Not So Tiny Power Meter

Posted in Projects, Software Libraries, Tutorials by Bill
19 Dec 2010
Not So Tiny Power Meter

The Kill A Watt is an awesome product; it measures volts, amps and power factor of an individual appliance which can be used to calculate power, cost to run, etc. It’s also quite hackable. But I wanted something that would give me the same data for my whole apartment. After some Googling, the best I could find was this project from picobay, but I didn’t want to invest in an expensive network IO platform. There were also some off-the-shelf solutions, but they too were expensive and limited. Well, time to design my own solution then.

Enter what I call the ‘Not So Tiny Power Meter’. The catchy name comes from the microcontroller I used, ATtiny85, and some sizing issues I had with the enclosure.

I started out with a plan to use volt-meter current clamps just like the project I linked above (photo of clamp from picobay.com) and use a dedicated chip, the AD736, to convert the AC signal off the clamps to a DC voltage representing the RMS current value. The chips are expensive, tough to use as I found out, and still require external amplifiers to scale up the value to 5V ADC range. So I nixed that idea. Instead, I decided to use a single op-amp to scale up the AC voltage off the clamp and sample it directly with the ATtiny’s ADC. The circuit would be cheap and easy to design and I can convert the signal to RMS in code.

Then I had a thought. If I’m sampling directly, why not measure more than just amps? As an EE, I’d love to know more about my power usage, like power factor, frequency, and a more accurate measure of power by not assuming a voltage like most other projects; but I still wanted to keep the device simple. Then I had another thought: Why not measure voltage through the same transformer that’s giving my circuit power? After a few tests, I found that a properly designed rectifier and regulation circuit wouldn’t distort the source AC waveform too much. They key was to keep the values of the capacitors before the voltage regulator (circled in red) to a minimum, just enough to support a stable DC voltage. Anymore and the inrush when the rectifier diode starts conducting severely distorts the AC wave form.

My design is simple. An AC transformer powers the circuit and a voltage divider drops the source voltage down to ADC range for measuring. A dual sided half-wave rectifier and regulation circuit provides +5V and -5V rails. The AC signals off two AC clamps are scaled up using two op-amps. I planned on using trim potentiometers to calibrate the gains of all the measurement circuitry, but found it was easier to just use transfer functions (found with experimentation) in code.  Everything is measured with an ATtiny85, and transmitted out of the breaker panel by a cheap RF transmitter. Since all sources are AC, the ATtiny could only read the positive half of the waveforms. When the signal would go below the ATtiny’s GND, the protection diodes and input resistors would protect the ADC pins from damage.  With this design, I can measure voltage & frequency off one phase and current & power factor off both phases.

(Circuit Diagram, Click for Full Size)

The theory of operation is simple. First, the ATtiny85 will repeatedly sample the volts ADC pin for over a full period of the 60Hz sine wave. The peak value of the samples is remembered. Repeat for both current clamp ADC pins. After the max values are captured, the ADC clock is increased for faster sampling, with higher errors. To measure frequency and power factor, I used a 8 bit timer that  I extended to 16 bit in software. Using the timer, I measure the difference in time between two peak values of the voltage waveform. Then, I measure the difference in time between a peak value of the volts waveform, and a peak value of the current clamp waveform. Repeat for the second clamp. After all these measurements, some conversions are done to convert the peak values to RMS, times to frequency and power factor, run through a transfer function to account for various gains in the circuit and transmitted out via software serial as ASCII sentences with checksum.

Comparing to real measurement hardware, my project had respectable measurement accuracies of:

  • +/-  1 Volt,
  • +/-  1 Amp,
  • +/-  2% for frequency,
  • +/- .03 power factor when current is above 10 amps.

The circuit is designed for 120 Volt, 100 Amp mains, but can be adapted for other systems.

I ran into a few issues through the course of my project. The first issue was with the enclosure. All good projects should be protected by an enclosure, especially when installed into a breaker panel. First my poor planning resulting in a enclosure that was too small to house all the banana jacks for the current clamps. Then the second enclosure didn’t match the mechanical drawings provided by the manufacture. What stinks is I already had PCBs made to the spec of the drawings before I received the case. O well, time for double stick foam to mount the PCB instead inside an overly large box (part of the irony of the name).

The second issue was the quality of the signal off the current clamps. When using function generators for testing and programming, I could measure frequency, power factor and max value with great accuracy. The noisy signal off the current clamps is another story. Really, power factor measurements with currents less then 8~10 amps are very noisy.

Third, I originally used 434Mhz radios, until I realized it is the same frequency as my external temperature/humidity probe for my clocks. I quickly changed the radios to the lower frequency versions.

Anyway, I got the project built, tested, and installed into my breaker panel. Everything is internal so nothing extrudes. Right now, the data is received by another one of my projects, an Arduquee display. The display just shows live power usage. I plan to experiment with data loggers to log the data and/or play with the Google PowerMeter API to send the data into the cloud. This project was just to build a sensor to get the data out of my breaker panel.

Here’s some photos from the build, and the installation into my breaker panel. Click for larger pics.

Quite a workbench

Checking the signals

Installation into panel

 

The temporary LED display showing power (Watts), now hanging on my wall. Notice the RF receiver on the left.

