New Open Source Split Phase Energy Meter that uses EmonCMS

Tags: #<Tag:0x00007f880656a820>
(Robert Wall) #33

This is a property of all transformers, so I’m afraid you’re likely to be stuck with it. The magnitude of the impedance might be somewhat less with a higher-rated transformer, but other factors, like phase shift, could easily spoil everything.

Not really. If you do filter (excessively), you’ll end up not measuring the power in the harmonics, which will introduce a deliberate error. It ought not to be a large error, because the power in the harmonics should, if your loads obey the rules, be small. But you don’t know that they do and whether it is or it isn’t negligible.

(Robert Wall) #34

In part, also by every other Tom, Dick & Harry who has exactly the same sort of power supply (except probably using a full-wave rectifier) hung on the system and all pulling current just at the peaks.

If I look at my UK mains waveform directly (×10 'scope probe), then I see pretty much the same degree of flattening as I do from the TDK a.c. adapter.

[Picture is below.]

(Bill Thomson) #35

Here’s a scope pic of my mains. (10x probe) How does it compare with your mains waveform?

(John) #36

Thanks Bill. You’re right, I read that incorrectly, thanks for explaining. That makes sense, but I probably cant go with that low of a cap without causing other problems with the SMPS.

(Robert Wall) #37

Here you are:

The yellow trace is the raw mains, the blue is a standard TDC UK a.c. adapter, unloaded.

It is very clearly “flat-topping” on the mains supply itself.

(Bill Thomson) #38

Given we typically have 1 to 6 houses on a single transformer vice the 100 or more you have in the UK, would you say that’s why my waveform isn’t flat-topped? (although it clearly is distorted)

(Bill Thomson) #39

Application note Basic Calculation of a Buck Converter’s Power Stage (Rev. B) says:

6 Input Capacitor Selection

The minimum value for the input capacitor is normally given in the data sheet. This minimum value is
necessary to stabilize the input voltage due to the peak current requirement of a switching power supply. The best practice is to use low-equivalent series resistance (ESR) ceramic capacitors. The dielectric material must be X5R or better. Otherwise, the capacitor loses much of its capacitance due to dc bias or temperature. The value can be increased if the input voltage is noisy.

The data sheet says: Input Capacitor

For most applications, 10 μF is sufficient and is recommended, though a larger value reduces input current ripple further. The input capacitor buffers the input voltage for transient events and also decouples the converter from the supply. A low ESR multilayer ceramic capacitor is recommended for best filtering and should be placed between VIN and PGND as close as possible to those pins.

(Bill Thomson) #40

Why not use the third channel for PV monitoring?

(John) #41

I know, but I have a half rectifier before the SMPS input, so anything less than 100uf and theres not enough current to power everything.

I actually designed a 3 channel that can fit an esp32 on top, and a 6 channel:

That can have add-on boards stacked

(Kobuki) #42

This looks nice. A few questions:

  • What is the expected accuracy of the device? Will it support calibrated CTs?
  • Is it applicable to a single phase + inverter system on 230V?
  • Will there be an option to add a separate supply for powering electronics (so the transformer is only for waveform capture)?
(Robert Wall) #43

That is likely to be a significant factor. The fewer disturbing influences where the fault level is the lowest, the better (from the point of interference and waveform purity, etc.).

(And I think it could be hundreds in some cases.)

(Bill Thomson) #44


Oh, BTW, that’s half wave rectifier, vice half rectifier. :wink:

(John) #45

This looks nice. A few questions:

  • What is the expected accuracy of the device? Will it support calibrated CTs?
  • Is it applicable to a single phase + inverter system on 230V?
  • Will there be an option to add a separate supply for powering electronics (so the transformer is only for waveform capture)?
  1. Assuming it’s calibrated to the CTs that are being used, it is 99% accurate.
  2. For 230V you’ll need to use an AC transformer that can bring down the voltage at least 12v
  3. I don’t currently have a plan to add an option to power electronics only. However, if you really want to you could just take off the diode right next to the power jack, and hook up the 3V3 pin to power.
(Kobuki) #46


  1. I think you meant 1% accuracy :slight_smile:
  2. The transformer is not an issue, the question is, will it support a 230V system? That would need other calibration values, etc.
(John) #47

Yes, it’ll support a 230V system. It would need to be calibrated for the transformer, correct. That procedure is outlined here

There is a parameter in the library for 50hz power as well.

(dBC) #48

You might struggle to get 1% accuracy without including some phase angle calibration. Those VTs and CTs each introduce phase shifts in the signals you’re measuring. The IC supports being able to add delays in the V or I down to a resolution of ~488nsecs to compensate for that, but the library you’re using appears to just hard code all those registers to 0 (unless I’m missing something that happens elsewhere?).

  CommEnergyIC(WRITE, PQGainA, 0x0000);     // Line calibration gain
  CommEnergyIC(WRITE, PhiA, 0x0000);        // Line calibration angle
(Robert Wall) #49

And those phase shifts may well vary with the amplitude of the quantity being measured. Though this is very much a second-order effect, it could be important with small, low power factor loads in particular.

(dBC) #50

I’ve not drilled down into the details, but one of the headline features in the datasheet suggest they’ve even considered that:

.Flexible piece-wise non-linearity compensation: three current (RMS value)-based
segments with two programmable thresholds for each phase. Independent gain and
phase angle compensation for each segment.

(Robert Wall) #51

Now that’s interesting.

Considering only our “shop” c.t. and v.t., the c.t. would need to use all 3 segments (in a ‘bathtub’ shape) while the v.t. is close to a straight line over the operating range.

(dBC) #52

That might also be helped by the lower voltages. Looks like full deflection on the inputs is at 720mV (RMS).