In This Article
Manufacturing an IGBT or SiC power-module cold plate is not simply a matter of machining internal channels. The mounting surface, hydraulic passages, joined structure and sealing interfaces must remain stable within one dimensional and process chain. Production CTQs include mounting-face flatness, flow distribution, post-joining distortion, cross-hole burrs, internal cleanliness, leak and pressure integrity, and long-term corrosion in the specified coolant.
Define the manufacturing boundary first
In PCS, SVG, wind converters and other high-power electronic systems, the cold plate is part of the thermal-resistance chain. Heat from the power semiconductor passes through the module baseplate and TIM into the cold plate, then into the coolant.
| Interface | Function | Main CTQs |
|---|---|---|
| Power-module interface | Transfers heat into the plate | Flatness, roughness, hole position |
| Fluid interface | Carries heat away | Channel section, distribution, pressure drop |
| Structural joining interface | Closes the fluid path | Joint integrity and distortion |
| External piping interface | Connects to the cooling system | Threads, O-rings and fitting seals |
These interfaces must be designed as one system rather than optimized independently.
Put heat sources and datums on the same drawing
DFM starts with the number and location of IGBT or SiC modules, baseplate size, fastener locations and TIM footprint.
| Feature class | Examples | Control strategy |
|---|---|---|
| Functional CTQ | Module surface, locating holes, seal grooves | Strict datum and inspection control |
| Hydraulic CTQ | Main channels, branches, ports | Flow/pressure-drop and cleanliness validation |
| General structure | Outer profile and non-functional relief | Normal machining tolerances |
The goal is not to make every feature high precision. It is to identify the few features that control thermal, hydraulic and assembly performance.
Why mounting-face flatness comes first
TIM fills microscopic gaps; it should not compensate for macroscopic warpage.
Warpage -> locally thicker TIM -> higher contact resistance -> higher local temperature -> larger module temperature spread -> more thermal-cycle stress
Flatness must therefore be reviewed together with module-baseplate condition, TIM thickness, fastener pattern, tightening sequence, plate wall thickness, channel geometry, joining method and operating-temperature deformation.
The key question is whether the mounting face remains stable after joining, finishing and assembly.
Channel design: smaller is not automatically better
Finer channels can increase heat-transfer area and velocity, but they also raise pressure drop, pump power and blockage risk.
| Channel feature | Main risk | Manufacturing concern |
|---|---|---|
| Width | Pressure drop and blockage | Tooling and consistency |
| Depth | Weak bottom wall or higher thermal resistance | Depth and remaining wall |
| Turn radius | Local loss and stagnant regions | Tool path |
| Parallel branches | Flow imbalance | Section consistency |
| Inlet diffuser | Poor distribution | Transition geometry |
| Outlet collector | Recirculation and pressure loss | Collector geometry |
Power-electronics cold plates require a balance among thermal resistance, temperature uniformity, pressure drop and manufacturability.
Manifold flow distribution must be validated with the plate
A system can meet total-flow specification while individual branches remain badly unbalanced.
| Problem | Possible cause | Response |
|---|---|---|
| Large branch variation | Branch-hole diameter variation | Controlled tooling |
| Low far-end flow | Poor main-gallery sizing | Rebalance sections |
| Local blockage | Cross-hole burrs or particles | Deburr and cleaning gate |
| Lot-to-lot drift | Internal dimensional variation | SPC on hydraulic sections |
| Unstable test | Plug or fitting leakage | Standardized sealing interfaces |
Multi-module systems may require flow-versus-pressure-drop or branch-flow validation in addition to dimensional inspection.
The real risks of CNC channels
Machining from solid is flexible for prototypes, but production risks include residual-stress redistribution, thin channel bottoms, trapped burrs, datum drift across multiple setups and loss of flatness after joining.
A more robust route is often:
material stability check -> balanced rough machining -> stabilization -> semi-finish channels -> joining -> datum restoration -> final module-surface machining
Vacuum brazing versus FSW
| Item | Vacuum brazing | FSW |
|---|---|---|
| Principle | Braze alloy forms the joint | Solid-state plasticized joining |
| Multilayer structures | Strong advantage | Limited by weld access |
| Large covers | Feasible | Feasible |
| Thermal input | Whole furnace cycle | Local thermo-mechanical input |
| Distortion | Furnace cycle, filler and fixture | Path, clamping and local plastic flow |
| Design freedom | High | Tool access required |
| Post-machining | Often required | Often required |
A better sequence is:
thermal/hydraulic requirement -> channel architecture -> joining access -> distortion budget -> final machining allowance -> joining-process selection
Joining must be part of the dimensional stack-up
Finishing the mounting surface before joining can create:
finish machine -> join -> thermal cycle/local shrinkage -> flatness drift -> rework
DFM must define pre-join machining stock, joining fixtures, datum restoration, joint location, thermal stabilization and the final machining sequence.
Simple sealing interfaces often create leaks
Production leaks can come from O-ring grooves, scratched sealing faces, fitting threads, plug faces, post-join port distortion, material porosity or inconsistent assembly torque.
