On Sun, 23 Jan 2000 13:29:11 -0800 barry lotz  writes:
>Hi
>   I'm kind of entering this thread in the middle. ...continued.

The problem with all this stuff is differentiating drift from legitimate
signal. In the example you cited, you simply have an indicator which
alerts the human (you) when attention MIGHT be required. After all, those
LEDs might be blinking due to a large quake.

My feeling is that with proper materials, reasonable periods (10 -15 sec)
and good use of damping and filtration, drift can be minimal with little
to no effect in performance. You will note also that I use the "solenoid"
type pickup which has tremendous advantages over the typical "edge
driven" coil in that there are no inverse square problems (as coil is
moved to and from magnet) and no internal cancellation (far side of coil
fights the near side). But the main difference is that the sensitivity of
the detector is quite flat across a +/- .7" deflection. I can usually
tell when the coupling is drifting too close because the gain will
increase quite dramatically when the coil is less than .05" from the face
of the magnet, where normally it is "centered" about .75" from the
magnet. Drift of +/- .5" can be ignored, and that range should be readily
achievable with a sensible and low cost design.

HOWEVER, since not all are interested in such designs, let me tell you
how I WOULD do it if I thought it was worth the effort:

It is logical to assume that, over a long period, a mechanically centered
boom will make as many excursions to the left of center as to the right.
There will be short term exceptions much as a bridge with double tolls
one way and toll free the other will carry "free" travelers who never
return, and paying travelers who always return via some other route. With
this in mind, I would simply put a "null" differentiator with adjustable
hysteresis (one quad OpAmp) on the "pen motor" signal (signal in your
design that is around 0 to +/- a volt or two) and have it clock a chip
(such as 74191) up and down. If the chip count is preset at 1/2 its total
(depends on how many bits) the perfectly centered boom will stay around
that count. In other words, each time it moves left you increment and
right you decrement. In addition you would need to clock for a permanent
shift (no return across center) by adding one increment per X minutes the
boom is off center. When the counter clocks UP to maximum or DOWN to
minimum (again, easily done with a couple NAND gates) you then increment
a stepper motor (one step) with a lead-screw that moves a lead weight
side to side either on your seismo base or on the slab itself. Since you
would be doing this for a very long period instrument the weight and
deflection required are probably less that one would think. By adjusting
the hysteresis (possibly to ignore six second noise) and the count total
(through use of divide by "N" counters), as well as the pitch and/or
gearing of the stepper you would have a simple and cheap method that
closely approximates the human operator.

In normal operation the count may reach max/min in a couple days of slab
tilt, adjust the weight a little and then again a couple days (or weeks)
later. If you put the new piano in the room overhead and the boom swings
permanently to one side, the "X" minutes increment will eventually roll
over enough motor steps to correct for the condition.

You want to try it? Go knock yourself out...

_____________________________________________________________________

Public Seismic Network Mailing List (PSN-L)