Diagnosing Battery Sag in ArduPilot Flight Logs

TL;DR: Battery sag in ArduPilot logs is the gap between BAT.Volt (measured voltage under load) and BAT.VoltR (resting voltage estimate). A small gap during high-current segments is normal; a large gap that doesn’t recover when you reduce throttle indicates worn cells or an oversized current draw. Calculate per-cell sag by dividing the gap by BATT_VOLT_NUM_CELLS — under 0.1 V per cell per amp drawn is healthy; over 0.3 V per cell signals an aging pack on the way out.

What "sag" actually means in lithium chemistry

A lithium-polymer cell’s terminal voltage drops below its open-circuit voltage when it’s delivering current. The drop is proportional to the cell’s internal resistance times the current. A fresh pack with healthy cells has low internal resistance (5–15 mΩ per cell typical) and small sag; a worn pack has higher resistance and larger sag for the same draw.

ArduPilot’s battery monitor computes the resting voltage in real time using a model of cell impedance — effectively asking "what voltage would this pack read if no current were flowing?" BAT.VoltR is that estimate; BAT.Volt is what the ADC actually measures.

The fields that matter

BAT  TimeUS  Inst  Volt  VoltR  Curr  CurrTot  EnrgTot  Temp  Res  RemPct  H  SH

• Volt — measured voltage under current load.
• VoltR — resting voltage estimate.
• Curr — current in amps.
• Res — estimated whole-pack internal resistance in ohms. Rising means aging cells.
• RemPct — remaining capacity percentage. Computed from CurrTot against BATT_CAPACITY.
• Temp — pack temperature, where available, scaled by 100 (e.g. 23.4°C is logged as 2340).
• SH — state-of-health percentage. 0 if not estimable.

Reading sag in Mission Planner

Plot BAT.Volt, BAT.VoltR, and BAT.Curr on one chart:

• The Volt trace dips below VoltR when current rises. The gap is sag.
• During hover at constant current, the gap should be steady. Step changes in current produce step changes in sag.
• If Volt recovers to within ~0.1 V of VoltR when throttle drops, the pack is fundamentally healthy. If Volt stays well below VoltR for extended periods, the cells are worn or the current model is wrong.

Per-cell math — the rule of thumb

Divide the gap (VoltR − Volt) by the cell count to get per-cell sag. Per-cell sag thresholds at the current draw at the time of measurement:

• Under 0.05 V per cell at low current — excellent. Fresh pack or well-sized for the load.
• 0.05–0.15 V per cell during sustained current — normal for healthy packs in the middle of their life.
• 0.15–0.30 V per cell at design current — pack is aging; capacity will already be reduced from new.
• Over 0.30 V per cell — pack is approaching the end of safe life; expect failsafe trips and reduced peak performance. Replace soon.

These thresholds are normalised against the current draw. A 6S pack reading 0.5 V of sag at 50 A is similar per-cell health to one reading 0.25 V at 25 A — both are around 0.08 V/cell normalised.

Sag versus depletion

Two related but different failure modes:

• Transient sag — voltage drops momentarily under high current and recovers when throttle reduces. The pack still has capacity; you just hit its current-delivery limit. Fix is more pack (parallel cells, higher C-rating) or a less demanding flight envelope.
• Depletion — both Volt and VoltR drop together as the flight progresses. The pack is genuinely running out of energy; this is normal at the end of a flight and is what failsafe is for.

The diagnosis matters because the fixes differ: transient sag needs a better-rated pack; depletion needs longer flight time, smaller missions, or a less demanding payload.

Pack resistance as a fleet maintenance signal

BAT.Res trends across flights tell you which packs are aging:

• Capture BAT.Res from steady-hover segments (not transients).
• For a fresh 6S 22000 mAh pack, expect whole-pack resistance around 15–30 mΩ.
• Twenty percent above the new value is a warning; fifty percent above is replacement territory.
• The trend matters more than the absolute number — different chemistries and current sensors give slightly different estimates.

For an operator with fleet logs, plotting BAT.Res as a per-pack time series identifies aging packs before they cause an in-flight failsafe.

Confirming it in MAVExplorer

MAV> graph BAT.Volt BAT.VoltR BAT.Curr
MAV> graph BAT.Res
MAV> param show BATT_VOLT_NUM_CELLS BATT_CAPACITY BATT_LOW_VOLT BATT_LOW_TIMER

The parameter dump tells you what thresholds the autopilot was configured to react at. If failsafe fired at a voltage you didn’t expect, the parameter snapshot is the first thing to verify.

Fixing the underlying problem

1. Use a higher C-rated pack if sag is high but cells are fresh. The pack’s peak current delivery limits the autopilot’s authority.
2. Replace aging packs if BAT.Res has climbed significantly across a few months of use.
3. Tune failsafes for the worst-case sag. BATT_LOW_VOLT should be set against the pack’s sag profile, not its resting voltage. A pack that sags 0.5 V at peak current needs failsafe thresholds that account for the dip.
4. Reduce peak current draw — flatter throttle envelope, lighter payload, smoother manoeuvre patterns. The autopilot’s rate-loop tuning influences peak motor current; an over-tuned rate loop produces unnecessary sag.
5. Check sensor calibration. If BAT.Curr reports significantly different values than a clip-on ammeter, the current sensor is mis-calibrated and the sag calculations are wrong.

When this isn’t actually a sag problem

• Power module fault. If BAT.Volt drops sharply but the airframe is still performing normally, the measurement might be wrong — check POWR.Vcc and see our power module post.
• Connector corrosion. Intermittent voltage drops at random times that don’t correlate with current draw point at a loose XT60 or worn bullet connectors, not a pack problem.
• Cold weather. Lithium cells deliver less current at low temperatures and read lower voltages. Warm packs to 15°C+ before flight in cold conditions.

When LogHat helps — and when it doesn’t

LogHat plots BAT.Volt versus BAT.VoltR with the per-cell sag computed, and trends BAT.Res across consecutive flights of the same airframe so a degrading pack is visible before it triggers failsafe. What we can’t do is tell you which pack of the three you charged this morning was actually loaded on the drone — that needs an operator-side pack-tracking process.