But I personaly think the best room is no room, especially over headphones where we don't expect one. do we really want the room? or do we want that floating stereo traingle?
I found it rather tricky with "no room".
I see the problem with room coloration and the question about the "right room“. An anechoic HRIR looks like the logical, purist answer.
But stereo is a system built upon psychoacoustic effects and approximations. Anechoic stereo (speakers as well as HP+HRIR) produces comb filtering with mono signals among other things. And this is not the same in all recordings as it depends on recording technique (XY versus A-B etc.)
And stereo is mixed in 99% of all cases with listening room reverberation in mind (a form of hopefully benign "spatial distortion").
When I built an anechoic crossfeed with the Neumann-HRIR and Anaglyph I was not satisfied with the spectral balance. With this FR (mono) of the HRIR this does not come as a surprise.
[These FR are not smoothed, it is the HRIR that has been smoothed.] This is the out put of R+L at the left ear.
So I equalized in such way that a mono signal (L+R) is reproduced with neutral FR.
But when listening the result had an extrem emphasis in the midrange.
I checked the FR of left and right (speaker) channel separate (again for the left ear) and found this.
What the heck?!?
This is the aforementioned comb filtering of course. The bumps (1.8 kHz) do cancel in a mono signal, but with music (stereo) it sounded unbearable.
It became clear to me why the FR of the HRIR looked so warped.
Now, how would one EQ that? I chose to flatten out the bumps (more or less of course and with the same filters in both channels) and try. I compared spectral balance of the EQed HRIR with the direct signal and tried to match that. Now the mono FR is off but it sounds ok. The result is instructive: it depends on the recording. Sometimes spectral balance is very close to "direct" sometimes not. With mono recordings it is not.
So there is no "right HRIR" (or EQ) either, so it seems. I guess stereo is just not (fully) compatible with binaural.
And then there is the mentioned problem with the "wrong" ears. In Anaglyph there are several HRIR from individuals (measured at IRCAM). They all sound off to me. In REW the responses look quite ragged and with rather big differences left-right. Maybe that is the reason.
The KU100-HRIR from Cologne are quite symmetrical, processed, somewhat smoothed and to me it sounds much better. But still with somewhat fuzzy localization. I guess that fuzziness might disappear with my own HRIR.
one could add reverb to the roomless binaural IR. first reflection algos also do exist.
I like the idea of a synthesized room with ideal characteristics that can be tweaked. But first reflections might not yet do the trick.
In Anaglyph there is a room module that can add a collection of rooms with some adjustment options. I did only a very short check and to me it seemed like not giving the same spatial effect (realism) as the WDR BRIR. So I did not investigate further, maybe it would be worth it.
Actually the WDR Controlroom1 does not leave much to wish for me. It is rather dry ( I like dry) but still has certainly enough reverb to reasonably eliminate the stereo problems from above. The difference between mono FR and single channel FR is much smaller. [Maybe some of the directivity discussion is about that?]
It sounds beefier ;-) For me coloration (in comparison to direct) does not change that much between recordings. And localization is definitely better for me, too.
So I would say: yes, I want the room! To me it is just better.
[Though I do not like the idea of adding "spatial distortion" at all, however benign it may be. I would prefer a purist solution.]
