Headphones, Sonarworks and Harman Curves

[LevityProject]

I've been harping on a few things in the forum, and I'm noticing that I am either missing something huge (which I'd be grateful to understand better) or I'm onto something, but I need more information. Perhaps it's a bit of both.

Here's some context.

I use Sonarworks to calibrate my headphones for mixing on the road. One thing I did that changed a lot of how I perceive audio was I bought a pair of custom Tuned headphones from Sonarworks as well as a handy little perceptual plugin called tonal balance control.

The goal was to be able to accurately hear what tonal balance control is telling me. Eg if there is missing 2k information, I should be able to hear that issue in the headphones. What I've noticed is that I hear it more with some sort of Harman type curve. So my mind has been blown this week looking into how to perfect a curve for my ears.

What I know for sure is that these headphones I've got are utterly and completely flat. And what I am learning now is that in headphones, "flat does not = perceptually neutral).

So I need to clarify terms. I see "flat response" as a hardware technical term relating to the frequency response of a given speaker/headphone not coloured by our ears.

Neutral is a qualitative description of the tonal shape of a speaker.

Does this make sense?

What I want to do now is add in a smooth Harman curve to these flat response headphones to see if I can create a really smooth and neutral, natural speaker-like sound out of them.

As a though experiment, let's say a pair of headphones measured a straight line across the frequency spectrum. What would need to be done on an EQ to give these theoretical cans a Harman curve?

Second, I see two variations of Harman curves ... One is for headphones .. like this (I think we use a newer one?)

the other is for speakers, which seems to be somethin like an HF rolloff that starts at 1khz and is more than a few db down at 24khz relative to 0db (ala Bob Katz).

Am I missing something with that, or is the Harman type curve different from headphones to speakers? Is the speaker curve even a Harman curve?

I also have speakers that I've tuned to a flat response. So I'm wondering if the curve for those would look something like a tilt EQ that boosts below 100hz and starts rolling off at around 1khz, or for less bass boost, just a gentle rolloff that starts at 1khz.

I was a slowish convert, but the data is pointing me in the direction of mastering the harman curve, and therefore hopefully transforming my audio production practise.

 

[Galliardist]

It should work roughly like this:
If you have speakers that measure flat with an anechoic measurement, and you put those speakers into a decent room, you will hear through the effects of the room and get a good tonal response. There isn't a Harman target curve for speakers: there is a measurement that is the result of measuring good speakers in a good room.

The headphone curve is based on mimicking that in room response, and is a target curve. The curve is modified from controlled testing with a decent sample of different listeners. So the 2018 curve has more bass than previous iterations as a result.


WIth speakers, you can use EQ to mediate the room response which will affect bass, and also to mediate for poorer equipment. Seriously, though, it is better to start with well measuring equipment, especially for professional work.

...or maybe not. Mixing engineers, good and bad, get up to all sorts of things like using objectively poor speakers and overly dead rooms, as shown repeatedly in this forum and elsewhere. I assume there are reasons for that: but understanding how audio works should be part of the job, surely? It's probably time you read Toole's book as a grounding, a much better bet than asking on forums, even this one!

 

[staticV3]

What I know for sure is that these headphones I've got are utterly and completely flat. And what I am learning now is that in headphones, "flat does not = perceptually neutral).

What I want to do now is add in a smooth Harman curve to these flat response headphones to see if I can create a really smooth and neutral, natural speaker-like sound out of them.

If you're working with target curves, then you first needs to elaborate on what you (or Sonarworks) mean by "flat" when referring to those custom headphones.

If you mean flat response when compensated to Sonarworks' target curve, then adding the Harman curve on top of that will be catastrophically bad. It'll sound atrocious.

If they mean flat when measured on a GRAS HATS, like for example the LCD-4:

then correcting that headphone to the Harman OE2018 target should result in drastically better (and more realistic) sound quality.

The Harman target curve is only valid for headphone measurements created using a GRAS rig with anthropomorphic pinnae.

Without GRAS measurements of your headphone model, the Harman target is useless.

Btw, Sonarworks' in-house target is very close to Harman.

If you have custom Sonarworks calibrated headphones, then correcting them to Harman instead should not make a big difference.

 

[LevityProject]

If you're working with target curves, then you first need to elaborate on what you (or Sonarworks) mean by "flat" when referring to those custom headphones.

If you mean flat response when compensated to Sonarworks' target curve, then adding the Harman curve on top of that will be catastrophically bad. It'll sound atrocious.k here's where I'm coming from. This is what I mean by Flat.

Ah ok, here is where I'm coming from. this is what I mean by flat ... One note, the Pioneer headphones have a God-awful frequency response out of the box. Hyped in all the wrong places. This is custom-tuned, which from what I now understand simply pushes them to what they are capable of fully reproducing across the frequency spectrum.

The tuned sound is quite focused and has excellent bass and top end, but lacks upper midrange clarity and feels a bit thick.

I think this makes sense because sensitivity to upper midrange frequencies in human hearing is a bit less (-2db at 3.5khz at 70dB SPL) than the rest of the spectrum and we have sensitive hearing in the low-mids, according to the Munson sensitivity curves. Low frequencies also take more power to drive, so that might factor in here as well but I'm not sure how.

source: https://www.izotope.com/en/learn/what-is-fletcher-munson-curve-equal-loudness-curves.html

The Harman curve suggests that something is happening to sub-bass (below 100hz) when it comes to headphones. I'm guessing that since headphone drivers are designed to vibrate ear canals, not entire rooms, a lift in the sub-bass is pleasing because it can deliver the right amount of vibration inside our ears. This can also explain why some like more and others less - different shapes and lengths of canals can lead to different needs. But I'm going to bet it's all within a couple of dB at the most. This phenomenon also seems to extend into upper bass (100-300hz). The Harman curve in my original post shows this clearly if I'm not mistaken.

