Two phase setup + negative current measurement

I’m on a grid with 230VAC @50Hz. The local grid has two phases (L1 and L2) at 230VAC referenced to ground. There is also a “neutral” wire which is 0V referenced to ground.
I have two meters. Meter A is connected to both phases and the neutral wire (2F+N). No fancy stuff for meter B, this one is connected to one phase and the neutral wire (1F+N). The schematic shows the phases (black/brown) and the neutral (blue) wiring.

With a volt meter, I do not measure 400VAC between L1 and L2 (as would be the case for 3F+N systems). The volt meter measures 0V difference between the two phases. So quite surely, they are not 120 degrees apart. So, I was inclined to think these two phases are actually “in phase”.
After measuring these phases with an oscilloscope, I saw they are not in phase, and neither are they 120 degrees (6,6ms) apart. They are about (3,3ms) apart, i.e. 60 degrees.

I have an EmonPi and EmonTx with 9VAC reference on L2. This happens to be the phase the wall socket is on. On L2, I normally get a positive consumption, unless I activate an induction cooking plate (connected to both L1 and L2). If I switch on the induction plate, the measured power on emonpi current transformer 2 goes negative.

Anyone has an idea why this faulty measurement occurs? Something wrong in the way I’ve connected CT2 of the EmonPI? Or anything else in the setup?
I’m quite sure the CT’s are connected on the phase wire, not on the neutral. Perhaps real vs. apparent power has something to do with the (inductive) load?

Something there is not correct, because you cannot have two waveforms 60° apart, yet measure 0 V between them. Your oscilloscope picture does show a voltage difference - it is the vertical gap between the two traces. Are you absolutely certain that 0 V was the true voltage, and not just a bad connection with your meter?

I can understand the 60° part. That will come about if you have a 3-phase, 3 wire system. Then, treating one phase as the reference, the other two will indeed show a difference of 60° between them. But I’m not aware that Belgium uses that system. [Edit - Further digging reveals that this configuration was used in Belgium, and might still be present in very old systems. Is that the case for you?] As far as I know, only Norway uses a 230 V 3-wire system, but there, each line is 132 V to earth.

The 60° also does not make sense for your induction cooker - unless there is also a neutral connection and the two phases are treated as two separate feeds.

I shall believe for now that you do have what looks like a 3-phase, 3-wire system, but with one ‘phase’ being the neutral. (I cannot call it a 2-phase system, because that would imply 90° between phases.)

Your error is using the voltage of L2 for the reference for L1 current. That will never give the true real power reading because of that 60° difference between L2 and L1.

What is missing is your reason for connecting the emonTx and the emonPi the way you have. I would suggest you use the emonPi exclusively for the black (your L1 - conventionally L1 is brown, L2 is black, L3 is grey) phase, so PV2 & Distrib B, and move the voltage reference onto the black phase also; and use the emonTx exclusively for the brown phase (Brown Use & PV1)

You can put the c.t on the neutral - it must simply face in the opposite direction to make it correct.

That is true, but if your induction cooker is properly designed, it will not be very reactive, but power factor less than 1 is almost inevitable.

[Just for information - the ‘discrete sample’ 3-phase sketch time-shifts the voltage to provide a reference for the second and third phases. The way the time shift is calibrated is to put all 3 c.t’s on the same wire, and adjust the time/phase calibration so that the powers in L2 & L3 are exactly -0.5 × the power in L1. I think you might be seeing much the same thing.]

If you have a 3-phase transformer with a star-connected secondary winding supplying you, and the star point is not connected but the grey phase winding is connected to neutral and earth, then this is what your phasor diagram will look like:


The centre of rotation is the tip of the grey phasor.


I repeated my measurements between both waveforms: I now measure 230V AC between both phases. It was indeed a bad connection with the probes of the meter. I could also measure 11V and 25V depending where I touched the contact point with the probes. This confirms your analysis: I have a 3-phase, 3-wire system with one phase being the neutral.

