UCC28634: AUX Winding Ringing

Hi Larry,

Thanks for reaching out. From the waveform, it looks like there is ringing during the MOSFET off time. Could you try snubbing the switch node more? This should help attenuate the ringing on the switch node. This blog post might be helpful for determining snubber parameters:e2e.ti.com/.../calculate-an-r-c-snubber-in-seven-steps

Best Regards,
Ben Lough

Hello Larry,

Definitely, the aux waveform should not ring this much during the MOSFET on-time.

High-voltage transformers usually have more turns than typical low-voltage off-line designs.
The turns-ratio from primary to auxiliary windings appears to be about 10:1. Since there is a significant overshoot on the aux voltage at the falling edge of the primary voltage, it suggests that the transformer has a lot of leakage inductance on the aux winding that is ringing with the Aux cap for the controller's VDD input. Also the ringing frequency looks like about 1us in period, which may correspond to a reverse recovery time of a standard P-N diode. An ultrafast diode should be used for this application.

Can you please check the part number of the aux diode and determine if it is a standard-recovery or an ultrafast part?
If it actually is an ultrafast, then some other source of series or parallel capacitance must be present somewhere in the aux path.
Can you please provide more details about the auxiliary winding circuit of your application, such as a partial schematic of this portion of your design, plus the transformer parameters?

Regards,
Ulrich

Hi Ulrich,

Thanks for your response! The diode used has a 25ns recovery (Diodes Inc. ES1D-13-F) so that seems fast enough. I measured the leakage inductance of the transformer to be ~35uH (shorting AUX and Secondary coils). Below is the AUX supply schematic:

I was also wondering about the main switch snubber circuit, which I had obtained from another reference design. (Note that the R38 symbol should be a resistor, not a capacitor as shown):

Also, here is the transformer data sheet. Thanks very much for your help and please let me know if you need any additional information or measurements.

 G194012LF r 0.pdf

AUX Supply Circuit:

Main Snubber Circuit:

Hi Larry,

Thanks for the additional info. Your AUX supply section shows a 1000pF cap C9 across the AUX winding. I recommend to remove that cap and see how the waveform changes.

I realize that the application schematic of Figure 44 in the UCC28634 datasheet includes such a cap. I think that is a special case that helped clean up some noise in that particular design, but I don't think that it is applicable to all designs. It's value was chosen, in part, based on the transformer inductances of that application. The author(s) probably should have made that clear. After removing it, you can assess your waveforms and determine whether or not you need anything there at all.

If you think that you do need one, the value for your C9 would be selected to work with your design's transformer.
I recommend to remove what you have first, and evaluate the results.

Regards,
Ulrich

Hi Ulrich,

I removed C9 and unfortunately it did not have a significant affect on the waveform. I have also tried increasing the gate resistor on the main switch to see if that would help the ringing but that has not helped either. Here are the gate drive vs. drain waveforms for reference.

Regards,

Larry

Hi Larry,

I'm sorry that the C9 idea didn't work out. But thanks for the new waveform; it provides another clue that might be the answer.
Given that the MOSFET stays on ~3X longer than the gate drive pulse indicates the primary MOSFET has massive Coss. This Coss can be ringing with the leakage inductance during the extended on-time and even afterward.

I suggest that your MOSFET may be oversized for the application, maybe by a factor of 3 (for lack of a better number). On-time extension due to Coss is generally unavoidable, but this is an extreme case in my experience. I suggest to try one with 3xRds(on) and 1/3xCoss of what you have now. They may not scale exactly inversely, but the main thing is to significantly reduce Coss. A side effect is that you should see your switching losses go down to make up for the increase in conduction losses.

Regards,
Ulrich

Hi Ulrich,

Thanks for your comments! I have tried two FETs so far - one was a GaN device with a Coss = 41pF (TP90H180PS). The other was a traditional FET with a Coss of 140pF (STF20N95K5). Both have yielded similar results. Would anything else be causing the FET to stay on?

Regards,

Larry

Hi Larry,

There can be a few things to account for the extended on-time.
1. Huge Coss in the MOSFET.
2. More capacitance added to the switched node for EMI suppression or other reasons.
3. Delay through a gate-drive buffer (if you have one) due to anemic or under-designed drive current available to discharge Vgs of the FET.
4. Very low peak primary current in that operating condition. (Although simplified, for a flyback it is basically: Ipk/C = dv/dt of the drain voltage, where C is the total capacitance on the switched node.)
5. Various combinations of any or all of the above.

The Coss ratings of those MOSFETS are specified at 100V or higher. If you take a look at the Coss curves of each, you’ll see that the STF part exceeds 5000pF for Vds < 5V. The TP part is similar. So the peak current at turn off has to charge up 5nF+ before the Coss drops enough for Vds to really start rising quickly. And the lower the Ipk for that cycle, the slower everything is. In fact Ipk is still rising even as Vds is rising, until Vds > Vbulk. This is worse in the light load condition, of which the waveform appears to be, given the narrow on-time and wide switching period. Not as pronounced at full load.

Both of these parts have relatively low on-resistance for a 900V+ rating, which makes them quite large MOSFETs. I still suggest to try using smaller devices to see if the on-time ringing abates.

That is still the main issue to solve: eliminate the on-time ringing. Ringing comes from an undamped L-C tank, either series or parallel. The FET Vds looks solidly on. The AUX reflection of Vds shows ringing, so the L-C tank is more likely to be associated with the AUX winding. (Changing the MOSFET is mainly a shot in the dark.) You removed C9 and it didn’t help. There must be some other reason for the ringing. I request of you to post the schematic of your power stage or enough portion of it to see what other circuits may exist that could contribute to this problem. Also the major parameters (Vin, Vout, Iout) to give it some context.

Finally, I suggest to scrutinize your prototype to ensure that the parts and connections on the board match those of the schematic diagram. The schematic may work fine, but if the board doesn’t match it, you can have problems. Especially if there exist significant parasitic components not represented on the schematic.

Regards,
Ulrich

Hi Ulrich,

Below is the primary side schematic of the converter. Note that there is one error that has been fixed in the layout (R19 is shown connecting directly to R18, but is has been move to connect between R18 and R17). The secondary side is using a synchronous rectifier arrangement. Vin is as shown, secondary side has 2 isolated 32V outputs, both around 0.4A max output.

Regards,

Larry