ProjekinAja. Embedded Systems Stage 4 · Power & Wireless Lesson 4.1
● Stage 4 · Lesson 4.1

Batteries: Chemistry, Capacity, and the Lies of mAh

1000 mAh is a capacity at a specific current, temperature, and cutoff. Change any of them and the number changes.

⏱ ~24 min read 🔋 Power system 📐 Worked numbers

🎯 By the end you'll understand

Why a battery's printed capacity is a claim made under conditions you do not control, and how to convert it into the number you actually own: usable capacity above your regulator's minimum input, at your temperature, under your pulse load. You'll be able to read a discharge curve, compute voltage sag from internal resistance, explain why a 40 mA BLE burst is harmless on a LiPo and fatal on a coin cell, and justify Beacon's 1000 mAh LiPo over a CR2477 with numbers rather than instinct. You'll also know the four things that make lithium cells dangerous and which part on Beacon's board guards against each.

KEY INSIGHTCapacity is a measurement, not a property

The tag on the bench ran nine days on a "1000 mAh" coin cell. The firmware was fine. The battery was fine. The number on the battery was the problem — it was measured at 0.2 mA into a 2.0 V cutoff, and the tag drew 40 mA into a 3.0 V regulator.

A battery does not have a capacity the way a resistor has a resistance. It has a discharge curve — voltage versus charge removed — and that curve bends differently depending on how fast you pull charge out, how cold the cell is, and how old it is. The manufacturer picks a gentle current, a room temperature, and a generous end voltage, integrates the curve, and prints the result. Your product picks none of those. So the first job of a power-system designer is not to read the number on the cell; it is to re-derive it under the conditions the product will actually impose.

🔑 Hold onto this

Usable capacity is the charge you can remove while the terminal voltage stays above your system's minimum input, under your worst-case pulse. Three things eat into the nominal number: the cutoff (charge below your minimum voltage is stranded), temperature (cold raises resistance and slows chemistry), and internal resistance (every pulse subtracts I × R from the voltage the regulator sees). A "1000 mAh" that becomes 600 usable is not a defective cell. It is the normal outcome of reading the datasheet number as if it were yours.

CORE CONCEPTThe discharge curve, and the window you actually get

Start with the one picture every battery decision comes back to. A lithium-ion or lithium-polymer cell (LiPo is a packaging variant of the same chemistry, not a different one) sits at about 4.2 V fully charged, spends most of its life sloping gently through 3.7 V, its nominal voltage, and then falls off a cliff somewhere around 3.0 V, which is where the cell is considered empty. Below roughly 2.5 V you begin doing permanent damage to the anode's copper current collector, which is why every protection IC has an under-voltage trip in that region.

4.2 3.7 3.3 3.0 2.5 2.0 cell voltage (V) charge removed (% of nominal) 0 50 100 3.3 V — LDO needs ≥ this + dropout 3.0 V — ESP32-C3 module minimum LiPo · 3.7 V nominal LiFePO4 · 3.2 V nominal (flat) CR2477 · 3.0 V nominal (Li-MnO₂) stranded Room temperature, gentle current. Shapes are representative; use your cell's datasheet curve for numbers.
Three discharge curves and two horizontal lines that matter. A 3.3 V LDO strands everything below about 3.4 V (the LiPo's last ~5 %). A LiFePO4 cell sits above 3.3 V for only the first few percent of its discharge — its plateau is at 3.2 V — so a 3.3 V LDO strands almost the entire cell. A CR2477 starts below the ESP32-C3 module's 3.0 V minimum before you draw any current.

Now draw your system's minimum input voltage on that curve. Everything below the line is stranded: charge that is physically in the cell and that you will never use. With the XC6220 LDO on Beacon's breadboard, the minimum input is 3.3 V plus dropout — call it 3.4 V at Beacon's currents — and the LiPo curve crosses that at roughly 95 % discharged. Five percent gone before you have done anything. With the TPS62840 buck on the PCB, the regulator keeps working down to a 1.8 V input, so the cutoff moves to the cell's own 3.0 V floor and you recover most of that. This is the first, and least appreciated, reason the buck wins in Lesson 4.2: not efficiency, but reach.

