Archived Forum

Are you simulating the load & P.V exactly as in the drawing here: http://openenergymonitor.org/emon/mk2/build

If you are not measuring the nett current, the control loop is not closed and the triac will continue to fire. That is the most likely explanation.

You could try running any of my Mk2a or Mk2i sketches in Tallymode.  This will give you a good idea of whether the measurement side of the system is working correctly. You will also be able to calibrate your system accurately. 

Until the measurement side is known to be working correctly, I wouldn't connect any dump load. 

OK, tried a number of things ...

with the CT unconnected and transformer unpowered I get 2.5V on pins 1,2 & 3 of the LM358.

with the transformer powered at 240v I get 10.2v across the 150K & 15K resistors in series and 0.7v across the 15K resistor.

with the CT unconnected or shorted I am getting 0v on the voltmeter and using the rawsamples tool, the ADC values are centered at 509 ~ 511.

Voltage readings are spread nicely from 186 to 835 in the raw samples tool.

Using tallymode in mk2i rev4, I had to adjust powercal to 0.021 to get a sensible reading from a 500w light bulb as my known source. Using a 60w light bulb powercal it needed to be 0.01. I have kept a value of 0.021.

With the phasecal tool, even numbers around 3 or 4 are only giving a power factor of 0.4 for my 500w light bulb, which I would have thought was a purely resistive load (?). I have kept phasecal at 1.

I have tried running the bench test again with mk2i rev4 and I am getting the same results. I am certain I have set up the CT correctly, measuring the net current. As I increase the simulated PV load, (60 w lamp) by adding more turns, the rate of triac firing increases but triac firing continues with the pv load switched off.

thoughts again ?

thanks

Some of your results look fine, so there can't be too much wrong with your setup. 

I have to go out now but will send you a PM about this.

Calypso

I ran phasecal tool with a 3kw heater and it needed a phasecal of 10 to get a power factor of 0.92.

Using the raw samples tool, a 60w light bulb as a resistive load showed a small but discernable variation in the CT output, a 500w light bulb gave a full spread wave form but the 3kw heater seemed to overload the CT, sending values from 0 to 1023 with only 1 or 2, non extreme readings. 

The CT is OK as I have plugged it into emontx and it is giving good readings. Is it likely to be the LM358, too much gain ?

"The problem which I can't figure out is that the triac is firing with no simulated PV. IT fires even if I disconnect the CT."

You need to reverse the CT on the wire to empty the bucket, simulating export. What you are doing with the CT is not simulating PV, but home consumption and if you are importing or exporting.

Even I did the same thing till the penny dropped!

canary50, I'm sure you're going down the wrong route.  I've just sent you a PM which will hopefully guide you in the right direction.

All of the sketches that I've posted should work equally well on either hardware platform.  I've just build up an emonTx-based rig, and Mk2i_rev5 is working nicely with everything fully functional. 

thanks Robin, fitted the 150R resistor and everything is working on the test bench as it should be. Power factor 1 on a resistive load, current and voltage traces evenly spread, phasecal now set to 1 and powercal at 0.08.

I am using the SCT 013 CT and for some reason I was under the impression it was fitted with a burden resistor so didn't fit the additional one ... but I have now.

On to the next step ....... I notice in one of your videos you have a changeover switch for solar to constant. Are you using a mains relay ?

And I might do battle again with IP addresses to get the nanode talking to emoncms ....

thanks again !

Yes, the CT really does need a burden resistor, otherwise there's nothing to develop a voltage across for the Arduino to measure at its ADC input. 

With the standard YHDC CT, and a 150R burden, the full scale of your measurement system should be around 4 kW, which is fine for most PV systems.  If you wanted to measure the consumption of an entire house, your burden would need to be correspondingly smaller.  A burden of 18R sets the full-scale value at around 25 kW.

No, there's no relay, just two switches in parallel in the Live path; one manual and the other electronic.  The triac doesn't seem to mind working in this way. 

A current transformer behaves as a current source. That means (subject to its power rating) that it will generate whatever voltage is necessary to drive the current it needs to - in the case of the 100 A: 50 mA one, one 2-thousandth of the load current. If the ADC had a current input, that wouldn't be a problem. As it is, the ADC input is a voltage input and it needs the burden resistor to convert that current into a voltage.

