Monday, July 26, 2010

Listening Comparison Of Comb Filtering in Boundary Mounted & Free Air Mics


One of the criticisms of placing microphones next to boundaries is that it produces objectionable comb filtering. This comparison test demonstrates that any comb filtering produced is very subtle and occurs at very high frequencies. The use of moving pink noise  for detecting the phenomenon was suggested on the Nature Recordist list. I performed many "3 O'clock" to "12 O'clock" pans with a boom box speaker around the mics in an arc at a distance of 5 ft. and 10 ft. I selected one sweep from both distances where all three mics exhibited smooth amplitudes.  I used a 744T recorder with 45 dB of gain to record all three mics at the same time.  I had to use a mix of AT3032 and AT 4022 mics which are known to be very, very similar in terms of performance. 

A blind version of the QuickTime movie comparison is available for downloading (3mb .zip)  to see if you can match the samples with the right mic configurations. I'll soon post the ID's below. 

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = =


Here's the comparison test the samples identified [ download .zip 3mb]
Here's a larger image of the sonogram with more detail.


Follow-up discussion can be found on the NatureRecordist list as well as in Paul Jacobson's assessment below. Curiously, the Perpendicular to Boundary capsule positioning produced the least amount of comb filtering in this test. 

The guesses I received  suggest that any artifacts pink noise sweeping might reveal are difficult to accurately link to the arrays:


Sample A Guesses:  Free Air, Free Air, PBB, Perp2Air, Perp2Boundary
Sample B Guesses:  Perp2Boundary, PBB, Perp2 Boundary, PBB. PBB
Sample C Guesses PBB, Perp2Boundary, Free Air, Free Air, Free Air

3 comments:

pj said...

Hi Rob,

This is a very interesting comparison.

I don't think that Long and Wickersham's patent on pressure zone devices suggests that you'll necessarily get audible comb filtering with the Perp2Boundary array. Rather they show a notch filtering effect that is dependent on the frequency and angle of the sound source. The proximity of the mic to the boundary means that we only see/hear the first notch of a comb filtering effect.

The main issue for Long and Wickersham is that with Perp2Boundary type array the cancellations vary across the face of the microphone depending on the angle and distance of the sound source.

With the Perp2Boundary style array the first cancellation notch is down at around 8khz when the sound source is at 90 degrees and increases in frequency as the sound source moves around toward 0 degrees. For our purposes 40 degrees or 1:30 using the clock position the cancelation moves out of audible range.

This can be seen on the Perp2Boundary sonograms as a narrow dark band increasing in frequency as the sound source moves away from 90 degrees.

One advantage of the PBB2 design is that it positions the mic diaphragm sufficiently close to the boundary to move any cancellations well out of the audio range. Having the the diaphragm 1mm above the boundary moves the first cancelation to around 52khz. In practical terms this should mean the PBB2 has a smoother polar pattern and less coloured sound than a Perp2Boundary array using the same mics.

cheers
Paul

pj said...

One further comment. The proximity of the mics and possibly a wall has resulted in some reflections which may have left visible artefacts in the sonograms. The "lines" which are apparent near 0 degrees position in the PBB2/Flush2Boundary and Free2Air sonograms are most likely a direct result of these reflections.

cheers
Paul

Rob D. said...

Hi Paul-- Thanks for the succinct clarification.

Several guessers commented they could positively ID which two samples were boundary mics by their tonal balance (mid range lifts). These are distinct boundary traits yet folks had a hard time discerning which was which. It would be logical to see if you and I came up with independent, consistent descriptions of the performance differences between HSPB array to PBBx sometime. Rob