The birch (Patreon)

I spent the last couple of days trying to optimise my entire recording/monitoring chain by laying at the ground and listening to the wind noises created by the moving birch leaves above me. It’s extremely hard to capture these noises precisely and also reproduce them faithfully with a headphone.

I want to eliminate any possible coloration from my binaural recordings caused by my ears at least to an acceptable degree, I was not satisfied with all the compensations I achieved so far. But trying to reproduce this natural spectrum with my binaural microphones helped me to find issues in both my current tuning for the HD600 which is used as flat reference to also create the necessary compensation for my recordings but also in my current IEM tuning for the Free Pro 2. There will be definitely a revision in future as I hear some remaining problems with the current “reference tuning”.. 

After lots of unsuccessful attempts I finally also managed to achieve a really usable compensation for my binaural recordings which you can hear from the attached audio file. It should sound most realistic when played back over the HD600 with my EQ (you can find the updated version here) it’s not perfect yet but definitely much better than anything else I managed until now. Of course having an as flat as possible headphone equalized towards my ear helped to avoid any additional colorations, thus the file shouldn’t sound that “off” when listened over a pair of flat studio monitors especially in the most critical upper midrange, it sounds pretty smooth and natural. Something which I couldn’t achieve with any of my previous versions so far. It won’t sound “correct” over any other headphone, but probably just noticeable to a trained and demanding listener. Neither does it sound perfectly correct over the Free Pro 2, that’s why I want to have another look at this to improve my tuning even more. Unfortunately the high sample variation of the Free Pro 2 makes my attempts rather a waste of time, the only solution would be to offer custom tuned units with a tuning optimised for their raw response and your ear canal resonance. The latest unit I received some days ago sounds completely off with my current settings and would need optimised settings.

The file was recorded with me laying flat at the ground to avoid any stronger wind rumble looking up into the sky to the tree above me. This allowed to avoid any sound coming from behind me, thus the entire “sound stage” is located in front of me, to the sides and also to the bottom and to the top, but the birch was exactly in front of me. Thanks to the distance and rather big and diffuse extension of the sound source I was able to get a nicely constant spectrum which was impossible to achieve with speakers as the spectrum will change continuously if you change the distance to the speakers or the distance between them etc causing holes and weird interferences in the midrange.

To explain the problem a bit better: In order to create a true sounding binaural recording which represents the real sound I need to suppress my own HRTFs from the recording somehow. Below you can see the measurement of my equalized HD600 done with my open inear mics for left and right ear. The measurements are averaged for several repositionings of both microphones and headphone.

Aside from some slight imbalance in the bass region for my unit, you can see that up to 1khz the response is basically flat. Thus any flat signal entering my ear will also sound flat through the headphone or should be captured flat in my recording, therefore I don't mess around much with the response below 1khz too much as it's already granted that how I perceive that area from my reference headphone this should represent "neutral" which of course can vary from "neutral" as perceived from a speaker inside a room. But above 1khz the human ear distorts the signal especially depending on the angle of the entering sound and both my left and right ear measure differently and need to be processed separately for best results. Easiest approach is to equalize this response to flat and use this a compensation for my recordings. I tried manual EQ, auto EQs, also matchEQ towards a neutral reference spectrum, unfortunately the results did not sound realistic like this. Comparing the recorded sound to the real thing always resulted in a very resonant and metallic sound full of single peaks. My attempts to optimise and correct for the flaws manually by ear never sounded convincing and ended in lots of frustration. I tried various strenghts of smoothing and additional averaging of the responses and got best results by using psychoacoustic smoothing instead of the usual octave based smoothing which you can see below:

Using this equalized to flat as base for further manual adjustments by ear but also verifying the final achieved spectrum with a reference spectrum I managed to record using a standard mic I could finally achieve my current result which I am quite satified with. To avoid any issues with left/right my additional corrections are only applied globally for both left and right side.  I am currently still trying to optimise any remaining colorations, but also my speaker recordings sound already way more faithful now and I am currently reencoding my latest UBOOM L video with the new compensation for a hopefully even more realistic sound. My upcoming videos will already use this or an improved version of the current compensation. As next I will need to sit down and readjust my tuning for the Free Pro 2 again because I can detect the colorations way better when listening to a recording of a real sound source and comparing this to the real thing, than comparing 2 headphones with pink noise to each other.

I have yet to test how my current compensation would work for recordings of headphones, as right now I concentrated on optimising recordings of more distant frontally placed sound sources. This is the reason I had paused with headphone reviews recently as I couldn't guarantee realistic sound samples with my current recording setup.