Arrakis : Leaf Thickness and Compass Test


Leaf Thickness Validation

The goal of this project is to validate the leaf thickness protocol on the MultispeQ v1.0 device. These devices were the first 20 manufactured by Lectronics in Saline MI to be used for testing and validation.


  • The MultispeQ leaf thickness is an accurate (thickness = multispeqThickness * .99 + .01, across 11 devices and 209 samples), inexpensive, and simple field based leaf thickness measurement.
  • Strong magnets or large pieces of metal close to the device can impact thickness calibration.
  • Leaf veins should be avoided when clamping if a very accurate leaf thickness measurement is desired.


The Leaf Thickness sensor on the MultispeQ v1.0 device works by placing a magnet on the bottom part of the leaf clamp, and a hall effect sensor on the top part (see image below). Hall effect sensors measure the strength of a magnetic field. As the clamp closes on the leaf, the magnetic field becomes stronger (ie denser field lines), producing a higher reading on the Hall Effect sensor.


Hall effect sensors are very inexpensive and can measure very small changes in the magnetic field, making them ideal for this application. The MultispeQ is calibrated to pieces of plastic of known thickness, from .05mm to 2.74mm (for those interested, a 2nd order polynomial fit is applied to fit the hall effect response to the actual thickness). However, differences in the magnetic field (like very nearby pieces of certain metals or other magnets) can change the Hall effect values and cause calibration values to be off.


11 devices were calibrated it the project 4b- v1.0 FACTORY CALIBRATION. Six pieces of plastic of known thickness were used to calibrate the raw Hall effect output (a value from 0 - 65525) to a thickness in mm. The calibration thicknesses were 0.05, 0.125, 0.19, 0.6, 1.016, 2.74 mm.

The 11 devices then remeasured each of the pieces of plastic of known thickness, and compared to the actual thickness. A linear fit between actual and MultispeQ thickness was created for each device. The ideal equation for the fit is actual.thickness = 1 x multispeq.thickness + 0 (ie actual thickness equals multispeq thickness).

The fit for all 11 devices and 6 thicknesses was very good. Below is the graph of MultispeQ thickness on the x-axis versus actual thickness on the y-axis.


The equation for a linear fit is y = 0.99x + 0.01 (r2 0.997)

Sources of error

While the above fit is close to the ideal fit, there is some variation from device to device. Below is the graph and linear fit equations for each device.



Another way to look at device to device variation is to see, at each thickness, the standard deviation across all 11 devices is. Below is a table showing the standard deviation, actual thickness, and predicted thickness.


Another source of variation can come if a magnet or large piece of metal is too close to the device tip. Avoid magnets from within 4 - 5 inches of the device tip, and be aware if you are within an inch or two of very large pieces of metal (a large metal table, for example).