Archived Forum

I think you have done nothing wrong.  What you are seeing is noise from the digital part of the processor getting in to the ADC input. Remember that your input is scaled to read 100 A - that is 23 kW, and you are reading less than 0.2% of that. There's a full explanation of how a small amount of noise is amplified by the conversion process here: http://openenergymonitor.org/emon/buildingblocks/measurement-implication...

Because you are not measuring the voltage, the sketch rectifies the noise and it registers as the power you are reading. You will have much more accurate values when you measure a real, larger current (and if you add a voltage monitor).

When starting myself using without ac voltage sensor it took me some time to realize that I had used ct.currentTX() to initialize instead of ct.current() otherwise i got very strange and inconsistent readings.

Completely another note is how much ADC range is avail for the recommended burden resistor which to my maths is not practical for household use (but my maths can be wrong too). My main fuses are 50A and practical loads per phase never exceeds 8kW.

On my Summary Page, there are some tools which can show how your input sensors are behaving. 

MinMaxAndRangeChecker displays the current and voltage data for each of the four analogue inputs on an emonTx.

RawSamplesTool_4ss_2 displays the current and voltage signals as a waveform.  This shows how much of the available input range is being used.  As posted, the input for the current sensor is set to Analog pin 1, which is CT 3 on an emonTx.

Min and max range checker works (after it stabilizes in around 10 loops) without AC voltage sensor voltage  input but how is it with RawSamplesTool without AC voltage sensor input ?

Below is a very minimalist script to test that your CT setup works.

/*
EmonTx CT only simple example
*/
#include "EmonLib.h"
EnergyMonitor ct; int power;

void setup()
{
  Serial.begin(9600);
  ct.current(3, 19.6); //YHDC TA12L-100 with 100 Ohm Burden
}

void loop()
{
    power = ct.calcIrms(1480) * 232;
    Serial.print("W="); Serial.println(power);
    delay(100);
}

ps. I take back with if(settled), just realized that had been using ct.currentTX for initializing - should have used ct.current which had caused me problems in the past in one of the test scripts. Have edited the previous post to avoid confusion.

Did some more testing with 40W load. After settling down:

1) normal SCT-013 with 18ohm burden and calibration value 111.1 - tolerance is within 10%
2) normal SCT-013 with 150ohm burden and calibration value 13.3 -  tolerance is 0% after settling in

According to my maths can run 150ohm burden up to 10kW but maybe will loose some accuracy then with higher loads ?

MinMaxAndRangeChecker simply displays the sample values that are being generated by the ADC.  If the input sensors take some time to settle, then that will be apparent in the displayed results.  The sketch itself does not need any time to settle and will work OK without a voltage sensor.

RawSamplesTool, and the improved _4ss_2 version of same, do need a few seconds for the system to settle.  This is because the logic needs to identify the start of each new mains cycle.   If there is no AC voltage signal present, then this tool won't be able to display either waveform because it will not know where each mains cycle starts and ends.