Why High-Frequency Transformer Prototypes Fail EVT Tests: 7 Checks Before the Next Build

Short answer: When a high-frequency transformer fails EVT, the root cause is often not one bad component. The winding construction, converter operating point, PCB parasitics, clamp circuit, insulation requirement, and thermal test setup interact. Before ordering another sample, compare the failed unit against seven items: actual switching conditions, magnetizing inductance, leakage inductance, winding capacitance, core flux margin, copper loss, and the insulation/mechanical build.

Replacing the transformer without identifying which parameter moved usually produces a second sample with a different problem.

Start With the Failure Symptom, Not a New Core Size

A typical EVT review starts with a message such as:

• “The MOSFET drain spike is too high.”
• “The transformer reaches 105°C after 20 minutes.”
• “The board passes function testing but fails conducted EMI.”
• “The auxiliary output rises when the main output is lightly loaded.”
• “Two samples work, but the third one saturates during startup.”
• “The transformer passes hipot but does not meet the required creepage distance.”

These symptoms do not point to the same correction. Increasing the core size may help one thermal problem while making the PCB layout worse. Interleaving windings may lower leakage inductance but increase primary-to-secondary capacitance. Adding turns may reduce flux density but increase copper resistance and winding height.

The first step is to record the failure under a repeatable operating condition. The second is to connect that symptom to the transformer and circuit parameters that can actually cause it.

Which Transformer Problems Produce Which EVT Symptoms?

This table is a triage tool, not a substitute for measurements. Several faults can appear at the same time.

Check 1: Did the Prototype Run at the Conditions Used for the Magnetic Design?

Before changing the transformer, compare the real converter waveforms with the assumptions used to design it.

Record:

• Minimum, nominal, and maximum input voltage
• Continuous and peak output load
• Actual switching-frequency range
• Maximum duty cycle
• Peak primary current
• Startup and overload behavior
• Burst, skip, valley-switching, or frequency-foldback operation
• Ambient temperature, airflow, and enclosure condition

A transformer designed around a nominal 100 kHz switching frequency can see a different flux swing if the controller drops to a lower frequency during startup or overload. A part that stays cool on an open bench can overheat in the sealed enclosure. A sample approved at nominal input may saturate at minimum input, where primary current is highest.

Capture the drain voltage, primary current, gate signal, and relevant output waveforms on the same test record. Do not send only a thermal image or a photograph of the oscilloscope without probe scale, timebase, input voltage, and load.

Power Integrations warns in its first-time flyback bring-up notes that breadboards and uncontrolled layouts introduce parasitic elements and make creepage and clearance difficult to control. EVT measurements should be taken on the intended PCB or a layout that accurately represents it.

Check 2: Is Magnetizing Inductance Correct Under the Agreed Test Method?

Magnetizing inductance controls current ramp and stored energy in flyback designs. A value written as “450 µH” is not enough to compare two samples.

Confirm all of the following:

• Measurement frequency
• Test voltage or current
• Series or parallel measurement mode
• Which windings are open
• Temperature of the sample
• Allowed tolerance
• Whether the measurement is taken before or after varnish, potting, and final core clamping

Then compare the LCR result with the actual primary-current waveform.

A transformer can meet a small-signal inductance specification yet have inadequate saturation margin at the converter’s peak current. Air-gap tolerance, core permeability, gap contamination, assembly pressure, and turns count all affect the result.

If samples vary noticeably, ask for individual measurements rather than only a lot average. For a gapped core, the air-gap process and inspection method deserve the same attention as the winding turns.

Check 3: Is Leakage Inductance Driving the Drain Spike?

Leakage inductance stores energy that is not transferred through the ideal coupled path. When the primary switch turns off, this energy contributes to the voltage spike and is then dissipated or recovered by the clamp circuit.

Measure leakage inductance using a defined method:

• Measure from the specified winding
• Short the specified secondary windings
• State the test frequency and voltage
• Use the same lead length and fixture
• Record a maximum value, not only a typical value

Commercial transformer datasheets often state exactly how leakage is measured. For example, Coilcraft specifies primary leakage with secondary windings shorted on its MA5475-AL flyback transformer page. That level of test definition is necessary when comparing samples from different builds.

