Calibrated the numbers under some load and am getting readings in the ballpark of my smart meter.
However, there is no current and even when I remove the CT from a cable, the emonTx is still registering between 2 and 4W.
Is that normal? - is that down to calibration? - is there anything I can do about it?
It’s annoying now i’ve moved the clamps to my solar generation instead of grid feed, so it looks like i’m generating a few watts over night, when in fact it’s not.
So, you’re seeing 2 - 4 W from an instrument capable of reading to 24000 W, and you’re worried about it? What does that work out as, as a percentage?
I’m afraid, for an analogue instrument, you’re expecting far too much. You’ll have to remain disappointed, and subtract a few watts in emonCMS, or get creative with the input processing and mask values below 4 W.
It might be worth looking up what the starting current for your smart meter is. I’ve looked up a non-smart L&G one and the base current for a Imax of 100 A is 20 A, and the starting current for that one is 0.004 Ib. In terms of power at 240 V, I make that about 19 W (resistive load). That means that tariff meter won’t record less than 19 W.
What surprises me about your complaint is normally, people complain when they see their PV consuming overnight - but of course that’s genuine, inverters do that to the tune of a few watts.
That article is actually about the non-linearity that becomes apparent due to the discrete nature of the voltage steps of the ADC. I think that’s not quite the same as @ichilton’s complaint. What he’s seeing is most likely noise, picked up on the input cables or coming from the digital part of the processor itself.
Agreed. But I think the first and last sentences of that article make it pretty clear that it’s still relevant to his situation.
I’m more trying to manage expectation. CT clamps are generally incapable of measuring zero current stably, and anything on the low end is always going to be suspect. But as you’ve said, the proportion of that ‘iffy current’ is usually so low, it’s best accepted and ignored, for our purposes.
That is a very well-known problem in engineering - it comes down to “What is zero?”. And there’s no answer to that, other than “Something less than I can measure with what I have to hand.” Personally, I think the emonTx performs amazingly well, all things considered, to have a spurious reading of only 0.0167% of it’s full rated capability. If the emonTx’s software only reported in increments of 10 W, and rounded the value to the nearest 10 W, most people in @ichilton’s situation would be more than happy, and in most cases would be oblivious of the fact that the result was inaccurate and only good to the nearest 10 W.
I guess you mean split-core c.t’s. When you’re talking about errors at these currents (relative to the rating), it’s probably worth being specific. It’s likely that an otherwise equivalent ring-core c.t. will perform better at these very low currents, but that still won’t have any influence on how the ADC performs when the signal is only a very limited number of samples peak-peak.
Yup. Implicit in the ‘clamp’ epithet, I thought It’s hard to get the ring type on to a pre-existing supply cable without disconnecting it, and they don’t really ‘clamp’, per se.
We’re splitting hairs and having fun, tho, I hope…
TBH, I didn’t interpret @ichilton’s comment as a complaint, really - more of an ‘enquiring minds want to know’, which I think is legit. But you are right - the emonTX is an excellent bit of kit.
Caveat
Though I confess I don’t use one - I rigged up my own bit of homebrew pulse-counting hardware with a phototransistor, and use it with an old Pi and the EmonCMS software. I lack the granularity that a CT would give me, but I gain the per-minute precision from the pulses, so everything’s a trade-off. And now I have access to Octopus’s API, plus a Zappi2 charger (which has its own API and a CT measuring the grid), so I’m slowing combining, organically, all the bits I was missing in my setup. IMHO, the whole OEM/Emon project is commendable. It’s giving my kids a future.
As I wrote to begin with, it is normal to see a small reading even when no power is there - even if the c.t. is unclipped and a long way from any cables. The problem is there’s a small amount of noise - most likely coming from the digital side of the processor - that finds its way into the analogue circuitry.
Obviously, the best solution is to remove the source of the reading if it’s external. What I forgot to ask was, is the spurious power still there if you leave the c.t. plugged in and move it away from any electric or magnetic fields.
If the reading persists, then I’d suggest you look removing the offset using the input processing in emonCMS. If the value is relatively steady, the easy and probably the most accurate way is to subtract it using + [a negative value]. If it’s variable over a small range, then I’d suggest subtracting but leave always a small positive value, then use ‘If >, skip next’ and then make the next operation ‘Reset to zero’. Unfortunately, the only way I can think to handle this if you have a bi-directional power flow is the first suggestion, to subtract a constant value.
It’s hard to get the ring type on to a pre-existing supply cable without disconnecting it, and they don’t really ‘clamp’, per se. 
