At 08:27 PM 11/11/97 -0500, Sean Breheny wrote:
>My design was originally for use in physics demonstrations or velocity and
>acceleration, so my goal was resolution of about 3 cm or so over about 7
>meters, about the same as your robotics idea. I have worked on the design
>of a similar project before, and we then experienced significant difficulty
>in working over 15 feet reflected distances (30 feet total round trip) when
>indoors, presumably due to reflections. The other guy who was the main
>designer of the project (who is also on this list, Louis Marquette), has
>since tried placing the transducers in cylinders and I think that he had
>good success.

I can't claim complete credit for this idea, myself, either.  I had heard that
 people had successully done this with the Polaroid transducers.

>Previously, you said that you wanted to avoid using an AGC. Theoretically,
>if we design the comparator/clipper circuit well enough and provide enough
>gain, it should respond equally well to weak signals and strong ones. You
>may not need to adjust the gain for longer distances. The main thing you
>need to do is adjust the timing and guard against the multipath.

Agreed.  The main design goals are low cost and reliable operation.  If we can
 make this device work with a simpler circuit, then the costs stay low.  Also, I
 don't want the end user to need to "fine tune" anything. I want it to work when
 assembled.

>You might also have difficulty with getting the circuit to work with very
>close ranges, like several inches, for a few reasons:
> 1.) The transmitting transducer as well as the whole device will vibrate
>during the TX period. They will continue to vibrate slightly at 40KHz for a
>few microseconds or so after the TX time. If you start listening for an
>echo  soon enough to hear one which comes from only a few inches away, you
>will most likely hear the vibrations instead of the echo.

True.  This is a big problem with the Polaroid unit, from what I hear.  Since
 the same transducer is doing both transmit and receive functions, it is even
 more susceptible to false triggering if its "blanking interval" (the time
 period it doesn't look for an echo) is set too low.

I was thinking that the transducers should be fairly snugly mounted in some sort
 of damping medium.  Maybe something along the lines of cork or rubber, or maybe
 some sort of styrofoam would work.  Also, placing them in tubes to increase
 their directivity should help.

>2.) The two transducers are probably separated by a few centimeters on the
>device. This means that the signal path does not go straight out, but
>instead must bounce off at an angle and return, forming an equilateral

Isosceles, actually, but I understand your point.

>triangle, not a straight line. Since the wavefront spreads out over long
>distances, and the length of the path becomes long compared to the distance
>between transducers, this effect is negligible at several feet. But I
>imagine that it may become pronounced at very short distances.

Although the absolute accuracy may drop off for very short measurements if you
 assume that the return time is a linear function of distance to the target, the
 measurements would still be useful, in any case, as a way of distinguishing
 "closer" or "farther".  Also, if accuracy is needed at close range, consider
 the following diagram:

             D
____________________________ Wall or other object
             ^
            /|\            A & B are the transducers.
           / | \           CD is the distance from the sensor
          /  |  \               array to the wall
         /   |   \         AD + DB is the distance measured (AD=DB)
        /    |    \        By the Pythagorean theorem:
       /-----+-----\         2     2     2
       A     C      B      AD  = AC  + CD

Since we know AD and AC, we can solve for CD as follows:
                  __________
                 /   2     2
           CD = V  AD  - AC

>I have also seen in several places recommendations that a specific pattern
>of pulses be used rather than a striaght 40KHz burst. This allows you to
>throw out noisy readings because the return pulses will not fit the
>expected pattern. I imagine that the pattern must be carefully designed so
>that it is still recognizable after being passed through the transducers,
>which act effectively as a narrow 40KHz band pass filter.

Probably a good idea, but well beyond the scope of a "first pass" at the
 project.  This might make a good area to experiment with in the future,
 however.  Hopefully, there won't be much 40KHz noise in the environments this
 will be operating in.

 - Rick "Once we finish the ping, next we do tracert" Dickinson


---
 You can help design a Serial Sonar Unit for model robots!
  Vist http://www.notesguy.com/notesguy for details, or
   send mail to Rick Dickinson at rtd@notesguy.com