Xref: utzoo comp.graphics:11694 sci.electronics:12106
Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!umich!samsung!think!linus!saint!lewis
From: lewis@saint (Keith Lewis)
Newsgroups: comp.graphics,sci.electronics
Subject: Re: Signal/noise and sampling...
Message-ID: <109567@linus.UUCP>
Date: 25 May 90 13:52:03 GMT
References: <1022@rna.UUCP>
Sender: root@linus.UUCP
Reply-To: lewis@saint.UUCP (Keith Lewis)
Organization: The MITRE Corporation, McLean, VA
Lines: 22

In article <1022@rna.UUCP> dan@rna.UUCP (Dan Ts'o) writes:
>
>	I wish to test several cameras for their respective signal/noise
>characteristics (S/N). The only DIGITIZER I have at the moment is spec'd
>at 8-bits (and only provides 8-bits output). I wish to infer the S/N of
>the CAMERA, which may or may not be better than 48db (8-bits). What is a
>good procedure ?
>
This test depends on the digitizer being up to spec.

Point the camera at something with a smooth intensity countour.  A shadeless
lamp 1 foot away from a SMOOTH white wall will do (keep the lamp itself 
out of the frame).  Grab a few frames in exactly the same position.  
Without noise, the frames would be identical, right?

In reality, if the camera IS better than the digitizer, there will be bands
where all the frames agree exactly and bands where they often differ.  The
thinner the differing bands are, the better the S/N of the camera.
- - - - Keith Lewis	lewis@saint.mitre.org, klewis@mitre.arpa
engineer-musician-network_manager-agitator-peacemaker-lover-philosopher
(703)883-5873	Disclaimer:  These opinions are not mine; I found them.
		If they are yours, you may claim them within 30 days.