Alt-BEAM Archive
Message #05806
To: beam@sgiblab.sgi.com
From: Terry Newton wtnewton@nc5.infi.net
Date: Fri, 20 Aug 1999 09:39:55
Subject: [alt-beam] Re: TMOSFET sensor madness (stubborn anyone?)
At 04:30 AM 8/20/99 -0700, Sean Rigter wrote:
>Terry Newton's circuit and your detector both use "random leakage"
>effects. The value of the gate voltage is the sum of all leakage
>currents and totally random and unknown. By connecting a very high
>value resistor between the gate and a voltage source, the gate voltage
>can be "set" and the FET operation can be made more predictable. This
>resistor should be an order of magnitude smaller than the stray leakage
>"resistance" and as a result will stabilize operation at a reduction in
>the sensitivity.
>
> [very good analysis]
Y'all found the electroscope circuit, that's an old fun one dating back
to the dawn of electronics (back then it was a tube). I would recommend
using J-Fet transistors (or op amps) as they do not suffer from zapped
gates like mosfets... the gate is just a junction no insulation to zap.
Still put a big resistor on the gate to limit inflow just in case.
That doesn't work with mosfets, > 100v its zapped period. 100V can
be generated by moving your hand... at least use (low leak) clamp
diodes if you have to use a mosfet.
I have found that the signal is too unreliable to use as a
navigational sensor, the reason I used it was to produce a signal
when approached, waking up the robot. Reliability isn't required.
Just kind of neat how someone can shuffle feet on carpet 6' away
and it wakes up... e-fields are enormously complex and even move
with the breeze. I forget the numbers but it's something like 100
volts per foot natural gradient, we walk around with our heads
in a 500 volt field but we short it out, distorting the gradient
to some distance as all objects do. Reality looks very different
when viewed in static, things are there that aren't really there.
The table sensor idea is interesting... to test feasibility hook
up an electroscope w/led to a robot (can't be held by a person since
that distorts the field) and observe the output as it approaches
environment features. Tweek the circuit between runs. If anything
repeatable shows up, design circuitry to do whatever with the signal.
If a clear signal can be obtained only when approaching an obstacle
or if the antenna is beyond the table edge, then it would be a very
useful thing. Good luck!
Terry Newton
Home