📖 C-rate, and why the datasheet number is measured at 0.2C

Discharge rate is expressed as a multiple of capacity: 1C on a 1000 mAh cell is 1000 mA (empties in one hour); 0.2C is 200 mA (five hours). Datasheet capacity is nearly always the 0.2C figure, sometimes lower for coin cells (a CR2477 is rated at a standard drain of a fraction of a milliamp). At higher rates you get less, because the I × R drop reaches the cutoff earlier and because the chemistry cannot move ions fast enough near the end. For Beacon this cuts the other way: an average of 38 µA is 0.00004C, so the LiPo delivers at least its rated capacity — the losses come from elsewhere.

COMPARISONFour chemistries you will actually choose between

You will meet many battery chemistries in catalogues; in a small battery-powered product you will realistically pick from four families. The numbers below are datasheet-class values for typical cells — ranges, because vendors differ, and because you must read the sheet for the cell you buy.

ChemistryNominal / full / emptyCurve shapeInternal resistanceSelf-dischargeRechargeable?Best for
Li-ion / LiPo (LiCoO₂, NMC)3.7 V / 4.2 V / ~3.0 VGentle slope, cliff at the endTens of mΩ (small pouch ~50–150 mΩ)A few % per monthYes, ~300–500 cycles to 80 %Anything with pulses; anything rechargeable
LiFePO4 (LFP)3.2 V / 3.6 V / ~2.5 VVery flat, then a cliffTens of mΩA few % per monthYes, 2000+ cyclesLong-life, thermally safer; can feed a 3.3 V-tolerant MCU directly
Primary lithium (Li-SOCl₂, e.g. ER14505 AA)3.6 V / 3.65 V / ~2.0 VExtremely flat for yearsTens of ohms; rises with passivation~1 % per yearNoDecade-life meters at microamps; needs a capacitor for pulses
Coin cells (Li-MnO₂, CR2032 / CR2477)3.0 V / ~3.2 V / 2.0 VSlow decline, cliff at the end~10–40 Ω fresh, rising as it depletes~1–2 % per yearNoSub-milliamp pulses; disposable tags

Three things in that table decide most designs. First, energy density: Li-ion at roughly 150–250 Wh/kg versus LiFePO4 around 90–160 Wh/kg means the same energy in an LFP cell is a bigger, heavier cell — the price of a safer, flatter, longer-lived chemistry. Second, the nominal voltage decides your regulator. A 3.2 V LFP cell cannot feed a 3.3 V LDO; but an ESP32-C3 accepts 3.0–3.6 V, so some designs run the MCU straight from an LFP cell with no regulator at all, accepting that the last few percent (below 3.0 V) are unusable and that the cell's 3.6 V charge peak is right at the edge of the module's absolute maximum. Third, internal resistance, which is the difference between a coin cell that works on paper and one that works on the bench. That deserves its own section.

💡 Temperature changes all three numbers at once

Cold slows the ion transport that makes a cell work. At 0 °C a typical Li-ion cell delivers something like 80 % of its room-temperature capacity, and its internal resistance roughly doubles; coin cells are worse on both counts. Heat does the opposite in the short term (slightly more capacity, lower resistance) and the opposite of what you want in the long term (faster aging, and for Li-ion, faster self-discharge). Beacon's R9 says 0–45 °C indoors, which is benign — but the "≥ 6 months at 22 °C" in R1 is deliberately conditioned on temperature because the number would not survive a cold warehouse.

Cycle life is the number that rechargeable-cell buyers under-weight. A LiPo that keeps 80 % of its capacity after 300–500 full cycles sounds fine until you do the arithmetic for a product charged every week for three years: 150 cycles, so the cell ends its life at perhaps 90 % of new. For Beacon, charged twice a year, cycle life is irrelevant and calendar aging dominates — a LiPo stored at 4.2 V and 40 °C loses capacity noticeably faster than one stored at 3.8 V and 20 °C. If you want to be kind to the cells in a pilot fleet that sits in a drawer, ship them half charged.

HOW IT WORKSInternal resistance: the number that kills coin-cell designs

Every real cell behaves, to first order, like an ideal voltage source with a resistor in series. That resistor is the internal resistance, Rint, and it has two consequences. It wastes power as heat (I² × R, negligible at Beacon's currents), and — the one that matters — it drops the terminal voltage by I × Rint whenever current flows. The regulator, and everything behind it, sees the terminal voltage, not the open-circuit voltage.

