Ceramic panel with four patterned circuit areas and curved conductor paths
Supplied circuit example. Material grade and process are not verified by the photograph.
Invented calculationResultLimit
2mJ ×20k event pairs/s40WSwitching contribution only
50A RMS squared ×1mΩ2.5WNot a conductor current rating
80A RMS squared ×1mΩ6.4WResistance assumed at temperature
0.30K/W ×60W +0.05K/W ×40W20K rise at AIllustrative linear thermal model

Translate the inverter topology into a substrate problem

A direct bonded copper layout for a solar string inverter must connect a particular switching circuit, not an abstract collection of power chips. Establish the converter stage, topology, device identities and module boundary before drawing copper islands. A substrate for a boost stage has different connections from one supporting a three-level output leg, even when both belong to the same inverter. The DBC process guide explains the construction context. The layout task is to place devices and conductors so that electrical paths, heat removal, isolation and assembly requirements work together over the intended operating range. A visually tidy pattern is only the beginning of that assessment.

Specify the operating points that drive placement

Prepare a set of operating conditions that represents the intended photovoltaic and grid-side behavior of the converter. Include the relevant DC voltage range, phase current, switching strategy, frequency and cooling conditions. If reactive-power operation is required, include it explicitly rather than assuming that the rated active-power condition bounds every device loss. Distinguish the maximum system voltage from the voltage applied to each semiconductor and insulation boundary. Assign responsibility for the loss calculations to the circuit design team and preserve the assumptions with the layout revision. Without this information, the substrate designer cannot know whether a compact placement serves the worst thermal condition or merely a convenient nominal example.

Map losses by device position

Infineon's NPC1 explanation illustrates why inner and outer switching positions can have different conduction and switching demands. That specific example supports a broader design practice: calculate losses by position and operating condition instead of distributing total inverter loss equally among devices. Do not transfer the example's device ratings or loss split to a different modulation strategy. Create a table that identifies each switch and diode, the loss components included and the condition producing the relevant temperature. This table becomes the heat-source map for the DBC assessment. It also exposes when a proposed device substitution changes the thermal problem even though its footprint fits the existing copper pattern.

Check event accounting in switching-loss estimates

As an invented arithmetic example, suppose a device has a combined turn-on and turn-off energy of two millijoules at a stated electrical and thermal condition. If that event pair occurs twenty thousand times per second, its estimated switching contribution is forty watts. If a different operating sequence produces ten thousand equivalent event pairs per second, the contribution is twenty watts. Neither calculation includes conduction loss, and neither permits the energy value to be reused at arbitrary voltage, current or temperature. State how events are counted for the selected topology and control scheme. An incorrect event count can misplace the thermal priority before the DBC geometry is even considered.

Draw each important commutation path separately

A three-level circuit can have more than one relevant commutation path. Use switching-state information from the circuit team to identify which devices and connections participate in each transition. Trace the complete path through internal conductors, terminals and the local DC-link connection rather than highlighting only a short copper segment near one die. Rank the paths using the actual operating conditions and component requirements. A layout improvement for one transition may lengthen another connection or alter device placement. Keep those tradeoffs visible in separate annotated drawings so a reviewer can see which path motivated each placement decision and which paths still require electrical verification in the assembled module.

Include external terminals in the layout decision

A low-inductance internal pattern can be undermined by an unfavorable connection to the rest of the converter. Review terminal placement with the intended busbar and capacitor arrangement before freezing the substrate outline. Semikron's SKiM technical explanation discusses internal terminal symmetry and the importance of the DC-link connection in its module context. It should not be read as a performance guarantee for a custom solar substrate. Use the principle to define a complete comparison boundary. When requesting parasitic extraction or measurement, identify the terminal planes and included conductors so the reported result describes the assembly that will actually be installed rather than an isolated favorable fragment.

Separate geometric symmetry from electrical equivalence

Equal-looking copper islands do not establish equal current sharing between parallel chips. Connection lengths, bond arrangements, terminal access and device characteristics all belong in the assessment. Conversely, a topologically asymmetric leg may intentionally use different device areas because positions have different loss duties. Decide where symmetry is required and where functional differences justify different geometry. For parallel positions, document the evidence used to assess sharing under the relevant static and switching conditions. For different switching roles, preserve their separate loss and temperature results. This prevents a cosmetic requirement for identical shapes from overriding the circuit's real needs, while still making unintended connection imbalance a clear review item.

Use resistance calculations with a defined current waveform

Consider an invented copper path represented by a resistance of one milliohm at its operating temperature. A root-mean-square current of fifty amperes gives a resistive loss of two and a half watts. Using the same resistance with eighty amperes gives six and four-tenths watts. The increase follows the square of current in this simplified model. The calculation is not a current rating for a particular copper thickness or island shape. It shows why a neck or terminal connection should be assessed using the current it actually carries. State the temperature associated with the resistance and use the appropriate waveform rather than inserting an average phase current into every conductor calculation.

