Newsgroups: comp.robotics
Path: brunix!uunet!noc.near.net!howland.reston.ans.net!darwin.sura.net!news-feed-1.peachnet.edu!umn.edu!staff.tc.umn.edu!rogers
From: rogers@staff.tc.umn.edu ()
Subject: Re: Position - how to know/keep? + found some "goodies"
Message-ID: <C6DnBv.EHr@news2.cis.umn.edu>
Sender: news@news2.cis.umn.edu (Usenet News Administration)
Nntp-Posting-Host: staff.tc.umn.edu
Organization: University of Minnesota
References: <10474@blue.cis.pitt.edu> <C6BM4t.FoI@news2.cis.umn.edu>
Date: Sun, 2 May 1993 02:06:23 GMT
Lines: 112

In article <C6BM4t.FoI@news2.cis.umn.edu> rogers@staff.tc.umn.edu () writes:
>In article <10474@blue.cis.pitt.edu> fmg@alpha.smi.med.pitt.edu (Filip Gieszczykiewicz) writes:
>>
>>	Greetings. I was thinking about the position control of an
>>	under/water (U/W for short) robot and have hit a wall: how
>>	does a tethered robot know where it is? I came up with a
>>	number of solutions (some doable now, some longterm)
>
>
>>	Oh, perhaps my original plans for max depth of 250 feet were a bit
>>	optimistic... is it just me or do Murphy's Laws apply to depth ->
>>	It seems that pressure is expontially related to depth ;-)
>
>   As for your depth problem, have you thought of filling your
>U/W robot with a fluid?  If you filled up the robot with a fluid about the
>density of water (hopefully non-conducting and non-corroding) you should
>be able to go to very extreme depths and not worry about hull implosion.
>Individual hermetically sealed components would pose a problem, but plastic
>and ceramic parts (without a lid or window) would be quite happy.
>
>
>   The Sonar is the way to go and I think all you really need is a nice
>array of transducers you can read and write to well, and a DSP filled with
>good code.   Analog Devices has a DSP book and they spend a chapter or
>section on this problem.  (It is a hardcover book I got at an Analog Devices
>DSP seminar.) (I can get the details of it if your interested, it is in my car).
>
>
>  I have given up my sonar thoughts to concentrate on Lasers, Because at a
>Linear Technology seminar they covered an inexpensive way to time events
>with 100ps resolution.  (Light travels about an inch in 100ps)
>Previously I didnt consider timing light echos to be an option, but this
>picosecond resolution timer has given me lots of new Ideas. (It can be built
>for Tens of dollars instead of Hundreds of dollars)
>
>
>-- 
>Brynn
>        rogers@staff.tc.umn.edu    Varitronic Systems




  A number of people want more info one the 100 picosecond resolution
timer I mentioned in a posting, so I guess I better post it.

   Also for those interested in sonar (multi transducer) a book that has
a whole chapter one the subject is "Digital Signal Processing Applications
Using the ADSP-2100 family"  Prentice-Hall 1990, ISBN 0-13-212978-7.
The book was written by Analog Devices to push their DSP's, but actually
is useful for anyones DSP's, and the book is designed to be a collage
textbook.


  Now for a Picosecond timer.  The technique is to start filling a small
capacitor with charge on the start signal, and to stop charging the capacitor
on the stop signal.  An A/D conversion is then done on the voltage present
in the capacitor, which is proportional to the time you spent charging the cap.

  You need to use a high input impedance ADC and your current source that 
you charge the capacitor with needs to be good and linear.

  The circuit given in the seminar has an 12 bit ADC (LTC1282 or LTC1273) 
which I would guess is about $12-15 in thousands, $30 or so in 1's Q.
Of course they used a Linear Tech part (it was a Linear Tech seminar).
A couple of Nand gates and a flipflop, 3 transistors 4 diodes a current source
(LM134) and a handful of 1% resistors and some caps. 
  The speaker mentioned that this circuit is NOT temperature compensated, which
would be a problem.

  

  To make a real Laser rangefinder or better yet, imager, you need some more
stuff that was not mentioned.

You need a Laser that can give you a pulse of light, pulse size is probably
not much concern.  

Also needed is a detector (A photodiode) that is both very fast responding
to light and very sensitive.   This might be the hardest part of the circuit.
From what I can tell, an Avalanche PIN photodiode might be the best type of
component for this job, but I just dont know [anyone know where to find
avalanche photodiodes?].


Finally you need a mechanical system to aim your laser and detector and 
know with pretty good accuracy where you are aiming.

Then you just aim at a spot, fire a Laser pulse and start your timer.
Stop your timer when you detect a return pulse and do the AD conversion.

Save the direction and time somewhere.

Move your spot and do it again.


There is still a large design project here.  I was so interested in the
timing circuit because that part of the circuit is at least 10X cheaper
than doing it with a digital counter (10 Gigahertz digital design).

Also it means you really only need one tracking mirror instead of the
two that I was looking at for a triangulating system.

I had worked on a triangulating system for a while but there are a number
of advantages to doing it this way. [its cheaper for a number of reasons,
but it still is gonna be in the 100-300 dollar range instead of $1000+]

Anyone have some suggested improvements?

-- 
Brynn
        rogers@staff.tc.umn.edu    Varitronic Systems
