Isolated Gate Drivers: Checking CMTI in a Fast Half-Bridge

Quick answer

CMTI is a dynamic common-mode immunity quantity. It describes the common-mode rise or fall rate that an isolated driver can tolerate while preserving the required output state, under a stated test condition. It is not the same as isolation withstand voltage or normal working voltage. For isolated gate driver CMTI half bridge dv dt, review the barrier references, polarity, input state, output load, temperature and measurement position together. The safest purchasing conclusion is conditional: request a matched evidence set, keep the measurement or thermal scope explicit, and do not rank or approve a module from a single family number. The records should state what was measured, under which conditions, and what remains outside the evidence. Clamp functions need the same discipline; see separating Miller clamp claims from evidence.

1. Define the decision before reading a datasheet

Start with the decision the buyer must make. Is the goal to compare two switching events, review a pulsed thermal limit, qualify an isolated control interface, interpret a low-load symptom, or request a suppression design? Each goal needs a different record. Write the selected part number, revision, topology and application boundary at the top of the review. A family page can identify a commercial candidate, but it does not establish the selected die, package option, waveform or system approval.

2. Build a condition register

A compact condition register prevents false precision. Record the electrical quantity, its unit, the time interval, the temperature definition and the instrument or source document. For a switching comparison this means bus voltage, current definition, device temperature, driver rails, gate resistance, parasitic/layout description, probe model, bandwidth and delay correction. For a thermal review add pulse width, period, duty, loss-power evidence, initial case condition and the chosen Zth scope. For CMTI add common-mode amplitude, rise/fall direction, input state and correct-output criterion. For an SCR review add the selected IL or IH definition, gate-pulse record, load-current trace and the event that caused the observation.

3. Keep similar numbers in separate columns

The most common review error is placing unlike quantities in one ranking column. A current peak is not a current slope. Isolation withstand is not CMTI. Steady thermal resistance is not transient impedance. A family IT(RMS) label is not a selected latching or holding current. Create separate columns and require a source and condition for each. If a row cannot be completed, mark it as an evidence gap rather than interpolating from a neighbouring model.

4. Use an explicit, limited calculation

A small calculation can expose missing information without pretending to qualify equipment. As an illustrative example only, a 100 A change in 10 microseconds has an endpoint-average slope of 10 A/µs, while the same 100 A change in 2 microseconds has 50 A/µs. Those values say nothing about local overshoot, waveform shape, device limit or safe operation. Likewise, if a hypothetical loss of 200 W were paired with a hypothetical 0.05 K/W transient impedance, the arithmetic rise would be 10 K; it is not a vendor curve or a selected module result. Label assumptions beside every derived value.

5. Read source scope and limitations

Use manufacturer or standards material for the principle, then return to the selected product documentation for the actual limit. The Infineon double-pulse article explains why bus inductance, driver differences and instrumentation affect correspondence; it does not qualify either RITOKS family. Infineon’s thermal note distinguishes resistance from time-dependent impedance, but its OptiMOS examples are not an XI900HB120N3L, XS004HB120 or SCR module curve. TI’s CMTI application note explains measurement-position effects, including a documented room-temperature example, not a universal probe rule. ST’s AN303 defines latching and holding concepts with temperature and gate-pulse conditions; its TRIAC examples cannot be copied as MTC/MTA limits. Littelfuse’s thyristor compendium discusses current spreading, recovery and dv/dt in a high-power disc context, not a generic RC or gate recipe.

6. Turn the review into a buyer-ready request

A useful request names the selected order code and asks for the missing correspondence. Ask for the exact datasheet revision, topology, electrical quantity definition, reference temperature, repetitive or non-repetitive scope, gate or driver condition, waveform/raw-record provenance, instrument correction and the owner who can close an open item. Link the request to the relevant RITOKS family page so a supplier can identify the intended commercial range, while keeping the requested proof separate from a marketing description. Do not ask a supplier to confirm an approval that the available evidence cannot support.

