Magnetic Gravimeter: Best Results

First created: Sun Feb 25 2001
Last updated: Sun May 27 2001

Three Days Of Data: 2001-05-[23-25]

I ran this experiment for 3 consecutive days, May 23 to 25, 2001. These are my first data collected with some kind of temperature control circuitry in place. The float signal and the temperature signal now stay in range for days at a time, which is a big improvement.

Original data

I collected the following data. The temperature reading are in black (rising temperature causes a decreasing signal) and the float position is in red (rising temperature also causes a decreasing signal).

Residuals after subtracting temperature effects

Data, temperature and residuals. The black and red curve are as before. The green curve is a linear regression of the the float signal to the temperature signal. The blue curve are the residuals. The means of the original temperature and probe signals were removed to improve the legibility of the residuals.

Probe(t) = 0.0038 + 2.3*Temp(t + 11)

The statistics of the fit (very high R-Squared) indicates that temperature variation explains nearly all the probe variation.

Residual Standard Error=0.0694

          Estimate Std.Err 
Intercept   0.0038  0.0011
X           2.3009  0.0017

Frequency analysis of residuals

Because the correlation with temperature is so tight, a frequency analysis (Discrete Fourier Transform or DFT) is hardly necessary. If contributing frequencies are cutoff at the first serious dip (9 on the plot, corresponding to a frequency of (9-1)/3 times a day) the smoothed fit to the residuals shows a 12 hour variation. The fit is less than convincing, considering the scatter of the residuals. Moreover if the 12 hour variation were taken seriously, it matches the theoretical east-west projection of the tidal forces, in an instrument designed to capture the vertical projection!


The temperature variation explains just about everything. For a chance to see the tidal signal, the temperature control would have to be improved still and sensitivity of the probe would have to be increased. But the improvements would probably require a new design. The temperature measurements are already amplified 100x and are quite sensitive, witness the scatter of the temperature points, creating the scatter in the residuals. The probe movement is also already amplified 100x. I am still using feedback on the probe movement, removing it could increase sensitivity.

In short, no tidal signal and little prospect that this device will ever see one.