# Calibration curve of an instrument

In the calibration procedure of an instrument (e.g. a pressure transducer), when we want to determine the calibration curve, is it a good choice to force the passage of the calibration curve through the origin?

• What do you mean by forcing the curve through the origin? Aug 7, 2017 at 13:41
• I mean imposing that for an input equal to zero the output of the instrument is set arbitrarily to zero. Aug 7, 2017 at 16:30
• I could see this being done if the variable you want to work with is gauge pressure. You would want the signal at atmospheric pressure to correlate to 0 psig, as an example. You may want to gather the calibration data and see if forcing the curve to go through the origin appreciably affects the accuracy at the higher end of range. Aug 7, 2017 at 17:44

This depends on the sensor. In some sensors, an offset it to be expected, and should be included in a calibration curve. In others, there is a good physics reason that the actual output should be zero when the thing being measured is zero, and you should force it through the origin.

Lets look at two examples:

1. A half-bridge strain sensor with two active elements. This type of sensor consists of two strain gauges stuck on a metal beam such that when the beam bends one is compressed and one is tensioned. The two are connected in series, between the supply rail and ground (lets say 5V and ground for arguments sake). The zero position will be when neither is stressed, and the output will be close to 2.5V, but if one strain gauge is slightly different from the other due to manufacturing variations, then it will be slightly off. In this case, you should not force the fit through the origin.

2. A current-sensing shunt resistor. These are low value resistors designed to allow you to measure the current through them without dissipating too much energy. The voltage measured is directly proportional to the current thanks to ohms law. If there is no current flowing through, then there cannot possibly be a voltage generated, and any voltage you do measure is probably a fault in the measurement kit not the shunt. So in this case, you might well choose to force the calibration curve through the origin.

In general, you would need to look at the sensor's datasheet (if you bought it) or think carefully though how it works (if you made it) to decide whether you expect an offset. If you tell us what type of pressure transducer it is and how it works, then we might be able to tell you which category it falls into. Mechanical pressure sensors based on a movable membrane will usually (but not always) fall into the first category, where ionisation gauges would usually fall into the second.

Either way, after doing the fit, you should check whether the observed offset is what you expect, and that the fit to the data is good.

• In my case, due to the calibration system used I'm not able to have a value of pressure lower than 9 bar. So the feasible options are to either force the curve through the origin or allow for an offset. The sensor are the "GE Sensing PTX 500" Aug 8, 2017 at 6:45
• With that sensor you should not force the fit through the origin. You should do a fit then check the offset - if it is more than 0.05mA then you should check that the sensor is working properly or consider adjusting it. Aug 8, 2017 at 13:04

If you're developing the calibration curve for a sensor (a device like a pressure transducer), since your measurement system will account for the conversion of the signal from an electrical signal in V or mA to Pressure using the sensor's calibration curve, then there's no reason to do so unless the sensor is really giving a strong bias at zero input. If it is, then I would suspect a defective sensor.

Let's say I have a load cell and I want to develop the calibration curve for it. I will set up my calibrator which applies a precise load in N and reads the electrical output of the transducer in V. I'll do this at 0, the max load I expect and a few points in between. If my sensor has a linear response, I should have a nice straight line graph with an equation like this:

$$F_{Newtons}= m*Volts + b$$

Where $m$ has units of $\frac{Newtons}{Volt}$ and $b$ has units of $Newtons$

If $b$ is large, relative to the $m*Volts$ term, I would question the integrity of the load cell.

• Yes, I totally agree. In my case, due to the calibration system used I'm not able to have a value of pressure lower than 9 bar. So the feasible options are to either force the curve through the origin or allow for an offset. Aug 8, 2017 at 6:45