Application boundary: When thermal design decides bank reliability
Capacitor banks in DC links, power factor correction, and UPS output stages share one characteristic: they operate under continuous ripple current, so the internal temperature settles at a steady-state value that depends on the balance between joule heating and heat removal. In benchtop prototypes with open airflow, this balance is easy to achieve. In production cabinets, the bank is enclosed, air velocity is low, and neighboring components contribute their own heat. The consequence: the capacitor core runs hotter than the datasheet's reference condition, and both lifetime and capacitance drift reflect that difference.
This article applies to banks built from aluminum electrolytic or metallized polypropylene film capacitors in power conversion equipment operating between 25°C and 70°C ambient. It does not cover supercapacitor or ceramic capacitor stacks, where the governing constraints differ. The goal is to give you a decision procedure for selecting capacitors and cooling geometry before the design goes to thermal testing.
Decision thresholds for key thermal parameters
Ripple current vs. rated current
One threshold to check is whether the actual RMS ripple current per capacitor approaches the rated ripple current quoted at the reference temperature (typically +85°C or +105°C for electrolytes, +70°C or +85°C for film parts). If the calculated ripple exceeds 80% of the rated value at the bank's minimum ambient, you need either a larger can size, a lower ESR part, or forced airflow. Pushing ripple above the rating without cooling shortens service life in a way that cross-reference tables do not capture.
ESR at the operating frequency
ESR falls with frequency for both technologies, but the relevant value is the ESR at the switching frequency of the application, not the 120 Hz value printed on many data sheets. A 10 kHz ESR that is 30% lower than a 120 Hz ESR changes the heat calculation by the same proportion. Measure or request the frequency-specific ESR from the vendor before the thermal budget is fixed.
Core temperature limit
For aluminum electrolytics, the hotspot temperature is roughly 10°C above the can surface temperature at equilibrium. Film capacitors run closer to the surface temperature because the heat source is distributed through the dielectric, but the case temperature limit stated on the datasheet remains the binding constraint. If the maximum case temperature at full load exceeds the rated case temperature by more than 10°C, the cooling design is inadequate.
Lifetime impact
A general approximation for aluminum electrolytics is that lifetime halves for every 10°C rise in core temperature above the rated value. That is an approximation, not a guarantee, but it explains why a bank designed for a 5°C margin at 50°C ambient may fail early if the actual internal temperature reaches 75°C in a production cabinet.
Comparison of cooling approaches for capacitor banks
| Cooling method | Typical convection coefficient (W/m²·K) | Practical temperature rise reduction vs. natural convection | Constraints |
|---|---|---|---|
| Natural convection, perforated cabinet | 5-10 | Baseline | Low cost, but limited to low ripple densities |
| Forced air, axial fan | 20-40 | 30-50% lower case temperature | Air velocity of 2-4 m/s across the bank; filter cleaning |
| Air duct with fan per bank row | 30-50 | 40-60% lower case temperature | Duct design, pressure drop, acoustic noise |
| Heat sink contact on the can end | 10-30 (via interface) | 20-40% lower case temperature | Flatness of the can surface, thermal interface material |
| Liquid-cooled cold plate | 100-500 | 60-80% lower case temperature | Pump reliability, coolant loop complexity, cost |
Forced air is the most common compromise in industrial cabinets. A bank in an open rack with 2 m/s airflow dissipates roughly three times more heat than in still air. The ratio is not linear with velocity; most of the gain appears between 1 and 3 m/s.
Step-by-step selection procedure
- Calculate ripple per capacitor. Measure or simulate the total RMS ripple at the bank bus, divide by the number of parallel capacitors, and multiply by an imbalance factor of 1.1 to account for ESR spread between parts.
- Estimate the heat source. P = Irms² × ESR at the switching frequency. For a 2 A ripple and a 100 mΩ ESR, that is 0.4 W per capacitor. Multiply by the number of capacitors to get the bank total.
- Set the target case temperature. Use the lifetime target to determine the maximum allowable case temperature from the datasheet's lifetime-temperature curve. Do not use the absolute maximum rating; use the temperature at which the bank is expected to reach the required service life.
- Calculate the allowable temperature rise. ΔT = Tcase,max − Tambient,max. If ΔT is below 10°C and the total dissipation exceeds 20 W, forced air or a heat sink is needed.
- Size the airflow. Estimate the required air velocity from the convection coefficient in the table. For a 40% temperature rise reduction in a 300 mm deep cabinet, an axial fan delivering 2-3 m/s over the bank face is a reasonable starting point.
- Verify with a thermocouple. After prototype assembly, measure the can surface temperature at full load and worst-case ambient. Compare with the calculation and adjust the fan or capacitor ESR rating if the measured rise exceeds the target by more than 5 K.
Cross-reference check for replacements
When replacing capacitors in an existing bank, the thermal budget is a function of the replacement part's ESR and heat sink geometry, not just its capacitance and voltage. A replacement part with the same capacitance and voltage rating but a higher ESR at the switching frequency will raise the case temperature. Before finalizing a cross-reference, request the ESR at the application frequency and recalculate the temperature rise with the procedure above. This step prevents a replacement that meets the electrical spec but violates the thermal budget.
A capacitor bank that stays below its rated case temperature by a deliberate margin will outlive one that runs at the rated limit, regardless of the brand. The cost difference between a slightly larger can and an early field failure is not close.

AKKN Electronics


