3
\$\begingroup\$

Introduction

I have a project about measuring the voltage, current and power factor of an electrical system. I had designed a system that allows the microcontroller to measure the peak of an AC with DC offset signal. The whole system relies on making an ADC conversion right on the middle of the half positive cycle of the signal. I have two AC with DC offset signals, one that comes from a current transformer and other from a voltage transformer. When I finally got both ADC measurements I can calculate how much is the actual voltage, current flowing through the system, and the power factor between signals.

enter image description here

This is my circuit schematic. It’s a bit large so you might want to expand it: enter image description here Explanation of the circuit: enter image description here It has a sort of “Load selector” because since the current transformer is a current supply I can change the voltage range by changing the load resistor. (I’m using a 120 ohm for measuring >15 A and 1 kohm for measuring <1 A)

I’ve simulated the whole circuit on VSM Proteus and the results were flawless. I can correctly measure voltage, current and power factor with my microcontroller.

enter image description here

The problem

The problem comes with the real circuit. I had designed and built the PCB for the shown circuit. The resulting output signals were expected. The zero-crossing signal from the current and voltage transformers had a 10 ms HIGH and 10 ms LOW logic levels, this means a 20 ms period signal, thus 50 Hz signal; seems PERFECT! But, when I tried to measure the middle of the positive half cycle the results were not as expected, on the other hand the results on the simulated circuit were ideal, steady output numbers. On the real circuit the ADC measurements were “oscillating”, this might imply that the middle of the zero-crossing signal doesn’t point to the peak of the original signal.

Simulated results:

enter image description here 2.869-1.696-2.874-2.874-2.874-2.874-2.874-2.874-2.874-2.874-1.701-2.879-2.879-2.879-2.879-2.879-2.874-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-1.701-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-1.706-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879-2.879

Real results:

enter image description here 2.801-1.740-2.400-3.182-2.278-1.686-2.776-3.089-1.940-2.053-3.079-2.581-1.730- 2.527-3.196-2.136-1.706-2.933-2.898-2.019-2.248-3.177-2.351-1.691-2.674-3.167-2.297-1.887- 3.065-2.708-1.716-2.444-3.148-2.239-1.696-2.835-3.040-1.926-2.414-3.099-2.522-1.691-2.615- 3.187-2.107-1.730-2.957-2.889-1.799-2.283-3.148-2.341-1.691-2.698-3.143-2.058-1.916-3.030- 2.796-1.711-2.468-3.177-2.239-1.716-2.845-3.025-1.916-2.107-3.104-2.498-1.691-2.576-3.172

Thanks for reading, any hint on this would help me! If you need more information just ask. P.S: I'm not a native speaker so feel free to correct any grammar errors.

\$\endgroup\$
4
  • 1
    \$\begingroup\$ Why is there such a large time constant (10ms) at the output of the comparator? Another thing that's not visible: if you use a transformer before the ZCD, if it's a small one it will not consume much but it will risk distorting the waveform due to magnetization current, might not; better check with oscilloscope against the primary (careful with chassis, not many are aware). \$\endgroup\$ Commented Oct 28, 2018 at 17:14
  • \$\begingroup\$ @aconcernedcitizen Shouldn't the half cycle of a 50Hz signal be 10mS length? I'm not aware of the distortion of the waveform due to magnetization current, it can be radical? I've measured it a week ago and it seemed clear. \$\endgroup\$ Commented Oct 28, 2018 at 18:18
  • \$\begingroup\$ See this, for example. While the current is the one that suffers most, you should make sure with an oscilloscope. What you say about the period, and what I was talking about, the RC time constant, are two unrelated things here. Your comparator should give you pulses at the output, as clean as possible so that their edges can be used as proper triggers further on. You are using an RC load, which, due to its time constant, rather obliterates the purpose. Eliminate R4 and C11 or, if the comp needs open colector (don't remember now), only C11. \$\endgroup\$ Commented Oct 28, 2018 at 18:39
  • \$\begingroup\$ @aconcernedcitizen It's not an exact ZCD, it just sends a logic 1 all along a positive cycle and a logic 0 all along a negative cycle. Well, if I eliminate both capacitors from the comparators I get a bit noisy signal, BUT via programming, I can discriminate that little noise and get the exact edges of the signal, thus calculate the resulting peak. Did that and the result were clean, despite a bit of deforming on current transformer. Thanks a lot, if you want to post it as an answer I will accept it. \$\endgroup\$ Commented Oct 28, 2018 at 20:45

1 Answer 1

Reset to default
3
\$\begingroup\$

The comparator from the zero-crossing detection has an RC time constant at the output that defeats the purpose of having a clean pulsed output. If there is noise and needs to be filtered, you can still keep the RC but with a much lower value for the capacitor. As it is, the time constant is 10ms, which is too high, since it is comparable to the input's period. This is what happens to the signal:

tau

The green trace is with \$\tau\$=10ms. red with 1ms, blue with 0.1ms, and black with 10\$\mu\$s. Depending on the present noise, you can reduce the time constant to a minimum such that it leaves the pulses as clean as possible, to be able to provide the necessary edge triggers for the next stage, while still functioning as it was intended: a zero-crossing detector.

One minor thing to remember is that the LM311, as open collector as it is, has a source/sink current limit for its output stage, so care should be taken that, when going low, the capacitor doesn't discharge more current that it can handle. The datasheet says 50mA, so a quick solution is this (consider the 2N3904 the open collector):

bla

Note that I didn't change the resistor/capacitor values so the time constant shown is double. Change according to your requirements.

\$\endgroup\$

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.