I have a step-up flyback design with the following parameters:


The transformer is not ideal with 1:7 and the inductance is a bit off, but it is what i made and soldered in a way it is difficult to remove. with a \$Vo< 180V\$, regardless of load, the transistor stays cool to the touch. Above ~180V the system starts to make an audible buzz and the transistor becomes hotter, ~60C.

After doing a lot of investigation, I believe it is related to the reflected output ringing going below ground and turning on and off the internal diode? You can see the ringing in the attached image.

Some issues I have ruled out via investigation:

  • The transformer is made well from the stand point I do not believe there is any arc over etc.
  • It also has an adequate core gap and should not be saturating, this is also reinforced by the linear looking current ramp that is visable using the FET \$R_{ds}\$

The noise may also be unrelated to this, I was thinking sub harmonic oscillation, but It looks like my duty cycle is typically below 50% and this should be operating in DCM reguardless.

Is there any way to reduce the ringing without reducing the reflected voltage thus keeping the FET cooler?

enter image description here

Below is the gate image, everything looks alright there. I tried to measure current through the FET, using the 0.042m Ohm feedback my scope just sees noise. Unfortunately everything is soldered in kind of tight and with power planes there is not a great way to insert a different resistance somewhere else in the chain.

I also attached a slightly modified schematic, the switching freq is different, and some of the capacitance is not identical, but it is close enough to the as built for you guys.

enter image description here

enter image description here

Below is also an earlier image of the board: enter image description here

  • \$\begingroup\$ Is the audible buzz not coming from the transformer due to higher peak currents? Which internal diode are you refering to?which IC are you using? \$\endgroup\$
    – Navaro
    Commented Mar 29, 2020 at 19:57
  • \$\begingroup\$ @Navaro - The FET body diode. As for the transformer, I am fairly sure the noise is coming from it, but not because of loose winding or ferrite. It is same volume regardless of load, and effectively starts at low frequency ~5Hz and builds up to ~1kHZ as voltage and load is increased. It is not the typical whine you hear from cheap power supplies. \$\endgroup\$
    – MadHatter
    Commented Mar 29, 2020 at 20:00
  • \$\begingroup\$ What have you done to prevent oscillations in Fet's gate circuit? Check Vgs with the oscilloscpe. Insert the schematic. \$\endgroup\$
    – user136077
    Commented Mar 29, 2020 at 20:06
  • 1
    \$\begingroup\$ The fact that the body diode is activated at low input voltage is a classic and should not bring higher losses in the transistor: the transformer is fully demagnetized and what you see is the energy stored in the capacitance lumped at the drain going back and forth between the primary inductance and this capacitance. What you should look at is the primary drain current at low input and calculate the conduction losses \$I_{drms}^2r_{DS(on)}\$. The buzz you hear could come from a loop instability: do you operate in open or closed-loop? \$\endgroup\$ Commented Mar 29, 2020 at 20:24
  • \$\begingroup\$ Closed loop with compensation, I'll have to post the schematic. \$\endgroup\$
    – MadHatter
    Commented Mar 30, 2020 at 20:41

2 Answers 2


So basically when the current goes to zero on the secondary side, there is a oscillation between the main inductance and the various capacitances (such as Coss of Mosfet and Tranformer).

𝑉ds(𝑑) = 𝑉𝑖𝑛 + 𝑛 βˆ™ 𝑉0π‘π‘œπ‘ πœ”0(𝑑 βˆ’ 𝑑0)

So you have a n of (1/7) .
Vin = 24V
Vout = 180
So around Vo is 180 V . The Equations leads to a minimum of :
Vds = 24 - 25.7 = -1.7 V

As This happens the body diode of the fet starts to conduct and the resistance limiting the current through it, is only the sense resistors.

There is not a lot you could do about the reflected voltage despite changing N, but what you could try is adding some resistance ( a few ohms) in the drain of the Mosfet to limit the current through the body diode.

Other option is to work in Continous conduction mode, as in CCM there is no oscillation in that operation mode. But that would mean that you probably require a different transformer.

  • \$\begingroup\$ Thank you, I'm glad to see someone else come to the same conclusion. It does seem like the only way to really fix this problem is to increase the ratio of the transformer which is something I'm doing on the next revision anyways. \$\endgroup\$
    – MadHatter
    Commented Mar 29, 2020 at 21:23
  • \$\begingroup\$ Be sure to verify the analysis first by just adding some resistance and see whether the Mosfet does not get hot. \$\endgroup\$
    – Navaro
    Commented Mar 29, 2020 at 21:35
  • \$\begingroup\$ With it peaking like 2+A, I'm not sure how much I can still add and get a reasonable equivalent output. \$\endgroup\$
    – MadHatter
    Commented Mar 29, 2020 at 21:49
  • \$\begingroup\$ Maybe you don’t understand my suggestion correctly but I am referring to adding some resistance in series with the drain if the Mosfet on the primary. A few ohms perhaps \$\endgroup\$
    – Navaro
    Commented Mar 29, 2020 at 21:58
  • \$\begingroup\$ Can you share your findings? \$\endgroup\$
    – Navaro
    Commented Mar 30, 2020 at 13:45

The most direct solution is an R+C snubber or damper across the transformer:

Schematic with snubber added

Set \$R = \sqrt{\frac{L_M}{C_P}}\$ and \$C \geq 2.5 C_P\$ to reduce the Q of the ringing to below 2 or so. \$L_M\$ being the winding's magnetizing (open circuit) inductance, and \$C_P\$ being the stray (equivalent 'P'arallel) capacitance on the winding: the sum of self-capacitance, transistor COSS and (reflected) diode CJO. (The latter are nonlinear parameters; use the figure near operating voltage, since that's what voltage the ringing is centered around.)

Snubbers may also be valuable for dampening the high-frequency ringing at Q2 drain (instant of turn-off) or D2 (instant of turn-on), which are the respective capacitances ringing with leakage inductance. These will have lower impedances so need smaller R's, and somewhat smaller C's. Multiple snubbers can be connected in parallel as needed (it's rare you would use more than one at a time, but certainly possible to do).

Choosing D2 for lower recovery time/charge, and lower capacitance, may also help. (ES1J is already a pretty small diode, so this is unlikely to be improved.)

Alternately, replace U2 with a quasi-resonant or active-clamp type controller (the latter needs more circuit changes, of course), and use the (at least partial) class E-like operation to your advantage. (I do not have any parts to recommend offhand, unfortunately.)


Your Answer

By clicking β€œPost 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.