[sdiy] Adding audio channels, to increase S/N

[Glen]

At 07:41 PM 9/4/03 , Ethan Duni wrote:

>I wonder if the original poster might expand on his reason for considering
>this stuff? To me, it seems that the effectiveness of any of these schemes
>depends directly on how you plan to handle the different signal paths. For
>instance, simply running multiple identical signals into cables that are all
>stuck together in a snake will not improve your signal to noise ratio.

I was only considering the effects of this technique where "random"
(typically thermal) noise was a significant issue. I know better than to
think that any aberration that existed identically in all channels would be
reduced by this technique. I also know better than to think the it would
have any noticeable effect on a simple audio cable or snake. 

I was thinking of those situations where it can indeed make a difference in
the S/N ratio. It certainly makes a difference in some analog processes
such as in analog amplification, analog signal processing, or lossy and
noisy analog recording mediums such as magnetic tape. (Yes, I have read
stories of people hooking up huge multi-track analog tape recorders as
either a mono or stereo recorder, to reduce the noise imparted to the
signal by the magnetic tape recording and playback process.)

The same basic principle is also used in photography to improve the S/N
ratio of some images. You basically take several consecutive images of the
same subject and average them together. The desired image (your actual
subject) remains relatively the same, but the random noise present in each
exposure will partially cancel out, leaving you with a cleaner image
(higher S/N). Recently, forensics experts have used a similar technique to
attempt a clarification of low-quality video surveillance tapes. The idea
involves using several still frames from the video, averaged together to
produce a sharper still image than any single video frame would have
produced on its own. This has enabled better facial recognition, and the
ability to read finer details in the images, such as otherwise illegible
vehicle license plate numbers. The exact details of this *particular* task
almost certainly involve more than a simple averaging of adjacent video
frames, but I believe the concept is very similar, if not fundamentally the
same.

I think it's funny that my little post has resulted in a thread with so
many replies.  :)

All I really wanted was a simple confirmation as to whether the theoretical
potential improvement was actually 3dB with each doubling of the "channels"
involved. (You can also replace the word "channels" in the previous
sentence with amplifiers, frames, images, transistors, tubes, active
elements, tracks--whatever is appropriate to the task at hand.) I assumed
that most people would realize that I probably knew when the technique
would work and when it wouldn't, and that I just had a fuzzy recollection
of the exact statistics involved. 

I truly appreciate the people that replied with a confirmation of the 3dB
statistic, and I find the number of people hinting that such a pursuit
might not be practical, truly amusing. It seems so common for experts or
engineers to remind others about what's not possible or not practical.
Where's their sense of adventure?  :)

(Remember that I'm the same fellow that recently said it's sometimes fun to
go where angels fear to tread.)   :)


thanks and take care,
Glen Berry