In brief:
Adjusting the Hall sensors and their electronics.
Measuring the accuracy of the setup.
Verification of the performance.
This procedure can be done in the Lab or Workshop. When the TopMount is at its final destination no additional justage should be necessary.
Assumed here is that when the magnet is placed the voltage level on the Hall sensor outputs will be lower than the midscale level they have when the magnet is not present. As the amplifiers are inverting the readout is increasing when the magnet approaches a sensor.
1/ Lock the magnet to the center with the locking tool.
2/ Verify with a non magnetic ruler that the distances between the magnet and the Hall sensors are very much the same for all 4 sensors. If not, re-adjust the PCB.
3/ Set the gain to maximal with the Gain potmeters.
4/ Adjust the Offset pots for a reading of 128.
5/ Lower the locking tool.
6/ Move the wire to the North, such that the 6 mm tube just touches the edge of the hole in the lower plate. Reduce the North Gain potmeter for a reading of 250. Repeat this for South, East and West.
I was able to measure the accuracy of the Hall system with an old A3 plotter from which I only had the mechanical part (plotting bed with stepper motors) and for which I realized some new electronics to control the steppers.
Alternatively one could use (parts of / or a modified) 3D printer or a laser engraver. The essence is that you have a thing which can be moved in X- and Y- direction by computer control.
Fig 1. The A3 plotter with the wire attached.
The plotter was centered below the top mount at 134 cm above the plotter bed and the wire was attached to the plotting head with a small spring.
A specially written program moved the plotting head from -200 to + 200 mm in NS and EW direction in steps of 20 mm, and registered the readings from the Hall sensors. After that the readings were graphically presented.
To prevent the magnet to wobble the movements were done very slowly, after a movement during about 2 seconds no data were registered and after that the data of 2 seconds were averaged. One complete measurement lasted around 25 minutes.
Some results below:
Fig 2. Measurement without serious calibration.
The red circles represent the measuring locations, the green circles are the measured results after some scaling. The green lines indicate which result belongs to which location.
Fig 3. After the adjustment procedure above.
The large errors in the corners result from the magnet stem touching the lower plate. A small rotational error results from an angular misalignment between the plotter and the topmount. I have accepted these errors for the time being.
When I map the amplitude / height ratio of the pendulum in my home to this setup I find the dotted circle to represent the actually used part of the range.
Yes, it is possible to derive a calibration matrix from these data an do interpolations for a stil better accuracy.
Verification.
In january 2025 a question was raised: "are we sure that the measured pendulum behaviour reflects the actual behavior".
So, below I provide more direct evidence with two short video's. In the first video you can see the bob move over a centimeter tape when it has about the maximum amplitude of the minor ellipse axis. I made two screenshots from this video, for the two directions of the movement.
Click on one of the stills to see the video.
Still 1: The edge of the bob passes at 132 mm. Click to see the video.
The video shows that the rotaton goes CCW, that is the positive direction.
So the peak to peak amplitude is 132 - 73 = 59 mm. Amplitude minor axis = 29.5 mm.
The Log file reports at that time an amplitude of 27 mm. So there is aa small scaling error.
VIDEO of the compass display on the logging laptop. It shows just over 1 period of the pendulum.
The dots come from the position measuring system with the Hall sensors and the magnet. They arrive at a rate of approximately 10 Hz.
Again we see the CCW direction of th ellipse.
Chain of events:
Log files from 2025-01-19.
Around 12:00 the measuring tape was attached. For this the bob had to be taken away and was relaunched.
Around 15:19 the video's were made, when the minor axis was about maximal..