← Back to the chapter · Previous detail · Next detail →

Technical detail · Hardware Optimization Strategies

Sleep-rail efficiency and daily charge

Regulator efficiency is not one number; it changes with load. A buck is excellent at the high currents of a radio burst, but at the microamp currents of sleep its own controller quiescent current can dwarf the load. Suppose a buck's measured battery current is 25 uA for a 10 uA, 3.3 V sleep load, including 15 uA of quiescent current. At 4.2 V input, its power efficiency is (3.3 V × 10 uA)/(4.2 V × 25 uA) = 31%. A low-quiescent LDO drawing 11 uA from the same battery for the same load is (3.3 V × 10 uA)/(4.2 V × 11 uA) = 71% efficient. At light load the LDO wins in this example.

Over one day, that sleep choice alone is visible. A 25 uA buck-fed sleep rail for 23.76 hours consumes about 594 uAh. An 11 uA low-Iq LDO sleep rail consumes about 261 uAh over the same interval, saving roughly 333 uAh per day. If the product reports rarely, that sleep-rail saving can be larger than the active burst optimization; if it reports constantly, the high-current buck path may dominate instead.

This matters because a duty-cycled device spends almost all of its time asleep, so the sleep regulator, not the active one, often dominates the energy budget. A common mistake is to select a single high-efficiency buck for the radio peak and then pay its quiescent current through months of sleep. The better designs pick the regulator for the regime that dominates the energy: a buck for the high-current active phase, and a low-quiescent LDO or a converter with a light-load pulse-frequency mode for the long sleep. The same finish-and-sleep logic explains hardware accelerators: a crypto or DSP block that completes an operation in far fewer cycles than the CPU lets the whole system return to that low-power sleep sooner, so the accelerator is an energy win even when its instantaneous power is higher.

← Back to the chapter · Previous detail · Next detail →