Fluxgate Magnetometer tests: the FGM-3h (continued)

Thermal Effects
Long Term Drifts

Further Results of FGM-3h tests and comparisons, January - February 2000

Last changed April 27 2003 (Some links updated)

"Lab notebook" web page of Darrel Emerson (continued)

See also: Fluxgate Magnetometer tests: FGM-3h (December 1997)

Email to Darrel Emerson:

These tests describe measurements made an FGM-3h fluxgate magnetometer and its associated electronics; the effect of temperature changes on the magnetometer system, and residual long term drifts over a period of about 2 months were studied.

The sensors are made by Speake & Co., supplied in the USA by Fat Quarters Software and Electronics. More information may be obtained from Erich Kern. I have no personal connection with either company. The magnetometers are particularly sensitive, and are designed to measure small fluctuations in the geomagnetic field. The results presented here are of my entire system, including home-made electronics for DC voltage stabilization for the magnetometer, and the data acquisition hardware and software. The drifts I detect may come from this supporting electronics rather from the magnetometer itself; I'm still looking into that.

  1. Introduction
  2. General description of setup
  3. Measurements of Magnetometer drifts due to temperature: daily averages, corrected data, and a comparison with USGS data
  4. Conclusions


The earlier notes, Fluxgate Magnetometer tests described the Speake FGM3 series of fluxgate magnetometers, intended to be used to measure small variations of the geomagnetic field, or other small magnetic effects. The FGM-3h in particular is very sensitive - so sensitive that it will saturate if subject to the full geomagnetic field; in normal use, the FGM-3h is aligned E-W, at nearly right angles to the main geomagnetic field. Used in this way it is sensitive to the "Y" component of the earth's field. Fluctuations in the "Y" component are almost exactly proportional to variations in the "D" term, or the magnetic declination. Any book about the geomagnetic field will give the details.

These earlier measurements showed that, with reasonable care, the residual noise or error on measurements with the FGM-3h of magnetic activity, over a few hours, can be less than 0.5 nT (or 0.5 gamma) rms. This is provided that slow drifts, with a timescale of several hours or longer, can be ignored.

In the earlier measurements, the dominant cause of drift in magnetometer output was found to be temperature changes. I do not know how much of this temperature sensitivity is a feature of the magnetometer itself, or of my own electronics that supplies power to the magnetometer, and measures the pulse frequency from the sensor.

This document describes the result of monitoring the temperature of the FGM-3h magnetic sensor, and using the temperature measurements to derive a correction for measured magnetic field readings. This correction is quite successful. There are still weak residual long-term drifts in the output of the sensor, but sufficiently small and slow that for most purposes they will be insignificant.

General setup

The general measurement scheme is exactly as described in the earlier Fluxgate Magnetometer tests document. The FGM-3h magnetometer requires a stablized 5 volt supply, and gives a stream of TTL-compatible pulses, whose period is approximately proportional to the magnetic field. Corrections for slight non-linearity of this relationship is discussed in the earlier document. For zero field (i.e. with the magnetometer aligned exactly along the geomagnetic E-W axis) the pulse frequency is approximately 60 kHz. The responsivity was measured and plotted in the earlier test document, but is around 1 Hz change in frequency for a 1 nT change in magnetic field.

For the earlier measurements, the output of the magnetomete was fed to a frequency counter (an Optoelectronics model 3000) with an RS232 interface, so that results could be logged to a computer (a dedicated 286 laptop). The counter was set to a gate period of 10 seconds, so the frequency of around 60 kHz could be measured to about 0.1 Hz precision. For the measurements described here, a circuit using a Basic Stamp II was used. The schematic (which is more complicated than it needs to be) of the Basic Stamp II data acquisition system is available here. Some "tricks" were necessary to get the best out of the Stamp circuit - including making it monitor its own temperature, with a thermistor, and to apply corrections accordingly. One day, I'll write those details down in another web document. Throughout the measurements described here, the Stamp circuit and the original Optoelectronics counter were both collecting data simultaneously, logging data into independent laptop computers. Although data from the Stamp circuit were used for the analysis, both sets of data were regularly compared, and were found to be almost indistinguishable from each other.

