L6389ED Half-Bridge Board Design and Validation

L6389ED Half-Bridge Board Design and Validation

Validate an L6389ED half-bridge through bootstrap budgeting, gate-drive measurements, timing checks, and controlled board bring-up.

Last updated: November 2026

Bottom Line: Build an L6389ED half-bridge around three checks: a reliable floating gate supply, gate-drive timing verified on the real switching devices, and a layout that keeps high-current gate loops and switching-node coupling under control. The driver combines high-side and low-side functions, but its voltage capability and internal protection features do not establish the rating or robustness of the complete power stage. Begin with a current-limited reduced-voltage assembly, validate startup and shutdown, and then repeat the critical measurements across the intended operating conditions. Approve the production design only when the waveforms, thermal behavior, and fault response satisfy documented limits.

Establish the device baseline before drawing the schematic

The L6389ED is an STMicroelectronics half-bridge gate-driver ordering option in an SO-8 package. ST's L6389E product page describes its floating section as supporting a rail up to 600 V and identifies separate high-side and low-side outputs. That figure is a device capability under the manufacturer's conditions, not a recommended bus voltage or a system isolation specification.

Use the current L6389E datasheet to control pin assignments, electrical limits, and timing assumptions. The manufacturer identifies source and sink capabilities of 400 mA and 650 mA respectively. Those peak current figures do not mean a gate can be charged with constant current throughout its entire switching transition.

The public catalog entries for L6389ED and L6389EDTR are useful sourcing references. Check the complete ordering code and packaging against the assembly process, rather than treating the suffix as an electrical upgrade. Put the controlled drawing revision and the required shipping format on the bill of materials.

Before routing, create a small verification sheet containing the chosen switching device, gate-charge information, driver supply, PWM behavior, operating frequency, and expected switching-node excursions. Separate values guaranteed by the component documentation from assumptions that the assembled board still needs to prove.

Design the floating supply for the complete operating sequence

The bootstrap supply must deliver enough charge for every intended high-side interval and must recharge when the switching sequence permits it. The internal bootstrap function reduces external component count, but it does not eliminate the charge budget or the need to validate the operating sequence.

Start with the charge relationship C >= Q / deltaV. As an illustrative calculation, if the total required charge during a high-side interval is 30 nC and the allowed bootstrap droop is 0.5 V, the ideal minimum capacitance is 60 nF. That total must include the switching-device gate charge and the other relevant charge consumption over the interval.

The calculation is a starting point rather than a component recommendation. Account for effective capacitance under bias, temperature, tolerance, leakage, recharge impedance, and startup conditions. Use the switching device's gate charge at the actual operating conditions instead of assuming its input capacitance is an adequate replacement for the charge curve.

Validate the longest intended high-side on-time, the shortest available recharge interval, and every low-frequency operating mode. A design can work at its nominal PWM pattern yet fail during startup, burst operation, or a slow control-loop transition. Record the bootstrap voltage relative to the switching node so the measurement represents the supply that the floating output actually sees.

Check the initial charging sequence explicitly. If the control system starts by requesting a high-side pulse before the floating supply has charged, the resulting behavior may differ from steady-state operation. Make the startup sequence part of the controller requirements and verify it again after any firmware change that affects PWM initialization.

Size the gate path around charge and switching loss

Gate resistance and loop impedance should be chosen from measured switching behavior. The driver output, external resistor, switching-device internal resistance, and PCB inductance form one path. Changing only the external resistor can reduce ringing while also increasing switching loss, so the final choice needs both waveform and temperature evidence.

A first-order transition estimate is t approximately Q / I. For an illustrative 20 nC charge interval and 0.2 A average drive current, the estimated time is 100 ns. The Miller region and the driver's changing output impedance make this an approximation; use it to check plausibility rather than predict an exact edge.

