Symptom: Venting, Bulging, and a Drive That Trips on Undervoltage
You are seeing it on the bench and in the field: a 450 V electrolytic DC-link capacitor in a 7.5 kW motor drive that has been in service for 14 months. The case is bulged at the vent score, and the drive occasionally trips on DC-bus undervoltage during regenerative braking. The replacement unit passed incoming inspection, but the same failure pattern reappears within 600 operating hours in a 55 °C cabinet.
This is not a random component defect. The root cause is almost always a mismatch between the capacitor’s rated ripple current and the actual thermal budget of the inverter’s DC link. The failure sequence is predictable: ripple current generates internal heat, the core temperature rises above the rated maximum, the electrolyte evaporates, capacitance drops, ESR climbs, and the resulting higher ripple current accelerates the cycle until the vent opens.
Layered Root-Cause Checks: From Cabinet to Core
When a DC-link capacitor fails prematurely, do not replace the part and re-run the machine. Work through this sequence in order.
Layer 1: Verify the Actual Ambient and Airflow
Measure the air temperature at the capacitor surface, not the cabinet thermostat reading. In many compact drives, the DC-link cap sits directly above the IGBT heatsink. The local ambient can be 15 to 20 °C higher than the air intake. Check whether the cooling airflow path is blocked by wiring harnesses or busbars. A common error is assuming a 85 °C rated capacitor operates safely when the local ambient is already at 65 °C and the ripple current is at the datasheet maximum.
Layer 2: Calculate the Core Temperature, Not the Case Temperature
The core temperature (hot spot) is what determines electrolyte life. For a screw-terminal electrolytic, the core is typically 5 to 10 °C hotter than the case at rated ripple. The relationship is:
T_core = T_case + (I_ripple² × ESR × R_th)
Where R_th is the thermal resistance from core to case in °C/W, typically 2.5 to 5 °C/W for a 35 mm diameter part. If you measure 75 °C at the case and the ripple current is 4.5 A with an ESR of 0.35 Ω, the internal heat is roughly 7 W, which at R_th = 3 °C/W yields a core temperature near 96 °C. Against a rated life of 3000 h at 105 °C, the lifetime halves for every 10 °C drop, so at 95 °C core temperature you should expect around 6000 h—close to the observed failure time.
Layer 3: Measure the Actual Ripple Current with a Current Probe
Do not rely on the calculated value from the datasheet formula. The ripple current in the DC link of a PWM inverter has two components:
- Low-frequency component (2 × line frequency for single-phase rectifiers, 6 × line for three-phase): typically 20–30% of the rated load current.
- Switching-frequency component (IGBT switching frequency, e.g., 4–16 kHz): amplitude depends on the DC-link inductance and the output current. This component can be higher than expected, especially with long motor cables or a low-inductance busbar design.
Measure the combined RMS current with a true-RMS current probe at full load and at the worst-case operating point (often just above motor base speed, where the modulation index is high). Compare the measured value against the capacitor’s maximum allowable ripple current at the actual ambient temperature. Many suppliers provide a 105 °C rating of 2.8 A, but derate to 1.9 A at 85 °C and 1.2 A at 60 °C. If your measured value exceeds the derated limit, the failure is a ripple-current overstress, not a component quality issue.
Layer 4: Check ESR and Capacitance at the Operating Frequency
Use an LCR meter at 100 Hz and 10 kHz. Record these acceptance values for incoming inspection and for field-return verification:
| Parameter | Measurement Condition | Acceptance Limit |
|---|---|---|
| Capacitance | 100 Hz, 20 °C | Within ±20% of rated value |
| ESR | 10 kHz, 20 °C | Below 1.5× initial datasheet maximum |
| ESR change | Compare to 100 h aged sample | Less than 20% increase |
| Leakage current | At rated voltage, 5 min | Below 0.03 × C × V (in µA, C in µF, V in V) |
A return unit that shows capacitance at 60% of rated and ESR at 2.8× the initial maximum confirms end-of-life by electrolyte evaporation. A unit that still shows rated capacitance but has bulged cases points to excessive internal pressure from gas generation—a sign of overvoltage or reverse polarity, not normal aging.
Prevention Checklist for New Designs and Replacements
When you source DC-link capacitors for an inverter design or a replacement program, use this checklist before you approve a part:
- Define the worst-case operating point. Calculate ripple current at the highest ambient temperature and the highest modulation index, not at the nominal rated condition.
- Derate for temperature. Apply the manufacturer’s derating curve. As a general rule, if the local ambient exceeds 60 °C, choose a capacitor with a rated ripple current at least 1.5× the measured worst-case value.
- Verify ESR at 10 kHz. A lower ESR part reduces internal heat for the same ripple, but verify that the ESR is stable over the expected life. Film capacitors have a flat ESR curve and no electrolyte loss; if size allows, consider a film capacitor for the DC link when ripple content is high and ambient exceeds 70 °C.
- Check the surge voltage rating. The capacitor’s rated voltage should be at least 110% of the maximum DC bus voltage including transients. For a 400 VAC input, the DC bus is about 565 V, so a 600 V rated capacitor is marginal; a 630 V or 700 V part provides more surge margin without sacrificing reliability.
- Confirm compliance documentation. Request RoHS and REACH declarations from the manufacturer before purchase. For export shipments, confirm the part is not on any restricted substance list and that the supplier provides a certificate of conformity with the shipment.
The core lesson is to treat the DC-link capacitor as a thermal system, not a passive component. Measure the local ambient, calculate the core temperature from measured ripple and ESR, and compare against the derated limits. When you do this before the failure occurs, you shift from forensic repair to predictable design.