On Tue, 07 Apr 2009 09:12:18 -0400, Joseph Gwinn
wrote:

In article ,
Don Foreman wrote:

On Tue, 07 Apr 2009 00:02:58 -0400, Joseph Gwinn
wrote:

In article ,
Don Foreman wrote:

On Sun, 05 Apr 2009 13:42:14 -0400, Joseph Gwinn
wrote:

In article ,
Spehro Pefhany wrote:

On Sat, 04 Apr 2009 19:50:29 -0400, the renowned Joseph Gwinn
wrote:


As for manufacture by a HSM, either approach is practical, but the Wyler
approach is far simpler mechanically. In either case, the electronics
part is dead simple by present-day standards, but was a big deal back
in the 1960s and 1970s.


Joe Gwinn

Have you considered the liquid type? They claim 5 arc second
repeatability.

Yes, I did consider electrolytric types (a bubble level vial with
electronic readout). The pendulum units are a factor of ten better in
resolution and repeatability, 0.1 or 0.2 arc seconds versus 5 arc
seconds. Long term drift is also an issue. It sure would be easier if
the electrolytic sensors were better than they are.

Joe Gwinn

Have a look at
http://www.spectronsensors.com/datas...S-105-2704.pdf

Not bad. The resolution is up there, but the repeatability is 0.0008
degrees, or 2.9 arc seconds, versus 0.1 arc seconds. Better than 5 arc
seconds, but still. This is why people bother with pendulum tiltmeters.

Joe Gwinn

I thought it was better but you are correct.

If one were to build an electrolytic level, this would be a good sensor
for it. The electronics are pretty simple. But I assume that one must
use only AC currents for sensing angle, a slight complication.

Very slight. Recall that liquid crystal displays (as found in a $1.99
multimeter from HF) also require AC excitation.


2.9 arc seconds is about 14 microinches per inch or .00014" in 10".
Will you post your pendulum design and results as time goes on,
please? Interesting project!

If I actually do it, sure.

This is one of those few areas where a dead simple gadget can achieve
such precision.


An odd thought: There was a burglar alarm design sold in the 1970s that
used a silicon strain gage glued to the underside of the floor
(especially the joists) to detect a person walking through. The maker
soon disappeared; don't really know why. May have been too fragile or
too temperature sensitive to be practical. (Silicon strain gages were
new then.) Or, too hard to install in an existing building. Or, too
sensitive to wind induced building sway. Or merely too hard to explain.
A precision tilt sensor could do the same thing, and could be installed
in a closet somewhere. The big advantage was that the sensor was not
easily found or approached undetected, and that one could set the
sensitivity to ignore small critters scurrying about and most pets.

Joe Gwinn

I designed stuff like that in those days. Signal processessing to
reject false alarms was indeed the trick before microprocessors became
common. It's still a good trick, but a lot easier to implement now.

Another approach was seismic. The sensors were geophones developed
for oilfield exploration, basically a dynamic mike with a slug instead
of a diaphragm. We made devices that could sense a man walking from a
distance of several meters, a vehicle from several hundred meters or
a chopper overhead, and it could tell the difference. It used a very
small number of CMOS logic chips (less than half a dozen) to do it.