Sunday, December 05, 2010

John Hartog's "Wooden Bowl" Array with Curved Boundaries

With the possibility of making front to rear transitioning smoother, John started with a large, wooden bowl cut into pieces [download construction video 1 mb  .zip]. Using high density close-cell foam for substructure, he assembled the pieces into an array with proportions that echo a Crown SASS-P body. There are several photos of the finished array  on his website.

Click on graphic to enlarge
John's idea was that curved boundaries might "wrap" the pressure effect evenly around the array creating a more even gradient in the polar pattern of each boundary-mounted mic. Mic orientation after construction was more forward-facing at 45 degrees compared to 70 degrees for the Crown SASS. He design even includes s partially-covered baffle of high-density open-cell foam quite similar to the SASS P's except its nose is narrower.

He took the array to a large open field in the Mill Creek Wilderness near Prineville, Oregon and conducted two localization tests, compared frequency response with the same mics in free air and produced an example recording at the site for us to evaluate.

First, I looked at the free air comparison to evaluate the tonal nature of the lift created by the curved boundaries:

Click on movie still to enlarge
Download QuickTime Comparison Movie [ 13mb .zip ]

I used excerpts from John's 30 foot localization test of air "presence" with embedded footsteps a the same spots. The compared sounds are very similar but not identical. Pink noise would be more conclusive, but we have been able to detect a frequency response shift from boundary lifts using airy location presence in previous tests.
To the best of my ability, I cannot hear a boundary effect "lift" in the wooden bowl sample. The audible change between the samples appears to be increased left-right separation from the addition of the baffle between the mics. Pressure layers should be formed with the curved, hard boundaries and the flush-mounted mics should be affected; I can only guess that the build-up is quite small compared to flat boundaries. The reason would be interesting to explore, but Its sufficient for our purposes to observe that there is negligible frequency change produced under 2000 Hz by the curved boundary arrangement in the wooden bowl array. The sonogram made from calibrated samples seems to support this conclusion as well.

Next, I looked at John's localization test made for his new array at distance of 30 feet:

 Download QuickTime  Movie [ 25 mb .zip ]

Monitoring with speakers in my studio, clicks at 11, 12 and 1 imaged in "phantom" positions between my left and right speakers and clicks 2, 3 and 4 and 8, 9 and 10 seem to be stationary at the speaker positions. There is a distinguishable drop in level from 3 to 4 and an increase in level from 8 to 9. Positions 5, 6 & 7 image between the speakers and with decreasing tonal brightness making it easier to sense that these positions are to the rear. Taken as a whole, front to rear imaging is distinct but the goal to image many evenly spaced positions from front to rear was less successful.

The recording John provided made of a distant Coyote chorus and a Great Horned Owl is excellent for considering diffuse field performance [recording in entirety .mp3 ].  

 Download QuickTime Comparison Movie [ 13mb .zip ]

I went through my usual experiments with equalization and stereo field balance.  There's more noise in the recording than I'm accustomed to with AT 3032 mics. After quite a bit of very narrow notch filtering, the upper harmonics in the coyote calls were coming through with good clarity. I suppose we could come up with an explanation for the additional noise if we went through all of the factors. Perhaps I'm getting used to the noise performance advantage of arrays with larger flat boundaries and the 3032/4022 AT mics.

I've found that increasing the playback level of the mid range often improves clarity and depth imaging in diffuse field recordings. With John's recording, I created a smooth, +6 dB curve with Firium centered about 1200 Hz and tapering gently to about 4.5K Hz.   Using the left channel of parametric EQ inserted into an M-S chain, I attenuated some of the lower mid range rumble in the center of the stereo field (see yellow dips from 50-250 Hz in far right EQ graph). There was a little more bass rumble on the left which was addressed with the yellow cuts centered at 90 and 220 Hz  in the middle graph. I added a tad more mid-range lift to the sides and center of the field and boosted the overall level of the center 2.5 dB. 

I believe that some of the traits noted in the localization test can be heard in the stereo imaging of the distant sound sources.  With the center boosted and rumble reduced, the echoes have longer decay and provide greater sense of depth. Note, however, that the distant echoes seem to be concentrated in the center of the stereo field with a pronounced gap between the center and the right speaker. Very likely there were reflections originating between these two angles that are not well-represented. 

The recording responded to adjustments in post pretty well. I would have to look into the noise question before I could tell what direction to take the design.  Its hard to tell from these tests how the array shapes crucial mid range frequencies coming from the rear. Even if it proves to perform well in this regard, one would have to weigh this advantage  against the gain loss of the curved boundaries. I would never have guessed that the spherical shaped boundaries would behave so differently.  Maybe we can find add a sphere shape to a sound wave animation applet to see what happens in the pressure zone where the mics are flush-mounted.  Rob D.