They aren’t using a headphone when performing the measurements, so it doesn’t have to be redone for each headphone. It’s basically measure speaker FR in ear, measure speaker FR free field, calculate HRTF based on the difference.
If the technique weren’t useful for headphones, I’m sure oratory1990 would say so. 
Yes, but they’re using a headphone when applying the HRTF information to… correct the headphone. 
That works out better if you include information about the characteristics of the device you’re starting with, especially wrt. the size of its membrane and its position relative to the eardrum during use. Griesinger’s method bakes this in without special efforts or extra devices. In fact, bjorken22’s method could get there too: he just needs to measure in-ear when listening to speakers and in-ear when listening to the targeted headphones, and subtract those two from eachother. Then there shouldn’t be any need to add the -10 dB bass tilt manually (he initially gets it in both his measurements but then subtracts it away), or do any other guesswork-based manual corrections.
That isn’t saying much, Oratory is someone who has done staggering amounts of work based on the flawed premise of a single headphone-specific EQ profile (with no listener personalization) being able to provide a “flat” FR, or “neutrality”, or “transparency”. But as Griesinger has shown by comparing equal-loudness tunings made by different people, all with the same headphone, individual ear shape differences change the sound so much it’s nothing short of ridiculous to claim one EQ profile could achieve neutrality for everyone.
They’re using it to determine the FR target when using the IE microphone to measure the headphones. So yes, the measurement of each headphone needs to be done. But the procedure to determine the target does not.
It’s a common misconception, but oratory1990 has never claimed that his EQ presets provide a flat/neutral/transparent experience or that they do not need to be personalized. In addition he is a trained audio engineer who designs headphones for a living.
I didn’t mean to be dismissive of your statement about the IE microphone’s position though. For sure, ideally you’d be measuring at the eardrum. So the target, as calculated, will need to be adjusted. Just as bjorken22 noted, it is a starting point.
OK, so at @Resolve’s suggestion to “keep pressing Blaine on this”, and since I now know more about who has what username around these parts , I’ve actually read the chapters of that Theile paper and, as I initially suspected, it absolutely does not debunk Griesinger’s EQ method.
What it shows is that you can’t get through the perceived-loudness comparison method a (re)usable, numeric, graphically representable form of someone’s frontal speakers based HRTF-at-the-eardrum, of the same kind that you get with in-ear microphones, i.e. like a measurement of perception-independent SPL at the eardrum.
But that is not at all what Griesinger’s method is doing, or is intended or claimed to do! It precisely aims and claims to make the headphones produce the same perception of loudness envelope from both headphones and speakers, not the same SPL-at-the-eardrum! And amusingly and surprisingly enough, I only had to look as far as the next chapter of the very same paper to find further support for this approach:
Recently Sahr [I] suggested that just these psychoacoustic effects, which experimenters had been trying to avoid, have to be included in the measurement. He considers in a hypothetical model the path of a sound signal which has to travel up to the final judgment in the human brain and concludes:
One should assume that comparisons of loudness would be the most exact method of judging headphones. With this method, the sound of headphones is really judged as it is perceived by the test person. On the other hand, it might be noted that the measurements by probe microphone occur during the relatively early stages of the process.
However, this suggestion has not yet been examined.
Wilkommen im 2017, herr Theile! Yes it has. By dr. Griesinger. 
And his claim is that test subjects had near-universally(?) positive impressions of the resulting timbre.
If the goal is to present a good result to the consciousness process in the brain, as the output of the perception process, stopping your analysis at the level of physical air vibrations against the eardrum and referencing everything to that is too soon, too upstream. There’s still more that’s personal and variable that has to be included. And Griesinger’s method includes all of that.
I feel like you aren’t looking at the case for the diffuse field response as a baseline here, and that’s what’s leading you astray. It’s true that both that the SLD effect impacts the equal loudness response of a frontal loudspeaker, and that a frontal free field measured HRTF is a subjectively highly non-preferred response, but these factors do not necessarily counter each other out.
Griesinger has a fair number of anecdotes, but if there’s a work of his that includes a listening test which actually compared an individualized frontal loudness-matched equalization to…any headphone target which we know can score subjectively better than frontal FF, I’m unaware of it. On the converse side, we have results from both Lorho and Olive that indicate that a frontal FF-derived response will be highly non-preferred.
The best-case scenario for the body of work that we have would be that the SLD effect was precisely inverse to the differences between FF and the preferred targets, but if you look at Sahr and Theile’s work, you’ll find that this has not been demonstrated to be the case. Barring a “two wrongs make a right” situation, we find ourselves in a quite odious position of using an improper method (sighted loudness comparison between unlike sources, which we know produces incorrect matching) to achieve an incorrect target (frontal FF, which we know is the least-preferred of all HRTF based targets tested in every test it is present in).
Yeah, that’s the unfortunate weakness in my line of argumentation so far, I’d like to see more preference comparisons with his method included.
I… don’t understand. Are you actively trying to feed me more material in favor of my position and against yours? The abstract at your second link literally says:
For both headphones the new target based on an in-room loudspeaker response was the most preferred target response curve.
