Appreciate it!
I’ll try both methods and see which one works better. I find confidently identifying peaks takes some practice.
As for the bass shelf, so far I’ve just ended eyeballing it so that it conforms to the headphones natural shelf shape - usually just attenuated. I was just wondering if there was a silver bullet I wasn’t aware of.
Anyway, big thanks for the advice!
I have a question.
If I measure a set of headphones on a rig where I have the HRTF, then measure those same headphones on someone’s head, can I use the delta between the two measurements to derive that person’s HRTF?
I found this video regarding frequency response, which has some advice regarding equalization:
What is (Not) important in a Frequency Response?
The explanations regarding measurements were quite revealing. The take away for me was to take measurements with large grains of salt. The measurements on a dummy head are not going to be the same as on a given end user.
@Resolve sorry if this has been discussed before, but based on your latest experience with real ear measuring, what is the best way for an end user like me to do proper EQ for open back headphones at home? Especially between 1-10kHz? Where it matters most…
So, here’s where things get interesting. The in-ear measurements do not not take the outer ear into account. In the video referenced in my last post, the presenter talks about the HD 820. The drivers are angled (not like the OAE1/2, but angled none the less). The video talks about how the fit on the head causes the actual sound heard from the 820 may not match what the measurements show via direct ear injection. So, after watching the video, decided to give the HD 820 a spin with no EQ applied.
Listening to Frank Sinatra (A swinging affair) with PGGB at 352 KHz to a Canever ZeroUno out to the GR Research OTL headphone amp, the sound had a slight midbass boost, but overall, the music sounded very good. The midrange was clear, concise, with a great soundstage, and excellent detail. Horns sounded like horns, an area where many many headphones fall short. The treble was smooth, extended, without being harsh. The sound is a LOT better than many of the reviews would have the gentle readers believe.
In summary, believe your ears over everything else.
This isn’t correct. In-ear mic measurements do take the outer ear into account, that’s primarily the information you get. Open canal mics = Pinna effects in the presence of the ear load. Blocked canal mics = Pinna effects in isolation.
So whatever angle the driver is relative to the ear, that’s included in what gets captured by the measurement. What Grell is talking about has to do with frequency response - and how that is different for individuals compared to what you may see on a measurement rig, because of these variables. This is nothing separate from frequency response, just that this isn’t always measured in the same conditions.
It would seem that the frequency response heard by the listener would have some variance with phase and amplitude based on the position of the drivers compared to the position of the mic in the ear. This may be why some people don’t hear the dips/peaks with measurements?
I would think that what Mr. Grell was referring to.
Frequency response contains both phase and magnitude information.
And in-situ response variation is the reason why some people don’t hear the peaks/dips as others.
I tried the suggested method with a tone generator sweep and it wasn’t that hard. You wont get exact numbers but decent result in the end and at no cost at all. I use the APO/peter eq grafics
Mike
I don’t think I’m bad at EQing either. Using the same method for years (something I picked up while EQing speakers with live RTA measurements). But I want to have objective results. Measurements I can look at and try to understand what’s going on. I was hoping to get some directions from @Resolve 
We’re working on something, but at the moment my suggestion is to use a tool like Owliophile. Don’t worry too much about getting things perceptually flat, focus more on identifying peaks and reducing them.
Found this set of explanations about frequency response. It explains the issues and potential differences one can run into when listening and evaluating headphone playback.
Thanks for the suggestions. Looking forward to seeing what you are doing. In the meantime, I might pursue some in-ear mics to measure headphones on my head.
And this is precisely what we’re talking about. There are other non-sound related factors that can influence a preference, like the occlusion effect or high acoustic impedance. One thing that I think Jermo got slightly wrong here, though I can understand the assumption, is the idea that because a more preferred headphone deviated more significantly from Harman than a less preferred headphone, these other perceptual factors are the reason why. I’m not sure he said this specifically but that’s the implication I get from this dialogue.
Blaine on our team at the time, gave a presentation that showed this idea is actually congruent with the research where frequency response is being tested using the same headphone, where non-sound related perceptual factors remain constant.
So the assumption that “closer to Harman = more preferred” is demonstrably incorrect, even when we’re just talking about frequency response. The study Jermo cites also shows that, but you can’t conclude from this the reason is primarily to do with non-sound related perceptual factors. In fact, in Dr. Olive’s own subsequent work with IEMs, he showed the same people rate similarly very different sound signatures, and this again indicates preference is a range, not a singular fixed result. And so the idea that closer to ‘known good’ = more preferred is something to move away from.
Have you heard or tried the cross feed function of the Topping D900? Using the Convolution setting and studio room option, the results are simply amazing. Closed back headphones suddenly sound a lot more balanced. The Grell OAE2 sounds like a multi-kilobuck rig.
Here’s an brief explanation of the convolution setting: " Convolution Mode : This is the preferred and more sophisticated option. It uses advanced digital signal processing to simulate Head-Related Transfer Functions (HRTF) and room acoustics for a more natural, speaker-like soundstage."
This software does an excellent job of creating a binaural experience with standard recordings. It would be a very interesting endeavor to see what differences would be generated regarding EQ settings between no cross feed and using the Topping cross feed with the optimum settings.
So I was thinking about this: The -10dB downslope concept comes from spinorama data correct? But is there a technical or research basis (I think I read that there is?) that shows that that this ideal speaker total power curve is also justified for the listening position?
It’s derived from Harman’s headphone preference literature most directly, given they found a roughly 9-10 dB bass to treble delta to be preferable across studies. This just happens to also align fairly well with their speaker preference research.
Yeah that makes sense. I knew about the delta in their research, I just thought it was more of the 105 +6 2500 -3 kind of delta rather than a straight downslope, but maybe I’m remembering that wrong?
2018 was like that, while 2013 was closer to a 105 +4.5 2500 -4.5 delta, but considering the speaker lit usually doesn’t have such a distinct shelf shape we feel okay using something a bit less “subwoofery,” if that makes sense. The blind testing we’re doing internally right now shows that some will not prefer the distinct shelf shape, while others may have a preference for a more distinct shelf (but that shelf’s corner frequency will change based on the person).
I don’t think the shape matters quite much as the level at low magnitudes, while at high magnitudes the shape becomes quite important. Just my $0.02.
Cornering frequency would correlate with listening room size preference or something right? Interesting. And high frequency shape would be a linear downslope?
Also, I’m just thinking about the -1 db/oct downslope right now: a linear downslope, I guess, just makes sense as a byproduct of the logarithmic physics of scale. It’s what the real world does and its what we expect. Even despite cycle-of-confusion issues, I would bet that its a common sound engineer best practice to mix and master with this in mind.