Evaporometer Updates

Evaporometer Updates

---

November 17th, 2019 - Back to Basics (Part 1): The Need for Load Cell Reliability Testing on EvapLite

Author: Marissa Kwon

This update is the first in a series of "Back to Basics" updates, because it is time to review the fundamentals of the Evaporometer hardware - starting with the load cell. The EvapLite is an environmental sensing unit with only load cell measurements and an ambient temperature and humidity sensor. This version of Evaporometer was recently tested and deployed in the summer of 2019, and will be redesigned with hardware and software changes to provide more accurate load cell readings over time.

Show More

A New Approach

Results shown that each new Evaporometer suffers from the problem of unreliable of load cell data. From the perspective of previous lab members, the solution depended upon a new enclosure design - a way to isolate the strain gauge element from the drastic diurnal temperature changes. This resulted in significant design improvements in enclosure performance, production cost, and aesthetic appeal. However the underlying issue persists... meanwhile the hardware has seldom changed from the original design.

Description

This update aims to summarize possible solutions - not causes - related to inaccurate readings from sensors over long periods of time. This assumes that the sensing equipment are of reliable quality and added noise is to blame for unreliable readings at the level of precision required.

When additive noise to a signal causes the output to vary too much from the pure signal value these are two main approaches to the problem, from an electrical engineer's point of view:

• When sampling: Isolate the measurement system hardware from sources of noise. In our case this could be a reference voltage with substantial ripple, or a consequence of not properly grounding the INA125P amplifier.
• After data is received by the uC: Apply a Low Pass or other digital filter to give a higher weighted value to samples that statistically represent reasonable data. Using a simple exponential filter, for example, will give more or less weight to each new sample. This minimizes the impact of quickly changing sample values by design. This filter will also provide some lag before the input reaches the actual value. A future post will discuss how this filter's performance can be adjusted to suit our application.

With a known measurement value (i.e. a load cell with no additional weight applied) we can observe how much the input signal changes with respect to time, temperature change, and other possible sources of noise we are trying to identify. All of this information will be used to perform a series of tests on the redesigned EvapLite hardware and software.