Part 2: The Calibration Every Battery Warning Rests On

EdgeTX Cockpit Voice, part 2 of 9. Making a RadioMaster GX12 speak its own telemetry, so a low battery is something I hear instead of something I forgot to look at.

‹ Part 1: Why a Drone Should Talk to You · Part 3: Three Buttons, Three Colours, and the AND Gate › · Start at part 1

Every battery warning in this series is a comparison against a number. So before any of it, the number has to be true. This part is about the setting that decides whether it is, and it is the one I have most obviously failed to get right.

The whole ladder rests on a calibration you have probably skipped

I need to put this immediately after the previous section, because everything that follows depends on it and I do not want anyone building this on a bad foundation.

Your battery warnings are exactly as good as your voltage calibration.

That sounds obvious written down. It is not obvious in practice, because a miscalibrated voltage reading does not look broken. It looks like a perfectly plausible number that happens to be wrong by 200 mV, and every threshold in the ladder above inherits that error silently.

I have two aircraft that are miscalibrated right now, which means my warnings on those two fire too late. Not "slightly imprecisely" — too late, in the direction that costs you a pack. I know this and I have not fixed it yet, which is the kind of admission this blog exists for.

The knob is vbat_scale in Betaflight. It corrects the ADC divider ratio for the actual resistors on your board, which vary between boards, and it is set to a generic default that is right for nobody in particular.

The 3S-to-4S trap

The specific way this bit me is worth spelling out, because it is a natural thing to do and there is no warning.

I had aircraft set up and flying on 3S, then moved them to 4S for testing. Nothing in that transition tells you your calibration is now costing you more. But it is, for a compounding reason.

report_cell_voltage = ON means the FC divides pack voltage by its detected cell count. And that detection is itself derived from the measured pack voltage at power-on, the FC divides what it reads by a maximum-cell-voltage constant and rounds. So a voltage error propagates twice:

1. Directly, into the reported per-cell figure.
2. Potentially again, by pushing the detected cell count to the wrong integer.

That second path is the nasty one, because it fails silently and plausibly. If a badly-scaled 4S pack reads low enough that the FC decides it is looking at 3S, then it divides by three instead of four, and hands the radio a per-cell number that sits comfortably in the normal range while being completely fictional. Every threshold in my ladder would then be measuring a quantity that does not exist, and the ready self-test would happily fire, because a wrong number above 4.2 V is still a number above 4.2 V.

The self-test I was so pleased with earlier in this post checks that the signal path works. It does not check that the number is true. Those are different claims and I want to be clear about which one I have.

The regression in the new configurator

Here is the practical annoyance, and it is the reason this is getting its own post rather than a paragraph.

The way I used to calibrate was to spin the motors up to a modest load — something drawing on the order of 2 A from the pack, and then switch to the calibration tab with the motors still running, so I was calibrating at a realistic operating point rather than at idle. That matters: you want the reading trustworthy where you actually use it, under load, not just at rest on the bench.

In the current Betaflight configurator you cannot do that any more. Leaving the tab cuts the motors. The workflow is simply gone.

I have not yet worked out the right replacement procedure, so I am not going to invent one here. That is the next post: proper voltage calibration with the current configurator, what changed, and how to get a trustworthy reading under load without the old trick.

One honesty note about a number earlier in this post

The 3.065 V/cell sag figure I quote further down, from an 83 A punch-out on my 3-inch, carries this same dependency. It is what the flight controller recorded, and its accuracy rests on that aircraft's voltage calibration being sound. I have not independently verified that particular airframe's vbat_scale against a reference meter. Treat it as a strong indication of the shape of the problem rather than a metrologically clean measurement.

If you build the warning system in this post and skip the calibration, you have built something that will confidently tell you the wrong thing in a calm voice. That is arguably worse than a number in the corner of the screen.

Two of my aircraft are still telling me the truth late. I know which two, and I have not fixed it, which is the sort of thing that belongs in a lab notebook rather than a tutorial.

Series: EdgeTX Cockpit Voice, part 2 of 9. Making a RadioMaster GX12 speak its own telemetry, so a low battery is something I hear instead of something I forgot to look at.

‹ Part 1: Why a Drone Should Talk to You · Part 3: Three Buttons, Three Colours, and the AND Gate › · Start at part 1