Phenomena and Risks
Under-voltage (Brown-out) occurs when the supply voltage falls below the minimum specification of the device but is not completely zero. Logic may enter a metastable state, RAM contents may be corrupted, and Flash/EEPROM writes may experience bit flips. Cellular modules have large current steps during burst transmissions; if the input source has high internal resistance or inadequate decoupling, local rail voltages can momentarily drop, causing watchdog resets or module abnormal restarts, manifesting as "random disconnections."
Hardware Countermeasures
Place sufficient low-ESR ceramic capacitors and appropriate electrolytic/polymer capacitors near the module's power pins to provide pulse current. When powered by batteries, choose cells with lower internal resistance or add supercapacitor buffering. Wire gauge and connector contact resistance should be considered in the calculations. Some designs use ideal diodes or OR-ing circuits to prevent backflow, but voltage drop should also be considered.
Monitoring and Software
MCUs and power management chips often integrate a BOR (Brown-Out Reset) threshold, which triggers a reset when voltage drops below the threshold to prevent execution of incorrect instructions. The threshold should be higher than the minimum writing voltage for Flash. Software should avoid large-scale NVM writes when battery voltage is near critical; it can first check the voltage threshold before entering low-power mode or delaying transmission.
Testing Method
Use a programmable power supply to simulate the battery discharge curve, while observing the module's transmission current and rail voltage ripple. Keep oscilloscope probe ground lines as short as possible and measure across the near-end capacitor. Record whether there are voltage spikes below the threshold before reset.
Engineering Practice Supplement
Starting from "Undervoltage and Reset," common on-site issues often focus on "specs meet lab standards but hard to reproduce in the field." Besides RF and protocol, check for power-ground bounce, connector contact, and firmware race conditions. On the cellular side, it is suggested to collect RSRP/SINR, cell reselection, and PDP reconstruction records; on the application side, there should be clear strategies for retries, idempotency, and queue persistence. If data security and compliance are involved, write the collection scope, storage location, retention period, and third-party sharing into privacy policies and contract annexes to avoid high costs of patching after deployment.
The following are additional tips related to cellular IoT project delivery: it is recommended to keep screen recordings of typical scenarios, screenshots of key indicators, and summaries of masked packet captures during the acceptance phase; in the mass production phase, change records should be kept for software and hardware versions, calibration parameters, and key component batches to facilitate problem attribution, customer audits, and long-term maintenance. When payment or personal information is involved, also check whether log masking and key rotation strategies are executable.
How far the rail falls at the transmit instant
Undervoltage means the supply is already below the level at which the device is guaranteed to work correctly, but it has not yet fallen to nothing. Logic can stop in a half-valid state, memory contents go stale, and a flash write in progress picks up bit errors. A cellular transmission raises current in a very short time. Supply resistance, connectors, and thin traces all contribute voltage drop in that instant. The conditions are enough low-impedance storage near the module supply pins, a reset threshold set above the safe flash-write voltage, and software that refuses to start a new write when the voltage is near the threshold.
Sizing the supply from average current misses the transmit spike. The oscilloscope has to see the waveform at the module pin, at the moment of transmission. A long probe ground lead measures the dip as shallower than it is. After the cell ages and internal resistance rises, a design that used to pass starts restarting at random. In the log it looks like a software crash.
The reset threshold must let the flash write finish
Setting the reset threshold very low, in the hope of avoiding restarts, lets flash write halfway at an unsafe voltage. Setting it above the normal transmit dip restarts the device on every upload. The window between those two has to be opened by storage and layout, not by repeatedly changing a software delay. A watchdog reset and an undervoltage reset should be distinguishable in the log. Otherwise the field keeps investigating the protocol stack.
Near the end of battery life, the policy should be to postpone transmission or enter a low-power state, not to keep writing the configuration area. The drop across the protection circuit also belongs in the margin. Chemistry and pulse capability are in IoT battery chemistry. If the dip appears only at power-on, go back to inrush and soft start rather than lengthening the reset time further.
Read the restart cause before writing a log
If a watchdog, a software assertion, and an undervoltage reset share one abnormal-restart counter, the field cannot tell whether to investigate the supply or the protocol stack. Read the boot reason before the next flash write, and report it with the heartbeat. A restart that appears only at the transmit instant is the one worth capturing with a scope on the supply pin. If the device also restarts while idle, check the protection circuit and grounding first, rather than adding more storage capacitance.