A pressure sensor datasheet gives its resistance as 5 kΩ +/- 1 kΩ or 3.5 kΩ +/-0.5 kΩ (optional version). The full-scale output is 150 mV +/- 30 mV at 100 psi, 5 V bridge voltage Vb.
The way I imagine the pressure sensor works is that when pressure changes, two resistors contract and two expand in length by a certain percentage, changing their resistances by the same percentage.
So if the initial resistance at zero pressure is 5 kΩ, the output voltage sensitivity to the excitation voltage Vb and pressure is the typical 150 mV/5 V/100 psi, then if 100 psi and Vb = 5 V are applied, this can be modelled by adding/subtracting 150 Ω to/from the 5 kΩ resistors. The 5 kΩ resistors change by 3% and the output voltage Vo is 150 mV.
However, if the initial resistance at zero pressure is now 4 kΩ and the same pressure and Vb are applied, the 4 kΩ resistors should change again by 3%. This can be modelled by adding/subtracting 120 Ω to/from the 4 kΩ resistors. The 4 kΩ resistors change by 3% and the output voltage Vo is 120 mV.
Is the initial resistance tolerance of 5 kΩ +/-20% the cause of the full-scale output variation of 150 mV +/-20%? This idea seems to work well for 5 kΩ resistors, but not for the 3.5 kΩ +/-0.5 kΩ (i.e. +/-14%) option.
Or is the full-scale output variation another independent inaccuracy caused by different production resistor batches contracting/expanding differently at the same pressure? In other words, should the datasheet full-scale output variation be modelled on top of the initial resistance tolerance?