Model OCV R_int terminal (+) terminal (−) V_term = OCV − I × R_int Terminal voltage during a 45 ms, 40 mA burst LiPo, 80 mΩ: sag 3 mV CR2477, 15 Ω: sag 600 mV 3.0 V minimum → brown-out pulse on
Same 40 mA pulse, two cells. The LiPo does not notice. The coin cell's terminal voltage drops by 0.6 V — from below the module's minimum to far below it — for the entire burst. The MCU resets, the reset restarts the burst, and the tag spends its life in a reboot loop that looks, from the outside, like "flaky firmware".

Run the numbers for Beacon's active window: 40 mA average for 45 ms.

1

LiPo, 1000 mAh pouch, Rint ≈ 80 mΩ

Sag = 0.040 A × 0.08 Ω = 3.2 mV. Invisible. Even at 0 °C with resistance doubled it is 6 mV. A LiPo could source this pulse ten times over without the regulator noticing; the pulse is simply not a design constraint.

2

CR2477 coin cell, fresh, Rint ≈ 15 Ω

Sag = 0.040 A × 15 Ω = 0.60 V. Open-circuit 3.0 V becomes 2.4 V at the terminals — already below the ESP32-C3 module's 3.0 V minimum before we account for the regulator. And "fresh" is the best case: as a coin cell depletes, its resistance climbs into the tens of ohms, so the sag gets worse exactly when the open-circuit voltage is also lower.

3

The capacitor "fix", sized honestly

The usual advice is to put a bulk capacitor across the coin cell so it supplies the pulse. Charge per burst is Q = 0.040 A × 0.045 s = 1.8 mC. To supply that from a capacitor while letting it droop only 0.3 V, you need C = Q / ΔV = 1.8 mC / 0.3 V = 6 mF. That is not a ceramic; it is a supercapacitor or a hybrid-layer capacitor, with its own leakage current, cost, and board area. It can be made to work — plenty of coin-cell BLE tags do — but only by shrinking the burst (a few packets, no sensor heater, no LED) and accepting a large, leaky part. The honest lesson is that the coin cell's chemistry fixes the pulse budget, and Beacon's 45 ms at 40 mA is far outside it.

🧪 First-order capacitor model you can actually use

With a capacitor C in parallel with a cell of resistance R, the pulse divides between them with time constant τ = R × C. For a pulse of duration t, the cell's share of the current is roughly t / (t + τ). CR2477 (15 Ω) with a 1000 µF ceramic bank: τ = 15 ms, so the cell still carries 45 / 60 = 75 % of a 45 ms pulse — the capacitor barely helps. With a 10 mF hybrid-layer capacitor: τ = 150 ms, the cell carries 23 %, sag falls to about 140 mV. The widget below uses this model; it is crude, but it is the right crude.

SEE ITDischarge explorer: turn a nominal mAh into a usable one

Pick a chemistry, then impose Beacon's conditions on it: the average current, the pulse current, the temperature, and the minimum voltage your regulator needs. The curve redraws with the open-circuit voltage (solid) and the voltage during a pulse (dashed). Usable capacity is everything to the left of where the dashed line crosses your cutoff. Try the CR2477 first, because that is the design everyone wants to build.

⚙ Interactive · Discharge explorer

Deterministic first-order model: piecewise open-circuit curve per chemistry, resistance rising with depth-of-discharge and cold, capacity derated by cold. Shapes and resistances are representative datasheet-class values, not a specific cell's sheet.

LiPo · 1000 mAh
38 µA
40 mA
22 °C
3.00 V
none
charge removed (mAh) 0 500 1000 cutoff 3.00 V — open-circuit voltage - - voltage during pulse
Usable capacity
—
Of nominal
—
Pulse sag (fresh → 80 % used)
—
Life at average current
—

—

—

Things to notice. On the LiPo, the dashed line sits on top of the solid one — the pulse is irrelevant, and what limits you is the cutoff: raise it to 3.4 V (an LDO) and watch the usable fraction drop by about a tenth. On the CR2477, the dashed line starts below a 3.0 V cutoff at zero charge removed: the tag brown-outs on its first transmit, and no amount of capacity fixes that. Add the 10 mF capacitor and the sag shrinks to about 140 mV — and it still fails, because a cell that starts at 3.05 V cannot stay above a 3.0 V floor under any load at all. Drop the cutoff to 2.8 V (a buck-boost, or a chip that runs from a 1.8 V rail) and the coin cell finally gets a usable window at room temperature — a small one, on the order of 120 mAh; then drag the temperature to 0 °C and watch it halve. That is the whole argument of this lesson in one slider.

