2
\$\begingroup\$

I am selecting a GaN FET for a buck-boost DC-DC converter application. The Vgs of the part I'm looking at combined with the gate charge and gate resistance lets me switch the FET on in about 5.3 ns. This would be 0.53% of the period with 1MHz switching frequency. Is this fast enough for 1MHz? This is a sensitive application and the converter efficiency has to be as high as possible.

Here's the schematic of the buck-boost converter (Fig. 1. from H-Bridge Buck-Boost Converter with Dual Feedforward Control)

Circuit configuration of H-bridge buck-boost converter

Source: K. Muro, T. Nabeshima, T. Sato, K. Nishijima and S. Yoshida, "H-Bridge Buck-Boost Converter with Dual Feedforward Control", 2009 International Conference on Power Electronics and Drive Systems (PEDS), Taipei, Taiwan, 2009, pp. 1002-1007, doi: 10.1109/PEDS.2009.5385893

\$\endgroup\$
3
  • \$\begingroup\$ If power efficiency is paramount why go for such high frequency? \$\endgroup\$
    – tobalt
    Dec 3, 2023 at 20:22
  • \$\begingroup\$ @tobalt I was under the impression that 1MHz would be a good middle ground between physical size of the components and efficiency. \$\endgroup\$
    – entieda
    Dec 3, 2023 at 20:42
  • \$\begingroup\$ @entieda well depends on power, too, I guess, but to me 1 MHz seems to be on the high end. So if efficiency is important, then you could opt for more/larger/better passives, more cost and higher efficiency by reducing the switching frequency. \$\endgroup\$
    – tobalt
    Dec 3, 2023 at 20:45

2 Answers 2

3
\$\begingroup\$

the gate charge and gate resistance lets me switch the FET on in about 5.3 ns. This would be 0.53% of the period with 1MHz switching frequency. Is this fast enough for 1MHz?

Short answer: yes.

However, the speed of the gate drive is just one of many factors that influence the overall efficiency of a power converter, and in many cases, it would have a minor effect compared to other power loss mechanisms. I would expect that power loss in the FET, which is comprised of conduction loss, and switching loss (including internal diode turn-off loss), would be much larger than losses caused by a gate drive of 5.3ns. Other components having significant power loss will be the magnetic components (inductors, power transformers), and the capacitors.

Switching loss can be reduced by implementing "soft-switching", which can greatly reduce energy lost per switching cycle. This usually comes at a cost of increased conduction loss in the FET and other components (the degree of increase depends on the method used to achieve soft-switching), however the switching power loss saving is usually greater than the extra conduction loss, thus giving an overall net benefit. Soft-switching also reduces EMI, particularly if diode turn-off loss is greatly reduced.

An introduction to the two main types of soft-switching, Zero-Voltage-Switching (ZVS) and Zero-Current-Switching (ZCS), is provided here:
https://patents.google.com/patent/US5742495A/

\$\endgroup\$
2
  • 1
    \$\begingroup\$ Buck and boost don’t lend themselves to soft switching without ripple currents >= load currents. Unless you’re going for crazy frequency it’s very unlikely to be a winner especially with gan fets which can hard switch with lower losses. A better way to reduce losses in this example is to reduce switching frequency and take the size hit. \$\endgroup\$
    – asdf30
    Dec 4, 2023 at 4:52
  • 1
    \$\begingroup\$ @asdf30 Agreed. However, the question was quite specific, regarding just switching time, it was not a more general "What are the design trade-offs to achieve high efficiency". The answer to that questions is far more complex and nuanced. Cheers. \$\endgroup\$ Dec 5, 2023 at 21:46
2
\$\begingroup\$

In accumulated coverters (like boost) more important is transient when inductor starts discharging. A more energy in inductor is in this time and you need to transfer it to output maximum efficiently.

Btw, I would more care about passives parasitics. Cap ESR is a big problem at 1MHz frequency. Also winding a good inductor and dumping the leakage energy.

Regarding 5ns rise it should be ok.

\$\endgroup\$
1
  • 1
    \$\begingroup\$ I'm having trouble making sense of this, could you rephrase it a bit please? \$\endgroup\$ Dec 3, 2023 at 20:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.