h a l f b a k e r y
Professional croissant on closed course. Do not attempt.
idea:
add, search, annotate, link, view, overview, recent, by name, best, random
meta:
news, help, about, links, report a problem
account:
Browse anonymously,
or get an account
and write.
Login
Create account.
|
|
|
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
A light shines through a gradiated lens onto a suspended, super thin reflective diaphragm. This reflected image hits a CCD sensor. The CCD output goes to a computer for analysis.
The diaphragm could be varied in thickness, possibly leading to superior wide range frequency response.
Rather than
a single waveform from a traditional mic, many more vibrational variances are picked up as an 'image' and digitally processed. DSP chips are used for processing in self contained units (drop-in replacements for current microphones), and/or inexpensive versions are available where the CCD output goes a tv card for processing on a host PC.
Traditional video editors can be used to 'see' and manipulate the sound. Existing visual filters (darken/lighten, enhance edges, etc.) can be applied with new, interesting results when output as sound.
The Microflown
http://www.microflown.com/
Recent advance in mic. technology [csea, Apr 18 2005]
Annotation:
|
| |
In practice, is this much different than the optical sound used for film, particularly 16mm distribution? |
|
| |
Possibly legend: The Soviets were once caught using the windows of the Pentagon as microphone diagrams. They were bouncing IR lasers off of the window surfaces and reading the modulation in the reflected beam as audio. |
|
| |
How were they caught?
Wow, that sounds like a setup for a punchline. Any takers? |
|
| |
I like the underlying idea of an optical mike but I think that some of the details need working through. Although recorded sound usually takes up less space than recorded video, it does need a faster sample rate. If you sampled at the speed that video is recorded, you'd lose all but the low frequency sounds. Do a google for the words "Nyquist Frequency" for a bit more information. |
|
| |
I am aware of the laser reflection audio pickup techniquie (shhhh, don't talk to my ex girlfriend). I was thinking of this more as a higher fidelity mic, not for spying. |
|
| |
st3f makes a great point though, the sampling rate needs to be taken into careful consideration -- I didn't think through that aspect far enough. For my method, each 'pixel' of audio would need to be processed -- a full 'video' frame sampled at only 60 fps wouldn't be sufficient. |
|
| |
I haven't double checked, but I think a CCD effectively integrates the intensity of light. When sampled at 60 fps, it gives the result of the total illumination over the last 1/60th of a second. |
|
| |
If the illumination could be distributed over the CCD in a raster-like way, with minor 'Y' variations corresponding to the signal, could it work? |
|
| |
Each sample would then contain many horizontal trace lines, and each line would look like an audio waveform. |
|
| |
On another track; if a laser beam was bounced off a 'microphone mirror' and subsequently bounced off the surfaces of two opposing parallel mirrors, then the beam would have the angle doubled each time (like an amplification). |
|
| |
Try to separate the requirement for
optical input from the notion that CCDs
are probably the most handy way of
doing it. A charge-coupled device is
really built for spatial optical sampling,
which is probably a red herring in this
instance. Perhaps all you need is an
optically focussed sample of some sort
of detectable excursion of reflectivity
value, relying on nothing more than
deviation from the norm of a single
light level. Given that, a lens +
phototransistor arrangement pointing
at a single flat light level might prove
completely sufficient, and would
allow sampling without top-end limits
of frequency of a video device, even
exceeding the
Nyquist point. |
|
| |
Though it's not optical, you might want to check out the Microflown [link]. |
|
| |
As well as the frequency response issue, any optical sensor using a CCD element would have to contend with signal-to-noise issues. Typical CCD elements have SNR of around 50dB. Quality microphones require up to 130dB SNR, pretty hard to achieve. |
|
| |
The CCD SNR ratio is for direct measurement of intensity of light, and not for measurements that depend on position of the light? |
|
| |
Interesting link. I never came across one before. |
|
| |
[Ling], you are right, but I think positional measurement techniques would also be hard to achieve due to inaccuracies in the optical path. Could a sufficiently large (# of elements) and small (for practicality) CCD show linearity to one part in 2^16 or 2^20? I don't know. |
|
| |
I could take a guess, I suppose, and reason that 40,000 samples per second divided by 1024 horizontal pixels divided by 60 CCD samples per second = 0.6, so (less than) only one horizontal scan line would be needed per frame (imagine the sound waveform is recorded as a wavy line across the frame).
In the Y direction, there are 2048 pixels (portrait). This is 2^11. Not enough resolution, so you are right. |
|
| |