measuring tilt

http://www.signalquest.com/
http://www.spectronsensors.com/tilt.html

There products seem to either be just mechanical devices, which I think
won't be accurate enough, or accellerometers with a serial interface added
(which is utterly useless to me - in fact serial would be a pain to deal
with)


These also seem to be the sort of fluid based sensors that I'm hoping to
avoid. I just have this picture in my mind of the vibrations causing teh
fluids in such a sensor to just bounce and whatnot causing very inaccurate
readings...

-Michael

If your application would allow it, you could try to use inexpensive IR
ranging sensors mounted to the bottom of the platform. So, for example,
a single sensor mounted on the bottom of one side of the platform could
be used to derive tilt on one axis by measuring the distance of the
bottom of the platform to the "ground" reference. Add another sensor for
a second axis. There is at least one simple balancing robot that has
been built this way.

The Sharp GP2D12s or similar might work (analog output voltage that
varies with distance), although a lot depends on the expected distance
with respect to the ground over the range of platform tilt. The Sharp
sensors are pretty insensitive to ambient light, so the system would
probably work under a variety of lighting conditions.


--
(Replies: cleanse my address of the Mark of the Beast!)

Teleoperate a roving mobile robot from the web:
http://www.swampgas.com/robotics/rover.html

Coauthor with Dennis Clark of "Building Robot Drive Trains".
Buy several copies today!

I'm not sure you can get very much better without very much more
sophistication. "Tilt" is nothing more than an attitude against a
gravitational mass.

The mercury blob in a tube potentiometer is probably your best bet. I would
suggest a small very low-pass electronic circuit to stabilize the signal.
What sort of response time do you need?

Sharp makes some indexpensive tilt switches (about $2 each from Digikey)
that you can mount on an arc to measure discrete tilt from, say, 0 to 90
degrees. The extents depend on the number and angle of the switches you
mount. These switches are really optical slotted switches, with a ball
in the slot. They're pretty cool.

If you need more accuracy, there's always an accellerometer, like the
Memsic units ($29 from Parallax), or the slighly more expensive ADXL202
module that lots of folks sell. You can still get the surface mount
version of these directly from Analog, I believe, on a samples request.

--  Gordon

The sharp sensor would be fairly sensitive to vibrations and sudden
movement, correct?

An accellerometer sounds like it might be a good idea. Correct me if I'm
wrong - but I'll need a 3 axis accelerometer chip, right? It's
unfortunate there seem to be so few 3 axis chips on the market -
Analog's only one is delayed till september apparently - which is too
long of a wait for me. Kionix makes a very nice looking chip:
http://www.kionix.com/Product%20Sheets/KXP74%20Series.pdf  - seems the
primary challenge will be finding a distributor for the bugger. (already
e-mailed all the ones listed on their website - here's to hoping I
actually get a response from one of them) It even uses SPI which is
perfect for me as my whole machine is wired for SPI communication
between everything, so it'll fit in quite nicely.

How sensitive would an accellerometer be to vibrations and sudden
movements?

Thanks Gordon,



-Michael J. Noone

It's hard to say because the Sharp sensors are single axis, so
vibrations and movement not in the axis will likely have little effect.
The Sharp datasheet might indicate g-force specs if this is critical for
you.
 

Well, Z-axis measurement isn't mandatory for measuring "tilt" so you've
added a complication here. If you need three axes, add a second chip.


Usually very sensitive, which is what they're designed for (airbag
deployment, for example). You need to reference the spec sheet to
determine what type of accellerometer you need for your application. You
might find a higher-g chip is needed if your application involves a lot
of vibrations.

-- Gordon

@NOgmccombSPAM.com:


Hi mr. McComb - keeping things simple is definitely a plus in my book. So a
2 axis sensor is all that is needed to measure tilt? Thanks,

-Michael

How about jsut using the "pendulum on a pot" concept.  If the frequency
of change of tilt wont be too high, you could jsut stick it somewhere
where it can swing around. The shorter the pendulum, the less youll have
momentum error.  Total parts cost probably 50 cents.

Andy P


Michael Noone wrote:

Problem is I expect that would end up being fairly large and heavy... This
robot is very very small and very very lightweight... budget is also not a
concern.

    Accelerometers will get you a good down reference for a stationary
object.  If you need a reliable down reference while something is
moving around, you need gyros.  A full attitude reference requires
three rate gyros, three accelerometers, and some computation.
The Crossbow AHRS is a good example.

    If you're not moving around much, you can get decent tilt
info with one two-axis accelerometer, like the Analog Devices
ADXL320.  ($3.50 each in quantity).   Or a single-chip
inclinometer (the ADIS16201) for about $35 in quantity.
Note that because two accelerometers get you a vector
quantity, you can do some correction for moving around,
or at least tell if you're moving.

    If you're building a balancing robot that runs or an aircraft,
you have to have rate gyros.  Inclinometers will not work.

                    John Nagle
                    Team Overbot