Interaction between CT sensors

[arnoldh (Arnold)]

My question is whether two CT sensors monitoring current on the same cable will affect each other.

I’ve a Microchip PIC based electricity monitor that’s been running for 2 years. Now I want to add gas meter monitoring. I’ll probably have to redesign the PCB, and certainly make firmware changes, so I’ll have to do extensive testing of the upgraded monitor side by side with the existing one. The accuracy on the existing monitor is good, about 0.1% on the PV line, and I don’t want to lose any accuracy on the ongoing measurements while testing the new monitor.

There are 4 sensors, three CT sensors on the grid, PV and immersion heater circuits, and one AC voltage. It would be easiest to connect two separate sets of sensors to the two monitors. However the meter tails for each circuit are short and each pair of sensors would necessarily touch each other. I suspect each transformer would disrupt the measurements of its adjacent one. Can anyone confirm this?

An alternative is to use only one set of sensors and split the output from each sensor into two, one to the existing monitor, one to the one under test. The sensor lines go to four ADC inputs on the PIC microcontroller. So each sensor line will be connected to an ADC input on two PICs, one on the existing monitor, one on the test monitor. I think this should work OK. Does anyone see any problems with it?

If you think there may be problems with both these approaches, do you have any other suggestions?

[Robert.Wall]

I think there may well be interaction between the two c.t’s, but as they’re both measuring the same current, it should (only) be a question of calibration - which of course is something you want to avoid. The c.t. will be susceptible to external fields, how much so will depend on its construction, so I’d expect a toroidal wound ring-core to be far less affected than a bobbin-wound split-core one (you don’t say which one you’re using). There are numbers for the effect of an adjacent conductor in the tests for the SCT-013-000 (Issue 6) - it indicated 160 mA with 100 A touching. Those numbers represent the worst-case position for the interfering conductor, which might give you some indication.

I’ll try to figure out a way of checking your situation. It’s going to be tricky because the stability of the public electricity supply is likely to be a lot worse than the difference I expect to see.

@arnoldh

An alternative is to use only one set of sensors and split the output from each sensor into two

Without seeing the circuit diagram of the front end of your PIC, I wouldn’t like to say what, if any, interaction there would be. If it followed the configuration of the emonTx, you could share the c.t. between the two. Clearly you’d have only one bias network and one burden resistor active, and you’d share the burden voltage together with its bias offset with the (presumably new) second monitor. When you’d concluded the test period, you’d need to physically move the burden resistor to the new monitor to preserve its calibration (because the burden is in the calibration chain).

The obvious way to get around this is to keep the gas metering separate! You don’t say what happens after you’ve got the raw numbers, so I can’t guess. But presumably, you’ve rejected that approach for a reason.

[arnoldh (Arnold)]

@Robert.Wall

Without seeing the circuit diagram of the front end of your PIC

This is the schematic from Eagle

The three CT inputs are on the left. The voltages across the burden resistors are between the mid point of the 3V ADC range and the sensor inputs.

However I hadn’t thought about the burden resistors properly. Only using one would mean I couldn’t leave both circuits intact while I’m testing, and calibration would be a mess. This approach will just not work.

@Robert.Wall

The obvious way to get around this is to keep the gas metering separate! You don’t say what happens after you’ve got the raw numbers, so I can’t guess

The measurements are saved in flash memory until the PC comes on line and requests them, when they’re sent wirelessly to the PC. The PC reformats and saves them to a database. I don’t want two separate modules talking to the PC side.

Seems I’ll have to go for the first solution. I’m using SCT-013s. I’ll look at the test you mentioned.

[Robert.Wall]

I have figured out a way to test that, involving 3 c.t’s. Here’s what I did:
I used the test rig that I use for all c.t. tests. I put all 3 c.t’s on the 20-turn coil, with ~ 50 A test current (2.5 A × 20). C.T.No.7 (from the report, if you want to check its properties) was the reference. The test c.t’s were nos. 8 & 9. I was using the STM32 & emonTx Shield, with dBC’s “Shield6” sketch (from the STM thread), simply because that was set up. I let it run for 10 measurements, dropped the numbers into a spreadsheet and worked out averages and the ratios between them. The numbers:

C.T’s touching:
C.T. 8 with respect to C.T. 7: 1.0020811327
C.T. 9 with respect to C.T. 7: 1.0022962249

C.T’s 50 mm apart:
C.T. 8 with respect to C.T. 7: 0.9997844494
C.T. 9 with respect to C.T. 7: 1.0004381468

The ratio touching : apart
C.T. 8 1.0022971784
C.T. 9 1.0018572643

Then I did it again but with C.T. 8 & C.T. 9 swapped places.

C.T’s touching:
C.T. 8 with respect to C.T. 7: 1.0015706239
C.T. 9 with respect to C.T. 7: 1.0024897279

C.T’s 50 mm apart:
C.T. 8 with respect to C.T. 7: 0.9994110972
C.T. 9 with respect to C.T. 7: 1.0004914282

The ratio touching : apart
C.T. 8 1.0021607993
C.T. 9 1.0019973182

So of the two, C.T. 8 seems to be more sensitive to its neighbour than its neighbour is to it.

The effect is about a 0.2% increase in the indicated current when the C.T’s are touching. That’s significant to you because it is twice the inaccuracy you see against your tariff meter.
I’m not sure how far we can trust those numbers, the raw numbers for current are around 1000 with two decimal places, therefore the rightmost 5, possibly 6, digits should probably be discarded. (It’s not calibrated in engineering units - the value is in ADC units.)

@arnoldh

However I hadn’t thought about the burden resistors properly.

Sorry about that

@arnoldh

Only using one would mean I couldn’t leave both circuits intact while I’m testing, and calibration would be a mess. This approach will just not work.

Can’t it be made to work? You build Serial No.2 without burdens R4-R6 and without R3 & R7. Link J1-2, J1-3, J1-4, J1-5 & GND to the same places on S/No.1, then do what you need by way of testing. When you’re happy, you lift R3 - R7 off S/No.1 and solder them in S/No.2.

You could have temporary resistors on a spare plug for J1, along with the connection to VMID.
I’m assuming the software removes the bias offset.

[arnoldh (Arnold)]

@Robert.Wall

The effect is about a 0.2% increase in the indicated current when the C.T’s are touching. That’s significant to you because it is twice the inaccuracy you see against your tariff meter.

These numbers are reassuring. I didn’t know whether I could expect 10%, 20% or whatever. I can do much of the testing later in the day when there’s no PV generation. The grid measurement accuracy is around 1%, so that’s less of a problem.

@Robert.Wall

Only using one would mean I couldn’t leave both circuits intact while I’m testing, and calibration would be a mess. This approach will just not work.

Can’t it be made to work? You build Serial No.2 without burdens R4-R6 and without R3 & R7. Link J1-2, J1-3, J1-4, J1-5 & GND to the same places on S/No.1, then do what you need by way of testing. When you’re happy, you lift R3 - R7 off S/No.1 and solder them in S/No.2.

You could have temporary resistors on a spare plug for J1, along with the connection to VMID.

Yes, you’re correct, I could do it this way if I had to.

@Robert.Wall

I’m assuming the software removes the bias offset.

Yes, it does

Thank you very much for doing this analysis. It gives me the confidence to do the testing with two complete systems side by side - always my preferred option.