Average human size ear canal
This is an interested topic because surprisingly not everyone agrees that ear canal is necessary to make the most realistic record. You can find sources telling that all directional information is captured at the entrance of an ear canal or just few millimetres deeper inside. But at the same time measurements from the labs show that there is directional dependancies just above around 6kHz. Paper link here. Problem is that measurements I have found are taken on real human after inserting probes into ear canals. These measurements are very inaccurate at high frequencies. Another thing is that previously I had a binaural head which I tested with an ear canal and without it. Results I got with the ear canal were better. More realistic image. Record made with ear canal had bigger difference between front and back sound compared to the record made without ear canal. So I decided to take some measurements to see if ear canal really has direction dependant characteristic. Some info about it can be found here.
Pic.1 Measuring rig for different angles
Pic.2 Measuring of a blocked ear canal
My measurements in Pic.3 show ear canal transfer function from blocked ear canal to eardrum of my binaural head. It’s important that this is an artificial head so accuracy is very high compared to the human volunteers with inserted probes. Measurements are taken for five different sound directions 0 to 180 degrees in 45 degree steps in horizontal plane (horizontally). If there is no ear canal dependency on sound arriving direction these five curves should overlay perfectly. Below 4kHz differences should be ignored because it could be affected by reflections. Below 1.5kHz differences are surely from reflections. Above 4kHz the reflections are completely eliminated with frequency dependant windowing. This is good because I could take measurements in my small workshop, and not so good because aggressive gating smoothes out the differences at higher frequencies. But even with that smoothing differences exist. So there is a plus for ear canal.
Another plus is that signal is amplified up to 6dB at 3kHz with ear canal versus blocked entrance. So signal to noise ratio of the capsule increases. It’s just a bonus! How much in dBA terms? It is very interesting. Let’s make some calculations. I take the part of the recorded signal from the environment where it is so quiet that I see self noise of the capsule. That will be my reference point. Now I take the average of transfer curves (Pic.3) and create compensation filter which looks like this in Pic.4. So I take the reading of the level in dBA before and after equalisation. Before it says -89.66 dBA and after it is -92.57 dBA. So I gain almost 3dB in signal to noise ratio of the capsule just by using ear canal. These are the benefits.
Some may say that ear canal differs from the listener's canal and related errors are produced. There are some errors theoretically. But the transfer function of the ear canal will be perfectly equalised back by applying EQ file for the head.( this is true if arriving angle of sound source is known at the playback moment ). Recording and equalisation are done on the same head so it is reversible. This will be easy as the phase for this transfer characteristic is minimum phase. Pic 5 shows a beautiful part of the excess phase which indicates minimum phase characteristic. The benefits from ear canal are taken but everything else will be about the same as for other binaural heads with closed ear canals.
Importance of ear canal is accepted among best binaural head's manufacturers. B&K's new HATS Type 5128 includes anatomically accurate ear canals. Also GRAS has Anthropometric Pinna with ear canal.
You can read a short compendium on binaural theory here. The author of that sections is independent researcher who made ear replicas with anatomically accurate ear canals which you can use if you want to make your own DIY binaural microphones. This is the link of ear replicas.
Pic.3 Ear canal directional dependence
Pic.4 Ear canal compensation filter
Pic.5 Ear canal transfer function amlitude and phase for 90 degree angle.