Consider alternative statistical method of computing PT calibration values.

[harrisonCassar]

Description/Motivation

Currently, we take advantage of an assumption that our PTs behave linearly, utilizing a linear best-fit line to produce calibration values a,b of the form a*x + b for a given raw voltage value of x. We've observed in the past that this results in perhaps non-zero produced PSI readings for ambient pressure/depressurized system (0 PSI), as well as perhaps lower-confidence in the outputted PSI readings for extreme PSI readings (close to minimum/maximum values rated for a specific PT).

Proposed Solution

To fix this, we could perhaps consider a polynomial (or just non-linear) best-fit to produce calibration values for our PTs.

Some pros/cons to consider: Pros:

• Could better capture non-linear behavior of PTs throughout entire operating range (perhaps most expected at extreme operating values).
• Avoids assumption that PTs' output react linearly to changes in pressure. Cons:
• Could perhaps overfit to calibration data, which may have inherent noise that makes it stray from actual behavior on average.
• More computationally expensive (how much though? not sure until we test!)
• Strays from assumption that PTs are linear (as this is how they should behaving).
• Have to learn/compute an additional calibration value as a part of the calibration process.

We could, also, instead do simple linear interpolation? Lots to consider!

Related to #10, #11.

[harrisonCassar]

Labelling as low-priority due to reasonable assumption of linear-behaving PTs.

During operation, we'll be sure to be aware of this limitation for the time-being.

[philipdoucla]

Through operations, PTs and LC demonstrated to be strongly linear.

At the current moment, linear best-fit proves to be the best option, and thus we move this issue to closure until future investigation.

~TRC 1/23/23