IN PRACTICESafety: the four ways lithium goes wrong, and the part that stops each

Lithium cells store a lot of energy in a flammable electrolyte next to a thin separator. They are safe in ordinary use because a handful of limits are respected, every time, by hardware that does not depend on your firmware being correct. Here are the four failure modes and Beacon's answer to each.

Failure modeWhat happens physicallyGuard on Beacon
Over-charge (above ~4.25 V)Lithium plates on the anode as metal; the cathode releases oxygen; the cell heats and can vent or ignite.MCP73831 regulates to 4.20 V; DW01A trips over-charge independently at about 4.25 V and opens the charge FET in the FS8205 pair.
Over-discharge (below ~2.5 V)The copper current collector dissolves; on the next charge, copper dendrites can short the separator.DW01A trips under-voltage around 2.4 V and opens the discharge FET; firmware treats ~3.0 V as "empty" long before that.
Over-current / shortHeat in the tabs and separator; a full short on a small LiPo can reach tens of amps.DW01A senses the drop across the FS8205 FETs and trips at a few amps; the reverse-battery PMOS (2.6) prevents a backwards-plugged pack from doing it in the first place.
Charging in the cold (below 0 °C) or hotBelow 0 °C lithium plates instead of intercalating — invisible damage that raises resistance and short risk. Above ~45 °C aging accelerates.The MCP73831 has no thermistor input; Beacon's indoor 0–45 °C environment (R9) and USB charging in an office are the mitigation. A product charged outdoors would need the MCP73833 or a separate NTC check in firmware.

Two practical notes. Swelling is a LiPo's way of telling you it has been over-charged, over-heated, or is simply old; a puffed pouch has generated gas inside and must be retired, not squeezed back into the enclosure. And the connector is not standardized: "JST-PH 2.0" describes the housing, not which pin is positive. Packs from different vendors arrive with the red wire on either pin. Beacon's reverse-battery PMOS from Lesson 2.6 exists precisely because at 25 units someone will plug one in backwards, and a PMOS costs less than a board.

⚠ "The pack already has a protection board"

Most LiPo pouches ship with a small protection circuit module (PCM) — usually exactly a DW01A-class IC with a dual MOSFET — soldered to the tabs. Beacon puts its own on the PCB anyway, for two reasons: you cannot see or verify the vendor's PCM, and pilot units will be built with whatever pack was in stock that week. Two protection circuits in series are harmless. Budget their combined quiescent draw (a microamp or so) in Lesson 4.3.

BEACONBeacon's cell: a 1000 mAh LiPo, and why not a CR2477

Requirement R1 says six months at one report per minute at 22 °C, on a 1000 mAh LiPo. Now that you can read a discharge curve, the choice can be defended rather than asserted. The candidate everyone proposes first is a CR2477: it is also "1000 mAh", it needs no charger, no protection IC, no USB-C, and it would delete R6 from the requirements table entirely.

Question1000 mAh LiPoCR2477 (1000 mAh)
Open-circuit voltage vs the module's 3.0 V minimum4.2 → 3.0 V; the whole curve is above the floor3.0 V fresh; the whole curve is at or below the floor
Sag on the 40 mA, 45 ms burst≈ 3 mV at 80 mΩ≈ 0.6 V at 15 Ω, worse when depleted or cold
Capacitor needed to survive the burstNone beyond normal decoupling≈ 6 mF for a 0.3 V droop — a supercap with its own leakage
Average current 38 µA — is the cell happy?Yes (0.00004C)Yes; the average is fine, only the pulse is not
Paper life at 38 µA1000 mAh / 38 µA ≈ 26,000 h ≈ 3 yearsUndefined — brown-out on the first transmit
Cost and partsCell + MCP73831 + DW01A + FS8205 + USB-C + TVSCell + holder
Field servicingPlug in USB-C twice a yearSwap the cell (if it lasted)

The coin cell loses on physics, not on preference. Could it be rescued? Yes, by redesigning the product around it: a radio that draws 10–15 mA rather than 40, a burst of one packet rather than three, no SHT40 high-precision mode, the LED deleted, a hybrid-layer capacitor, and a firmware that measures the terminal voltage during a burst rather than at rest. That is a different product with a different requirements table. Beacon's answer to the pulse problem is to pick the cell that does not have one, and to spend the saved engineering on the charger and protection circuit in Lesson 4.4.

