I wanted to provide some positive feedback on real-life accuracy of the TX, as opposed to the 'basic' testing at http://www.openenergymonitor.org/emon/emontx/accuracy.
Having seen a significant variance in the values of reported power usage from commercial, clamp-based monitoring systems, against actual (known) power, I had been slightly sceptical about the accuracy of emonTX. These commercial systems seemed to show a variance of 10-30% (and of course, dont report 'reverse' flow /phase change) against real power.
Converesely, I have validated that these same commercial solutions with pulse-counting options are actually very accurate, even at low power usage levels (despite some members views documented on the forum). Given that many electricity companies use the pulse counting method in electricity meters to generate bills, I have used a commercial pulse counting solution on my house meter as a baseline of accuracy.
Very simply, and using the excellent functions at open.sen.se, I have been able to overlay data from emonTX against the commercial pulse counter, to demonstrate how accurate emonTX actually is.
I have connected the commercial device to another open project (Home Automation Hub), which allows reporting of power usage at 6 second intervals (as opposed to the commercial offering of 5 minutes), and submitted to COSM.. The emonTX also publishes to COSM, but at nearly real-time rates.Open.sen.se then pulls both feeds into one graph for comparison over the same time interval.
The ability of emonTX to allow calibration in software has even allowed me to obtain some really cheap CTs (£2 each) http://www.taehwatrans.com/
The two graphs attached show how close the commercial and TX measurements track each other, at both high and low power usage.
Summary excellent confidence in a correctly calibrated TX, allowing near perfect reporting.
Sorry for the big pictures, but I wanted folk to be able to track the closeness of each plot. (Note that the greyed-out/non display plot is the commercial clamp value, or 5 minute averaged value; the HAH line is the commercial pulse count reported value).
.png)
.png)
Re: Positive results on accuracy
Am I right in thinking the last graph goes down to 40 W?
A slightly off-topic note that there are some interesting comments about electricity meters in the letters here: http://eandt.theiet.org/magazine/2012/08/your-letters.cfm
Re: Positive results on accuracy
Yes Robert, confirming 40W reading on my PV circuit (evening time)
The IET's articles make interesting reading. I think i would challenge one of the contributors comments that "'the accuracy of 'all' electronic electricity meteres should be questioned' although I sympathise with possible degradation due to age ( and I will undermine my own argument in a minute....). None of my house meters are smart (and the lack of global standards noted is also worrying) but are Elsters, which are supposed to be 'certified' for their accuracy.
And therein reveals the reason for my desire to obtain an accurate external baseline. I have 'felt' that the 'leccy supplier has been charging us more than I feel we are using; i do know I am a high user (c15KWh - 20KWh daily) though. I also have a background load of about 300W which ive not yet traced. Estimated readings contribute to the uncertainty, although i supply corrected readings regularly.
So, my next activity is to build enough TXs to cover all the feeds on my two consumer units, and see if a) they equate to total house usage, and b) find the pesky background loads.
Actually, if anyone wants to lend me their pre-built, or sell at a reduced price, any spare TXs, that would help save me some time, as I need about six.
{appolgies for missed capitials and typos, as on iPad}
Re: Positive results on accuracy
Thanks for that. Just from the theory alone, it's around the 40 W area that emonTx accuracy becomes dependent on unknowns - how much noise is injected into the measurement, where the 'zero' sits in relation to the ADC steps, etc. It would be really interesting to see compable data in the sub-40 W region from numerous 'Txs.
Any yes, I agree that sweeping statements like the one you quoted aren't too constructive, but it only takes a few well-publicised incidents to spread alarm, despondency and mistrust far and wide.
Re: Positive results on accuracy
I'll see what the sub 40w looks like over the weekend, although im sure i put code in the Base to filter out noise of less than 20W
Re: Positive results on accuracy
Andygodber: I also have a background load of about 300W which ive not yet traced.
If you were to turn everything off except the socket for your voltage sensor, there should be minimal current flowing. Then you could switch on each circuit in turn to find the culprit(s).
Our house consumes around 150-200W depending on the state of the fridge etc. At one stage this morning, the system was so well balanced that the immersion's neon was flashing slowly when the washing machine was resting, but not when the drum was turning :)
Re: Positive results on accuracy
Andy,
can I ask , which CT did you source from the site you mentioned above? I'm thinking of getting some TS9Ls to try.... This is because I'm doing a trial and have 17 CTs in the main board; if the trial is going to go anywhere, I'll need to be able to close the cabinet, safely, after installation so I'm looking for smaller CT's to test with.
thanks,
Eamonn
Re: Positive results on accuracy
Eamonn - I used the TS10L - also quite small, as i had the same problem of trying to get a lot into my consumer unit.
I got mine from a chap on ebay, who may still have some. Otherwise, I bet the manufacturer would supply to you for the quantity you need.
In one of my other posts, ive described how I got three TXs into a small box, for 9 CTs - two boxes would do for your 17....
Re: Positive results on accuracy
Andy, thanks for that - I've contacted the supplier so we'll see. Nice write-up on the multiple emontx build. Very neat. I must publish my own builds.