Looking at the pv data below for 11:38:42 the solar voltage on string 1 (273.5) multiplied by the string 1 current (1.9) is 519.65, yet the portal shows that as 543. And similar with string 2. I have looked at other days and times and find similar . Why the difference ? I 
Maxwell I have similar
Taken from Inverter Data spreadsheet downloaded.
30/06/2024 12:10 NORMAL 232.9 232.6 1.5 1.4 359 347 706
30/06/2024 12:05 NORMAL 234.3 232.5 1.9 1.8 446 441 887
30/06/2024 12:00 NORMAL 234 230.7 1.5 1.5 361 357 718
30/06/2024 11:55 NORMAL 234.5 232.6 1.5 1.4 352 346 698
30/06/2024 11:50 NORMAL 234.3 230.9 1.6 1.6 386 374 760
30/06/2024 11:45 NORMAL 231.5 229.5 1.9 1.9 452 452 904
30/06/2024 11:39 NORMAL 237.1 229.7 2 2.1 498 496 994
30/06/2024 11:34 NORMAL 231.7 232.7 2.2 2.2 521 517 1038
30/06/2024 11:29 NORMAL 234.4 238.3 3.1 3 742 732 1474
Same issue 12:00 234 x 1.5 = 351 230.7 x 1.5 = 346
I wonder if this is sometihing to do with the difference I reported in thread
"PV Generation Day Total - Report vs Inverter Data different" ?
I realise now the nice mothly bar charts you can get from the Energy Graph tile on web portal use a number that is NOT the same as the "PV Generation Today" column (at just before midnight) in downloaded spreadsheet.
I do not know where the number in barchart comes from !
From bar chart 30/6 14.97 from spreadsheet 14.8
From bar chart 29/6 35.41 from spreadsheet 34.3
From bar chart 25/6 36.37 from spreadsheet 34.0
I have been using the smaller number for reporting to pvoutput dot org
Maxwell I'm assuming there's some rounding going on as its only to 1 decimal place. That gets you a lot nearer. The other reason may well be a slight delay between each value being read. The feedback loop in MPPT inside the inverter will be very rapid at adjusting PV voltage up and down but I doubt the values will be constantly streamed back to the registers in real-time so there could well be lagging/stale data between a Voltage, a Current and a Power calc as there is a lot going on in there. Even if its just a second or so.
If you have a day of a perfect clear blue skyell do the values get closer together in accuracy?
hoggy I keep seeing reference to registers, what exactly are they?
Robgy basically a little box with a value in it inside it's brain. It's where the data is stored in the inverter as to what it's up to or target it's aiming for.
Things like the portal, GivTCP and so on can read these.
A bit like a spreadsheet I guess would be a way to describe:
Box A1 holds a value of 100
Box C2 holds a value of 2000
If you know what the "box ID" purpose is & also exactly how that data value is stored (i.e. is it in mV or V or a % - all from the internal documentation) you can then read that info and combine both together and output that info such that you now know:
Box A1 is for PV String 1 Voltage, measured in Volts, so now we know it's seeing 100V on PV String 1
Box C2 is String Power, measured in Watts and therefore it's producing 2,000W.
hoggy I only installed home assistant last week so very new to all this, are you saying that the strings with givTCP..xyz... are the register values?
Robgy Pretty much - in HA they are more of a "friendly name" for each register in the inverter as you can change them if you want to something less techy if you need to, although most of them broadly mirror the documentation and follow the register naming word for word.
There are hundreds of them - HA doesn't show them all and some are quite dangerous to mess with so there is also some protection built in to make sure you can't overwrite something that would cause harm to your system.
(There are a couple in GivTCP that are calculated from a bunch of registers together but not many, most are the direct output of the inverter)
hoggy thank you for the explanation, it would appear I've been reading registers for the last few days without realising it.🤪
hoggy Not sure that rounding plays much in changing these. So, the first image shows the portal snapshot and the second picture (excel sheet) shows that even rounding up to the nearest whole number (no decimal places) the ge reported v & i that the differences remain between calculated and what ge show.
Now, i must confess i don't really understand the nuances of how those mppts work. So, if rounding doesn't explain these differences then i can only assume that the measurements of v, i are done independently which introduces these errors. Or is some "feature" in their coding which is then introducing some timing error/bias which mis-reports the p for each measured/stored v & i ?
I meant more that the system is already rounding the raw data up/down but calculating the power on the pre-rounded values.
Take the 14:56 timestamp
239.2V | 3.1A | 749W (741W If V x I)
But if the current had been rounded down to 1DP and was higher originally:
239.2V | 3.132A | 749W (matches portal)
I've not done all of them just a few but does that theory hold at all if you play around with it?
hoggy So, i found a day that looks like it had good solid solar, first image
And i looked again at the ratio of string 1 power to string 2. This is an 8 & 6 panel split. So, i'd expect about .7. I worked that out between 10:00 - 20:00 figures outisde of these times can get misleading. And those are lower than i'd expect.
And you can see similar differences in calculated vs shown power figures
And for comparison. I found a day just over a year ago where the solar profile is similar
And the corresponding p1/p2 graph which shows values that i was expecting given the string split. And to my knowledge we don't have nextdoor with a big roof extension or big change in plant growth between snapshots. So, don't really see the first case being shading related. So, either faulty panel or inverter issue. How to work out which it is ?
I'm sorry but I'm no longer following, am I right in thinking we are now discussing two things:
The first is / was - the discrepancy between Volts, Amps & Power not following normal convention (V x I = Power) which I believe to be a decimal place issue.
The second is now the difference between two strings performance relative to one another (with panel numbers taken into account)