Immersion Heater - PV electricity

[suntrap]

I'd use a washing machine mains filter, which have the current rating and can be bought cheap.

 

[pmcalli]

I've been out of this thread for a few months while I followed the Nanode-based control solution that was discussed in earlier posts. I have done the assembly and mimicking and am about to connect to the 5V version of the same controller - MCPC2450A

pmcalli, do you have the output filtered? If so what have you used?

The I/P is filtered to stop fast turn on edge causing interference. I have used a standard emc filter from rs stock number 7024019

 

[DickiePhitt]

Thanks guys.
pmcalli, hope you dont mind but I pm'd you with a more detailed series of queries as I think you've done what I next need to do on my project

 

[sharpener]

I have now completed testing of meter behaviour and have very bad news for you. The meter measures true power and your analog multiplier measures apparent power so you will be feeding back uneccesary power. I have now tested using a synchronous demodulator and it correctly follows true power as measured by the meter. The other advantage is the circuit is unchanged whether you use one ct on composite meter tail or two cts on seperate meter tails.

No, the bad news is the other way round. If you are just doing synchronous demodulation of the current waveform you are not using any voltage waveform information (other than the zero-crossing points), so you are assuming that its waveform is the same as the current. This is only true if you have a pure resistive load, which may not be the case.

As suntrap has pointed out already in posting #388, what I am doing is multiplying instantaneous current by instantaneous voltage and then averaging the result over (at least) one cycle.

That this will work with any load or combination of waveforms to give true rms power is a well-known mathematical result, if you are not familiar with it then read this paper

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.134.5837&rep=rep1&type=pdf

or these articles

Power Measurement Article

Voltage, Current, Power & Energy : Definitions :: Electronic Measurements.

 

[pmcalli]

No, the bad news is the other way round. If you are just doing synchronous demodulation of the current waveform you are not using any voltage waveform information (other than the zero-crossing points), so you are assuming that its waveform is the same as the current. This is only true if you have a pure resistive load, which may not be the case.

As suntrap has pointed out already in posting #388, what I am doing is multiplying instantaneous current by instantaneous voltage and then averaging the result over (at least) one cycle.

That this will work with any load or combination of waveforms to give true rms power is a well-known mathematical result, if you are not familiar with it then read this paper

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.134.5837&rep=rep1&type=pdf

or these articles

Power Measurement Article

Voltage, Current, Power & Energy : Definitions :: Electronic Measurements.

Many applogies I didn't read your post properly I mixed you up with another poster who is using an analog devices true rms device fed from current taps which will only deliver apparent power. Thanks for the links best explanation I've seen. However your statement is not 100% correct. If you synchronously demodulate the current waveform with the voltage zero crossings you do get true power. A simple simulation with sine waves out of phase proves this. What happens with non sine waves I havent analysed. I have been running some tests against my meter which is of the spinning disk type and found firstly that the meter reads 4.6% low yip yip hooray but also creeps at 0.1kwhr when the controller is running the immersion at 2 to 3 kW. In my two ct circuit this can be compensated but not in the single ct without applying a higher offset which is not desirable for winter use. The unfortunate conclusion is that if I had been measuring apparent power then this problem would not occur as the current would be over reported. I suspect my meter is responding slightly to apparent power. Can you remind me which analog multiplier you are using and I will try it out as a measuring device to identify the cause of my creep.

 

[inie meanie]

I found found that to deal with the problem I have had to use two vairiable gain amps wich have allowed me to tune out the problem at the 2 to 3 kw point. I know 0.1 was only a small input but I didn't want to pull anything.

 

[pmcalli]

I found found that to deal with the problem I have had to use two vairiable gain amps wich have allowed me to tune out the problem at the 2 to 3 kw point. I know 0.1 was only a small input but I didn't want to pull anything.

