How accurate is the Peacefair PZEM-004T AC Comms Module? 

Sorry, I did not keep the data.

I am not sure I understand the question. If you are concerned if the module can measure power draw from noisy devices, like SMPS, the answer is yes. I have never seen any issues

No, unfortunately not. It does not care about which way you orientate the current transformer. You would need something that is made for solar installations, I suspect the majority of power/energy meters do not support direction.

The same situation applies to me. Have you found a method to measure energy to prevent energy going back to the grid?

@@irfansensoy5901 shelly 3EM

Its a 100W mono audio amp board using a TPA3116D2 chip from Ebay. It is actually surprisingly good for the price. It needs a fairly beefy power supply if you draw a lot of current on the output or it will start distorting the wave form. Otherwise it is very efficient (class D). It is linear from 60Hz to 2kHz, The 3db points are at

@@TheHWcave thx for the reply. I have a JOY-iT JDS6600 60Mhz signal generator, and that one has also a very low current output, but now i have a idea to amplify it. A have one more tip, you mentioned that it can be dangerous to switch between x10 and x1 , for a workarround, you can use the multiple turns as fixed primary, and switch the burden resistor between x10 and x1. For the x1 position, you can use a fixed burden resistor value, and for the x10, you can use the switch to parallel a lower resistor value. That way you only switch the secundairy side of the current transformer, that carry a low current, and also the x10 and x1 use the same amount of primary windings and the accuracy will become better for the x1 position. Grtz

@@BjornV78 A good idea and and I actually did something like that before ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Lf5nhBLmJ-c.html . but in this case there are some problems. Firstly the x10 resistor needs to be 10x higher not lower. Since a current transformer must never be operated without load, you need to wire the x10 mode resistor hard in (0.5 Ohm in this case) and use the switch to add the lower resistor (a 0.05555 ohm resistor in this case) in parallel. Tricky to find / make such a value (e.g. from resistance wire) and then the unpredictable contact resistance of the switch is already in the same region. A better way is to use a make-before-break switch with a 0.5 Ohm (x10) and the stock 0.05 Ohm (x1) resistor. The switch contact resistance may still spoil it, but this may work. Unfortunately the chip is calibrated for the 0.05 ohm resistor and the filter network, so you can't just use higher resistor values. Without these constraints, your idea is definitely valid.

From the datasheet the V9881D seems to collect and process data at speeds of several tens of kilohertz to do all kinds of signal processing and zero-crossing detection and parameters such as volts, current and power are internally available at least every 10ms or even more often. However, the firmware doing the serial connection over USB and sending the data using the MODBUS protocol with checksum etc is much slower and sends updates only about every 1 to 1.5 seconds. That means a fast change may be registered but you may not learn about it until up to 1.5s later. Worse, if a spike comes and goes within the 1.5s window, you may not get any information about that event.

You are correct in suspecting that the PF shown by the PZEM-004T is not very accurate for very small PF values. The reason is hard to explain in a RU-vid comment, but I try. Please check the V9881D datasheet. In chapter 10 it explains its inner workings very detailed. The signal processing part of the chip measures active (P) and reactive (Q) power but it does not measure or calculate PF. This is done in the firmware. PF can be calculated using pf = cos(arctan(Q/P)). So, how does it measure P and Q? There is a lot going on in the signal processing part of the chip but in essence the trick is that for active power P, it multiplies volts and currents while for reactive power Q, it first phase-shifts the current by 90 degrees before multiplying. If you consider the 2 extreme cases: first current and volts are in phase. The multiplication produces a (power) wave form that is always positive and the RMS average of that is active power. At the same time the Q calculation takes the same current and voltage and does the same multiplication but first phase shifts the current by 90 degrees. The result is a power wave that is always symmetrical around 0 (same amounts are positive and negative) so the average Q is simply zero. Plugged into the formula you get a PF = 1. If the current is phase shifted by nearly 90 degrees already, the then active power would be mostly symmetrical around zero with a tiny non-zero rest, and Q will now be a large negative average. This gives a large negative value of P/Q and an arctan of -89.9999 deg and cos(89.9999) is a PF of nearly zero. Obviously this method has increasing accuracy problems (because of arctan) the lower the PF gets but the P and Q values themselves are still very accurate. The problem in calculating very small PFs might well explain the discrepancies you see because both meters indicate show very small PFs.

@@TheHWcave Infinite thanks for taking your time to give me this splendid summary explanation. I had been checking the datasheet of this v9881d and also of the ADE7953 (www.analog.com/media/en/technical-documentation/data-sheets/ADE7953.pdf, they seem very similar. they apply an HPF before the multiplication (this is where I understand according to your explanation that it is where the 90º phase shift occurs due to the high pass filter?) and then an LPF to eliminate harmonics. What I don't understand is why it does for PF=cos(arctan(Q/P)) instead of something simpler like PF=P/S. If S = V x I and the P extracts it with the displacement trick. Also at the beginning of your explanation you say that P=V x I, wouldn't this be S? Excuse me if I'm talking nonsense, I'm starting to understand this fabulous world of AC! Thanks again for your effort to make me understand.

