Thermal design is the part of capacitor bank specification that is most often deferred to prototype testing. For a bank holding several electrolytic or film capacitors bolted to a chassis, the difference between a 10 K rise and a 25 K rise at the same ripple current is the difference between a component life measured in years and one measured in months. This article defines the boundary conditions where thermal analysis is mandatory, gives decision thresholds for the parameters that actually control junction-to-air temperature, compares cooling methods in one table, and outlines a selection procedure that works for DC-link, PFC, and AC filter banks.
Application Boundary: When Thermal Design Becomes Mandatory
Not every capacitor bank needs a dedicated cooling design. A low-power SMPS output filter carrying 2 A ripple in a ventilated enclosure can be sized by ripple-current ratings alone. The boundary shifts when any of the following conditions apply:
- Bank capacitance exceeds 1000 µF and the bus voltage is above 100 V (high energy density concentrates losses into a small package).
- Continuous ripple current per capacitor exceeds 30% of the datasheet value at the reference temperature.
- Ambient temperature inside the enclosure exceeds 50 °C for more than 8 hours per day.
- Multiple capacitor cans are mounted in a row with less than 10 mm spacing (mutual heating blocks convection paths).
- The bank is sealed in an IP-rated housing with no active airflow.
When any one of these conditions is true, thermal design must be performed at the parameter level, not by rule-of-thumb derating.
Key Thermal Parameters and Decision Thresholds
The thermal problem reduces to a single equation: core temperature rise above ambient equals the product of power loss and thermal resistance. Power loss is ripple-current squared times ESR; thermal resistance is the sum of internal (junction-to-case) and external (case-to-air) components. The parameters listed below are the ones to verify before selecting a replacement capacitor or a cooling method.
- Ripple current per capacitor (I_rms) — For aluminum electrolytic types rated at 105 °C, continuous ripple above the datasheet value doubles the core temperature contribution. A practical threshold: if I_rms exceeds 70% of the rated value at the operating frequency, either add one parallel capacitor per three units or select a higher-ripple series.
- ESR at operating frequency (mΩ) — ESR is frequency-dependent; datasheet values at 120 Hz are not valid for switching-ripple spectra above 10 kHz. Threshold: if ESR at the ripple frequency exceeds 15 mΩ for a 1000 µF/450 V class part, expect more than 1.5 W loss per unit at 10 A rms. The cooling target must then keep case temperature below 85 °C.
- Core (hot-spot) temperature — For electrolytics, keeping the core below 85 °C yields a projected life of roughly 20,000 hours at rated voltage; each 10 °C reduction in core temperature extends life by a factor of approximately two. The absolute limit is the rated maximum core temperature (typically 105 °C). For film capacitors, the limit is the insulation-class temperature (usually 85 °C or 105 °C); exceeding it degrades dielectric margins rapidly.
- Measured surface temperature rise — With natural convection, a case rise of 10 K above ambient is normal and acceptable. A rise of 20 K or more indicates that internal thermal resistance is too high for the ripple load; do not compensate by simply opening ventilation holes — recalculate the loss budget first.
- Ambient operating temperature — At ambient above 60 °C, standard electrolytic banks lose ripple-current headroom quickly; above 70 °C, film capacitors become the safer choice unless forced airflow of at least 2 m/s is guaranteed. Use manufacturers’ ambient-derating curves rather than linear extrapolation.
Cooling Options: Decision Table
Once the loss per capacitor is calculated, the external thermal resistance must be brought low enough to keep the hot spot inside the limit. The table below compares the four practical cooling methods for distributor-sourced replacement banks.
| Cooling method | Typical airflow / medium | External Rth reduction vs. still air | Construction impact | Appropriate for |
|---|---|---|---|---|
| Natural convection | 0–0.5 m/s free air | Baseline (reference) | Vertical mounting, 10 mm min spacing, no shrouds | Low-power banks; ambient ≤ 50 °C |
| Forced air (axial fan) | 1.5–3 m/s directed across can rows | 30–50% lower Rth | Fan placement, ducting, filter maintenance | Rack-mounted drives, industrial converters |
| Heatsink clamp (screw-terminal or snap-in with thermal pad) | n/a — conduction path | 40–60% lower Rth if Rth_clamp ≤ 10 K/W | Thermal pad compound, torque limits on cap terminals | Snap-in electrolytics; high ripple density |
| Liquid cooling (cold plate) | 1–3 L/min coolant | 50–65% lower Rth | Plumbing, galvanic isolation, sealing | Traction inverters, high-power PFC front ends |
Forced air is the most common retrofit in distributor stocks because it requires no mechanical change to the capacitor itself. Heatsink clamps give better results for screw-terminal electrolytics but require a thermal interface material with verified performance, not generic thermal paste.
Step-by-Step Selection Procedure
Use this sequence when qualifying a replacement capacitor bank or adding cooling hardware:
- Define the load spectrum at component level. Measure or simulate the RMS ripple current at the operating frequency (e.g., 1–20 kHz for DC-link, 50/60 Hz harmonic spectrum for PFC). Do not use the DC average.
- Calculate loss per unit. Multiply the square of the ripple RMS current by the actual ESR at that frequency. For a bank of N parallel capacitors, divide the total ripple by N before calculating individual loss.
- Compute the required external thermal resistance. Target core temperature (T_core_tgt) is 85 °C for long-life electrolytics or 10 K below the rated limit for films. The allowed external Rth is (T_core_tgt − T_ambient) ÷ loss per unit, minus internal Rth from the datasheet. If the result is below the natural-convection capability, move to the next cooling row in the table.
- Choose a cooling method and verify spacing. Select from the table based on the computed Rth and on enclosure constraints. Re-check that bank spacing (≥ 10 mm between cans) and orientation match the assumed airflow or convection path.
- Prototype-test with a thermocouple in the ambient path. Measure case temperature rise at full load for at least one thermal time constant (15–30 minutes for a 35 mm diameter can). Compare to the 10 K / 20 K thresholds above; if case rise exceeds 20 K, increase parallel capacitance or replace with a higher-ripple (lower ESR) series.
- Document the thermal budget. Record the measured rise, the ambient set point, and the projected core temperature so that future cross-references to alternative series remain valid without re-testing.
Applying this sequence at the sourcing stage — rather than after a field failure — keeps temperature rise inside the lifetime budget and gives the procurement engineer a defensible basis for requesting sample parts and datasheet verification before committing to volume orders.