To answer the question above simply, no, the in-ear response of speakers in a semi-reflective room would not be the same as a free field response curve, or a diffuse field curve.
Free field measurements are normally done in a heavily treated, echo-free room or enclosure. IOW, a room or sound field that’s free of any reflected sounds. So all you’re getting is the pure tone of the speaker or tone generator from a single direction, without the interference of any other reflected sounds from the room.
Diffuse field measurements, on the other hand, are done in reverberant rooms where the sound is coming more or less equally from all directions.
The direction of a free field sound source can also be varied relative to the measurement rig. So you can measure a dummy head’s response to a tone that originates directly in front of it, or behind it, or to the side, or even from above its head.
When you do measurements like this with a dummy head & torso simulator (HATS) in either an anechoic or diffuse room, what you’re actually measuring is the response or effect of the rig’s physical anatomy (ie its head, torso, and ears) to either a spectrally flat point source (free field) or spectrally flat diffuse sound field. IOW, the energy is usually equal at all frequencies in these two types of free and diffuse field sources. These measurements are done with a mic located at the ear drum reference point (DRP), inside the HATS rig’s artifical ears and ear canal. And the measurements are also known as the rig’s diffuse or free field head-related transfer function, or HRTF for short.
You can measure the HATS rig’s response to other kinds of sound sources as well. For example, you could set one up in your own home, and measure its response to a pair of speakers at the place where you normally sit and listen to your music. This is the type of sound field that I and Sean Olive (and others) are referring to when we talk about the in-ear response of neutral speakers in a semi-reflective room.
The rooms in your home are neither perfect reflectors, nor perfect absorbers of sound. So, semi-reflective. And the speakers that you use don’t just radiate sound in a single direction. The sound goes in all directions, but not equally so, and what you hear is a mixture of the speakers’ direct, reflected, and diffused sound in the room.
This also effects the spectral balance of sound, because most speakers tend to radiate more broadly in the lower frequencies. So even if your speakers measure flat across all frequencies from directly in front in a free field, when you put them in a semi-reflective space like the rooms in your home, the sound that you hear becomes tilted more towards the lower frequencies due to the attenuation of higher frequencies in the reflected and diffuse sounds.
The Harman over-ear target and earlier Olive-Welti curve are both rough approximations of the in-ear steady-state response of speakers in a semi-reflective space similar to your home, measured at the sweet spot or listening position (LP) in the room using different measurement rigs.
Here’s a photo of Tyll Hertsen’s Head Acoustics rig being used to measure the in-ear response of some speakers in Harman’s reference listening room btw. This was kindly posted on ASR a couple years ago by NTK.
This is probably neither here nor there. But since you asked, Luke, I’ll also mention another reason I like the DF+SP model better than some other models or ideas proposed here and elsewhere. The DF+SP model allows me to incorporate both the good and, if I choose, also some of the maybe not so good aspects of neutral speakers into my modeling of an in-ear response, via some more recent measurement data on both speakers and headphones.
If I don’t like the sound of HBK’s 1/3-octave 5128 DF with just a slope or bass boost, for example, then I have the option to look at some of the other recent sound power data on loudspeakers, and to incorporate some other things I might find there that might give better results with the content I listen to… such as a dip at 2k, or maybe a little boost at around 6k, or perhaps a little more rolloff in the sub-bass. As I described here and here, justifications for that sort of thing can often be found within the recent measurement data on speakers.
So the DF+SP model allows me to move a bit beyond the general confines of Harman’s old 3-knob preferred in-room response curve and the GRAS fixture measurements, and to experiment with some other useful and potentially better-sounding variations on the Harman theme, without straying too far away from either the general spirit or the letter of the Harman conclusions.