@@isabido Yes AC is complicated especially if it is no longer sinusoidal.. Anyway, no the 90 deg. shift is done only for the current channel and after the HPF and LPF which are used for both. Also there is a phase compensation (delay) circuit but that gets really into details. These chips were made to meet the standards for electricity (smart) meters so they are doing a lot of wonderful signal processing to get the best accuracy for P and Q. Note that apparent power S is not V x I. S is a vector (as are P and Q) and the magnitude of S is the vector sum of P and Q, in other words S = SQRT(P^2 + Q^2) (Pythagoras)

@@TheHWcave I knew of the formula S = SQRT(P^2 + Q^2) , but this would also be valid, right? S= Vrms x Irms. Thanks!

​@@isabido Yes that is correct

Not sure why it isn't as accurate as expected at

How strange. Mine works like clockwork. I had another viewer asking me about the update rate and I ran a test polling the module as fast as possible using only the AC_COMBOX.py command line interface with --time 0 parameter. This managed about 10 polls per second but we found that the firmware only sends new data about every 1.2 to 1.5 seconds (i.e. you get 10-15 times the same data packet if you poll 10 times /s). Normally I don't see timeouts or any other issue and I think that other viewer also did not see timeouts on his module. No I never attempted to change the slave address Could you try running just AC_COMBOX.py ? Try other USB ports, maybe USB2.0 instead USB3.x . Another possibility is interference. I do get a timeout or even sometimes a reset of the module if I use the x1/x10 hardware switch under load. In my built the current transformer with the x10 coil is right next to the module and can apparently cause some issues when the switch is moved. This should not be the case with a x1 and a single wire going through the core. In general it could be that your environment/mains is very noisy causing interference on the USB cable? If you are using the original PZEM-004T with the capacitor-dropper power supply there must be enough mains voltage to power the board and the opto-couplers (more than 60V-70V if I remember). Last tip, if you poll for example every second, you should be able to see the RX LED on the module light up periodically. If the message decodes correctly (checksum...) the TX LED should light at the same rate. If the RX doesn't light regularly at the poll rate, messages from the PC don't get to the module. If it does light regularly but the TX LED doesn't or not at the same rate, then the PC message contained errors and the microcontroller could not understand it. If they both light regularly then there is a problem in the PC

Sadly no. The module datasheet says 45 to 65Hz.. The chip datasheet itself says 35 to 75Hz and I measured that on my unit, 30 to around 400Hz is usable, but that may be an exception.

@@TheHWcave So from your tests anything under 30hz is unaccurate? Do you have an idea how get accurate measurements 15-100hz? With different circuit or ic? Thanks

@@DEMOxSPL I am not entirely sure what you are trying to measure. I am guessing AC voltage at 15-100Hz like for a VFD (Variable Frequency Drive)? The problem is the low frequency. Most meters do not go below 45 or 40Hz. From my own collection, the only 2 handheld meters that would work are the Brymen BM235 and BM869s because they have a VFD mode that goes from 10Hz. Bench meters go even lower. The Agilent 34401A goes down to 3Hz . The OWON XDM 1041 goes down to 20Hz. These are just meters I have. There are many more out there...

@@TheHWcave I want to measure power output from car audio amplifier (sub level range) I want to make my own circuit.

@@DEMOxSPL Ok, that's different then. Its an interesting project. If you have a reasonably smart scope, one way to do this would be to attach a resistor R as dummy load as speaker and use the scope to measure the voltage V over R and perform the math operation (V*V)/R on the signal which gives you a power graph

The original capacitor-dropper is rated 250VAC max 275VAC. Using the modification powering it instead from a safe(!) DC-2-DC converter and using multiple resistors for R17, it might be possible but its risky. I only tested it to 250VAC

@@TheHWcave thank you very much, I'm thinking of using it to measure energy from an inverter that reaches 275 VAC, and a maximum amp of 1.8

@@TheHWcave Mmm... do you think that I can only supply the meter with 110V or 220V and on the other hand expose the CT to 275 VAC? Do you think the measurement will works? I ask you if it has happened to you, anyway, I will have to try to see how it goes. Thanks, buddy

@@Carlos-id7my Inverters often don't have proper sine wave output, which means the capacitor dropper could well be overloaded even at lower voltage. For this use, you should definitely get rid of it and replace it with power from the fully isolated(!!!) DC-2-DC converter from USB. Once the dropper is gone, make sure R17 is split into multiple resistors in series to increase the overall voltage rating. One resistor is definitely not good enough, use at least 2, better 3 or more in series. With all this, consider this is mains voltage and you do this at your own risk