Designing for Standby: Quiescent Current Budgets, Load Switches, and Always-On Rails

A product "off" still draws current — and over a year, that invisible draw decides whether a battery product is premium or a warranty liability. Standby design is not one component choice; it is a budget that every rail, every switch, and every IC's IQ (quiescent current) must answer to.

Start with the number that matters: µA·days

A 2000 mAh cell must last 2 years in standby: average budget = 2000 mAh ÷ 17,520 h ≈ 114 µA. That is the entire envelope — MCU sleep, RTC, sensor duty-cycling, regulator IQ, and leakage combined. The immediate consequence: a regulator with 50 µA IQ consumes half your budget before the load does anything.

Where the microamps actually go

• Regulator IQ. DC-DC switching regulators trade IQ against efficiency; modern low-IQ bucks sit at 0.3–5 µA. At light loads, an LDO with 1 µA IQ often beats a buck whose switching overhead dominates — do the math at your real load, not the datasheet's.
• Always-on subsystems. RTC (0.2–1 µA), touch controller sampling (0.5–10 µA), BLE advertise bursts. Duty-cycle everything: a sensor awake 10 ms every 10 s divides its active current by 1000.
• Leakage you did not budget. Reverse leakage across blocking diodes (1 µA for a Schottky at 60 °C — check the temperature curves), capacitor self-discharge, and pull-up resistors that seemed trivial at 100 kΩ but live across the full battery voltage.
• Conversion overhead at light load. A buck with 2 µA IQ and 60% efficiency at 10 µA load can be worse than a 1 µA-IQ LDO. Graphs, not headlines.

Rail architecture: the always-on rail should be tiny

The professional pattern is a split architecture: a micro-power always-on rail (RTC, wake logic, memory retention) that never sleeps, and switched main rails cut off by load switches during standby. ETA（钰泰半导体） covers both sides: its nano-IQ LDO families (parts in the ETA3707 class run near 360 nA IQ) anchor the always-on rail, while its load switches and low-IQ bucks serve the switched domain. The key discipline: the switched domain's off-state current must be specified and tested — "load switch off" still leaks through the switch and, worse, back-feeds through downstream ICs if the rail architecture allows it.

Load switch selection details that decide success

• Quiescent vs shutdown current. Some switches draw IQ only when on; check that "off" truly means nA-class.
• Output discharge. A switch that floats its output lets downstream caps hold stale charge and downstream ICs stay partially powered. Active discharge costs nA and buys a clean off-state.
• Inrush control. Slew-rate-controlled turn-on keeps the 10 mF downstream bank from dipping the always-on rail brown-outing your MCU.
• Reverse-blocking. If a switched rail can be held up externally (USB plugged in while battery-powered), reverse current into the off switch drains the wrong source.

LED and display leakage: the sneaky budget killer

Status LEDs and always-lit displays are standby budgets' silent killers. A single 2 mA indicator LED exceeds the whole 114 µA envelope 17×. Solutions: high-side switch it (FMD's FT833-class load paths and similar parts), use a blink pattern at 1% duty, or pick LED-driver parts with sub-µA standby where the driver itself darkens the indicator except on events. FMD（远翔/Fremont Micro Devices） power and driver ICs are commonly deployed in exactly this role — cheap, appliance-proven parts whose standby specs are honest.

Verify the budget empirically

Budgets on paper fail on leakage nobody measured. The finishing school: a µA-range DMM in series with the battery, firmware stepping through each state, and a thermal chamber — leakage doubles roughly every 10 °C, and the spec you verify at 25 °C is not the spec your product ships with to a hot warehouse. Find the state that exceeds budget, fix the rail, repeat. Products that pass this loop honestly are the ones whose "2-year battery" claims survive contact with customers.