Calibrating individual circuits before permanent install?

[Overwatching]

After a bit of initial confusion, I have the device flashed and I can get readings.

I currently have the device set up on a single phase, with the Phase A clamp and all 16 circuit clamps checking the same circuit.
This circuit has a resistive heater fan attached to it, and it's all going through a Kill-A-Watt which I have measured to be off by +/- 0.25V.
I've done this to see how close the values for each circuit sensor are.

And they are quite a bit farther off than I was expecting.

Kill-A-Watt and Phase A clamp agree on 1375W.
But the individual circuits range from 1316-1421W.

Each [sensor+plug] is consistent though, so it seems like I could calibrate this problem away.

What is the proper way to do this?

Should I be putting a multiplier in each line of ct_clamps?
ex:
- { phase_id: phase_a, input: "5", power: { name: "Circuit 5 Power", id: cir5, filters: [ *moving_avg, *pos, multiply: 1.0845 ] } }

[flaviut]

Should I be putting a multiplier in each line of ct_clamps?

Yes, exactly.

You might find it easier to use the calibrate_linear filter to save yourself some manual math:

 - { phase_id: phase_a, input:  "5", power: { name:  "Circuit 5 Power", id:  cir5, filters: [ *moving_avg, *pos, calibrate_linear: ["0 -> 0", "1316 -> 1375"] ] } }

[Overwatching]

Changed

• calibrate_linear: {0: 0, 1405: 1376}
  to
• calibrate_linear: [0 -> 0, 1405 -> 1376]

and it worked.

[Overwatching]

This heating fan has a 1/2 setting too.

I wonder if I'm better off using calibrate_polynomial since I can get 0, half , and full power measurements.

[flaviut]

I believe the physical process here is linear, and using calibrate_polynomial will result in overfitting. You can pass 3+ values to calibrate_linear, and the system will perform a line of best fit on them.

[Overwatching]

perfect, I'll try that.

[Overwatching]

Note:

The energy values do not seem to be stored in RAM.
I left the device unplugged for 25+ minutes while I ate, and it still has the same values when it started back up.

Which leaves me with questions.
How much is getting written to the flash, and how often?
How quickly is this going to wear out the flash?

I guess I'll look through the code and see how much I understand.
Then I'll post a new discussion.