Guest blogger Jan Cumps is an active E2E community member, showcasing some of the best projects from around the TI Community
The rise of discrete gallium nitride (GaN) FETs has increased the need for more user-friendly interfaces while also improving efficiency. Half-bridge GaN power stages such as the LMG5200 have a separate drive input for the high and low GaN FET. The two inputs (pin 4 and 5 in Figure 1) give you the ability to optimize efficiency because you can tweak the exact point at which each FET switches on and off.
Figure 1: LMG5200 GaN Half Bridge Power Stage Diagram
But this option also comes with responsibilities. To prevent shoot-through during switching between the FETs, you have to provide a dead time between the switching off of one FET and the switching on of the other.
That’s why you shouldn't just use a pulse-width modulation (PWM) signal and its inverse to drive the high and low inputs, because you have to generate a wait time between the falling edge of one input and the rising edge of the other (Figure 2).
Figure 2: Inverted driver input signals with dead time delay
The evaluation board for the LMG5200 GaN half-bridge power stage uses the classic time-delay circuit to generate this tiny delay. Fans of the iconic book, “The Art of Electronics,” will recognize the setup as one of the practical examples from the first chapter. While the example given in the book generates a delay on both edges of a square wave, TI designers adapted the board so that it only inserts the delay on the rising edges. And that’s exactly the behavior you need.
The dead-time generator schematics
The user’s guide documents the full schematics of the evaluation board. Figure 3 shows the module that generates the high and low drive signals with the dead time.
Figure 3: Dead time generator schematic
If you abstract away the diodes, resistors and capacitors, the circuit generates an in-phase and an inverted signal from the PWM input. There will always be one side of the half bridge that’s on and one side that’s off. The passive components will generate that dead time for you, so let’s dig a little deeper into that part.
“The Art of Electronics”
In Chapter 1 of “The Art of Electronics,” the authors present a time-delay circuit. It uses an RC circuit – flanked by two complementary metal-oxide semiconductor (CMOS) buffers – to glue a short delay to both flanks of a PWM signal (Figure 4).
Figure 4: Comparison of time-delay circuit from “Art of Electronics” and TI asymmetric delay generator
In TI's reference design, a fast-switching Schottky diode in parallel with the resistor bypasses the RC delay during one of the edges.
The high and low signals with dead time
Let's follow the HI signal path of the evaluation board schematics (D4, R2 and C35 in Figure 3). Diode D4 bypasses R2 during the falling edge of input A (signal 1 in Figure 3 and Figure 5). At the output of the first inverter (signal 6), this input signal’s falling edge becomes a rising edge. The diode conducts and the R component of the RC circuit is almost 0. The RC constant is also almost 0. So the circuit adds virtually no delay to that flank (signal 7). After the second inverter, this becomes a falling flank again (signals 8 and 10).
Figure 5: Signal transformation from simple input to inverted outputs with dead time
When the input signal rises (again at signal 1), the signal after the inverter becomes a falling edge (signal 6). The diode D4 blocks and leaves R2 alone. You get a real RC circuit with 47Ω and 100pF. That will turn the edge into a ramp (signal 7). The Schmitt trigger front end of the second inverter will ensure that the output signal flips at a predictable voltage (and at a predictable time). So the result is that the rising edge at the output (signals 8 and 10) will lag compared to the input signal. And that was the effect I was after.
The LO signal path does exactly the same. An additional inverter at the input of this path ensures that both sides see and process an opposite signal: one to drive the high-side FET and one to drive the low-side FET. As a result, you’ll get two related signals, as shown in Figure 2. A signal only turns high after the other one switches off, and after that dead time has passed, which was our ultimate goal.
Figures 6 shows the capture of both inputs of the LMG5200 at the two transitions.
Figure 6: Capture of HI-LO and LO-HI transition of both driver signals.
You can vary the duty cycle and the frequency of the input signal as required. The dead time will stay constant.
By using a classic time delay circuit, you can safely drive a GaN half-bridge power stage with a simple PWM signal and control its output with variable duty cycle. What is your approach to generate dead times in your switching applications?
Thanks for the great blog entry! I have a couple of comments which may help a designer: First, the Schmitt trigger is used in particular to add hysteresis on the input of the gate following the R-C-D circuit. This prevents any noise coupling onto the slowly-rising edge from creating multiple edges at the output of the inverter. In addition, it is important to use a fast logic series, such as the *74LVC* series to minimize differential delay due to the extra inverter in the LI path.
Great post Jan Cumps!
Well done, Jan!
Michael Seeman , yes. It's key to choose fast logic.
The switching time of the gates should be neglectable compared to the short RC delay time. All should be implemented with fast logic, and in particular the speed of the first inverter in the LO path should be orders of magnitude smaller than 7 ns - since that's an extra delay that's only there in the LO part and should add as little time as possible to the RC delay.
Have you tested this circuit at different temperatures? I've seen these types of circuits vary greatly over temperature, because the threshold voltage on the input of the schmitt trigger / inverter varies with temperature, which results in a variable delay. If you're thinking of using this types of circuit in a production board, test it very carefully across temperature, and also with inverters from different batches, to ensure you always get the delay period you expect under all conditions.
Frank Van Hooft1 , I haven't tried this.
I'm at the limits of my test setup - measuring deviations from a 6-7 ns event is not realistic in my lab.
Maybe someone with precision measure gear can step in and try?
Jan,Blazing new trails as always! Really nice blog post.
Have you considered also trying the deadband generation circuit that is integrated with the ePWMs on the C2000 and Hercules launchpads?
Maybe a GaN booster pack ...Best Regards,Anthony
Anthony F. Seely , yes, I'm writing a project for the Hercules etPWM module.
I'm checking how I can use its xA and xB pins to deliver the inverse signals, and how I can use the delay property to affect the rising edge of both xA and xB.
I'm using a TMS570LC43x LaunchPad as the test bed.
Deadband generation with a Hercules microcontroller:
Any Hercules that has an ePWM module will work. If only someone could post a working NHET example...
Was Jan Cumps your favorite 2015 TI Community Helper? Vote for him now (daily through December 31st). 2015 TI Community Fan Favorite Voting now open!: e2e.ti.com/.../ti-community-fan-vote-vote-now-for-your-favorites
All content and materials on this site are provided "as is". TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with regard to these materials, including but not limited to all implied warranties and conditions of merchantability, fitness for a particular purpose, title and non-infringement of any third party intellectual property right. No license, either express or implied, by estoppel or otherwise, is granted by TI. Use of the information on this site may require a license from a third party, or a license from TI.
TI is a global semiconductor design and manufacturing company. Innovate with 100,000+ analog ICs andembedded processors, along with software, tools and the industry’s largest sales/support staff.