My “one-to-two days” almost up, consider this a relatively brief comment on the subject - I had hoped to get some additional data together, but a lot of fighting with production machinery stopped me from doing anything metrological this week. I’m a man of my word, however, particularly when @metal571 is concerned, so some commentary on my tinkering with in-ear microphones and my take on their general usability.
As folks who’ve talked to me before may be aware, proper measurements of headphones and potential problems with measurement systems is a concern of mine. For my own lab, I have laid my worries mostly to rest by acquiring an industry-standard, anthropomorphic measurement system (my beloved Brüel & Kjær 4128C Head and Torso Simulator), but such systems are priced far beyond what is reasonable for hobbyists, and frankly impose a pretty onerous cost on small firms, so I’ve long been interested in whether there were more economical alternatives.
A starting point for my projects was an older HATS also in my possession, the Cortex-made “MK2 HATS”, which has IEC959-compliant pinnae terminated with a 1/2" measurement microphone next to the canal entrance - I referenced this in my older post regarding measurement systems, and recently compared quite a few headphones on them, finding that the difference on a headphone-to-headphone basis and on-average was very similar, less the “noise” of positioning/placement. The fact that this system seemed largely comparable - the differences of its IEC60318-4-less ear canal’s transfer function being relatively “static” with respect to headphones, so it could be compensated with a single frequency response compensation file or EQ - was quite intriguing to me, and made me wonder how suitable measurements done using a microphone occluding a canal would be.
I fabricated a small in-ear microphone using a Primo EM258, a Comply T600, and a small 3D printed spacer glued into the tip to keep the microphone in place, and put this in my HATS. More measurements than this looks like later (note that each of those lines is an average of five measurements…), I had an average difference between the two that looked quite a lot like the response of a 60318-4 coupler (e.g. the lack of ear impedance didn’t seem to change much other than the lack of its consistent gain), and subtracting the averaged difference from the minimum and maximum differences per headphone showed that the two tracked essentially with the variation of my placements (last measurement is two sets of 4128 measurements compared using the same methodology).
Around this point, I recalled that Hammershøi & Møller 2008 had included data for blocked ear DF HRTFs (Table II, column “BE” under heading “∆LDF”), and became intrigued about how well a generic compensation and a $10 capsule would compare to the ear simulators and individual unit compensation of my 4128’s internal mics. I measured six largish over-ear headphones (averages each, of course) with both the ear simulators and the probe microphones, then compensated the ear simulator measurements with my HATS’ DFHRTF, and the probe measurements with Hammershøi & Møller’s population average blocked ear DFHRTF - you can compare the results for yourself, but I’m fairly pleased with them.
Do note that the positioning of the capsule relative to the canal entrance is quite significant - as an example, the HD800 with a few difference placements, not far apart, and the same with the flush microphone measurement subtracted. A microphone sitting in the concha bowl will not yield the same response, and cannot be compensated with Hammershøi & Møller’s data - additionally, there is some reason to suspect that measurements from in the concha bowl rather than the canal will vary meaningfully between headphones in ways that measurements in the canal would not; the sort of headphone-specific variations which make a single compensation vexing.
I had hoped to bring some additional new data - including in-ear measurements on some human subjects with the same compensation - to bear here, but I’m quite busy these days, so I’m afraid you’re stuck with what I’ve got for the moment. I can provide links to some freely-accessible academic information on this subject if there is interest, but I don’t have time to summarize at the moment, so I’ll hold off on that until tomorrow I think.
@antdroid this may be of interest to you.