It's quite possible that you can damage or destroy the Arduino analogue input if you leave off the burden resistor. A large c.t. when open-circuited can easily generate many thousands of volts and destroy itself by flashing over, not to mention the danger to anyone near. NEVER, ever, open-circuit a current transformer.

"I am using the SCT 013 CT and for some reason I was under the impression it was fitted with a burden resistor so didn't fit the additional one ... but I have now."

Good to hear you have the right numbers now, Were you using the SCT 013 000, not the SCT 013 030? (that has a built in burden),

It actually takes quite a lot to over stress a CT, if you have a 30A CT set up it may give a distorted O/P with a higher current, but in this application all it needs to do is tell what way the current is going and if there is enough surplus to fire the triac.

It actually takes quite a lot to over stress a CT, if you have a 30A CT set up it may give a distorted O/P with a higher current, but in this application all it needs to do is tell what way the current is going and if there is enough surplus to fire the triac.

The problem with a distorted output is that it is likely to be non-linear, getting progressively worse as the signal size increases.  A Mk2-type system will only be able to balance consumption and generation accurately if everything is linear.  When surplus PV is minimal or near maximum, the disparity of the energy content between mains cycles when the load is 'on' and 'off' is considerable, maybe 50:1 or greater. 

With the Eco Eye monitor, the voltage sensor is indeed only used to show which way the current is going.  That product assumes a standard value for the voltage so will be somewhat less accurate than any system which uses a measured value. 

Tell me what current you'd like to put through a SCT 013 000 (up to 250 A), what burden resistor you'd like to use, and I'll give you a picture showing the waveform distortion - with the primary current wave for comparison. There's a picture of the waveform at 250 A with a 15 Ω burden here. (There's also a picture with no burden, showing a massive phase shift and equally massive distortion at just 18 A.)

There's a picture of the waveform at 250 A with a 15 Ω burden here.

Nice report Robert.  One comment regarding:

The maximum phase error of 3º is insignificant (representing a power factor error of less than 0.0015).

That's only true for resistive loads right?  i.e. cos 3 / cos 0 is pretty close to 1.   But as the load gets closer to being purely reactive cos 88 / cos 85 is a long way from 1.

One of the things for which I use my system (which admittedly has a lot more dynamic range than emontx) is baseload tuning; answering questions like "Is it worth turning my microwave off at the wall?".  You can see in the attached screenshot when it's just being a clock, it's almost purely reactive.  An un-checked 3º phase error there would be quite significant.

That product assumes a standard value for the voltage so will be somewhat less accurate than any system which uses a measured value.

Indeed, especially if your background load is a nice reactive mix.  My party trick is to demonstrate how my currentcost display will drop by about 30W just by turning my microwave on at the wall.   Alas, it only works when the fridge is running.  The currentcost counts the fridge's reactive power as real, and the microwave's capacitors help correct the fridge's inductor.  Measuring current without reference to V is pretty useless in most cases.

It spawned a dodgy market in power factor correcting devices here.   Amazingly our consumer regulator managed to get their heads around it enough to shut them down:

http://www.accc.gov.au/media-release/federal-court-declares-consumers-mi...

The most power I've seen drawn from the grid as measured by the Eco Eye software was around 7.5KW, this was the washing machine, dishwasher and the wifes hair drier. Not thankfully a normal occurrence so using just I=p/v  thats around the saturation point of the CT, is it a problem? I don't think it is as I have a 4KW max solar system so in any case the Mk2 would not be attempting to divert, Would I be right in assuming anything over about 17A is irrelevant as thats the max the PV will o/p? Or would it be anything over 12.5A (what the immersion draws) assuming good sunlight and not much else going on so the triac will be fully on and still exporting power to the grid?

"It actually takes quite a lot to over stress a CT, if you have a 30A CT set up"  What I meant by this is that you aren't likely to to be drawing anywhere 30A at any one time, I do appreciate the importance of a good signal but as 4KW of PV or so seems to be the norm the diverter would be off. Does that sound right?

