level, could lead to reduced distortions. So yes it really depends on the headphone, and I'm tempted to think that the best ones(objectively) are likely to be the best candidates for massive EQ, for that very reason.

2/ We both tend to shrug when people talk about night and day impact of different low pass filters at 20kHz in DACs, but it's a different story when we bring ringing right inside the audible range. High amplitude gain and tight Q can introduce very audible ringing(with pre or not depending on the type of filter). I'm not so worried about phase shifts at an audible level when applied to both channels, but massive ringing is something we might like to avoid. And that isn't directly related to the headphone. I say directly because if we happen to compensate for the headphone's own phase shift, then we might just have improved things once again.

 

[bigshot]

I guess it depends!

 

[markanini]

With headphones I don't mind going up to 8 or 10kHz if I fell like it's needed. I rarely touch above that for a mix of reasons, some that you mentioned, but I don't worry too much. Not all headphones have wild FR changes with different placement. They still have some, and obviously shorter waves(higher freqs) can be affected more. But in practice I rapidly get a sort of automated routine when it comes to placing a given headphone on my head. It's pretty consistent after months of using it. Nothing like when I move a headphone around to take measurement on my miniDSP E.A.R.S.
So while you're absolutely correct that the accuracy of what we'll do in the upper range probably doesn't allow precise EQ, and that we should probably not rely on measurements for those areas because our own headphone probably doesn't measure the same and it was on some dummy head instead of our head, I don't see too much of an issue for some moderate EQ by ear in the upper mid and treble range. I still would advise against big spiky boosts to try and get some recessed area up. Because it probably won't work, and because it probably won't work, but also because once shifted in frequency from placement or because the pads slowly gave in over time, we can end up having created a big spike. Which is the worst thing we can have in a frequency response, and what could potentially be bad for the ear(we set the loudness for the overall response and end up with that spike being too loud, all the time and slowly damaging our air cells that resonate most with that freq).


Now with IEMs, the slightest change in insertion will shift some main resonances(initially near 3kHz, 5 etc) of the ear canal. One better be confident about the way he inserts his IEMs and that he does get the same final response in both ears(which probably never happens at high freqs TBH but sometimes the differences become huge because of different ear canals and insertions). That alone is probably enough to go easy on EQ in the upper range unless it's to tilt the entire curve with something wide like bigshot mentioned. That isn't much of a risk, so long as we confirm by ear that we do prefer it.

IMO the first work of EQing transducers is probably to be careful about what we purchase in the first place. Some stuff are easy to EQ, like too much bass, a certain overall shape in the low to mid, an overall bright headphone. Some aren't, like massive roll off, some frequency cancellation(most unnatural dips have a chance to be one), some very local and big spikes.
I have gotten the habit of looking for those in graphs before purchasing anything. That significantly changed the way I shop. Initially I would just go for what feels closest to what I like, and if there was a little area I didn't like, oh well, it's "almost there". Now I'd rather get something with perhaps way too much bass and treble for my taste, but with a fairly smooth curve otherwize, because I know that's going to be a clean EQ that could leave me with good extension. or I know what low end profile I enjoy, so if I imagine that it's going to be hard to get there based on what's measured, I might give up and look for something else. I know I could do it, but if I need 6 bands just to make the low freqs response I want, maybe that headphone is just not for me. It's a slightly different way to shop.

Headphones with questionable frequency responce get hyped up way too often. EQ might improve them but they should be weeded out. Same with IEMs, not all are equal in terms of how they handle varying insertion depths and reliable seal.

My guide deals with getting you started on trying out different targets and figuring out what you like. Because getting stuck on some of the pitfalls of high frequency EQ makes a lot of users dismiss EQ outright. At the same time you quickly hit the law of diminishing return while missing the big picture going about it that way. This is especially true if you already like your headphones, chances are they're a decent match above 2kHz. With some experience you can move on to trying EQ above 2kHz by ear.

 

[markanini]

Wouldn't the possible range of correction depend on the headphones? Most headphones aren't likely to be 10dB out of whack, but if you really wanted to make a curve like that, aren't there some cans that could do it (as long as you aren't clipping)?

I totally agree that one should try to get as close as they can to their ideal curve to start with though. I don't know why people don't do that. Maybe they like the brand or the color or something. My cans are nice because they don't require amping and they don't really require EQing either. Close enough for government work.

Adding harmonic distortion is an issue. REW is a pretty good tool for the task of headphone EQ overall because it doesn't correct for bass roll offs by default so you don't get extra bass distortion unless you raise those bands manually.

 

[bigshot]

I suppose it also depends on the Equalizer you’re using too. A 10dB correction is a pretty rare event I would imagine. I’ve never needed to get anywhere close to that. Maybe when EQing 78s... not sure.

 

[ADUHF]

Very nice writeup, markanini. Thank you for the post.

 

[markanini]

I suppose it also depends on the Equalizer you’re using too. A 10dB correction is a pretty rare event I would imagine. I’ve never needed to get anywhere close to that. Maybe when EQing 78s... not sure.

Some headphone have a lot off bass roll-off. If you were to use that much gain to fully match harman in the sub-bass region, you'd have to listen for distortion, and back off when it starts sounding detrimental. Or use better cans that don't roll off or distort too much.

 

[bigshot]

The bass rolloff might be because the bass just isn't there. You can do a huge correction to try and boost it, but it's going to be pretty distorted. If you want accurate sub bass, it's a lot better to use speakers rather than headphones. Sub bass isn't a strong point in headphones.