ACS712 strange values when measuring 3D printer power consumption

[dbemowsk]

@Sushukka said in ACS712 strange values when measuring 3D printer power consumption:

I have only one line going through ACS712 and then its splitted to multiple outlets with extension cord.

So with this, are you saying that the multiple outlets with extension cords are all connected through one ACS712?

[Sushukka]

@dbemowsk Yes, there are about 8 devices behind this line. But they rarely use over 600W together and are actually splitted after UPS. However, I can use all other devices but 3D printers at the same time and having quite reliable numbers. As I mentioned before when I plug only Prusa MK2 and nothing else the values goes crazy. Latest testing results:

• Only Arduino + ACS712 and Prusa plugged in --> ACS712 power usage: 630W, reference: 196W.
• Replacing Prusa with a hair dryer to the same setup --> ACS712 power usage: 435/849/1430W, reference: 443/864/1443W.

What's wrong with my Prusa? :anguished:

[dbemowsk]

@Sushukka I would plug in the printer on it's own with nothing else running and see if it goes crazy. Then turn on other devices til it freaks out. Once it freaks out, measure the current with an amp meter at the ACS712. You will need to disconnect the ACS712 to do this. That way you can see how much draw there actually is that the ACS712 is seeing.

[Yveaux]

@Sushukka Read and understand this: https://learn.openenergymonitor.org/electricity-monitoring/ac-power-theory/introduction

Post your sketch please...

[Sushukka]

@Yveaux Thank you for the link. Very good document which went straight to my favorites. :)
Here is the algorithm of the sketch:

/*
Measuring AC Current Using ACS712
This ACS712 algorithm is more precise than ACS712_2 algorithm but because of long while loop
cannot be used with animated displays or similar non-interruptable functionality
*/
const int sensorIn = A0;
int mVperAmp = 100; // use 100 for 20A Module and 66 for 30A Module


double Voltage = 0;
double VRMS = 0;
double AmpsRMS = 0;

void setup(){ 
 Serial.begin(115200);
}

void loop() {
 Voltage = getVPP();
 VRMS = (Voltage/2.0) *0.707; 
 AmpsRMS = (VRMS * 1000)/mVperAmp;
 AmpsRMS -= 0.45;     // Empty load correction
 Serial.print(AmpsRMS);
 Serial.print("A ");
 Serial.print(AmpsRMS * 230);
 Serial.print("W\t");
 Serial.print(Voltage);
 Serial.println("V");

 delay(500);

}

float getVPP() {
  float result;
  int readValue;             //value read from the sensor
  int maxValue = 0;          // store max value here
  int minValue = 1024;          // store min value here
  
   uint32_t start_time = millis();
   while((millis()-start_time) < 1000) { //sample for 1 Sec
     readValue = analogRead(sensorIn);
     // see if you have a new maxValue
     if (readValue > maxValue) {
         /*record the maximum sensor value*/
         maxValue = readValue;
     }
     if (readValue < minValue) {
         /*record the maximum sensor value*/
         minValue = readValue;
     }
   }
   
   // Subtract min from max
   result = ((maxValue - minValue) * 5.0)/1024.0;
      
   return result;
 }

I have also used another algorithms, but this seems to give now good enough results. This algorithm is taken from: http://henrysbench.capnfatz.com/henrys-bench/arduino-current-measurements/acs712-arduino-ac-current-tutorial/

Edit: Switched using Emonlib and used:

emon1.current(0, 9.5);             // Current: input pin, calibration.

Now all other loads seems to show correctly but Prusa is still showing too much (200W vs 300W). So it seems to be that the Prusa's power supply is creating lots of reactive/non-linear load affecting badly to the power factor (right?). Probably this is just that much you can do with these cheap sensors. Maybe need to set Prusa out of this power meter or build own Prusa power sensor with different calibration/correction routines.

[Yveaux]

@Sushukka The getVPP function samples for 1 second and returns the maximum peak-peak voltage of these samples.
For purely resistive loads it is correct to multiply this peak-peak voltage by 0.707 to get the RMS value.

If your load looks like the blue graph (reactive load, taken from the link in my previous post) you have to integrate the surface area to get the true RMS value (that's why true RMS multimeters are in general more expensive and are advertised as such). I think the emon lib does exactly that, but I would have to check.

As a test you could run a sketch which samples the voltage from the ACS712 as fast as possible and outputs these values to the serial port. Recent Arduino versions contain a 'serial plotter' which then graphically show the power usage.
A test sketch could be as simple as:

void setup() { 
 Serial.begin(115200);
}

void loop() {
 Serial.println( analogRead(A0) );
}

Flash it, select serial plotter in the tools menu and watch the output.

[Sushukka]

@Yveaux said in ACS712 strange values when measuring 3D printer power consumption:

void setup() {
Serial.begin(115200);
}

void loop() {
Serial.println( analogRead(A0) );
}

Cool, haven't thought you could use plotter like this. Hair dryer gave pretty sin/cos like load, Kossel 3D printer with a normal computer supply is quite close but Prusa's load was pretty spiky. It looked like:

                /|
               / |
              |   \
      /------/    |
----/             \----

Which probably underlines pretty well your point. :) Question is that is there anyway to tackle this when using emonlib?

[Yveaux]

@Sushukka said in ACS712 strange values when measuring 3D printer power consumption:

                /|
               / |
              |   \
      /------/    |
----/             \----

Cool ASCII-art :+1: Next time you could just attach a screenshot :hatched_chick:

I had a quick look at emonlib and it appears to measure the true RMS value; it takes a number of samples and sums the squared values: https://github.com/openenergymonitor/EmonLib/blob/master/EmonLib.cpp#L187
It is designed for specific hardware and applies some correction that might not match your sensor.
On the other hand, if your other (resistive) loads match the reference power meter the algorithm seems to work ;-)

Did you consider that your power meter could be off for reactive loads? :confused:

[Sushukka]

@Yveaux ASCII art + SID music ftw. :) I played around with the sampling time (changed from 1480 to 3000/6000):

double Irms = emon1.calcIrms(6000)

Now the values are pretty close together. Still had to manually do the zero load subtraction (around 30W) but after that ACS712 shows surprisingly accurate values even for low 10/20W loads. Now it seems that when Prusa is heating up I get constantly around 100W higher values than I should. But when I turn the Kossel or hair dryer on = adding load to the same line, the Prusa crap seems to fitted in (or filled in) so that the total load seems to be accurate. My "reference" power meter could be off for reactive loads, but not this much and moreover Prusa's power supply is around 200W and I get 300W readings from ACS712.

Anyway, I think I'll take the "so be it" step now and just have to live with the fact that Prusa will skew my power readings a bit. Usually there will be some other load in the same line so it should be partly corrected by that. Your help @Yveaux was pretty useful and I thank you for that. :) Also I usually have trusted to some homemade power calculation algorithms because OpenEnergyMonitor is so heavily based on CT3 but as I now saw how much better accuracy its algorithms give I'm going to switch in to that in future projects.

[moreau1991]

@Sushukka thanks for sharing!