Calibration current sensor SCT013

I needed to calibrate a current sensor to measure AC current until 15A. This current sensor is the SCT013 from YHDC. Theoretically 1 V AC is equivalent to a current of 15 A.

This sensor, output an AC voltage proportional to the current been measured.

I used an old hair dryer with three different speed/heat and an Amp Clamp to get different samples of current,

I cut one wire of the dryer and added a wire extension to use the Amp Clamp to measure the current.

My hair dryer was able to generate three different currents 4 A, 1 A and 2 A (approximately). To extend those ranges I put several loops of the wire inside the current sensor.

This is an image that shows the arrangement.

For each position in the dryer, I increased the loops inside the sensor until reach four turn measuring in each turn the current in the wire with the Amp Clamp, the RMS voltage (Vrms) and the Peak-to-Peak voltage(Vpp) of the sensor with the Arduino.I got four points for each position in the dryer switch, so the total points in the calibration were 12.

I did two calibration one with the Vrms and the other with Vpp.

Calibration with Vrms

To measure the Vrms output by the sensor I used the arrangement and the software described in Measuring RMS and Peak-to-Peak AC Voltage with Arduino.

I did the calibration without amplification and with amplification.

When I started measuring without amplification, I noticed that my Arduino gave me a reading of 0.37V when the dryer was switched off. The same thing happened using amplification. First, I thought that there were problems in the connection or in the circuit I used.

But then analyzing the samples values I got from Arduino, I saw that with zero current running through the wires there was an oscillation of 1 unit in the sample took by Arduino, and this is quite normal. When I analized the readings I saw values like 517 and 516. When I calculated the Vrms and then converted the sample values to voltage using the following formula, what I got was this:

V = Sample / 1023 * 5.0 + 2.15

Then S1 = 516 convert to: V1 = 0.3719 and S2 = 517 to: V2 = 0.3769.

If I suppose that this oscillation is the same in the period of time measured, then:

Vrms = SQRT( SUM(n) (V1^2 + V2^2) / 2n) = SQRT (n * (V1^2 + V2^2) / 2n = SQRT( (V1^2 + V2^2) / 2) = 0.37 V

You are going to see below, that this voltage is produced by a current of around 3.0 A.

Measurements below Vrms of 0.40 (to be sure) are in the normal range of oscillation of the Arduino, for this reason I discarded it when I did the regression.

The following images have the calibration values and the plot with the regression line, for the set without amplification. I did everything in Excel including the regression.

And this one is for the set of data with amplification

You can see that I discarded the values of Vrms less than 0.40 V for the reason we discussed above. Values discarded ara highlited in yellow,

From the three positions of the dryer switch, the data collected from one position produced a trace in the oscilloscope (see below) that showed that it was non-sinusoidal. These points are marked with “nosin” in the first column of the table. The non-sinusoidal trace that is coming from dryer is probable caused by a triac the dryer used to decrease the blower speed. The rest of the points are coming from sinusoidal traces.

In the calibration without amplification, it looks that 1.0 V RMS equal to 14.0 A.

For the calibration with amplification 1.0 V RMS equal to 8.8 A.

Calibration with Peak-to-Peak voltage

Below are the plot for the calibration against Vpp without amp and with amp.

The major difference I saw it was that using Vpp there were not outliers neither the problems with zero current. Also it looks that in both plots (without Amp and with Amp) the fitting is better than Vrms, you can see visually and also the coefficient of correlation is slighty better. In this 1.0 V Vpp equal to 5.4 A in the no amp alibration and for the no amp calibration is 1.0 V equal 3.4 A.

According to my data it looks that is better to use Vpp instead Vrms. Formerly I though the way arround, mainly in non-sinusoidal waves, but result that Vpp is better to estimate the current passing trough the transducer.