Power factor measurement

[dhouston]

You need to take several thousand instantaneous readings of both current and voltage per second and average their product over the time period to calculate true power (Kill-a-Watt does it this way). You can measure the current using hall-effect sensors and the voltage using an unregulated DC wall transformer (you'll need a calibration procedure). Apparent power is just Vrms * Iavg. TP / VA gives you PF.

Doing it this way will also handle loads like switching power supplies, CFLs, etc.

There used to be a PIC based example on Dr. Ed Cheung's website. He's a NASA engineer who built a device to measure the power used in his house. I'm not sure it's still there but he will probably respond to email, if not.

Of course, the easy way is to spend $25 for a Kill-a-Watt meter. I think I was the first person to buy one and review it (on comp.home.automation). Now you can buy them at every corner store.

[dhouston]

There's also an excellent discussion of PF on Wikipedia. I would have just added this to my earlier post were it not for the puerile editing rules used here.

[Charles Linquis]

You need to take several thousand instantaneous readings of both current and voltage per second

Not really. In most cases, you can do it slowly. Consider a 60Hz sine wave at first. You could statiscally get a sample at every degree if you sampled one time in each cycle for 360 consecutive cycles, as long as your measurement interval was not synced to the 60Hz. You could obtain those 360 samples in 6 seconds. So, you could measure voltage, then current, then wait for 15msec and do it again. Plenty of time for calculations. Of course, you don't get cycle-by-cycle results, but I don't think a KILOWATT does either. If the update rate is too fast, humans can't read it.

And you don't need to grab every degree, either. If the crest factor is not too high, you can get really good results by sampling every 10 degrees or so. If, by chance, the input was a pure sine (I agree that isn't too likely, but it is a starting point), and you sampled 5 degrees before peak and 5 degrees after peak, sin 85 = .996 That is a really small error. You might say that the waveform isn't changing very fast at the peak, and you would be right, the rate of change (COS) is greatest at zero-crossing. The sampling error would be highest there. But, the energy at zero crossing is low, so it doesn't contribute much to the RMS values.

[dhouston]

... I don't think a KILOWATT does either.

The Kill-A-Watt sampling rate is about 2kHz which is on the low side. I exchanged email with one of their engineers way back when it was introduced. And, I don't recall what the sampling rate was in the Microchip AN220 (I did read it long, long ago.) but believe it was even higher. With electronics, you can get some very non-sinusoidal loads. Harmonics are why you need high sampling rates. AN220 says, "The CS5460 power/energy measurement IC measures instantaneous voltage and current four thousand times a second."

IIRC, there were also some excellent articles on the Rensselaer Polytechnic Institute site and there was a really good article on non-linear loads written by a fellow in New Zealand whose name escapes me.

[Charles Linquis]

I still don't understand why they sample so fast. The power factors of most loads don't change over thousands of cycles. You can sample every single part of a waveform even if you sample it once a second. You just need to take a lot of samples.

I have built two wattmeters over the past few years. One did have to sample fast (about 10KHz). I needed the value of each half-cycle as soon as it reached zero-crossing.
I started the A/D, then did the calcs for the previous sample, then grabbed the result and started the A/D again.

The other sampled at a 2 mSec rate (500Hz). I sampled for 3 seconds (1500 samples). The waveforms I measured were definitely
not sinusoidal. The results were the same in both cases, and I actually used a KILL-A-WATT as a calibration reference.

[ardhuru]

I've seen a PC based utility that does this with an interseting approach; it uses the 2 sound card channels to track the voltage and current readings, does the math and displays all the parameters on the screen. The simple coupling hardware is also described in the help section. The company's since changed its name so it took some searching, but here it is. http://www.infinitespectra.com/freew...400/index.html

Since you can calibrate the software for your own setup, I'm sure the components arent too critical either.

And did I mention its free?!

Regards,

Anand

[dhouston]

Neat!

If you do this, make sure you use an isolation transformer between AC line and your test circuitry. Otherwise, you caan make the case of your PC hot.

[ardhuru]

Dave, the circuit does isolate both the channels already, doesnt it? The voltage sensing is by way of a 12 volt step-down, and the current sense is a coil wound around one of the power carrying cables. This might be the only difficult to procure component for those having no access to Digikey, but since there is an option to calibrate the unit, I suspect a home made hand wound coil should be fine.