Archived Forum

Hello Lloyd,

I did an emontx accuracy test here: http://openenergymonitor.org/emon/emontx/accuracy

These where the conclusions that I reached:

The error below 40W could not be measured reliably and below 100W it was worse than 10%. Above 100W it was better than 10%, above 150W better than 6%, above 250W better than 4% and above 500W better than 2%.

I would certainly expect a better power factor reading too than your getting there.

Are you using a standard emonTx, and which c.t? There is nothing inherently wrong with having a 2-turn primary for the c.t., but of course you need to change the calibration factor accordingly.

If I've read your tables correctly, two things stand out:
(1) The values differ from your reference meter by a factor of about 1.7, and
(2) you are reading proportionally higher than you expect at low currents.

No.1 does not tally with a 2-turn primary, and if you have the standard 100 A YHDC c.t. you either have a wrong burden resistor or there's something seriously wrong with the c.t., plug, socket or cable; or you have a completely wrong calibration factor. You can test the c.t. independently of the emonTx if you have a mutimeter that measures milliamps, you should get 0.5 mA per Ampere turn in the primary ( so 1 A in a 2-turn primary should give you 1 mA in the secondary). Of course connect the multimeter directly across the c.t., the multimeter is the burden.

I believe (2) is indicative of noise pick-up, possibly from the processor, though there are many other factors that also come into play at low levels, all connected with the process of the analogue to digital conversion and where the bias midpoint and the sample sit in relation to the steps of the ADC. Trystan and I have spent many hours pondering these. This gives a hint at just some of the factors.

If you have a 2-turn primary, you can halve the powers Trystan quotes above.

 [Edit] You seem to have two different calibration factors? One for the current only sketch and a different one for current & voltage?

Thanks for the replies. This is a standard emontx, single ct purchased through the store. There is also a temperature sensor attached, and the mains transformer voltage reference.

I've been pondering this during the afternoon, and although my understanding is limited, I wonder if trying to measure in this way is fundamentally flawed when using a proportional controller like the Crydom. If this is a phase angle control device, how will emonlib deal with large portions of the wave form being missing?

Lloyd

If you have calibrated it correctly (voltage, current, phase at unity power factor in that order) then because you should be getting 50 or so samples per cycle, it should be reasonably accurate. I assume that it treats positive and negative half-cycles the same, and let's say it is firing with a phase angle of about 90° (so near the peak). Up until then, you're measuring zero and the current is zero. Now you have just measured zero when it fires, you'll see no current for 1/25th of the half-cycle when you actually have peak current. Then on the next reading you again read the current that's actually present. So I think you've missed a block of current that is 1/25th wide and peak current high - is that an error of about 11% of the true value (4% * 1.414 / ½)? 

The easy way to determine the size of this error will be to make up a sine wave (just half is enough) in a spreadsheet and evaluate the 'correct' rms value (using a lot more than 25 points per half cycle) and compare it with the value you evaluate using exactly 25 points per half-cycle. For various firing angles of course.

All the time you have been assuming that your Brennenstuhl PM230 is accurate when measuring a chopped wave. This may not be a valid assumption.

I've done a quick graph. The green line is the 'true' rms value of the sine wave current at the firing angle shown (normalised to 100) and the red line is the value that the emonTx sketch would calculate for rms current if sampled every 7°.

Thanks Rob.  I calibrated current at nearly full load (~2.5KW) and  voltage to agree with the power meter.  I'm not sure how to calibrate phase without a 'scope.  

 I think I need time to digest your post, but my biggest lack of understanding is how the power factor can be calculated, espcecially at low power, when you don't have complete cycles.  How do you tie the current and voltage waveforms together to determine the phase shift?

Anyhow, I think measuring a 100W light bulb directly connected to the mains is the first test I need to do tonight.

 Thanks

Lloyd

Power factor has nothing to do with phase shift!  - - except in a lecture theatre where you only have a pure sine wave. The (phi) in P = V.I.cos (phi) is only phase angle when you only have pure sine waves - no harmonics, no chopping. The true and correct definition is the sum that the sketch does - cos(phi) = (real power) / (apparent power) and provided you calculate true instantaneous power and average it, and at the same time calculate I² and V² and average (and square root) those, then you can get the true power factor, whatever the waveform.

Hence to set up your phase correction, get a pure resistance (electric fire, immersion heater) and tweak Phasecal to give power factor = 1. As per the instructions.

 Results from simple tests bypassing ssr.  This looks to me like the CT and voltage source are both working.  The calibration is slightly off, and very different to the results above.

Lloyd

That's more like it! Now all you have to do is work out which one is wrong (or more likely how much both are wrong) at different firing angles of your SSR.

(Is it possible to relate firing angle to input command signal on the SSR? If so, you might be able measure both current and power against firing angle and produce curves similar to mine above).

 Sorry Robert, you've lost me.

I thought above that you said that the algorithm should cope with the output of the SSR. Perhaps I misunderstood you.  From my results above it looks like the system is behaving reasonably well without the SSR.  I can't see what I can do/should do to make an adjustment for the SSR in circuit. 

Lloyd

The emonTx algorithm should cope reasonably well, but it won't be perfect - for that you need an infinitely fast processor. I think (and I don't know because I haven't tested it) that the emonTx should be accurate to a few percent when measuring the SSR current (until you get down to very small conduction angles and very small powers of course, when the percentage error could be huge).

What I haven't got a clue about is the accuracy of your 'reference' meter when fed with a chopped wave. I tried to check but I found nothing that indicated what form factor it could cope with, what frequency it would operate up to, or anything that gave me a clue to its performance with anything other than near sine waves.

If you can control the firing angle of the SSR (i.e. does the manual give a graph of firing angle or power - or anything! - against input voltage?), then by manually controlling that and measuring power with the emonTx and power with the reference meter, you could get an estimate at least of how accurate both were.

Of course with access to a 'scope it would be easy. If you have a sound card in your computer, you could look at the test report for the YHDC c.t. to see how I used that as a 'scope. You could then calculate the power expected knowing the firing angle and the load, and relate it to the power actually measured. (N.B. the graph I posted showed rms current, not power).