Thank you for in-depth analysis. I appreciate very much your time and effort, especially since this is the first time when my readings are so much confronted and objectively analysed.
So, as I can understand:
- my system seems to be superior against earless DIY systems (good thing),
- even compared to 25k$ HATS, the system looks not that bad and my readings can be considered somehow valid (good thing),
- there is a little room for applying a proper compensation curves for this system (bad thing),
- I probably should stick with raw frequency response just as I am working with them right now (no need for changes?).
Thank you for being so receptive to my commentary - I hope this can prove fruitful to us both
Going line by line for clarity:
- my system seems to be superior against earless DIY systems (good thing),
Substantially superior, indeed - I have done some additional analysis of earless flat plates today, prompted by this post, which will show just how troublesome they can be.
- even compared to 25k$ HATS, the system looks not that bad and my readings can be considered somehow valid (good thing),
Assuming we attribute the big departure below 3khz to unit variation with the LCD2 Classic, yes - your system does not seem to vary all that much, and possibly a fair bit less than the EARS - it might even be comparable in this regard to the ear simulator less HATS I have used prior to acquiring my current, fully-compliant unit.
- there is a little room for applying a proper compensation curves for this system (bad thing),
This one is hard to be certain of. If your binaural microphone is, indeed, consistent in the degree to which it deviates from a HATS, it should be possible to apply the same compensation as for the HATS but compensated to the difference between the microphone and the HATS as well. I would not feel comfortable recommending that with the data we currently have, but if the differences were relatively stable (say, +.-2dB) between headphones, you might be able to somewhat safely apply a compensation curve. However, since your unit does not replicate a human head, you cannot characterize this curve via the conventional means of actually placing the measurement device in the sound field in question, it would need to be based on an estimate of difference, in the case of headphones, for your mic vs. a HATS with real compensation.
Of course, such a compensation would be unsuitable for really serious work - R&D, research, etc - but
if the differences with the mic are relatively consistent, it should be close enough to give readers some idea of what they will hear. However, to determine if this was the case, you would need to measure many headphones on both of the two systems (I would say no less than 10 for my own comfort), and if they are different physical units, you may be confounded by the variation of QC.
- I probably should stick with raw frequency response just as I am working with them right now (no need for changes?).
This is subjective - IMO, it is less useful to present readers with a raw FR...but it would be very difficult to compensate your current system well. At the least, it would be better to stay as you are now than to apply compensations intended for anthropomorphic HATS or derived from human measurements, and I would argue also better to stay with the raw data than the attempt to determine a compensation subjectively.
A huge and repeatable difference in ~1-4 kHz caught my attention here, where my system is getting quieter, without strong emphasis on this particular area. At the same time the peak in 6 kHz seems to be a little stronger. I am surprised mostly because I always thought about strong accuracy for this system up to 10 kHz and only from that point the readings should be less and less accurate. So much difference in 1-4 kHz really puzzles me. And reminds me of various target curves, like:
When I for example open my readings for LCD-2F, put them on and try to examine aquired data with my own ears, there is no such emphasis in 1-4 kHz. In HD800 - the same situation, strong peak in 6 kHz but no emphasis on 1-4 kHz range, both on my readings and when listening test signal (sine wave). Of course this way of verification is very inaccurate because of the human hearing curve, but I believe there should be at least some of the practical corelation between the tonality you are getting from synthetic readings and with your ears.
The explanation here is pretty simple: the point of comparison I am using is the raw response of my HATS, which features an accurate ear gain, which your mic does not. Because my HATS has an HRTF fairly close to the population average, you will see a rise into the resonance of the ear, whose specific shape (for a flat stimulus) will be determined in part by the sound field in question. Here are the HRTFs for on-axis free field and a diffuse field that B&K advertises, for example:
If we consider that subjective flat equates to diffuse field, we would subtract the diffuse field HRTF from the raw response to get the diffuse field compensated frequency response. The same is true if we consider that a more abstract target, like Dr. Olive's Harman target, is proper. I have used the raw data here simply because it was simpler, but if there was a consistent deviation in response between two systems, you could simply sum that with the HRTF to apply the same compensation to either.
I definitely would agree that it is ideal to present data with a compensation that is subjectively correct - which is why I favour diffuse field over the older but easier to validate free field - and don't usually support publishing uncompensated data, but since in this case it doesn't change things with regard to the suitability of your measurement system either way, I took a little shortcut
With 2C I agree that there is indeed an error on my side with proper placement. That is the reason for the graph getting huge dip in ~9kHz and ~16 kHz. With sub 1kHz differences, silicone ears often are getting in the way. They are quite stiff and can cause improper earpad sealing, which affect the amount of lows in readings. This thing can be in many cases corrected by holding headphones tight with velcro strap, so the silicone ears would bend easier. Not the most professional way, but quite effective.
One note with regard to placement: if you do not already, I strongly suggest you take averages of a number of positions or repeated placements (or both), which should cut down on eccentricities like that to some degree. Stiff ears wise, you're actually in the same boat as many of us there - accurate-stiffness pinnae are a "new" development (last ten years) in HATS, and so many people, myself included, have excessively stiff rubber ears for our dummies. I'd be careful with the placement of that strap, however, as it may be acoustically significant if put on the back of an open design.