My design is open-source. All the theory of operation is well documented in my code, and all code and Eagle PCB files are available to download:

Not So Tiny Power Meter files(zip)

I even have extra PCBs for any that what them, $5 plus shipping. Drop me a line in the comments if interested.

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.

Share

Warning: count(): Parameter must be an array or an object that implements Countable in /homepages/46/d285670699/htdocs/bill/wp-includes/comment.php on line 879
  1. 135 Comments.

    • jaqNo Gravatar says:

      Please keep me updated! Perhaps due to the excessive interest you could initiate a project page where all can collaborate

    • ChristianNo Gravatar says:

      I’m interested in a board too. Thanks!

    • Luis ViolaNo Gravatar says:

      Hi,
      Very nice project. Do you still have pcbs?
      Thanks

    • JoshNo Gravatar says:

      I’ve been wanting to monitor my power for a while. Please put me on your notification list. Thanks!

    • AndresNo Gravatar says:

      Bill,

      2 things..
      1-awesome site and the project is just awesome…definitely interested in getting a PCB (which I understand is the power meter reader!?) sorry for my ignorance.

      2-I would like to feature some of your work on my site, some people follow me for minimalistic life, saving on free apps and stuff like that… I thing this could help, I’m just planning on making reference, please contact me to get more details

      Regards,
      Andres L

    • LaurensNo Gravatar says:

      Hi!

      Seems like a very nice project, I’m interested in maybe buying a pcb! Is it easy to adopt it to work in Europe, on 230 volts and 50 hz?

      Thanks!

      Laurens

    • BillNo Gravatar says:

      @Andres

      Thanks!, and link away, I don’t mind.

      @Laurens,

      Should be easy to adopt. You still need an unregulated AC ‘wall-wart’ transformer to power the circuit at 12VAC. Then it’s just a matter of tweaking the gain equation in code.

    • Ian FountainNo Gravatar says:

      I’d like a PCB as well.

    • Bill LeederNo Gravatar says:

      Great Project!!!

      Do you have any PCB left? or do you have a kit for this Power monitor?

    • JohnNo Gravatar says:

      This project looks awesome and I’d love to do the same in my house. I’d like to put my name in for a PCB as well. Thanks!

    • MichaelNo Gravatar says:

      I’d love two PCB boards. A kit would be great if you get around to making one…

    • EricNo Gravatar says:

      I can’t see how this could yield very accurate (“accurate” is relative, I suppose) results, especially for measuring inductive loads. Besides the triac-chopping that Reza mentioned early on in the comment stream, it seems like your current and voltage waveforms are at best, loosely correlated. The problem (I think) is that you’re using two different circuits for measurement of each of current and voltage, and both circuit paths to the ADC are going to be phase-delayed differently. Since power is *instantaneous* current times *instantaneous* voltage, both have to sampled at the same time *before* they’re phase-shifted (in your case by two different RLC networks).

      it seems like the phase-delays would screw up both the power and PF numbers. Perhaps worse, It sounds from the explanation that you’re only using the full-cycle peak values for voltage and current, and throwing away all of the sampled data intra-cycle. This completely disregards the phasing of the current with respect voltage. I would expect this to put all of your power numbers off by something on the order of the *real* power factor — so for instance, an error of 20-40% just measuring the real power for something like a cheap CFL. Chopped waveforms like Reza’s triac example would also seem to be highly problematic.

      By the way, real power meters sample and compute/accumulate instantaneous V*I samples *many* times throughout one 60Hz cycle, and none would consider using less than an 18-bit ADC (and some around 22 bits). With a 10-bit ADC the best you can measure under perfect conditions is 1000:1, so I suspect that you can’t even see small-watt loads, depending on where the top-end of your scale is.

      Better meters don’t use anything but shunts, because of all of these types of phase-measurement, accuracy and non-linearity problems with CTs and Hall sensors. But shunts are clearly not feasible for most projects like this. The Kill-A-Watt clearly uses a shunt, for example, and is pretty accurate partially for that reason.

      I just wouldn’t be overly serious about accuracy claims. For whole-house coarse trending and consumption measurements, it’s ok for many uses though. Clearly a lot of us care about watching our power use, so any data about it is a good thing.

    • BillNo Gravatar says:

      @Eric

      I tested the circuit design with multiple function generators and a multiple channel Oscilloscope. (You can see all the equipment in one of my pictures). There was no phase shift of the respective channels presented to the micro-controller.

      Furthermore, once installed, I calibrated and verified the values by comparing to measurements done with a hand-held voltmeter and my multiple channel o-scope.

      You are right about the sampling, I am only capturing the max of V & I, but the error can be fixed by multiplying by pf. Though pf accuracy is the one I am most worried about, only due to the inaccurate internal clock source. I accounted for the drift in frequency in code by comparing frequency measurements to my o-scope, but I don’t know how well that will stay. I have read that the internal clocks can be 10% off the 8Mhz spec, but don’t drift much past the frequency they settle on.

      Next version I think I’m moving up to a ATmega with a more accurate external clock source, and ditching the Op Amps.

    • wcbzeroNo Gravatar says:

      How did you install the wall wart to get power to the board?

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.