One more consequence you can now compute. Under the LDO on the breadboard, the usable window is 4.2 V down to about 3.4 V — the widget puts that at roughly 95 % of nominal. Under the buck on the PCB, it is 4.2 V down to 3.0 V — about 99 %. On a 1000 mAh cell that is roughly 40 mAh recovered, which at 38 µA is five or six weeks. The regulator change in Lesson 4.2 is worth more than a month of R1's six before we say a word about efficiency.

🔑 The usable-capacity ledger Beacon carries forward

Nominal 1000 mAh. Minus the cutoff window (≈ 1 % with the buck, ≈ 5 % with the LDO). Minus a self-discharge allowance (a few percent per month, so up to ~15 % across six months). Minus calendar aging on cells that sat in stock. Minus the fact that the pilot lab is sometimes 18 °C, not 22. This is why the canon's "3 years on paper" is deliberately derated by 6× to demonstrate the 6-month requirement: the paper number is a ceiling, and the ledger only ever subtracts.

The firmware side is small but worth getting right now: Beacon reports battery voltage in millivolts in its BLE payload, and needs a "battery low" flag that does not chatter as the voltage bounces around a threshold. Measure at rest — before the radio starts — and use hysteresis.

batt.cESP-IDF · rest-voltage sample with hysteresis
// Called right after wake, BEFORE the radio or sensors draw current,
// so the divider sees the cell's rest voltage, not a sagged one.
#define BATT_LOW_MV      3400   // flag "low" below this (buck still fine to 1.8 V in)
#define BATT_OK_MV       3550   // clear the flag only above this: 150 mV hysteresis

static RTC_DATA_ATTR bool s_batt_low;   // survives deep sleep (RTC memory)

uint16_t batt_sample_mv(void)
{
    batt_divider_enable(true);           // PMOS on: 2 × 1 MΩ divider now draws ~2 µA
    vTaskDelay(pdMS_TO_TICKS(250));   // 5τ: 100 nF through 500 kΩ is τ = 50 ms — never a busy-wait
    int mv_adc = adc_read_calibrated_mv(); // ~half of the cell voltage
    batt_divider_enable(false);          // PMOS off: divider draws nothing in sleep

    uint16_t cell_mv = (uint16_t)(mv_adc * 2);
    if (cell_mv < BATT_LOW_MV) s_batt_low = true;
    else if (cell_mv > BATT_OK_MV) s_batt_low = false;
    return cell_mv;
}

Two details there carry weight later. The divider is switched by a PMOS so it draws nothing in sleep — Lesson 4.3 shows what happens when that switch leaks. And the flag lives in RTC memory, because on the ESP32-C3 ordinary RAM does not survive deep sleep.

WEAKNESSWhere this bites

✗ Measuring the battery while the radio is on

A firmware that samples the ADC at the end of the active window — after the BLE burst — reads the terminal voltage mid-sag. On a LiPo the error is millivolts; on a coin cell it is hundreds of millivolts, and the "battery low" flag trips on a half-full cell. Worse, a team debugging it on a bench supply (zero internal resistance) never sees the problem. Sample at rest, first thing after wake, or explicitly characterize the sag and subtract it.

✗ Trusting the nominal number in the battery-life spreadsheet

The spreadsheet says 1000 mAh / 38 µA = 3 years, so a six-month requirement looks laughably safe and nobody derates anything. Then the pilot runs in a 15 °C storeroom on cells that sat in stock for a year at full charge, with an LDO cutoff and a divider that leaks. Each of those is a 5–20 % haircut, and they multiply. The canon's 6× derating is not pessimism; it is the product of five ordinary haircuts. Put the ledger in the spreadsheet, line by line, and update it with measured values in Lesson 4.3.

✗ Skipping protection because "the charger is safe"

The MCP73831 regulates charge voltage and current. It does nothing about a short on the battery leads, a reversed pack, or a cell that is drained to 2 V by a firmware bug that never sleeps. Those are the protection IC's job, and its job is to work when the firmware is wrong. A 25-unit pilot is exactly the scale at which one unit will meet each of those conditions. The DW01A + FS8205 pair costs less than a coffee per board.

KEY INSIGHTHow this connects

Everything in this stage is a withdrawal from the usable capacity you computed here. Regulators waste some of it, sleep leaks some of it, every radio packet spends some of it. Keep the ledger open.