I also verified earlier measurements with reference to ground. L1 and L2 are obviously at 230V. The “neutral phase” (sounds really contradictory to me), is not at exactly 0V as I would expect it to be. I measured 3.7V AC. I wonder if that 3.7V AC is also 60° out of phase with the “real” phases. But I digress.

The induction cooker came with instructions on how to connect to 3F+N (400V) and 1F+N (230V) system. Since it has a total combined power of 11kW (48A at 230V), it’s too much to connect to just one phase. My main circuit breaker (40A) would blow at maximum power draw.

Lucky for me, the cooker is split into 3 zones. You can see at as three separate cookers. Each zone can draw a maximum of 16A @230V. If I recall correctly, I connected 2 zones to one phase (and neutral) and 1 zone connected to another phase (and neutral). So I treated the two phases as separate feeds.

Pure and utter laziness :slight_smile: Or just eagerness to connect things and start measuring. The wall sockets nearby all happen to be on L2. I would have to install an additional wall socket connected to L1.

For your information, in my initial setup, I had all CT’s around the same stretch of wire (L2). Just to verify the whole system was measuring and transmitting these measurements. Good thing I did, 'cause I learned that for the EmonTx, you have to flip the CT’s in the “wrong direction”. For the EmonPi, the arrows on the CT’s point in the same direction as the current flow.

Shuffling around the CT’s a bit would result in this design (yet to be implemented). EmonPi is referenced to L1, EmonTx is reference to L2.

This illustrates a very important point. NEVER believe a zero voltage reading when your safety depends on it.
You can use it to confirm that you have isolated the correct circuit, but I only believe zero volts when I see the voltage fall as I open the circuit breaker.

That is the voltage dropped by the neutral current flowing in the impedance (largely resistance) of the neutral conductor between your house and the point where the transformer is earthed. Its phase will depend on the proportion of the contribution to that current from each phase. It is quite normal. There’s some more explanation in ‘Learn’.

OK, that clears up another potential problem. If your cooker had a zone connected L1-L2 (which would be perfectly OK for the cooker), you could not have a meaningful measurement of the powers in L1 & L2 separately, because some of the current, hence power, in L1 would appear as negative current & power in L2. But the total would have been correct.

I think that will give you the correct answers. But why two c.t’s on the same cable?

Absolutely agree! That’s exactly how I go around when working on electricity. Even if there appears to be no voltage on the wires, I will even switch on the breaker first to see the voltage drop when I switch it back off.

For historic reasons :slight_smile: I initially connected a bunch of CT’s to the same wire. Just to validate measurements and a basic accuracy test. I have already moved a CT onto another wire and i will move another as soon as I see fit.

Even after yesterday’s change, I noticed negative current flow again on EmonTx CT1 (and CT2). There is no energy generating source on that wire. And even if the solar panels were on that wire, it’s a cloudy winter day and solar panels only output 20W each at the moment.

The induction cooker was off, the washing machine is the only suspect. It’s been on since around 10:25, shortly after, the negative spikes occurred. The program takes 2 hours to complete. I will check the wall socket of the washing machine once the program is finished.
Maybe the socket is connected to L1 and L2. Which, I guess, would be an explanation. Otherwise, I don’t have a clue what may cause these negative readings.
This is the situation since yesterday around noon:

As you can see, I disconnected the voltage reference for the EmonPi. Simply because I don’t have a wall socket with L1+N near the EmonPi yet.
After disconnecting 9VAC reference, EmonPi is giving me 0W reading on both CTs. I already rebooted the EmonPi to no avail. But that’s another issue.

Rebooted - as in rebooting the Pi, or powering down and powering up? The latter restarts the Atmel 328, which will then recognise no a.c. input and switch to reporting apparent power. The former won’t restart the emon part, which will continue to read zero power because of zero volts.

That could well be the case. I understand that the normal arrangement with the system you appear to have is one house gets “L1”, (230 V to earth), the second house gets “L2” (also 230 V to earth), and the third house gets “L1 - L2” (both lines 230 V to earth) and because of that, all houses have double pole circuit breakers. And the transformer sees a roughly balanced load across the three secondary windings.