If leakage is high, review:

• Primary and secondary winding order
• Number of layers
• Insulation thickness between coupled windings
• Margin tape width
• Lead routing around the bobbin
• Whether a split-primary arrangement is allowed
• Whether triple-insulated wire can reduce the required insulation buildup

Do not reduce leakage without checking capacitance and safety. Tighter coupling often brings primary and secondary conductors closer together, increasing common-mode current.

Check 4: Did Lower Leakage Create a New EMI Problem?

A winding structure can improve the drain waveform and still fail EMI.

Interleaving or sandwich windings generally improve coupling, but they can increase primary-to-secondary capacitance. That capacitance creates a path for high-frequency common-mode current. The result may be higher conducted noise even though the leakage spike is smaller.

When EMI changes after a transformer revision, compare:

• Primary-to-secondary capacitance
• Winding sequence and layer boundaries
• Shield winding position and termination
• Core grounding method, if used
• Y-capacitor value and return path
• Drain-voltage slew rate
• Common-mode choke and input filter behavior

Do not ask the winding supplier to “minimize leakage and capacitance” without ranking the priorities. The supplier needs to know which limit is critical and what trade-off the circuit can accept.

A useful revision request might say:

Reduce primary leakage below the agreed maximum while keeping primary-to-secondary capacitance below the EMI limit measured at the stated frequency. Do not change insulation margins or hipot construction without approval.

That is actionable. “Improve EMI” is not.

Check 5: Is the Temperature Rise Caused by Core Loss or Winding Loss?

A hot transformer does not automatically need thicker wire.

Separate the main loss mechanisms:

Core loss indicators

• Core temperature rises faster than the outer winding
• Temperature increases sharply with switching frequency or input voltage
• No-load or light-load heating is higher than expected
• Primary current waveform suggests excessive flux swing or asymmetry

Copper loss indicators

• Winding temperature is highest near dense layers or termination points
• Temperature rises mainly with load current
• DCR is above target
• High-frequency AC resistance is significant despite acceptable DC resistance
• Parallel strands do not share current as expected

At high frequency, skin effect and proximity effect increase AC resistance. Simply increasing conductor diameter can make AC loss worse if the strand becomes too thick for the operating frequency. Depending on current and geometry, the better correction may be multiple strands, foil, litz wire, a different layer arrangement, or a larger winding window.

Temperature data should identify where the reading was taken and under what conditions. Report ambient temperature, enclosure, airflow, input, output load, stabilization time, and whether the measurement came from a thermocouple, resistance method, or infrared camera.

Check 6: Does the Core Have Enough Flux and Saturation Margin?

Core saturation can appear only at startup, low line, overload, or elevated temperature. That makes it easy to miss during a nominal-condition test.

Review:

• Core material and effective area
• Primary turns
• Maximum volt-seconds
• Air gap
• Peak magnetizing current
• DC bias
• Reset condition for non-flyback topologies
• Minimum switching frequency during abnormal modes
• Core temperature

For a flyback design, the current ramp and energy storage must remain within the core and gap capability. For forward, bridge, and push-pull converters, flux balance and reset are equally important.

Do not approve a transformer only because it does not saturate in a brief room-temperature bench test. Check the worst input, load, frequency, duty cycle, and temperature combination defined for the product.

Check 7: Was the Insulation Build Designed for the Final Product Standard?

A transformer can pass a hipot test and still be unsuitable for the end product.

Hipot voltage is only one part of the insulation design. The review must also cover:

• Working voltage
• Basic, supplementary, or reinforced insulation
• Creepage and clearance
• Pollution degree
• Overvoltage category
• Maximum operating altitude
• Material group
• Bobbin construction
• Winding margins
• Wire insulation system
• Lead routing and sleeving
• Applicable end-product standard

IEC 61558-2-16:2021 addresses safety requirements and tests for switch-mode power supply units and their transformers. The final construction still has to match the product category and market requirements.

If reducing winding height or leakage requires a thinner insulation build, safety requirements must be checked before the revised sample is wound. This is not a cosmetic change.

What Data Should You Send for a Useful Failure Review?

A message saying “sample gets hot, please improve” starts another guessing cycle. A useful failure package can be assembled in one folder.

