EQ: Using Perceived Vs. Measured Flatness On Headphones For Mastering

[Audman71]

Hello,
 
I was wondering about the posibilities of EQ-ing my headphones, considering my current On-Ears are cheap to buy and don't currently sound the best for mixing. I have been experimenting with this off and on for a few years (I think I mentioned what I was up to here and there on this site), and I have concluded that the way I do it does not make any sense for causal listening. However, I am thinking about using the EQ techniques I use for mixing purposes. Here is what I mean:
 
As of right now, what I do is I use a digital audio workstation, a precise EQ plugin that is also an analyzer (ReaFir), and a test tone sweep (20hz to 20khz) to measure the lowest volume I can hear throughout the spectrum. I set the frequency lowest in amplitude (around 20hz) as the 0db point in my plugin and I equalize the rest of the frequencies along the spectrum to be the lowest I can hear them. Then I use the audio workstation to take existing songs and process them through it (for testing purposes).
 
The reason I ask the question of whether or not to use perceived vs. measured flatness for compensating the EQ for headphones is purely for budgetary reasons; aspects aside from headphone EQ notwithstanding, I want to know if I should buy a more expensive and inherently better pair of headphones than my current $20 ones and use those headphones for mixing my own tracks, or if I should use my technique on my $20 headphones and process my tracks to see if they are accptably mixed (I would, of course, bypass the plugin before rendering). I will say, however, I have a pair of Shure SRH940s and using my technique on my $20 On-Ears (the EQ adjustments make this a "perceived" response) impresses me more than when I don't use it on the EQ of the $300 Shure full-size headphones (which I will call "measured").
 
 
But is it an accurate technique? Is it a good idea to use it this way?

 

[PleasantSounds]

Interesting idea, but it has one major flaw: what we perceive as equal loudness changes with the volume level in a non-linear way.
Your mixes sound impressive because at very low volumes you have to boost the lowest and highest frequencies by much more than at any normal listening level. Google the equal loudness contour and you'll see what I mean.
Your Shure headphones are probably closer to flat and that's why they sound less flattering. But to be realistic, don't expect any headphones to be flat out of the box. 
Ideally you should check your mixes on a variety of platforms, starting from earbuds and ending on full size hi-fi speakers, at different volume levels. The best result is the one that sounds acceptable on all of them, even though most likely it is going to be a compromise.

 

[Audman71]

  Interesting idea, but it has one major flaw: what we perceive as equal loudness changes with the volume level in a non-linear way.
Your mixes sound impressive because at very low volumes you have to boost the lowest and highest frequencies by much more than at any normal listening level. Google the equal loudness contour and you'll see what I mean.
Your Shure headphones are probably closer to flat and that's why they sound less flattering. But to be realistic, don't expect any headphones to be flat out of the box. 
Ideally you should check your mixes on a variety of platforms, starting from earbuds and ending on full size hi-fi speakers, at different volume levels. The best result is the one that sounds acceptable on all of them, even though most likely it is going to be a compromise.

 
I've looked into that curve before and I just looked at it again. Wouldn't the sound appear to stay the same as the volume increases if I start equalizing with a sine sweep to make the volume for all frequencies from 20hz-20khz be at the threshold of hearing? It seems to indicate that to me, at least.
 
And I would actually argue that sounding flat sounds great. I like the sound of my Shure's a lot even though the bass frequencies are a little quiet, and their EQ curve is the most flat a built-in headphone EQ I have ever owned. In fact that's why I'm asking this question. My thinking is, if what I'm doing sounds flat, even if it is only flat to my ears, then I could use my technique as one of the baselines I use to mix audio. As I said before, I could bypass the plugin before rendering the track, thus not having my specific hearing curve affect the final sound.
 
I agree with your last point, though. It's something to think about.

 

[PleasantSounds]

OK, let's have a look at this together:
 

 
 
If you EQ for the faintest signal you can hear, the difference between 100 Hz and 2kHz is about 27dB (red dots). So to hear "flat" sound you need to apply +27dB at 100Hz.
 