You can expect to see a small “keep alive” power into your inverters overnight - usually in the low tens of watts.

Reboot as in
pi@emonpi(ro):~$ sudo reboot
What I did now:
pi@emonpi(ro):~$ sudo poweroff
Shortly unplugging the 5V power from the EmonPi (with the 9V AC adapter already unplugged) fixed it.

Correctly summarized. Everything should be connected to N and another phase (L1 or L2).

I just tested the socket of the washing machine by measuring between earth and both poles. You guessed it already: I get 230V in both cases, so this socket is not connected to N and another phase, but to phases L1 and L2. Works electrically fine off course, but causes negative readings. So next up is to fix this in the distribution box. I’ll do that when the sun is out again. Makes life a bit easier :slight_smile:

I’ll probably have to review/measure the induction cooker as well…

We’re getting there…

The only ‘problem’ with the washing machine wired as it is, is you cannot, unless you put a separate monitor on it taking a voltage reference L1 - L2, or generate a delayed voltage reference inside the emonTx, read it’s real power directly. But I repeat, adding together the powers from all 5 c.t’s will give you the correct total power.


Thank you for all your support this year. :slight_smile: I wish you a happy new year!

I hope the information thread benefits other people as well. With that in mind, I simplified the drawing to make it more generic.

1. Simple setup
Suppose there is only loads A and B, i.e. loads connected to either L1-N or L2-N. Just pretend load C does not exist. The easiest method to measure total energy consumption is to connect one CT of the EmonPi to each phase. CT1 one goes around L1 and CT2 goes around L2.
For simplicity’s sake, do not connect the 9V reference. The power measurements will be less accurate though, but it avoids some complexity.

In practice, we can’t connect just one CT to the neutral line, as both phases pass through the neutral line. In theory, I believe it could work with just one CT on the neutral line and Fourier transform the measured values in software. But I digress.

2. Intermediate setup
Still pretending load C between L1-L2 does not exist, we connect the 9V reference to L1-N. The CT sensors remains connected in the same way as before. Since the 9V reference is correct (in phase) with CT1, this will measurement will be correct, taking power factor into account. The measurement of L2-N done by CT2 will be incorrect since it will multiple the current through L2 with the voltage reference of L1.

The solution here is to generate a voltage reference for L2 in software. Although the amplitude of the “fake” L2 voltage reference may be off, the phase should be correct. Let’s say L2 lags L1 by 60 degrees, how can I create this L2 reference on the EmonPi? Do you have some pointers?

3. Advanced setup
If you don’t have loads between L1 and L2 (load C), setup 1 or 2 will do. Since I also have loads between L1 and L2, I guess setup even setup 2 won’t do?
In practice, do you need three CT’s to measure the total power of A+B+C?
Or in other words, is it even possible to do with just the EmonPi?

1. Simple setup
If you do not use the voltage reference (and hence only have an estimate of apparent power) then the ‘standard’ single phase sketch is all you need. If you have an emonTx, you must not use the three-phase sketch.
[That’s for anyone else reading this later. I know it doesn’t apply here.]

No, that wouldn’t work. The basic rule is: you need one less wattmeter than the number of wires. By ‘wattmeter’, I mean anything capable of measuring power. So you can have c.t’s on L1 & L2 and treat the N as common as you are doing, or L1 & N (treating L2 as the common wire), or L2 & N (treating L1 as the common wire). The third wire - whichever it is - carries the vector sum of the currents in the other two.

And that answers this question too:

No you don’t need 3 c.t’s, and yes you can do it with an emonPi, but not with the standard sketch.

It’s been done, and published - somewhere on the old forum. But you’ll have difficulties. It was written for an emonTx, and while that sketch will work, it needs much modification to include all the other things that the sketch in the emonPi does (like drive the display, receive the other nodes by radio, etc.). Then you must remember to use the “wrong” update method when you update your emonPi, else the sketch will be overwritten by the default one. (6.9 KB)