Newsgroups: comp.robotics
Path: brunix!news.Brown.EDU!noc.near.net!howland.reston.ans.net!agate!spool.mu.edu!uwm.edu!linac!att!att-out!undergrad.math.uwaterloo.ca!mwtilden
From: mwtilden@math.uwaterloo.ca (Mark W. Tilden)
Subject: Re: Cheap DRAM cameras?
Message-ID: <CFvup8.4xv@undergrad.math.uwaterloo.ca>
Sender: news@undergrad.math.uwaterloo.ca
Organization: University of Waterloo
References: <2b692b$2v7@news.u.washington.edu>
Date: Tue, 2 Nov 1993 21:07:56 GMT
Lines: 45

In article <2b692b$2v7@news.u.washington.edu>,
Phil Torre  KB7ZFH <ptorre@hardy.u.washington.edu> wrote:

>... recalled that years ago (early '80s) people started cutting the
>tops of of dynamic RAM chips and using the die as a cheap CCD image
>detector...

We did it and it's complex, hazard prone and not, alas, very useful.
Fragility, mask filtering, compartmentalization, mounting, and
alignment variations of the chips make it a nasty bit of work
mechanically, and then there are the electrical problems including
pixel luminance variations, bit scan speeds on the RAS and CAS clocks,
image re-encoding, controller electronics, and precision voltage
supplys.

And after all this, there's still the problem of what to do with the
data now you've got it.  There are entire books written about pattern
recognition, most of which demand excessive CPU power, but few will
tell you how to make sense of the visual data obtained, let alone how
to apply it.

Truth to tell, there was a small scale DRAM mentioned in a 1982 (?)
BYTE article that was useful, but it's long lost to technology.  Modern
memory chips suffer all the above problems, only more so.  As well,
there was the Fisher-Price Kiddycam, which used an ASIC and a
specialized DRAM to do the above.  The output is pseudo-NTSC and has
been used successfully in some robot designs.  Alas I haven't seen one
of these cameras for years.  Again, probably lost to history.

If you want some cheap distal sensors, look to modulated, focused
Ultra-brite LEDs and simple sensor clusters at the bottom of directed
culminating tubes (not lenses).  Replace the IR sensor on a Rshack
40kHz IR detector with a visible-light sensitive transistor and you'll
have a device which will respond to light you can see.  These modules
have a built in automatic gain control and notch filter which give them
high noise rejection.  It works.

Is all.


-- 
Mark W. Tilden.  "Gomi no Sensei des"       _    _    ________________________
MFCF, Un. of Waterloo, Ontario, Canada.    / \  / \  /________________________)
519/885-1211 <mwtilden@math.uwaterloo.ca> //\ \//\ \// ___o___________________
#include (standard.disclaimer);          //  \_/  \_/ (_______________________)