A problem there is that his method is fairly time-consuming and quite imprecise compared to in situ measurements. His best bet would be to do a direct comparison of subjective loudness matching EQs to in-ear microphone EQs, but since, memory serving, he has also advocated for frontal FF via in-ear microphones, I’m not expecting a lot here.
Do you not have AES access? I’m citing the fact that FF lost quite overwhelmingly
The “in-room based” targets are similar to DF in their HRTF “shape” but have a large bass shelf (RR) or a downward sloping response (RR_G) based on preferred in-room responses with loudspeakers
Nope, not a researcher, as I hinted at the first time you cited papers up-thread.
But I see my confusion now: strictly speaking Griesinger’s method is applied in a room, but not with an in-room-response with wall contributions, since he’s prioritizing frontalization and advocates the use of a smallish central speaker which would not generate much in the way of wall reflections and bass tilt.
For my use I’ve modified this method to prioritize timbral improvement, precisely because I found the bass to be insufficient, and the frontalization effect to be too selective with the instruments in the mix that would benefit, and too dependent on source material (basically a crapshoot with commercial stereo recordings). So my best results, which made me start recommending this to everyone, were obtained with a 30-degree stereo near-field setup (with monitors that go -3 at 49 Hz) rather than an anemic central speaker, and that’s why I tend to look at the in-room response result from that study as validating the method, even while I still keep calling it “the Griesinger method”, which in its original form corresponds more to the FF results.
So basically I should take care to clarify every time that I’m recommending and defending the Griesinger procedure, but applied to a reference of in-room full-range (or as good as you can get) stereo speakers.
Hey @abm0, I’ve been EQing my headphones using a very similar method for a while, then I went digging through the internet to see what other ways there were and I also stumbled onto Greisinger’s research.
One thing to note is that his method is focused on accurate Bi-aural recordings and playback. Since we’re mostly listening to material not accurately recorded in bi-aural format using his method of capture, I question the need to follow Greisinger’s method completely by matching frontal speakers’ EQ with the headphones’ EQ? The lack of bass response of following through with matching flat speakers is a concern.
That said, I found great success by making a multi-layered EQ profile using a method just like his (actually closer to what Linkwitz did to his headphones. link: Reference earphones). Use many sweeps or alternating tones, and leveling all the peaks and valleys in the FR, saving it as one layer of EQ. Then on another layer (easy to do in Equalizer APO with no addons by loading a second txt file for the current output device) adding back a personal preference curve using some of the main features of the Harman curve as filter frequencies and Q value (bass shelf, 3000hz, 6000hz, 8000hz, treble shelf), this way gave me a personalized sound quality that I simply haven’t been able to match using any other method so far (This method > Greisinger’s speaker match > EQing by ear and also fill in holes I see in review’s FR > EQing by ear by preference only > using 3rd party rig measurement based EQ to match “Harman”)
Cheers, and great discussions.
Thanks for the hint, but I don’t see why I would do that, the approach seems to lack a theoretical foundation and procedural rigor vs. Griesinger’s procedure which precisely specifies what you have to do to fix every single frequency band, it involves no error-prone tuning by ear (unless your speakers and headphones don’t have the same sub-bass extension - that’s the only region where I had to include guesswork).
I must be mistaken then, the Griesinger procedure involves him listening to tones by ear for both getting a HRTF curve off of flat calibrated speakers, and then listening to tones again by ear with headphones to get a HRTF curve off of the headphones, then applying the difference of the two to make a final EQ curve for the headphones?
From his own website http://www.davidgriesinger.com/ he said: About 10 years ago I realized that it was possible to test eardrum spectra using equal loudness measurements. Equal loudness measurements have been intensively studied. The key to make them work is an approximately 1 second alternation between a tone at a test frequency and a tone at a reference frequency. Almost everyone is capable of adjusting the perceived level of the test frequency to match the reference, with a reliability of about ± 1dB
You do whatever sounds best to you of course, so if this works, why not! However results from Harman research is pretty clear that most people much prefer a Harman curve instead of a free field curve. Page 29 of this here: https://www.listeninc.com/wp/media/Perception_and_-Measurement_of_Headphones_Sean_Olive.pdf
Also as mentioned earlier, there’s a good reason why Griesinger chose a Free field target, specifically for bi-aural content recorded and calibrated his way. So if you listen to anything else, there might be other EQ curves that you might prefer.
Do feel free to correct me though, I’m here to learn and not to be right! If you’ve got any tips on how you arrive at your perfect EQ let me know!
That’s not what “by ear” means in this context, it’s when you decide if you’ve reached the correct dB value by some arbitrary qualitative criterion like “liking what you’re hearing”. In Griesinger’s method you don’t apply arbitrary criteria, you judge pure loudness to reach the correct dB value. This has scientific rigor, “tuning by ear” does not.