1. Converter information

• Topology and controller part number
• Schematic revision
• PCB revision
• Input range
• Output rails and load conditions
• Switching-frequency range
• Current-limit setting
• Clamp or snubber circuit

2. Transformer information

• Drawing and revision
• Core and bobbin
• Winding turns and sequence
• Wire sizes
• Magnetizing-inductance result and test condition
• Leakage-inductance result and test condition
• DCR by winding
• Insulation and hipot requirement

3. Failure evidence

• Exact test condition
• Waveforms with scales and probe points
• Thermal image plus measured ambient
• EMI plot with limit line and setup
• Failed sample identification
• Whether the fault follows the transformer when units are swapped

4. Approval target

• Maximum drain voltage
• Maximum transformer temperature or temperature rise
• Efficiency target
• EMI margin target
• Required safety construction
• Mechanical envelope
• Date and quantity required for the next build

This information allows Chengpin Technology to review the winding structure against the observed failure instead of treating the transformer as an isolated black box.

How to Decide Whether the Next Sample Needs a Winding Change

The fastest route is not always a new transformer. Sometimes the correct action is a snubber change, PCB revision, probe correction, or controller adjustment. A supplier who immediately promises to “fix the transformer” without asking for operating data may be solving the wrong problem.

Frequently Asked Questions

Why does the MOSFET spike change when the turns ratio is unchanged?

The turns ratio sets the reflected voltage, but the turn-off spike is also affected by leakage inductance, current at turn-off, clamp behavior, PCB parasitics, winding capacitance, and measurement technique. Two transformers with the same turns ratio can produce different spike behavior.

Can leakage inductance be calculated accurately from the drawing?

It can be estimated, but production acceptance should use measurement. Power Integrations notes that leakage depends on the actual transformer construction and should not be assumed as a fixed percentage of magnetizing inductance; see its technical discussion on leakage measurement.

Will a larger core always reduce transformer temperature?

No. A larger core may reduce flux density and create more winding space, but total temperature also depends on core material, turns, conductor choice, winding arrangement, airflow, enclosure, and operating waveform. It can also increase cost and violate the mechanical envelope.

Why does the sample pass on the bench but overheat in the product?

Open-bench airflow, different ambient temperature, short test duration, cable losses, and enclosure heat recirculation can hide the real operating condition. Repeat the test inside the intended mechanical assembly or under an agreed thermal simulation.

Should the primary and secondary always be interleaved?

No. Interleaving can lower leakage inductance, but it may increase primary-to-secondary capacitance and complicate the insulation system. Use it when the electrical and safety trade-off supports it.

What is the minimum data needed for a transformer redesign?

At minimum: topology, controller, input range, output requirements, switching-frequency range, current limit, transformer drawing, magnetizing and leakage measurements, failure waveforms, thermal condition, insulation requirement, and mechanical limits.

How many failed samples should be returned for analysis?

Return enough units to distinguish a design issue from sample variation, and include matching test records and unit identification. The exact number should be agreed with the supplier based on the failure mode and available samples rather than chosen arbitrarily.

Before You Order the Next EVT Build

Freeze the test method before freezing the transformer revision.

Agree on:

1. The exact operating points used for approval
2. Magnetizing and leakage-inductance measurement conditions
3. Maximum drain-voltage and thermal limits
4. Winding-capacitance or EMI constraints
5. Insulation construction and safety requirement
6. Mechanical envelope and pinout
7. The first-article data that must accompany the samples

Then issue a controlled drawing revision and keep the failed sample results attached to it.

If your current EVT build has a drain spike, temperature-rise, EMI, regulation, or insulation problem, send Chengpin Technology the transformer drawing, schematic revision, operating conditions, and measured failure data. A focused DFM review can determine whether the next action belongs in the winding, magnetic design, insulation build, or converter itself.

Technical References

1. Power Integrations, Bringing Up a Flyback Supply for the First Time.
2. Power Integrations, LinkSwitch-LP Flyback Design Guide AN-39.
3. Coilcraft, MA5475-AL Flyback Transformer Specifications.
4. IEC, IEC 61558-2-16:2021 — Safety of Switch Mode Power Supply Units and Transformers.