Estimate the gate-drive power as P approximately Qg times Vdrive times frequency for each switch. With an illustrative 20 nC gate charge, 12 V drive, and 50 kHz operation, the charging energy rate is about 12 mW per gate. This is not the complete driver thermal calculation because internal consumption and the distribution of dissipation along the gate path also matter.

Choose an initial resistor that supports controlled bring-up, then adjust it using measured switching loss, overshoot, and false-turn-on behavior. If turn-on and turn-off requirements differ, evaluate a controlled asymmetric gate network and document the resulting current paths. Do not add a steering network without checking what happens when the driver is unpowered or the controller resets.

Use the gate waveform at the switching device as the acceptance measurement. A clean waveform at the driver pin does not prove that the gate-source voltage at the transistor is clean. The common-source return and the actual probe connection can make those two measurements differ substantially.

Verify deadtime together with the controller

The complete half-bridge must maintain acceptable non-overlap under the actual controller and switching conditions. ST describes interlocking and deadtime functions in the L6389E family, but these features do not remove the need to examine propagation, gate discharge, and switching-device behavior together.

Start with the intended PWM truth table and define what happens when the controller is starting, resetting, or deliberately disabling the stage. Test every allowed state transition, including transitions that occur at inconvenient points in the switching cycle. Avoid assuming that a software state named off immediately discharges every physical gate.

Excessive deadtime increases time spent in an unintended conduction path and can degrade efficiency or waveform quality. Insufficient effective deadtime can permit destructive overlap. The correct setting depends on the power devices, their gate networks, the load current, and the operating conditions rather than on a copied register value from a different board.

Measure both gate-source waveforms and the switching node with a consistent timing reference. Compare the results at the conditions where turn-off is slowest and unwanted turn-on is most likely. Include the selected transistor tolerances and the controller timing resolution in the documented margin.

Keep the power and gate returns deliberate

A compact gate-current loop reduces the parasitic voltage that corrupts switching measurements and behavior. Put local supply decoupling close to the driver and connect its return through the intended low-impedance path. Keep the switching-node copper away from sensitive logic routing where practical.

Route the high-side gate and its source reference as a controlled pair of paths. The floating driver does not reference quiet logic ground during switching. A layout that returns its current through an accidental shared path can create a misleading gate voltage or inject noise into the controller.

Review the low-side gate return separately from the main load-current path. Where the switching device provides an appropriate source reference connection, use it according to its drawing and the board architecture. Otherwise, minimize shared impedance and verify the remaining coupling on the assembled board.

Measure switching-node overshoot and undershoot using a probe arrangement designed for the measurement. Long probe-ground leads can create apparent ringing that the circuit does not produce. Compare the observed waveform with the instrumentation bandwidth and connection method before adding a snubber to fix a measurement artifact.

Choose a documented implementation path

Use the existing board requirements to choose among the exact ordering options and the broader driver family. The following catalog entries are useful starting points for evaluation; related names do not establish pin compatibility, identical input behavior, or a qualified substitution.

Implementation path Candidate Main advantage Main tradeoff Approval requirement
Controlled baseline L6389ED Preserves the selected driver design Requires complete board validation Verify the exact device drawing and measured waveforms
Production packaging review L6389EDTR Allows review of the alternate ordering format Packaging details still require confirmation Match order code, reel format, and assembly needs
Related-family redesign L6388ED Provides another documented family candidate May require schematic and control changes Compare its own datasheet before any substitution

For the controlled baseline, use L6389ED while the electrical design is still being established. Keep the initial gate network and controller settings easy to change, and reserve measurement access to the floating supply, both gates, and the switching node. This path minimizes simultaneous variables during early validation.

For a production ordering review, compare L6389ED and L6389EDTR against the manufacturer's ordering information and the contract manufacturer's handling requirements. Confirm that the proposed quote names the approved silicon and shipping format. Packaging convenience is valuable, but it should not silently change the component approval record or hide a supplier substitution.

