I am trying to design a flyback converter but the low side FETs suffers from high Vds voltage spikes.

The flyback with active clamp should convert 6 ~ 15V into 40 ~ 90 V. Power should be 100+ W. The FETs are driven by an Arduino at 62.5 kHz.

See a part of the schematic below:

enter image description here

The power supply is set to 6V at the moment and the duty cycle approx. 6%. I tried to wind the lowest leakage inductance but I couldn't get lower than 10%, that is why I introduced active clamp.

Scope picture voltage drain source of Q1 (low side FET):

enter image description here

Voltage across C27 (active clamp)

enter image description here

If I can make my conclusions:

  • Active clamp circuit works correct but the body diode of Q10 and Q4 will not conduct as fast as the voltage from inductance leakage will rise (approx. 50 ns).

  • The IR21844 have a deadtime where the minimum deadtime is too large for this issue. At the moment R2 is shorted to create a deadtime of 50 ns.

What is the best solution?

Most important part of the PCB:

PCB layout

  • \$\begingroup\$ C27-2 appears to be shorted to GND, or at least the polygons are drawn as overlapping. Please confirm. \$\endgroup\$ Commented Aug 18, 2023 at 18:11
  • 3
    \$\begingroup\$ Also can you show how the probing was done? The 1x probe setting is suspicious. Please confirm the bandwidth of the probes and settings, and where probe, scope and circuit (EUT) grounds are connected, if at all. \$\endgroup\$ Commented Aug 18, 2023 at 18:13
  • 1
    \$\begingroup\$ You might also ask about the underlying problems (not able to wind a transformer of low enough leakage; why using flyback at all at this power level when forward is more common; etc.), probably in a separate question. \$\endgroup\$ Commented Aug 18, 2023 at 18:16
  • \$\begingroup\$ C27 is in real life not shorted, this issue is resolved. Probing was done direct on the leads of the FETs. What kind of settings do I need to use? same picture when I use 10x instead. \$\endgroup\$
    – JBtowerway
    Commented Aug 18, 2023 at 18:21
  • 1
    \$\begingroup\$ Not the scope setting, the probe setting. Please check the manual for your probes: if they are 1/10x type, the 1x setting most likely has a bandwidth limit around 6MHz. Or is it isolated type, or what? \$\endgroup\$ Commented Aug 18, 2023 at 18:25

1 Answer 1


May I make the following comments:

Comment #1: If your scope can display more than one channel, then may I suggest showing as many voltage waveforms as possible simultaneously at the following nodes (refer marked-up diagram below):

OP schematic, marked up by commenter (me)

Comment #2:
Your concern seems to be: "Is the voltage across Q1 is excessive?"
My reply at this time would be: "Is this problem real?"

The method of probing this circuit will have a big impact on the quality of the waveforms shown on the scope, and will greatly help answer the question: "Is this a real problem or not?"

By "method of probing" what I mean is: "how the voltage signal is transferred from the PCB to the scope input". This includes the probes themselves, as well as the means by which the probe nodes are connected to the circuit node. And by "probe nodes" I mean both the probe tip, and its ground clip. There are many resources on this issue, but I guess as a minimum please apply the following rules. Glossary: "DUT" = Device Under Test:

  • Don't use x1 probes. Use x10 or x100.

  • Don't use switchable probes (x1/x10). If you only have switchable probes, then set them to x10 (or x100), and tape the switch in this position to avoid accidentally changing the switch as you move the probe.

  • Calibrate all probes before connecting to the DUT.

  • If your probe does not have correct tips for connecting to SMD parts, then temporarily solder one lead of a leaded 1k resistor to the DUT nodes you are probing (1/4W will be more than OK for this). Clip the probe tip to the other lead of the resistor.

  • Probe GND leads should be as short as possible (less than 100mm).

  • Connect all probe GND clips to exactly the same point on the DUT. If necessary, solder a temporary stiff wire to the DUT ground plane to make this easier.

  • To accurately capture very high-frequency signals in a noisy environment (as you have here with this flyback), you may have to use "ground springs", aka "pig-tails" on your probes.

More links to hints about scope probes here:

What do you use to probe small circuits quickly?

What are the differences between a x1 and a x10 osciloscope probe

Does the 1x oscilloscope probe setting slow down MHz digital signals?

How do you attach an oscilloscope ground spring?

Comment #3:
After you have gathered the waveforms accurately without excessive noise, which are presented correctly aligned in time with each other, and then presented with correct voltage scaling & timebase, then it may then be possible to share the results with the community to get help to answer this question:
"Is the problem real?"

If the answer is "yes", then there are many ways to address it, the simplest being putting a suitable voltage clamp in place - however, the best solution will depend on what is uncovered from further investigation. To help the community help you better, then may I suggest including the following information with your next update:

  1. Good quality waveforms, as discussed above.
  2. Include these (non-waveform) values at the same time that the waveforms are captured:
    (a) Output voltage, current, and power.
    (b) Input voltage, current, and power.
    (c) From (a) and (b) calculate power loss, and efficiency.
    (d) Are any components getting warmer than expected?

Comment #4:
Here are some comments/suggestions that may help make progress towards a solution:

If the circuit is operating as intended, then is the voltage across C27 supposed to be constant, or should it have the ripple that is being shown?

Are the gate drive waveforms as expected? (These are marked as nodes 1 and 4 in the image posted.)

What is the magnetising inductance of TR1 (measured at the primary winding)?

Is the flyback intended to operate in continuous or discontinuous conduction mode?


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.