7. What not to infer

Do not infer interchangeability from a shared package label, a similar current class or the fact that two devices are half bridges. Do not convert an isolation voltage into CMTI, a steady Rth into a pulse limit, a peak current into di/dt, or a low-load symptom into a single failure cause. Do not copy a competitor’s curve, certification or test condition to a RITOKS product. Do not turn a conceptual diagram into an energized test procedure. These boundaries protect both the buyer and the supplier from an apparently precise but untraceable conclusion.

8. A practical review worksheet

Use four passes. Pass one checks identity: part, revision, topology and source date. Pass two checks quantities: units, definitions, interval and reference condition. Pass three checks correspondence: driver, load, thermal boundary, measurement chain and repetition. Pass four checks the decision: what can be compared, what remains open and whether the next action is a technical question, a quotation request or a hold. The worksheet is successful when another engineer can reproduce the reasoning without guessing which number was intended.

Conclusion

For isolated gate driver CMTI half bridge dv dt, a defensible result is often a scoped request rather than a universal ranking. Keep the direct answer near the top, preserve the conditions beside every number and make unresolved evidence visible. That approach lets a buyer compare the XI900HB120N3L, XS004HB120N6, MTC or MTA families for the right technical question without turning a general source into an unsupported product guarantee.

The distinction matters in procurement because a supplier can answer a narrow question accurately while a reviewer silently broadens it. State the boundary in the request and keep the returned evidence attached to that boundary.

When two records disagree, first check whether the event definition, temperature and revision differ. A disagreement is often a scope mismatch rather than proof that one component is universally better.

A review can record confidence as confirmed by the selected document, supported by a general principle, illustrative only, or still requested. This makes later design changes auditable without pretending that a preliminary comparison is qualification.

Use the product page to identify the commercial family and the contact route, while keeping the selected technical report as the authority for limits and conditions.

For a quotation, put the unanswered field beside the consequence of leaving it blank. That makes a follow-up question easier to prioritize than a long generic checklist.

If the evidence owner returns a revised document, preserve the old revision in the review log and explain which conclusion changed. Silent replacement makes later comparisons unreliable.

A clean comparison also states what was not measured. Absence of a trace, curve or temperature record is a limitation, not evidence that the value was nominal or typical.

Keep conceptual graphics separate from installation instructions. Their purpose is to show relationships among quantities, not to authorize a live experiment or a protection setting.

A buyer can use the worksheet during supplier review, then attach the completed fields to an RFQ. The attachment should identify assumptions so the answer remains auditable after the project changes.

Where a standard or manufacturer note uses a special definition, quote that definition in plain language before using its abbreviation. Similar acronyms often hide different test boundaries.

For this topic, the first useful follow-up is the document that owns the limiting quantity, not a larger catalogue table. Ask for that document by model and revision.

A record with a precise unit but no interval is incomplete. Time, temperature and repetition often determine whether the number describes a transient, a repetitive event or a steady condition.

If a supplier supplies only a typical curve, retain it as context and request the selected-device limit separately. Typical information should not silently become a guaranteed acceptance value.

The worksheet separates a commercial choice from an engineering approval. It can narrow the next RFQ while leaving final application responsibility with the qualified design team.

Documenting the negative result is valuable: a comparison may conclude that the evidence is not comparable yet. That prevents an early procurement preference from becoming an undocumented design assumption.

A revision-controlled folder should keep the source PDF, extracted claim, condition row and reviewer comment together. This is more durable than copying a number into a spreadsheet without provenance.

For buyers working across vendors, use the same field names but do not assume the vendors define them identically. The definition and test boundary should travel with each value.

The strongest conclusion is often a conditional sentence: if the selected record matches the stated boundary, the comparison is informative; if it does not, request a new record before ranking options.

A conceptual visual can show that two curves share an axis while representing different scopes. It should not imply that their unlabeled shapes are measured data.

Finally, separate an information request from a field modification. The article supports better questions and clearer evidence; it does not authorize changing a gate, snubber, protection threshold or operating condition.

Source notes

Primary sources are listed in dossier.md; this local package does not authorize publication.