For a related-family redesign, assess L6388ED with its own documentation and a complete schematic comparison. Check logic polarity, input behavior, pinout, timing, startup, and protection operation rather than relying on the family prefix. A properly reviewed redesign can be useful when a different interface or procurement option is needed, but it requires validation as a new implementation.

Keep comparison measurements on a common fixture and load condition. Record the controller version, power devices, resistor values, bus condition, and instrument setup for each candidate. Otherwise, the comparison can accidentally attribute a layout or firmware change to the driver itself.

Common pitfalls and troubleshooting

Bootstrap collapse during long on-time should be diagnosed as a supply and timing problem before changing the driver. Measure the floating supply relative to its local return through the complete interval. Check effective capacitance, recharge opportunity, and leakage before selecting a larger capacitor that may still fail to charge during startup.

Unwanted gate turn-on often indicates coupling through the power stage and gate return. Observe the affected gate-source voltage while the opposite transistor switches, and compare it with the switching-node transition. Review the return path, gate discharge strength, and power-device behavior together instead of treating a gate resistor as a universal cure.

Unexpected heating can come from slower switching, excessive conduction in unintended paths, or a supply problem. Compare waveforms and temperature at the same load and operating frequency. Changing several gate-network values at once makes it difficult to identify which mechanism caused the improvement or regression.

Intermittent startup failures often appear only when the controller and power supplies rise in an unusual sequence. Reproduce cold starts, rapid restarts, controller resets, and interrupted input power. Capture the logic inputs and driver supplies together so the team can distinguish a control request from the driver's actual ability to produce it.

A clean low-voltage result does not complete validation at the intended operating condition. Switching-node movement changes capacitive coupling and can reveal a layout weakness that was invisible during initial bring-up. Increase the operating stress through a documented test sequence and recheck the quantities that establish margin.

Frequently asked questions

Does the driver's rail capability define the power-stage rating?

No. The finished stage depends on switching-device ratings, transient behavior, insulation design, PCB spacing, thermal performance, and the requirements of the final equipment. The gate driver's floating-section capability is one constraint within that design. Establish the operating envelope from the complete system requirements and verify the relevant device limits under actual switching conditions, including startup and abnormal operating states.

Can the high-side output remain on indefinitely with a bootstrap supply?

A bootstrap architecture requires a defined source of replenishment for its floating supply. Whether a particular on-time is achievable depends on total charge consumption, capacitance, leakage, and the available recharge sequence. Calculate the worst-case interval and measure the resulting floating supply. If the application requires an operating mode that cannot replenish it, review a different supply architecture rather than assuming the capacitor alone solves the problem.

Are L6389ED and L6388ED interchangeable on an existing board?

Treat them as separate approval candidates. Compare their complete drawings, electrical behavior, control interface, protection functions, and package information before deciding whether a replacement is possible. A shared family and package description are insufficient evidence. If the comparison reveals any difference affecting the board or firmware, document the change and repeat the relevant validation rather than approving the substitute solely through purchasing.

Which measurements belong in the production design record?

Keep the floating supply, driver supply, both gate-source waveforms, switching-node behavior, and thermal results at the operating conditions that establish the smallest margin. Include startup, shutdown, and the defined abnormal states. Record the measurement method and controller configuration with the results. The record should allow another engineer to reproduce the acceptance decision and identify which assumptions must be rechecked after a component or firmware change.

Release the board with a controlled component inquiry

A dependable L6389ED half-bridge combines a verified supply sequence, a measured gate-drive path, deliberate timing, and a layout that preserves the intended references. Keep the component documentation and the assembled-board evidence together so later sourcing changes do not bypass the original engineering decision.

Review available driver catalog entries and request a quotation with the exact ordering code, packaging, quantity, required delivery, and accepted documentation. Include any traceability or lot requirements that affect production release. That gives purchasing a concrete component definition while keeping the final design approval grounded in measured board behavior.