GLOSSARYKey terms introduced

Nominal voltage
The single voltage used to label a chemistry (3.7 V Li-ion, 3.2 V LiFePO4, 3.0 V Li-MnO₂ coin). It is roughly the mid-discharge plateau, not the full or empty voltage.
Discharge curve
Terminal voltage plotted against charge removed, at a stated current and temperature. The only honest description of a cell; capacity is an integral under it down to a chosen cutoff.
Cutoff (end voltage)
The voltage at which discharge is considered finished. The datasheet's cutoff (often 2.0 V for coin cells, 2.5–3.0 V for Li-ion) is usually lower than your system's minimum input, which strands capacity.
Internal resistance (Rint)
The series resistance in a cell's equivalent circuit. Multiplied by load current it gives the voltage sag; it rises with depth of discharge, age, and cold. Milliohms for LiPo, ohms to tens of ohms for coin and primary lithium cells.
C-rate
Discharge or charge current expressed as a multiple of rated capacity per hour. 1C on 1000 mAh is 1 A. Datasheet capacity is usually stated at 0.2C or lower.
Self-discharge
Capacity lost while the cell sits unused. A few percent per month for Li-ion (temperature dependent); around a percent per year for primary lithium and coin cells.
Protection IC (PCM)
A battery-side circuit (DW01A + FS8205 on Beacon) that opens MOSFETs on over-charge, over-discharge, or over-current, independently of the charger and the firmware.
Usable capacity
Nominal capacity minus the stranded tail below your minimum input, minus temperature derating, minus self-discharge and aging — evaluated under your worst-case pulse. The number that belongs in the battery-life spreadsheet.

ACTIVE RECALLCheck your understanding

Answer in your own words first, then reveal.

1A coin-cell tag reboots every time it transmits. The firmware team is convinced it is a software bug. What is the mechanism, and what single measurement settles it?›

The transmit burst pulls tens of milliamps through the coin cell's internal resistance (tens of ohms), dropping the terminal voltage by I × R — hundreds of millivolts — below the MCU's minimum supply or brown-out threshold. The MCU resets, restarts, transmits again, and loops. Settle it with an oscilloscope on the cell terminals (not the regulator output) during a burst: you will see the supply step down for the duration of the pulse and recover afterwards. On a bench supply the bug "disappears", which is the tell.

2Beacon's LiPo is "1000 mAh". Name three things that make the usable number smaller, and estimate the size of each for Beacon.›

(1) The cutoff window: with the buck's 1.8 V minimum input the cutoff is the cell's own ~3.0 V floor, stranding only about 1 % of a cell rated to 2.75 V; with the breadboard LDO the cutoff is ~3.4 V, stranding roughly 5 %. (2) Temperature: a lab at 15–18 °C instead of 22 °C costs a few percent of capacity and raises resistance (harmless for a LiPo's tiny sag, but real). (3) Self-discharge and calendar aging: a few percent per month over six months, more if the cells sat in stock at full charge. Each is a 3–15 % haircut; they multiply, which is why the canon derates the 3-year paper figure by 6× to demonstrate 6 months.

3Why can a LiFePO4 cell not feed a 3.3 V LDO, and why might it feed an ESP32-C3 directly?›

LiFePO4 is 3.2 V nominal and spends almost its whole discharge between about 3.3 and 3.2 V; an LDO needs its input to be above its output plus dropout, so a 3.3 V LDO is out of regulation for essentially the entire curve. But the ESP32-C3 module accepts 3.0–3.6 V, and an LFP cell's full-charge 3.6 V and its plateau both sit inside that window. So some designs connect the cell straight to the module, trading the regulator's cost and quiescent current for an unregulated rail, a stranded tail below 3.0 V, and the need to be careful that the charger's 3.65 V termination never exceeds the module's absolute maximum.

4Compute the capacitor needed for a coin cell to supply Beacon's 45 ms, 40 mA burst with no more than 0.3 V of droop, and say why the answer effectively rules the coin cell out.›

Charge per burst Q = 0.040 A × 0.045 s = 1.8 mC. For a droop of 0.3 V, C = Q / ΔV = 1.8 mC / 0.3 V = 6 mF. That is a supercapacitor or hybrid-layer capacitor, not a ceramic — it brings leakage current (a real fraction of Beacon's 8 µA sleep floor), cost, and board area, and it only works if the cell can recharge it between bursts. It is achievable for a much smaller burst (one packet, no sensor heater, no LED), which is a different product. Beacon's LiPo, with an 80 mΩ internal resistance, sags 3 mV on the same burst and needs nothing.

Next: regulators — the LDO turns the surplus into heat; the buck turns it into current →