Leak prevention should be managed as:
material integrity + joining integrity + machined sealing interfaces + assembly process
Leak, proof pressure and burst are different controls
| Verification | Question | Production use |
|---|---|---|
| Leak test | Is there a measurable leak path? | Can be 100% when specified |
| Proof pressure | Does the part remain intact and sealed? | Sampling or 100% per requirement |
| Burst test | Where is destructive limit? | Normally destructive validation |
Test medium, pressure, hold time, temperature and leak-rate limits must come from the product specification.
Cleanliness is a real production gate
Internal contamination may include chips, cross-hole burrs, abrasive particles, welding debris, brazing residues, cleaning chemistry or seal fragments.
A controlled process is:
deburring -> directional flushing -> ultrasonic/circulation cleaning -> filtration -> drying -> particle acceptance -> contamination-controlled packaging
Corrosion is a system decision
Power-electronics cooling loops may combine aluminum, copper and stainless steel. Material selection must consider coolant chemistry, conductivity, galvanic coupling, fitting material, joint metallurgy, cleaning residues and shutdown condition.
The decision is not simply aluminum versus copper by thermal conductivity. It is thermal + hydraulic + corrosion + manufacturing.
What becomes difficult in volume production?
| Prototype looks acceptable | Volume risk |
|---|---|
| One plate meets flatness | Material-lot stress changes distortion |
| Manual deburring works | Cross-hole residue varies |
| One leak test passes | Joining drift increases leak failures |
| Hand cleaning works | Higher takt leaves particles |
| Rework restores flatness | Large-scale rework is uneconomical |
| One hydraulic unit balances | Branch flow varies across lots |
Production release should convert CTQs into SPC, tool-life control, joining records, leak/pressure records, cleanliness gates, material traceability and flow validation.
A more robust process chain
incoming material verification
-> blank flatness/stress assessment
-> balanced rough machining
-> channel machining
-> deburring and pre-join cleaning
-> cover joining
-> post-join process verification
-> datum restoration
-> final module-surface machining
-> final fittings/seal grooves/critical holes
-> precision internal cleaning
-> leak test
-> proof pressure when specified
-> flow/pressure-drop validation when specified
-> CMM/flatness inspection
-> drying and clean packaging
RFQ inputs
| RFQ input | Manufacturing decision |
|---|---|
| Power-module model/STEP | Heat-source and mounting interface |
| Module count and position | Channels and distribution |
| TIM requirement | Flatness and roughness |
| Total heat loss | Thermal boundary |
| Coolant | Material and corrosion |
| Flow rate | Channel section |
| Pressure-drop limit | Hydraulic balance |
| Working pressure | Wall thickness and joining |
| Leak/proof/burst criteria | Validation plan |
| Fitting/O-ring specification | Sealing machining |
| Alloy and temper | Distortion and joining |
| Joining restrictions | Process route |
| Internal cleanliness | Cleaning process |
| Annual volume | Solid CNC vs extrusion/near-net shape |
| CTQ and Cpk/SPC | Production quality plan |
Conclusion
The manufacturing challenge is not one tight tolerance or one sophisticated flow path. It is maintaining the thermal interface, channels, joining, structure, sealing, cleanliness and corrosion controls through one repeatable production process.
A prototype can be adjusted until it works. Production requires stable post-joining flatness, repeatable flow and pressure drop, specified leak integrity, and an internal fluid path free of harmful particles.
That is the dividing line between machining a cold plate and manufacturing a reliable power-electronics liquid-cooling component.
FAQ
What are the most important machined dimensions on an IGBT or SiC power-module cold plate?
Individual dimensional tolerances are not enough. The most important controls are mounting-surface flatness and roughness, the positional relationship between mounting holes and datums, channel cross-section consistency, inlet and outlet sealing features, and whether these CTQs remain stable after joining. Multi-module plates also require control of flow distribution and temperature uniformity across heat-source zones.
Why is final machining often required after cold-plate joining?
Vacuum brazing, friction stir welding and other joining processes can introduce thermal cycles, local plastic deformation or residual-stress release. If the power-module mounting face is finished before joining, flatness and feature relationships may drift afterward, so production processes commonly reserve stock for final post-joining machining.
Should a PCS power-module cold plate use vacuum brazing or FSW?
There is no universal best choice. Vacuum brazing is suitable for multilayer, large-area and more complex internal structures but requires control of furnace distortion, braze materials and cleanliness. FSW is well suited to aluminum plate channels and provides a strong solid-state joint, but it is constrained by tool access, clamping, weld path and local distortion. Selection should be based on structure, volume, size, pressure requirement and final flatness.
Which CTQs should be clearly defined in an IGBT or SiC cold-plate production RFQ?
The RFQ should define mounting-surface flatness and roughness, mounting-hole position, flow rate and allowable pressure drop, working pressure, leak and proof-pressure criteria, coolant and material compatibility, internal cleanliness, fitting and O-ring specifications, joining method, final machining datums, annual volume, and any process-capability or SPC requirements for critical dimensions.