Assuming you will be actually listening some 50dB louder, that equal loudness difference drops to 10dB (green dots). But your EQ curve is set up on the hearing threshold, therefore you're getting additional 17dB of boost at 100Hz.

 

[Audman71]

Two things there:
 
1: What I'm looking at is a curve that shows the equal loudness curve compensating for human hearing, correct? My understanding is that is not how the average person hears, but that it is a curve that shows the counterweight to the way human beings hear frequencies naturally, which, like human hearing, is dependent on the volume of the source. Am I wrong?
 
A measured-to-be-flat sine sweep like the one I am using will by no means sound flat unaltered. I am trying to make it sound flat by equalizing the tone to my personal lowest threshold so as it gets louder the way I hear it is largely to completely unchanged. My surgical EQ and sine sweep tests seem to reflect this well.
 
 
2: I am not trying to alter the curve you show by using an equalizer. The test tone sweeps I used are flat and are not pre-calibrated to fit the perception of the average human ear.

 

[PleasantSounds]

You can't change the curve I have shown. To do that you'd have to modify your ears. It reflects the properties of human hearing and is created based on the following definition:
 
An equal-loudness contour is a measure of sound pressure (dB SPL), over the frequency spectrum, for which a listener perceives a constant loudness when presented with pure steady tones.
 
The effective SPL you're getting is as close to flat as your transducers allow, But that SPL should be measured by calibrated instruments and not your ears. Or you should treat this graph as calibration curve of your ears as an SPL meter.
 
Your method would be fine if all the lines on the graph were parallell - then the model created at the hearing threshold would be applicable to the whole loudness range. But as the volume increases, we start hearing more of the lowest and highest frequencies (in relation to mids). 

 

[Audman71]

Originally Posted by PleasantSounds /img/forum/go_quote.gif
  Your method would be fine if all the lines on the graph were parallell - then the model created at the hearing threshold would be applicable to the whole loudness range. But as the volume increases, we start hearing more of the lowest and highest frequencies (in relation to mids). 

 

 
Huh. After looking back at my tests, I see what you mean about the ends getting much louder. However, I just tried recalibrating my results and it seems to be working great now.
 
Would you recommend using this method for mixing, assuming it could be/is done exactly the right way? Is it worth spending a lot of money on headphones for mixing purposes if I can outdo it with an equalizer, a DAW, and a pair of cheap headphones (I mean, aside from the benefits of soundstage and other factors)?

 

[PleasantSounds]

Yes that can get you close to a good result. Think also about your audience and how they are most likely to listen to your mixes. There's no point doing all the production on a mega$$ setup if your audience use $5 earbuds. Try not to overdo the EQ-ing of your monitoring system. It should sound relatively plain. Better to add colour to your mixes.

 

[Audman71]

Alright, thank you! 

 
Believe me, these $20 headphones sound far from plain and far from lopsided when I made my most recent adjustments. As of now, I am listening to music through my equalized filter and there is bass aplenty without it being overpowering, treble is smooth while not piercing, and the mids, across the board, come through nicely. I'm very pleased. You might want to reconsider thinking of flat EQs as "plain", because to me it is very obviously not plain.
 
And yet my sine sweep sounds quite even. Go figure.

 

[PleasantSounds]

What I meant by recommending plain was that if your monitoring system has too much colour then you will be taking it away from the mixes and they in turn may come out pretty bland. Conversely, to make it sound good on a plain monitoring system you will have to make the mix much richer. Decide for yourself what works better. 

 

[Audman71]

Well, that's what this technique is for. The whole point is to equalize the signal so it sounds flat when using a sine sweep, which in the process of doing so corrects the headphone's EQ imperfections as well. Obviously the mix can be what I choose, but my goal is to have the best possible EQ to work with as a starting point.
 
The headphones under this technique aren't my only monitoring system, just my secondary one. I have JBL SRH305s for my primary monitoring system.