So, in putting Fletcher and Harman together, I think I've come up with a custom Sonarworks curve to help make these cans (and, if I'm right, any Sonarworks tuned cans) more neutral sounding and pleasing when not using room emulation. The caveat, as would be the caveat with any EQ moves that make a target curve more neutral and speaker-like: The louder you listen, the more of a trough you'll get between 2-5khz, due to human hearing sensitivities. At very loud listening levels, you'd need to double what I've done. But that will just sound bad

@staticV3 you are absolutely right that using a full Harman curve with Sonarworks fully tuned headphones sounded terrible. This confused me greatly until I realized that the sensitivity of our ears is different at different listening levels, and when a headphone is tuned flat, we are left with the Munson curves to consider how much or little to boost and cut to produce a neutral sounding response.

So what I decided to try is, well, a lot of things. I started by setting Sound ID to 50% wet, then adjusted with whatever EQ points I needed to create a Harman curve, but I did it to 50% of the Harman curve spec. I must confess that what I settled on didn't sound bad. But was a hassle to dial in.

*the middle red line is the curve that was created by the chaos around it .. haha.

Here's a poorly drawn line over the resulting curve for some clarity.

With that set-up, I then I set Sound ID to completely wet and set up a new curve with boosts and cuts relative to a listening level of about 70db. This brought out subtle details in recordings that I hadn't noticed before and gave them pleasing clarity without being harsh, and low-end presence and thump without being overblown. I also rolled off the high end -6dB at 20hz. This let a lot of nice sparkle through the masters that I auditioned and didn't seem to be too crispy.

Here's a visual. I'd be very curious to see how this preset compares to something like a pair of headphones that comply closely with the ASR target curve,

And here's the particulars:
100hz: +1.25dB (low shelf) ... I also added a 1dB cut at 51hz to tame the little hump, but that won't be needed on other headphones.
200hz: -1dB Q of 2
3.5khz: +2.25dB Q of 0.75
15khz: -5dB (high shelf)

This curve produced a neutral and really pleasing sound on every mix I threw at it vs the flat-line of sonarworks tuned. I liked this better than the complicated curve ...

My suspicion is that slight variations of this curve will work with every pair of tuned headphones with Sonarworks to give them a speaker-like tone, with only minor adjustments needed to compensate for ripples in the frequency response of the headphones that Sonarworks couldn't deal with.

 

[staticV3]

this is what I mean by flat ... One note, the Pioneer headphones have a God-awful frequency response out of the box. Hyped in all the wrong places. This is custom-tuned, which from what I now understand simply pushes them to what they are capable of fully reproducing across the frequency spectrum.

The tuned sound is quite focused and has excellent bass and top end, but lacks upper midrange clarity and feels a bit thick.

I think this makes sense because sensitivity to upper midrange frequencies in human hearing is a bit less (-2db at 3.5khz at 70dB SPL) than the rest of the spectrum and we have sensitive hearing in the low-mids, according to the Munson sensitivity curves. Low frequencies also take more power to drive, so that might factor in here as well but I'm not sure how.

Ah ok. That shows flat frequency response when compensated to Sonarworks' target response, which already takes the sensitivity of human hearing at different frequencies into account.

I'd be very curious to see how this preset compares to something like a pair of headphones that comply closely with the ASR target curve

From what I can tell, a GRAS measurement of the SE-MS5T does not exist.

This makes it impossible to apply the Harman target (=ASR target) as intended by the designer.

By applying a combination of Fletcher Munson curves and faux Harman curve to the already Sonarworks corrected response of your headphones, you may well have arrived at a response that suits your preference.

Conceptually though, there was no firm logic to this process, IMO.

That's totally fine! All that matters is the end result, which is having a sound that you enjoy.

I just don't want you to jump to conclusions.

 

[LevityProject]

Ah ok. That shows flat frequency response when compensated to Sonarworks' target response, which already takes the sensitivity of human hearing at different frequencies into account.

That's interesting ... In the software, It looks like Sonarworks uses an inverse eq curve of a Headphone's measured response to arrive at a "flat response".

At least, that is what all of the correction graphs show in Sound ID. So it's confusing if the "flat line" isn't actually flat.

if I'm understanding, correcting headphones with sonarworks produces a Harmon like curve even though it doesn't show that in the software?

 

[staticV3]

That's interesting ... In the software, It looks like Sonarworks uses an inverse eq curve of a Headphone's measured response to arrive at a "flat response".

That headphone response is not the actual raw response.

It's the error when referenced against their Sonarworks target response.

The inverse of that removes the error, correcting the (never actually shown) raw headphone response to the (never actually shown) Sonarworks target.

if I'm understanding, correcting headphones with sonarworks produces a Harmon like curve even though it doesn't show that in the software?

Correct.

Also, Harman*

 

[LevityProject]

Well I'll be a Monkey's uncle. Thanks for this clarification ...

maybe this explains why just a couple db of adjustment made the difference in the end. It was a thin line between pleasing and overcooked.

 

[LevityProject]

Conceptually though, there was no firm logic to this process, IMO.

Definitely was a flawed process based on a misunderstanding of how sonarworks measures their headphones. But a happy accident none the less to arrive at something that seems to work on three pairs of tuned headphones I tried it on. (One custom tuned, and two tuned to the SR average curves).