Under 'normal' loads with 'normal' power factors, 3° is indeed insignificant, but of course it does become significant at poor power factors (almost purely reactive loads). However, those loads tend to be very small so the overall effect still is it makes little difference. (Remember, down at the tens or hundreds of milliamps level, you're talking only a small handful of ADC counts peak-peak, so accuracy is not brilliant whichever way you look at it).

I once questioned one of the manufacturers of an "approximate" device: they admitted it multiplied current by a constant voltage and assumed unity power factor. I guess for anyone who knows no better, that's 'good enough' and they get some indication that they are saving energy. Whether they actually are or not, and the amount, is open to scrutiny.

The currentcost counts the fridge's reactive power as real, and the microwave's capacitors help correct the fridge's inductor.
So you see the supply voltage jump up when you turn the microwave on and the fridge is running? (You should do!).

Hmm... I don't think we're quite in the same ball-park as this device. (I think they mean 1 part in 10⁹, not 1 part in 109 )

However, those loads tend to be very small so the overall effect still is it makes little difference.

Unfortunately, when you're trying to nail down your last 200W of 4am  baseload, it is mostly made up of a whole lot of tiny loads, all with ugly power factors, all combined together to give a pretty nasty looking current signal.

(Remember, down at the tens or hundreds of milliamps level, you're talking only a small handful of ADC counts peak-peak, so accuracy is not brilliant whichever way you look at it).

I'm blessed with 24-bit high speed ADCs, one for each of 12 CTs and another for V, all running in parallel.  A combination of analog and digital filters give me an effective bandwidth of about 2kHz and a dynamic range of about 1000:1.  So if my full scale CT deflection is set for 20A say, I can accurately measure active and reactive power down to about 20mA (assuming your CT is up for it).  In that environment 3° matters.

So you see the supply voltage jump up when you turn the microwave on and the fridge is running? (You should do!).

I can't say I've checked, but I will.  What's the reasoning there?

This is turning into another picopence discussion-

"I'm blessed with 24-bit high speed ADCs, one for each of 12 CTs and another for V, all running in parallel.  A combination of analog and digital filters give me an effective bandwidth of about 2kHz and a dynamic range of about 1000:1.  So if my full scale CT deflection is set for 20A say, I can accurately measure active and reactive power down to about 20mA (assuming your CT is up for it).  In that environment 3° matters."

What sort of industrial environment are you measuring and why?

What's the reasoning there?

The supply is generally inductive, so adding a capacitor will "correct" the power factor (to the suppliers benefit, not yours!) and it's heading towards resonance so the voltage will rise. You can soon see the effect if you throw a handful of components - a voltage generator, a series resistor and inductor representing the supply, and a load resistor with or without a parallel capacitor - at your favourite circuit simulation program. This is how dip compensators work.

I have but one comment about the phase errors in your case: If you have a system with that kind of resolution, you should be looking at measurement grade current transformers at the very minimum, most probably ring-core c.t's, and not cheap split-core ones. And the same goes for the voltage transformer(s). If you can't do that, then you also need to consider the variation of the phase errors with load (for the c.t.) and system voltage (for the v.t.), and possibly correct for that variation too.

Actually, on the V side I use a precision divider and it all sits behind isolators that are good for 2.5kV, but yes, the overall system accuracy is only as good as its weakest link.  The trick is to find CTs with a nice flat amplitude error and phase error over a broad current range (and temperature range).  That and use an upmarket power calibrator where you can independently control V, I and Ø (that's the bit that gets expensive).

The supply is generally inductive, so adding a capacitor will "correct" the power factor (to the suppliers benefit, not yours!) and it's heading towards resonance so the voltage will rise.

Not enough for me to detect with a multi-meter.  Does it matter where I measure V?  Isn't the rest of the house and the entire grid it's connected to, low enough impedance to make it pretty difficult for me to have any impact on V?

Isn't the rest of the house and the entire grid it's connected to, low enough impedance to make it pretty difficult for me to have any impact on V?

Judging by the amount of interest that a 3 kW dump-load rapidly switching on & off has aroused, I think the answer to that one has to be No!

