Port Whine
The original reels of the Port Chester run of '71 were audibly tainted
due to a grounding problem. 4/5/71 also suffers from this problem and
probably other '71 shows which I have not heard yet. The problem
results in a tone in the original reels which hovers around 8000 Hz
with an amplitude of about +20 to +30 dB above the noise floor. This
manifests itself as a very annoying pitch, especially during quiet
sections, which is in the old-tv-tube or lousy-hard-disk or maybe even
the-sound-you-hear-after-you-smack-your-head-against-a-concrete-wall
class of 'whine'. However, because this interference is monochromatic
it's easy to get rid of if you have the processing power and a quick
FFT algorithm. I used the filter-shaping capabilities of the Sonic Foundry Noise Reduction
Plugin to tailor a very narrow notch filter which attenuated the
interference without affecting the rest of the signal. And it's all done
in the digital domain. Huzzah!
Actually, I actually ended up crafting many filters to process
these shows. The frequency of the interference actually drifted anywhere
from 7500 to 8100 Hz during a given show. This was probably due to a
slightly wandering reel speed. It made my work harder but when I tried
a much wider filter I started to be able to hear the attenuation in the
music. Hence, I used the narrowest filter possible over a reasonable
timeframe to attenuate the wandering whine.
Note/warning: There is an audibly inferior version of 2/18/71
going around. The filtering on this version was done with a "brute
force" technique (one very wide filter). For comparisons with graphs and
my own ruminations check out this
link.
Fig. 1
A snapshot of a Sound Forge clip of 2/24/71, set I. The
red markers indicate the frequency of the whine at the marker. As you
can see, the frequency of the whine drifted throughout the set, mostly
in one direction.
Fig. 2
An example notch filter. This filter was
used between markers 7710 and 7750 seen in Fig. 1 (not the highlighted
section, sorry, the region to the right of it). The frequency response
represents a mix of a few seconds of inter-song "silence" at 7710 and
7750, focussed on the the interference. The twin peaks represent the
interference at 7710 and 7750. I crafted the filter (the parabolic
shape with the nodes) under the assumption that the whine did not drift
significantly outside of the range of the peaks.
The end result of all this selective filtering is no more annoying
whine, and no audible attenuation of the signal, i.e., the show!
I also removed some of the nasty microphone feedback in the 2/18/71 Bertha
using this type of filter.
Send all comments to 
orf5.com