Balance local heat spreading against other requirements

Moving chips farther apart may reduce thermal interaction in one construction, but can change electrical connection geometry and substrate size. Increasing a copper island may assist spreading while reducing the available separation to another potential. Evaluate these changes together using the same loss map and cooling boundary. The thermal management guide provides a framework for distinguishing substrate resistance from the complete path. Do not infer a layout benefit from ceramic conductivity alone. A comparison should show the affected device temperatures and the assumptions behind them, including neighboring heat sources, attachment layers and the interface to the heatsink or baseplate used by the actual module.

Preserve thermal coupling in the comparison

For an invented two-source linear steady-state model, suppose the temperature rise at device A is represented by a self coefficient of 0.30 kelvin per watt times its loss plus a coupling coefficient of 0.05 kelvin per watt times device B's loss. With sixty watts at A and forty watts at B, the calculated rise is twenty kelvin. Ignoring B would give eighteen kelvin. These coefficients are illustrative and must come from an appropriate model or measurement for a real assembly. The example shows how a placement comparison can be misleading if it changes neighboring losses or silently removes coupling while claiming to compare only the DBC layout.

Assign insulation requirements to actual node pairs

Create a list of relevant node pairs and their electrical conditions, including the relationship to the cooling structure and accessible conductive parts. The nominal DC bus value is not a universal spacing instruction. Identify the applicable insulation requirements through the product's design process, then translate them into controlled geometry and verification for the module. Include the finished conductor edges, attachment material and other assembly features that may affect the separation. Do not copy a bare-substrate dimension into the completed module assessment without checking the surrounding construction. The purpose is to make the insulation boundary explicit so later copper enlargement or terminal movement is reviewed against the correct requirement.

Reserve space for the assembly process

A DBC layout must be manufacturable with the intended die attachment, interconnection and encapsulation operations. Confirm the usable attachment areas, bond-tool access, placement tolerances and inspection access with the assembly team. Copper that appears available in a top view may not be available to a tool approaching at the required angle. Likewise, a pad that accommodates the nominal die outline may leave insufficient allowance for the controlled attachment process. Record these requirements separately from electrical clearance so their origins remain visible. This allows a proposed geometry change to be checked against both circuit performance and assembly capability instead of discovering conflicts after the first patterned substrates arrive.

Specify finished copper geometry rather than artwork alone

Agree which dimensions are controlled on the finished DBC and how they are measured. Etched conductors have a manufactured profile, so the drawing should identify the relevant measurement plane and tolerances for critical gaps and attachment features. Ask the substrate supplier to review the proposed copper thickness and minimum features as one construction. A nominal artwork gap is not automatically the final available separation at every height. Keep the approved fabrication interpretation linked to the electrical and assembly models. That connection is especially useful when a supplier proposes a process adjustment that preserves the artwork but changes the realized copper profile or dimensional distribution.

Compare layouts with a controlled verification matrix

For each candidate, record the same operating points, device data, electrical connection boundary and cooling conditions. Summarize the predicted temperatures, relevant parasitics, conductor losses and manufacturing concerns without collapsing them into one unsupported score. Identify which findings come from calculation, which from extraction and which from measurements on representative assemblies. Where a candidate uses a different terminal or interconnect arrangement, include that change in the comparison rather than attributing all improvement to the ceramic pattern. The strongest layout choice is one whose tradeoffs are visible and tested at the important operating conditions, not simply the candidate with the smallest outline or lowest isolated thermal estimate.

Validate the assembled module before treating geometry as proven

Plan prototype measurements around the uncertainties that drove the layout choice. These may include switching behavior, current sharing, temperature differences between device positions and assembly consistency. Preserve the device, substrate, attachment and interconnect revision for every tested unit. If a measurement requires a probe or sensing feature that changes the electrical path, document that effect in the interpretation. A passed bare-board inspection establishes the inspected substrate requirements; it does not establish assembled switching performance or product qualification. Link the evidence from each stage so that the final design review can distinguish an accepted ceramic circuit from a module whose intended behavior has actually been demonstrated.

Release the design with a traceable set of tradeoffs

The released DBC package should contain the node map, loss map, finished geometry requirements, assembly constraints and the evidence supporting the selected layout. Include the operating conditions covered by the analysis and measurement, plus any limitations that affect future substitutions. A later request to move a terminal, enlarge a pad or change the ceramic should trigger review of the functions connected to that feature. This makes the layout reusable as an engineered design rather than a static drawing whose original rationale has been lost. For a solar string inverter, that rationale connects the changing electrical workload to heat removal, insulation and a manufacturing process capable of delivering the intended construction.

Engineering example

Numbers are invented engineering examples; no module rating, spacing requirement or supplier performance is inferred.

Before you release the design

  • Map losses by switching position and operating point.
  • Trace full commutation loops through external connections.
  • Evaluate intended symmetry and current sharing.
  • Retain coupling in thermal comparisons.
  • Control finished copper, insulation and assembly requirements.

Sources and further technical reading

Manufacturer references support the material and process context. Worked examples and checklists are engineering guidance; they are not test results or supplier guarantees.

  1. Infineon — NPC1 inner and outer switches; position-dependent loss example
  2. Semikron — SKiM63/93 Technical Explanations,rev1.53; electrical behavior and DC-link connection sections