Judging by the amount of interest that a 3 kW dump-load rapidly switching on & off has aroused, I think the answer to that one has to be No!

I'm not sure I see the connection.  You're trying (quite successfully by the looks of your youtube videos)  to match your power consumption to your power production right?

I think Robert is suggesting that a resonating LRC circuit in my house can change the voltage of the SEQ grid, but I suspect I've misunderstood what he's saying.

The 'test' impedance (I read that as the general worst-case) for a single phase supply for flicker is (0.24 + j 0.15)Ω in the line and (0.16 + j 0.10) Ω in the neutral at 240 V (you can of course lump the two together for this exercise). If you put those values into your simulation, add a representative resistive load, then add some parallel capacitance, you should see the voltage across the load rise. Of course, if your supply has a very high fault level, i.e. the impedance is lower than the test one, the effect will be much less than if it's marginal.

You won't of course achieve resonance - I wrote "towards resonance".

What Robin is saying is when he switches a dump load on and off, by virtue of the impedance of the supply, there is an additional voltage drop across that supply impedance and it causes the voltage you see to "dip". Repeated, frequently, it's called "flicker". What I'm saying is were that load to include sufficient parallel capacitance (and generally that's not practical but stay with the idea), then because the supply is inductively reactive, the interaction with the capacitive reactance you add can exactly cancel the voltage dip.

If you can see the power apparently reduce when you add a capacitive load, does it not mean that you aren't measuring the true real power in the first place? Sorry, but I think there is a reactive component in the quantity that you're measuring.

If you can see the power apparently reduce when you add a capacitive load, does it not mean that you aren't measuring the true real power in the first place? Sorry, but I think there is a reactive component in the quantity that you're measuring.

Yes, exactly.  If you re-read my post where I mentioned that you'll see I was talking about my currentcost display.  It has no connection to V at all and so can only measure apparent power.

My monitor has no such issues.. it even shows the fridge's reactive power on the right hand (inductive) side of the dial and the microwave's on the left hand (capacitive) side.  See screenshot above for the microwave.  I don't have one handy at the moment for the fridge.

Ah, I'd lost track and thought that you were saying that your hi-res monitor did it. I totally understand why the 'el cheapo' breeds tell lies.

Yes, as Robert has mentioned, I was simply pointing out that the local mains voltage will move slightly up and down as a relatively high-power load is turned off & on.  This is despite being connected to the low-impedance grid.  Various regulations exist which limit the amount of permitted flicker, and there has been much talk in these parts (and indeed some action too!) about this thorny topic.

I only mentioned it in reply to your claim that the grid is of a sufficiently low enough impedance to avoid there being any impact on the local mains voltage.

Ah, I'd lost track and thought that you were saying that your hi-res monitor did it.

That would be a worry, given its revenue-grade roots.

I totally understand why the 'el cheapo' breeds tell lies.

Yes, I only dusted off my old currentcost anecdote in response to richmc's bold claim that "in this application all it needs to do is tell what way the current is going and if there is enough surplus to fire the triac".  If that's all PV-dump were doing, I doubt it would come close to nailing that spinning disk meter like it does, especially when the background load is a nice reactive mix.

I was simply pointing out that the local mains voltage will move slightly up and down as a relatively high-power load is turned off & on.  This is despite being connected to the low-impedance grid.

Fair enough.  Your "judging by the amount of interest" comment was too cryptic for me to connect the dots.  I can imagine rapid fire switching of a 3kW load could be problematic.  The load being switched in my currentcost party trick is way way smaller (~3W real, 60VAR reactive).  So while the physics might dictate that a response will show up on V, in reality it's lost in the soup of Vgrid.

But speaking of grid impedance, it does show up as an issue here but typically in the other direction.  Because we only get paid if we export our PV, the winning strategy here is to turn everything off except the fridge, send everyone off to work/school and run all your discretionary loads at night.   On a sunny day this can result in areas of quite large production with very little consumption.  Throw in a bit of grid impedance and before you know it inverters start hitting their Vmax trip-points... typically in the mid to high 260's.  Voltage regulation on the grid seems to be very "chunky" and not very fast.