Sure, and from what data I’ve seen so far, it seems I’m using a Harman-ish curve because I reference my tuning to the response of full-range speakers with room gain, not to a free-field or anechoic chamber sound. I’m pretty much right where I should be: the most preferred response from the research, that of speakers + room contributions, and also personalizing it to my HRTF in one and the same procedure. I don’t see what I could further optimize about it. Maybe the speaker positioning, because I’ve been getting a suspicion that I’ve eliminated the sense of envelopment in some atmospheric electronic music that’s meant to “surround” you, and that usually uncorrected headphones do present as appropriately enveloping but maybe my HRTF tuned ones haven’t been so good at. I’ll have to check this hypothesis when I find the time. (Would have to revisit my opposition to the idea of diffuse-field if this turns out to be the case. )
I see what you mean now. I do comparison tones and sweeps to EQ the whole FR sound flat first before adding preference, so this portion of the EQ process has the same error margin as Griesinger’s procedure. However you’re referring to the final process of how to make the headphones sound right and what the final total EQ curve is, and you’re saying matching flat headphones is not arbitrary. I can see why you say that even though I don’t quite agree.
That’s the ultimate level really, as long as you’re feeling as content as can be, that’s what I also aim for as well!
So let me ask you this then: in your final phase where you apply your Harman-like personalization, how do you make sure you’re not combining dB values for fixing the headphones with dB values for fixing the mastering of the songs you use to judge the result? Or do you not use songs, is the final tuning also done with synthetic signals somehow? I don’t think you’ve clarified how you judge what the correct dB values are for your second curve.
Fair question. It’s literally just my preference. Just like the Harman curve is something that 65% of people like, and some people want more or less bass. I don’t believe there’s an absolute correct target that a personal listening device need to adhere to. I’m not using this to produce or master music after all, so I go for what sounds neutral to me.
Having heard over 150 headphones with all sorts of tuning, I would say my preference isn’t far off what Harman curve sounds like, just some deviations in the lower bass and around 5.5k.
I can see your reasoning for sticking to matching flat speakers, but even Toole and Olive advocates personalized bass/treble adjustments for sounds systems and headphones. So I feel no need nor guilt in tuning the way I want to.
All that said however, using something like Griesinger’s method is amazing for tuning out unnatural sounding treble and mids.
Hello, you are often talking about using in-ear mics for FR and HRTF measurements to EQ your headphones.
Which mics are you using or which can you recommend? And how does the process look like?
With regards to which mics we’re using, at the moment they’re custom ones made by Blaine - but to a certain spec because we want to be compatible with what certain researchers are currently doing. Listener can fill in the details on that space.
The process for how we’re actually achieving an ideal FR at our eardrums though may differ. And I think you could go about it in different ways.
I’m sure there are other methods of doing this and more utility to using in-ear mics. But with that said, the most obvious benefit is that for the information below like 900hz… you don’t need your HRTF. So it’ll reveal leakage effects and how the mids, low mids and bass actually performs on your head. This is also highly relevant if you wear glasses for example.
Thanks a lot for your answer! I’m aware that I have many questions but the one that came to my mind first was, what if we use binaural mics like the dpa 4560. They are placed befor the ear entry. (They’re the only ones easily accessible in my country.) What are the benefits and drawbacks?
And would it be best/good/make sense to make your own binaural recordings of things you know well (like a persons voice in a certain room/rain/etc.) with your own ears and use those as a reverence point? Because if the binaural recording is functioning right (recording with your ears with in the right way, whatever that means) and you listen to it with a headphone that’s “perfectly” EQed to your hearing, it should theoretically sound like your again were the recording was made, hearing it like your in that exact situation.
If so, would be the dpa 4560 suitable for that or have you another recommendation?
PS: Would you mind sharing a picture of your custom made in ear mics?
Thank you a lot! I really appreciate it. 
Hi, i read the hole treat true. And if i get this correctly, you use the Griesinger procedure with stereo “full range” speakers at 30° to get a flat “speaker in a room” response? I ges there are calibrated?
Can you explain me the procedure, so i can try it out my self?
Thank you.
Wrt. speaker calibration I didn’t do much since I already trusted my Tannoy Reveal 502 monitors to be reasonably well calibrated. But I did try to smooth the response out a bit just to check that step off the list, so I played white noise through them and looked to get a spectral envelope with an overall -10 dB tilt from 20 to 20k based on Tyll’s 2016 Innerfidelity article about the Harman work. No professional microphone either, just an inline mic off some Samsung earbuds that I was surprised to see painting an already close to ideal envelope over the whole audio band. Figured that was good enough for detecting if anything was really horribly far from flat. I ended up cutting some sibilance around 6k and filling in some gaps between room modes around 60, 200 and 800 Hz.
Some days this seems a touch too bassy, so if I were to redo it all I’d probably go for that Bruel & Kjaer quoted curve of ~9 dB down from 20 to 20k. But basically you pick an in-room target and then calibrate the speakers to produce that response at the listening position where you’re going to sit for the Griesinger procedure, you don’t need to follow more complex calibration methods that try to correct the response for some wide array of listening positions.
