Sealed roadside cabinet: derating, ripple, and an ageing output capacitor

Load. Camera 6 W + router 5 W + controller 3 W = 14 W, so 14 / 12 = 1.17 A continuous. Add the battery recharge current: a 7 Ah VRLA at 0.1-0.2 C draws roughly 0.7-1.4 A while recovering. Worst case during a transmit burst: 1.17 + 1.4 + 1.2 = 3.77 A.

Derated module output. The derating is 2.5 percent per degree C above 50 degrees C, and the cabinet is 12 degrees C above that threshold: 5 A x (1 - 0.025 x 12) = 5 x 0.70 = 3.5 A.

Conclusion. Normal continuous operation at about 1.2 A is comfortably inside 3.5 A, so the module is not thermally current-limited in ordinary service. The only over-limit case is the coincidence of a deeply discharged battery, a router burst, and a hot cabinet, which reaches about 3.77 A against a 3.5 A limit and is therefore marginal, not catastrophic. That marginality explains why reboots appear only in summer and only with the battery recovering. The dominant fault, though, is the tripled ripple at temperature, which is a capacitor and thermal problem, not a current-limit problem. Do not solve this by fitting the 10 A module alone, because it will dissipate similar losses into the same sealed 62 degrees C box.

Use the usual life-doubling rule: life doubles for every 10 degrees C fall in core temperature.

• If the core sits close to the 62 degrees C ambient: (105 - 62) / 10 = 4.3 doublings, so 4000 h x 2^4.3 = about 79,000 h, roughly 9 years.
• If the can is heated by its own ripple current and sits, say, 15 degrees C above ambient at 77 degrees C: (105 - 77) / 10 = 2.8 doublings, so 4000 x 7 = about 28,000 h, roughly 3.2 years.

The honest answer is a range of about 3 to 9 years, because the rule applies to core temperature and the core temperature is not being measured. That range straddles the unit's three years in service, so arithmetic alone cannot decide the case.

The measurement does decide it. Ripple rising from 60 mV to 150 mV purely as a function of temperature is the signature of a capacitor whose equivalent series resistance has risen sharply - end-of-life behaviour - not of a healthy capacitor in a hot box. After three summers at 62 degrees C the output capacitor should be treated as at end of life regardless of the calculated hours, and the 150 mV ripple is the credible cause of the router reboots, since the router sees that ripple on its supply during a 1.2 A burst.

Recommendation: move the conversion outside the cabinet. Mounting the 60 W AC adapter externally removes roughly all of the converter's dissipation from the sealed enclosure, so the interior temperature falls, and the adapter itself runs in cooler, better-ventilated air. It attacks the root cause (cabinet temperature) rather than a symptom.

Why not the 10 A module. It is the same family with the same topology; it will dissipate comparable or greater losses into the same sealed volume and will have the same ageing output capacitor. Its extra current headroom does not fix a capacitor whose ESR has risen, and it may not reduce interior temperature at all.

Why not the 4700 microfarad capacitor as the primary fix. It does buffer the burst: delta-V = I x t / C = 1.2 A x 2 ms / 4700 microfarad = 0.51 V, which is a real improvement at the load. But it leaves the aged module capacitor in place, adds inrush stress and can interact with the module's control loop. Keep a modest local bulk capacitor (1000 microfarad, 25 V, low ESR) as a secondary measure, not as the answer.

Additional design actions:

• Replace the module's output capacitor if the module is retained anywhere in the thermal path.
• Verify the external adapter's ambient rating, IP enclosure, cable gland and surge protection on the outdoor mains feed.
• Check DC drop on the longer 12 V run at 2 A; use at least 2.5 mm squared and keep drop under about 3 percent.
• Add a cabinet temperature alarm and log it, so the next thermal problem is visible before a router starts rebooting.