The sum of the 4 individual circuits increases by 956 W, whereas the total consumption (measured how?) has gone up by 366 W.
The c.t’s are CT9: 50 A, CT10: 50 A, CT11: 20 A, CT12: 20 A
What sort of lights are they, how many and what power?
I’m thinking it’s a wave shape or power factor problem (or both). If you have a particularly poor power factor for the lights in general (which is likely as IIRC, they fall outside the requirements for power factor correction), and the phase error compensation isn’t quite right for your c.t’s, then it’s inevitable that you’ll see a discrepancy. The c.t’s we use now are significantly better than the old 100 A YHDC (blue) one in this respect, but they are still not perfect - no c.t. is. How close to perfection you can get depends on how deep your pockets are.
I’m measuring total consumption across the system as the sum of grid import + solar generation.
The CT’s are 50,50,20,20 yes.
I don’t think it’s the lights. Two reasons, I just plugged a kettle in a basement socket and turned it on. Same thing happens across the 4 circuits. I also have the same lighting system elsewhere attached to CT6 and turning on the lights there doesn’t affect anything else.
As a hypothesis could it be a firmware issue with DxCore? I just used the latest version which at the time was 1.5.8 there have been numerous bug fixes since then. I have just rebuilt the firmware with the latest version and am considering reuploading that into the tx4. Maybe I should use a known working version and eliminate that possibility?
Here are two sets of output from the Tx4. The first with no lights turned on, the second with all the lights turned on Ch 6 and Ch 11. Ch1 = Grid Supply Ch 5 = Solar. Hope that’s helpful.
Am I or have I been misunderstanding all along? “Same thing happens” Does this mean you see an identical effect when you repeat the same test on each of the 4 circuits in question, or does it mean all four readings respond when you do a test on any one circuit only?
This is always possible, but I would doubt it. I used (and haven’t updated) V1.5.4 to develop the library and examples, so if you want to install this, I know every input behaves correctly with the test sketches that come with emonLibDB.
When I plugged the kettle into CT10 then CT9,CT11 and CT12 all responded as well.
I changed to V1.5.4 and the issue appears to have gone away after a brief test. I will continue to monitor until tomorrow but I’m 90% sure that was the problem.
Is this reciprocal - does putting a load on CT9, CT11 & CT12 separately affect the others likewise?
(If you can’t do it physically, can you swap the c.t’s around - the calibrations will be wrong if they’re different ratings, but the percentage power will remain correct – 200 W on a 20 A c.t. will show as 500 W on a 50 A – and it’s fine to swap the c.t’s live because they have internal burdens.)
This is quite serious if it proves to be the case.
I await your report.
I have done some more testing this morning and have gone back to basics. I removed all 12 sensor cables from the Tx4 to ensure that my home wiring wasn;t in some way to blame. I then tried one at a time each of the ports CT7 through CT12 with one of the sensors and with my kettle as the load.
Here are the results:
DxCore 1.5.4
With no sensors plugged in all circuits read zero as expected.
So in summary applying a load to any circuit affects it’s neighbours in some way including CT7 and CT8. Loading CT9 seems particularly bad for some reason.
When I did the test on CT7 I checked and CT1 to CT6 were not affected.
I have now upgraded to DxCore 1.5.10 the latest current version and performed the same test on CT7 as above. It produced the same results as 1.5.4. I have also plugged all 12 CT sensors back in on DxCore v1.5.10 and it doesn’t have the exacerbated issue of 1.5.8
I hope I just have a hardware problem. Someone else maybe could replicate the test? or I can send my Tx4 back for checking.
My initial thought is your GND connection between the extender board and the emonTx4 itself is suspect, but this doesn’t explain why all the extender channels don’t show the crosstalk more-or-less equally, it’s only adjacent ones and getting less bad the further away.
I think it has to be flux around the extender socket, or something physical that’s giving you a high resistance connection to the adjacent pin, because I never noticed anything like this in testing the library, and I would assume Trystan didn’t either. And you probably won’t read it with a multimeter because it’s effectively swamped by the bias resistors - the reading of the adjacent pin is about one hundredth, and the next a tenth of that. The bias impedance is 28 kΩ or so, but an empty socket is shorted to GND, so the impedance is 28 kΩ // 10 µF or 28000 -318j Ω at 50 Hz. You’re essentially looking for a resistance of around 30 kΩ between adjacent pins.
Can you see flux under the connector? If you can, it’s worth trying to clean it away. Here’s a page about how: https://www.conro.com/how-do-i-remove-flux-after-soldering-the-pcb-board/, but I’d advise extreme care and you must make sure you don’t move it from under the connector to under the ICs or the other components.
I reflowed the GND pin and I’ve used Isopropyl alcohol to clean the board. I can’t see anything under the connector at all on the top side of the board. Unfortunately CT9 to CT12 have now stopped working completely. Is there an emoji for tearing your hair out!
I’m not sure if you’re saying this is a test I can do or not?
I’m assuming these are the the pins I’ve soldered on the main Tx4 board for CT6-CT12?
Is this with the extension board attached or not?
I presume this is not with power attached to the board?
That was the resistance I estimated existed between adjacent pins (without the extension board connected) before you cleaned the connector, so “You’re looking for…” meant “Can you see anything that would give rise to this?” It was this resistance that would give the readings you saw on the adjacent channels. It’s immaterial now (but we might need to come back to this).
Do channels 7 & 8 still work as they should?
If they do, it might be time to negotiate with Gwil with a view to getting him look at the problem, because it’s quite difficult diagnosing faults by remote control.
If not, can you measure resistance (without power applied) between GND on the extension board and GND on the main board - the c.t. jack (front-most connection) and the body of the antenna connector would be good places.
Also can you measure resistance (without power applied) between 3.3 V on the extension board and 3.3 V on the main board (the header next to the RJ45 socket is a good place for the 3.3 V).
I suspect you could have lost the GND or the 3.3 V connection (or both) between the extension and the main board, either or both would stop all 6 inputs working properly.