The FGM-3h magnetometer used in these measurements is fairly well isolated. It is placed inside a vaccum (thermos) flask, which contains capsules of water to give thermal intertia. The screw-top of the thermos flask is replaced by a 2-inch styrofoam plug, of which about 1 inch pushes tightly into the flask. This gives better thermal isolation than the original screw top, and conveniently allows wires to the magnetometer, and to a temperature sensor probe, to enter the flask. The temperature probe is a thermistor, whose resistance is measured by the Stamp II microcontroller. The calibration from thermistor resistance to temperature is performed by the 286 computer. Long, thin wires were used to the temperature and magnetometer sensors, to reduce heat conduction to and from the outside world. This vacuum flask is inside a styrofoam picnic container, which is packed with freezer gel bags, to give more thermal inertia. This styrofoam container is placed inside a second styrofoam container, made of 2-inch thick styrofoam sheets. Inner and outer surfaces of the styrofoam boxes are coated with aluminum foil to reduce radiation transference of heat. The outer box is placed in a fairly secluded corner of my living room, against 2 outer stone walls of the house. All this thermal isolation quite effectively damps out the daily temperature variations that would otherwise badly affect the magnetometer readings. However, the inside temperature of the vacuum flask does change slowly, following the mean daily temperature, but with a time lag of a couple of days.

Magnetometer and temperature readings inside the flask are recorded by the Stamp II microcontroller every 10 seconds. The 286 computer calibrates and averages both magnetic and temperature data, recording the averages approximately every 2 minutes. The measurements described here are from the first 50 days of the year 2000.

Measured Magnetometer drifts

Below I show plots of my measured magnetic field, averaged in 24-hour blocks, for the first 50 days of the year 2000. I show matching plots of the temperature measured at the magnetometer; the correlation between changes of temperature and apparent changes in magnetic field is astoundingly good. The apparent change in field is entirely an instrumental effect, coming either from the magnetometer itself, or from the associated electronics. Because the apparent field change and temperature track each other so well, I believe that electronics outside my thermal enclosure cannot be responsible. More work is needed to locate the cause of this drift.

I have used the measured temperature changes to correct the measured magnetic field, assuming an empirical factor of 46 nT/deg F. Figure 3 shows the corrected values. Figure 4 shows measurements over the same period from the USGS Tucson Magnetic Observatory.

Although the steady drift in the USGS data (+0.5 nT/day) is smaller than my own (-2.3 nT/day), the residual rms noise in these daily averages, after subtracting the linear drift, is nearly identical in my own measurements to those from the Tucson Magnetic Observatory - i.e. 14 vs. 15.3 nT.

Figure 1

The plot below shows the temperature measured at the magnetometer, for the first 50 days of the year 2000. The temperature was measured every 10 seconds, but has been averaged over 24-hour blocks, 0h UTC to 2400h UTC for each of the 50 days.

Magnetometer temperature

Figure 2

The plot below shows the magnetic field measured with the FGM-3h magnetometer. The data have been averaged into 24-hour blocks.

Mag field drifts with temp

Figure 3

The plot below shows the corrected magnetic field, after subtracting a correction for temperature drifts. Changes in temperature were scaled by 46 nT/deg F to derive the correction. The straight line is a least squares fit to the data, and has a slope of -2.3 nT/day. The rms deviation of the data from that straight line is 15.3 nT.

Corrected Magnetometer readings

Figure 4

This plot (below) shows data over the same period taken from the USGS Tucson Magnetic Observatory. The data are available with 1-minute sampling, but have been averaged into 24-hour blocks to compare with my own measurements. The straight line is fitted to the data, and has a slope of +0.5 nT/day. The rms deviation from that straight line is 14 nT. Note that, from standard models, the expect drift in the "Y" component at my location (Tucson, AZ) and at this time of year is only about -0.02 nT/day.

USGS Tucson Magnetic Observatory


I've compared long term, day-to-day drifts in measurements from my Speake FGM-3h magnetometer with variations in temperature at the magnetometer. There is extremely good correlation. A simple correction, subtracting [(T-72)*46] nT from the data, makes a substantial improvement. There is a slow residual drift of -2.3 nT/day superposed on the residual. I don't know what causes the steady drift, but apart from that the residuals probably result from temperature variations of the electronics I use to collect the data.

Compared to the USGS Tucson Magnetic Observatory measurements, I'm not doing too badly. Their measurements show a much smaller drift of only +0.5 nT/day, but the rms noise of deviations from the steady drift is about the same magnitude (14 vs. 15.3 nT) on the USGS and my own data.

Finally, note that these measurements only show long term, day-to-day drifts and noise. Other measurements show that the short term noise of my FGM-3 h measurements, over a period of a few hours, is less than about 0.5 nT. Fortunately, most magnetic events of interest last from minutes to an hour or so, so the residual long term drift is of little consequence.

Back to top of document
Fluxgate magnetometer tests: Calibration of the FGM-3h magnetometer
Darrel Emerson top page