I thought I'd share a trace I've just captured of my evening background load (TV, computers, PVR, amplifier etc.)
The yellow CT is a heavy (~2300W) resistive load that I've included as an approximation of V.
The red CT is the feed to the UPS which powers the above list of equipment.
The scope has a nifty feature that lets you trigger on +ve zero crossings of its own AC mains supply, which was the trigger used for this capture. You can see the scope's trigger and the yellow CT line up pretty nicely (a very slight lag on the CT).
The red signal is pretty nasty. It leads V in parts of the cycle and lags it in other parts, and seems to manage two zero-crossings for each V zero-crossing. It's roughly a 400W load with a PF of about 85%.
Re: Picture of nasty mixed loads.
Here's an FFT of the red trace, pretty big spikes at 50, 150, 250, 350 and 450Hz.
Re: Picture of nasty mixed loads.
It's all those switched-more power supplies that don't have decent input filters!
All that harmonic current represents losses to the power companies and you can understand why they don't like it.
Re: Picture of nasty mixed loads.
Despite all the grief that you're giving it (red trace), your mains supply (yellow trace) has a remarkably good profile!
Re: Picture of nasty mixed loads.
I agree with Robin, a lot of "clean mains" have clipped peaks, yours looks very sinewave. practically all phone chargers and DC wall warts are switched mode nowadays.
Re: Picture of nasty mixed loads.
Yeh, here's the FFT of the yellow clamp. It's pretty pure, with very slight blips at the odd harmonics. I get the feeling that grid power quality varies a lot from jurisdiction to jurisdiction.
Re: Picture of nasty mixed loads.
Since it's real live data, and a fairly significant load (~400W) that makes up a large part of my electricity bill, I decided to suck the data out of the scope and into a spreadsheet so I could do some what-ifs. The scope was sampling at 100kHz, or 2000 samples per cycle. I think I read somewhere that emontx can do about 100 (V, I) samples per cycle.
So I compared energy calculations using all the samples Vs using every 20th sample (and multiplying the result by 20), i.e. 2000 samples/cycle Vs 100 samples per cycle. I calculated energy(*) for each of the 5 cycles and the biggest error I saw between the fast sampling and the slow sampling was only 2.64%.
(*) I didn't bother calibrating it to real units, but just left it in a-to-d-units^2 per cycle.
If anyone wants the raw data to do their own research on, just PM me. Another interesting one would be to see how sensitive an ugly load like that is to an uncorrected phase error between V and I.
Re: Picture of nasty mixed loads.
I think I read somewhere that emontx can do about 100 (V, I) samples per cycle.
Each analogRead() statement takes around 104 uS, and two are required per sample pair so that their power content can be evaluated. The max number of sample pairs that can be achieved per mains cycle is therefore 20000 / 208 = 96.15
In its free-running mode, my Mk2i PV Router code does 96 sample pairs per mains cycle, so that's as fast as the device can go. This is possible because all processing is deferred until the next time around. Although the ADC is working flat out doing back-to-back conversions, the main processor still has some spare time on its hands. This is largely due to the use of integer maths which is a lot faster than the standard floating-point equivalent.
"emonTx" is a hardware platform so does not as such have an associated sample rate. It does, however, come with some sample sketches such as emonTx_CT123_voltage.ino. These sketches invoke the library routine calcVI() which takes sample pairs at the standard rate for a limited duration. A standard pair of analogRead() statements, with subsequent in-line processing, reduces the sample rate to around 55 sample pairs per mains cycle.
Re: Picture of nasty mixed loads.
Thanks calypso_rae. I just halved the spreadsheet sample rate again, now down to 50 samples per cycle and the max error improved slightly to 2.18%. I'm guessing it just got lucky with this particular dataset. My samples are evenly spaced amongst the bigger set, which sounds like it may not be indicative of how the various emonTx sketches work.