GaN First Time Right™ Design Process

GaN FETs high performance emphasizes the need for good measurement techniques for high-speed circuits.

1. The ground loop should be minimized by utilizing a spring clip
   

   

2. Probing location should be kept as close as possible to the device being tested
   

   

   

If scopes or probes with insufficient bandwidth are used, then actual waveforms of a typical converter cannot be measured accurately. A 500MHz bandwidth is recommended for typical converters, and at least 1 GHz for some specific applications like LIDAR.

Of particular interest is the measurement of the high side gate in a typical half-bridge configuration. On top of the previous requirement in terms of bandwidth and measurement setup, this measurement presents additional requirements:

1. Galvanic isolation: although the math channels can be used to reconstruct the high side gate, this method is susceptible to noise and mismatch between the two probes. A differential probe is recommended.
2. Large comon mode rejection ratio (CMMR)
3. Common mode voltage rating > input voltage (Buck) or output voltage (Boost)
4. Large input impedance, preferably > 10 MΩ || < 2pF 

Test equipment manufacturers have developed high performance differential probes suitable for this: for example Tektronix IsoVu, LeCroy DL-ISO, and PMK Firefly probes.

This measurement method is commonly used to directly measure switching losses of semiconductor devices by using the math function of a scope to multiply the instantaneous voltage and current waveforms and then integrating them. The previous methods can be applied to measure the voltage, however measuring the current has these additional challenges:

• Bandwidth requirement: active current sensors struggle with the accuracy and bandwidth required, so current shunts are still the preferred method
• Current shunts require interrupting the power loop and inserting the sensor. The increase in power loop inductance can significantly change the measurement results

For these reasons EPC does not recommend double pulse testing, but rather utilizing Spice models (and a calibrated model if more accuracy is needed): EPC Device Models

Test equipment manufacturers are working on this topic, for example see the article Accurate Characterization of Low-Voltage, Small-Form–Factor GaN FETs.