Any idea to the cause. I have stupidly been concentrating on trying to get true power when apparent power would actually have given some margin to this effect. On my systems I see the effect on both a two ct balanced system and a single ct on composite meter tail. In the 2-3 kW range most of the power is resistive so I can only speculate that either the current taps are non linear or the meter reads the composite waveform incorrectly. I will try the suggested analog multiplier but not hopeful as I have had previous bad experience (over 25 years ago mind ) with this device beacuse of non linearity effects of up to 5%?. I will also try my demodulator circuit with added hysteris in the mains voltage squarer this will give a selectable phase shift and hence give the same effect as miss balancing the two ct circuit which was very successful. I am desperate to get a solution to this as I have a queue of people whoose installer used a spare way in the consumer unit making a two ct design difficult to implement.

 

[inie meanie]

I too have been battling with ever changing errors on the export rate. I have built and tested many many variations using different coils and approaches. I have noticed when inductive loads kick in ie. Washing machine and the fridge and freezer loads, these change the way the circuit will compare the power loads due to lag or lead on the current angle. It has all come to a compromise. I have now settled on the circuit that gives me adjustment on the buffering op amps. I have been told. Don’t let perfect be the enemy of good! I have been striving for perfection on the’ no export’, but Unless I spend a lot of money and map all this in a plc ( EMMA ) I am not going to get it. I am now happy to settle with ,< 120 watts export generally and up to 250 if the house has a heavy inductive load running. After all the testing and different approaches I am only marginally better than my original design. But hey it’s been fun. I now have to make some up for my friends who are all nagging me. Without this forum it would have been a lot harder.

 

[sharpener]

Many applogies I didn't read your post properly I mixed you up with another poster who is using an analog devices true rms device fed from current taps which will only deliver apparent power. Thanks for the links best explanation I've seen. However your statement is not 100% correct. If you synchronously demodulate the current waveform with the voltage zero crossings you do get true power. A simple simulation with sine waves out of phase proves this. What happens with non sine waves I havent analysed.

That's OK, no problem, at least unlike the true rms converter your synchronous demodulation will give the direction of power flow. Perhaps I should have said a linear rather than resistive load, because I agree that this technique will cope with sinusoidal current out of phase with the voltage.

But I still think it will fall down if there is any harmonic current. Another way of thinking of it is that your square demodulation waveform contains odd harmonics (1/3 3f, 1/5 5f etc) which are not actually present in the voltage waveform, so if there is any harmonic current at these frequencies it will give corresponding spurious terms in the expression for power. This is probably addressed in the paper above but I confess I have not actually looked it up.

Can you remind me which analog multiplier you are using and I will try it out as a measuring device to identify the cause of my creep.

I have sent you a PM about this as it is not the device I referred to in my original posting.

The I/P is filtered to stop fast turn on edge causing interference. I have used a standard emc filter from rs stock number 7024019.

Is this effective? I have fitted the even more expensive Schaffner multi-stage filter suggested by others, but there is still a 100Hz buzz from the unit, from the immersion heater itself and on AM radio. So I am now trying out an entirely different approach which may also work out cheaper.

 

[pmcalli]

That's OK, no problem, at least unlike the true rms converter your synchronous demodulation will give the direction of power flow. Perhaps I should have said a linear rather than resistive load, because I agree that this technique will cope with sinusoidal current out of phase with the voltage.

But I still think it will fall down if there is any harmonic current. Another way of thinking of it is that your square demodulation waveform contains odd harmonics (1/3 3f, 1/5 5f etc) which are not actually present in the voltage waveform, so if there is any harmonic current at these frequencies it will give corresponding spurious terms in the expression for power. This is probably addressed in the paper above but I confess I have not actually looked it up.

I have found the cause of my 5% error. I originally analysed the effect of crydom current shape but only the up to half power. Up to this point the demodulator gives a higher house power than actual so no problem I am not after 100%. I incorrectly assumed that this would be the peak but having simulated the full range the current shape gives a 7% under power. Limiting the integrator dc gain on a single ct implementation fixes the problem.
The multiplier arrived today and I have tried it out. Performance is as you described. I will be testing it over temperature in the next few days where I expect to have serious problems if its offset spec of 50mV is real. I should be able to compensate the circuit fairly easily. On the single ct after some analysis the scale error of the device will not matter as the peak value is not high when the currents are balanced. I too get a 100Hz buzz but no interference. Even with a multistage filter the current rise time is very high I think a different approach may be required to mitigate this effect.