Thank you very much Jim. Unable to conclude anything. I will repeat my experiments once my set up is ready. Will try to conclude whether this sensor can be used or I have to go for something better.
I would conclude that there is no definite correlation between the three series of data, within the limits of measurement accuracy.
I speculate that the apparently consistent drop in the water pressure sensor readings might be due to some gradual 'settling-in' process, e.g. a permanent slight deformation of some part of the sensor. Repeating the measurements, over a longer time period, might help to prove/disprove that theory.
Possibly because someone in an office far far away just picked some numbers out of thin air, without thinking of the practicalities of such a measurement.
While I am wearing my cynical hat, he also probably wanted the data logged at 1 second intervals.
Even Barnes Wallace did not need it that accurate! (The devil made me say that).
This minimum to maximum variation will be over a period, say a year. In a particular day change may be less than even 5 cm. 5 cm level change means a big change in the volume of water.
Note that the inflection points all generally-coincide. Temperature or turn-on/turn-off spike co-incident with the air conditioning turning on and off?
The % difference is well within the range of what might be considered barometric pressure.
So,in the absence of anything else, I'd say poor power supply regulation and that it reads absolute rather than gauge pressure. The shape also suggests barometric pressures.
The water and air sensors "track" somewhat shape wise suggest barometric pressure.
The meter suggests that you don't have enough accuracy to resolve the height resolution required. I would "think" that a 4-20 mA sensor that reads in gauge pressure and has a port to the air atmosphere is required.
4-20 mA allows the sensor to be powered by the same lines used for readout and the lines can be arbitrarily long.
Powering a sensor with a constant voltage supply over long distances, there is a voltage drop across the leads and the measurement usually has to be differential.
Click on Informative (i) icon lower right.. might be indexed somehow
1) you will always need a more stable reference Voltage 100 ppm, or an ADC with 14 bit accuracy
2) You must make a list of ALL possible error sources , simulate them or observe them and find the contribution to output result with at least 12 bit resolution preferably 14 bit.
e.g. Voltage, RF, air Pressure, Temperature, ripple, AC quality ( transients, brown out, surges , noise)
3) Learn Sensitivity Analysis Δx/Δy and Correlation techniques or curve fitting
Excel does 6 kinds of curve fitting easily, and you can hide data that ought to be excluded .
- plot xy data format then add curve fit and display formula)
- then use the formula in code for correction and calibration to compensate for any known sensitivity as long as it is stable. or make it stable.
4) thus your reference voltage or current source for the sensor must be equal in stability to your ADC or DMM.
5) write a design spec so you can test against your requirements. in future, do this 1st !!
6) Write a Design Verification Test (DVT) Plan with 1 page per test based on above specs. with rejection criteria. and setup.. in top and fill in results in bottom. with left side using photos or diagram.
7) After you follow this design methodology it is harder to fail. ( or as I was taught.. get it Right the First Time )