Thursday, April 02, 2009

Equalization Technique for Diffuse Field Recordings

At 2:27 PM +0300 4/2/09, Mike emailed me:
Hi Rob, Do you employ any structures around the mic rigs to equalize / compensate for the high frequency roll off associated with the diffuse field?

Hi Mike--

I'm not sure I get additional HF loss with the mics capsule oriented Perpendicular to Sphere in my rigs. However, I do find that all far-field recordings usually benefit from EQ if clarity or improved "realism" (based on naked ear experience) is the goal.
I've never used, "perfect mics/arrays;" please let me know if you find or make one!

As I've posted a few times on the naturerecordist list (with some dissatisfaction, I'm sure), I feel that each field recording or significantly time-separated event usually requires different EQ. Below is a step-by-step description of the way I tend to EQ field recordings:

[In performing the below steps, I am fond of the Eqium plug developed by Elemental Audio because its quite transparent and supports unlimited bands. Its now sold by Roger Nichols under a different name, Freqium. The EQ app/plugin must also have very quick response in "preview mode." You can use a simpified version of the below steps (fewer bands) if the paratmetric bands of the EQ plug you use aren't very narrow. The resulting tonal balance will probably turn out more natural than one without a procedure to address the biggest problems, first.]

(1) Set playback level at desired final playback level. Mark that point. Make sure the playback volume between the two tracks is balanced.

(2) Fade-up the volume to that point listening to frequencies that "stand-out" or have a drone quality. I calls this "Fade to Find." Make a mental note of the tone of the loudest one (the one you hear first as your fade-up the volume. Most of the time, they will occur between 125-700 Hz. If they are in the HF or VLF, ignore them for now and focus on the 125-700 Hz range.

(3) Use a sweeping narrow parametric band of EQ to locate that frequency between 125-700 Hz that is most exaggerated. After you find it, fine adjust "Q" width and and "tame" or gently attenuate it until the tone blends more naturally with adjacent frequencies. I never try to eliminate anything- this leads to overall imbalance. Basic rule, if it sounds good, cut the effect in half.

(4) Repeat steps (2-3) until the lower mid-range sounds more transparent-- meaning you can hear details that were previously "masked" by the sustained tones. Sometimes, there are as many as 6-8 bandwidths. I never try to eliminate anything- this leads to overall imbalance. Don't forget to use the "fade in to find trick" so you address the problem bandwidths in correct priority. [Note, sometimes the exaggeration is more pronounced on one channel, Eqium allows you address the channels separately].

(5) Explore boosting the upper mid-range (500-3.5K Hz) with 1-3 bands. This can add spatial clarity and depth.

(6) Perform narrow parametric on 16Hz-125Hz with same techniques as in step (2-3). Takes excellent monitoring. I use both phones and the best speakers I have to compare impacts. Don't forget to use the "fade in to find trick" so you address the problem bandwidths in correct priority.

(7) Address HF exaggerations-- most often with quite a few, very narrow bands of parametric EQ. The process is like "drilling out" very thin "zizz." Often its the "edgy" noise that comes from mic self-noise (which will show-up almost every time) and also HF distortion of content with lots of HF (like "sizzling" of leaves, grass in the wind). Done carefully, this method can preserve considerable HF "brilliance" while reducing the excessive HF content that make the HF in the space feel unnaturally close. Note that I've waited to do this step last because HF impression is very dependent on the space established by the mid range and LF.

(8) Adjust output level of the "mix" to get decent saturation-- no less than -15 dB peaks. Up to -.3 dB is no problem. For some reason, the effects of the EQ will be closer to what you heard in preview mode if the levels in the output file are robust. Determine overall dynamics on the EQ'd file in the mastering stage.

When exporting or mixing-out the file, "bump-up" to 24 bits the results if the recording is not already in 24 bits.

So, now you know why others dread this discussion. Rob D.