How to measure fast power switches

Silicon power devices in regulators and DC-DC power supplies will soon be replaced by GaN FETs. Compared with silicon MOSFET, its switching speed is much faster, and RDS(on) is lower. This will enhance the power efficiency of the power supply and bring benefits to everyone. If you are designing a power circuit with a GaN device, you need to know the switching speed of the device. To measure this speed, the oscilloscope, probe, and interconnect must be fast enough to minimize their impact on the measurement.

    Regarding device performance, the question I am often asked is "How fast are they?" Usually I will answer: They are very fast, but in fact I don't know how fast they are. To find out the truth, I used a 33GHz real-time oscilloscope and a high-speed transmission line probe to measure it. I will discuss the design constraints that affect the speed of the device and its future development prospects. After these measurements, I believe we will soon be able to design a power supply with a switching speed of 250MHz.

    Figure 1 shows the two evaluation boards used to make the measurements. Both evaluation boards are equipped with a gate regulator, a driver, a pulse regulator and two eGaN switches. The circuit board on the right is a complete DC-DC converter, which contains a Gen4 monolithic half bridge (both are switched on the same wafer), and contains an L-C output filter. The evaluation board on the left uses a separate Gen3 eGaN device in a half-bridge configuration without an L-C output filter. In both cases, the external pulse generator provides the PWM signal by soldering to the BNC connector of the pulse width modulation (PWM) input of the test board. In the case of input voltages of 5V and 12V, I measured the switch rise time on each evaluation board.



    Figure 1: Here only the half-bridge configuration is equipped on the circuit board on the left, and the circuit board on the right is equipped with a complete DC-DC converter. The banana socket can connect the test board to the electronic load. It can be connected to an external pulse generator through the BNC connector.

    Instrument and probe requirements

    In order to ensure that the instrument and probe will not have a significant impact on the measurement, we can assume that the rise time of the probe, oscilloscope, and half bridge can be added by the root-sum-square method. Although this method is not always correct, we can assume that this relationship holds in our initial estimation.

    The measured rise time of the half bridge includes the rise time of the oscilloscope and the rise time of the probe, and is:

    The actual rise time of the half bridge can be determined according to the following formula:

    In order to limit the measurement error to a certain percentage K, the rise time of the instrument can be correlated with the actual rise time:

    Solving for K, the ratio of the rise time of the instrument to the rise time of the actual half-bridge is:

    Therefore, for these two examples, if we want the measurement result to be less than 5% or 10%, the rise time of the oscilloscope and probe must be less than 32% or 46% of the rise time of the FET, respectively. In other words, the rise time of the instrument should be 3.1 or 2.2 times faster than the rise time of the FET, respectively.

    Measuring switch performance

    The oscilloscope used here is a Keysight 90000-X series 33 GHz oscilloscope with Teledyne LeCroy PP066 transmission line probe. The oscilloscope and the probe are connected through a 50 GHz Huber+Suhner Sucoflex-100 cable. The rise time of this setting is recorded using a 20ps fast edge pulse, and the result is shown in Figure 2. In order to ensure that the measurement is valid, the rise time of the oscilloscope and probe used to make these measurements is much faster than the above-mentioned values, so a "perfect measurement" can be achieved.

    Figure 2: Using a 33GHz Keysight Infiniium 90000-X oscilloscope equipped with Huber Suhner Sucoflex 100 50GHz cable and Teledyne Lecroy PP066 transmission line probe, the measured edge pulse rise time is about 20ps. The measurement result shows that the rise time of the test setup is less than 27.69ps, including the 20ps pulse rise time.

    The resulting 27.69ps rise time includes the 20ps pulse rise time, which can be subtracted by the root-sum-square method to determine the rise time of the oscilloscope, probe, and cable. In the case of subtracting the pulse edge, it can be completely determined that the device rise time is less than 27.69ps, so we can use it to make a conservative estimate.

    Based on previous calculations and using a conservative estimate of the instrument's rise time of 27.69ps, we can measure the rise time of the half bridge in the K% range.

    The measurement setup can measure 276ps with 0.5% accuracy and 619ps with 0.1% accuracy. The complete instrument setup is shown in Figure 3.

    Figure 3: Display of the complete instrument setup for the DC/DC converter. The input voltage of the test board is adjusted to 12V, and the power supply voltage of the gate drive regulator is 7V. The load is shown at the bottom right. Keysight 90000-X oscilloscope, Teledyne Lecroy PP066 transmission line probe and Huber Suhner Sucoflex 100 cable can also be seen on the figure.

    Measured performance

    Figure 4 shows the rise time of the DC-DC converter when the output voltage is approximately 1V and the load current is 20.0A. The measurement is performed under the condition that the input voltage of the test board is 5V and 12V.

    Figure 4: When the input voltage is 5V and 12V, the rise time measured on the test board is 682.33ps and 561.13ps, respectively. The working load of the DC/DC converter is 20.0A.

    The measured rise time is shown in Figure 5. When the half-bridge is measured separately, it is also carried out under the input voltage of 5V and 12V.

    Figure 5: When the input voltage is 12V and 5V, the measured rise times are 538.87ps and 332.68ps, respectively. This is only half bridge, so there is no load.

    According to previous calculations, under the condition that the rise time of the probe and the oscilloscope is 27.69ps and the measured fast rise time is 332.68ps, all four measurement results are within 0.5% accuracy range. The results are shown in Table 1.

    Table 1. Summary of test results

    These measured rise times are about 3 times faster than equivalent silicon MOSFETs, and RDS(on) is about 1/3. Normally, the efficiency of the final result is 3% higher, and the heat load is reduced.

    Design limitations

    Through these measurement results, you can see that the switching speed of these devices is extremely fast, but we still don't know how fast the devices are, and maybe we will never know. Given that we have just measured these speeds, how could this be? There are some key constraints that we cannot assess, at least not yet. One of them is the ringing caused by the resonance between the power loop inductance and the smaller GaN transistor capacitance, which is obvious in all rise time measurements. The capacitance value is fixed, and the inductance is at least partly (if not very obvious) due to the equivalent series inductance (ESL) of the input capacitor and the interconnecting PCB backplane.

    The drivers are connected by PCB traces, and the edge speed of the driver itself is about 1ns, which is much slower than the GaN PET switching speed. With the development of GaN technology towards the material limit (still several orders of magnitude), and driver performance enhancement, parasitic effects reduction and integration enhancement become a reality, the speed/performance will continue to improve. At the same time, the GaN FET output capacitance will continue to decrease, and the switching speed will be further improved.

    What does all this mean

    If the switching speed reaches 1%~2% of the switching period of the hard-switching application, you can see that the switching speed can be close to 50MHz. Now, the limiting condition is the parasitic element of the gate driver, which cannot operate at this speed. I think that when using a resonant switching topology, the switching speed of the DC/DC converter can reach more than 250MHz. Although the intrinsic limitations of materials cannot match the performance of GaN devices, silicon devices will continue to improve.