RC snubber for push-pull DC/DC converter

[Futterama]

Hi forum,

I'm building an isolated DC/DC converter using the LT3999 and a Würth 749196301 SMPS transformer:
https://www.linear.com/product/LT3999
https://katalog.we-online.de/pbs/datasheet/749196301.pdf

I need help for the design of the RC snubber circuit that is connected between the switching nodes, as shown in the schematic of the LT3999 demo board:
https://www.linear.com/solutions/5502

I haven't built RC snubbers before and I'm having a hard time finding good info for building one for a push-pull converter. From what I'm told, the RC snubber is calculated from leak inductance in the transformer and deadtime in the switches among other unmentioned things. I did find a guide on how to design RC snubbers using a trial approach which would be fine too (adding certain components and measure frequencies ect.). The guide is here:
https://www.maximintegrated.com/en/app-notes/index.mvp/id/3835

However, the guide above uses two RC snubbers, connected to each switching node and GND (figure 2), but the LT3999 demo board uses a snubber connected between the swithing nodes. What is the difference between the two? Which is better to use?

Thanks.

Best regards,
Martin

 

[ronv]

I can't stand to see a post go unanswered.
First, thanks for the links. I always just scope and play, scope and play.
I think it may have to do with how much dead time you have. If it's short you can use one. If it is long you need two to keep the voltage from rising before the others switch is on.

 

[Futterama]

ronv, thanks for answering my post.

I have 70ns deadtime, so at 800kHz, which is the frequency I will use, the deadtime is 11.2% of the switching cycle. Would this be considered "short"?

EDIT: Correction! The 70ns is referred to as "non overlap time" in the datasheet and happens 2 times every cycle. So at 800kHz the non overlap time is 22.4% of the switching cycle.

I will try to follow the guide I linked to, and see where that takes me. I guess the process is the same whether I use 1 or 2 RC snubbers? I'll make room on the PCB for both types and see what I can figure out.

 

[ronv]

Yes, I think it will depend on the frequency of the ringing.

 

[Futterama]

Hi again. I used the guide from the link, and I managed to control the ringing and voltage spikes. I ended up with 1nF and 32Ω on each switchin pin. However, this increased the load current with about 100mA which is unacceptable. What can I do about that? Use higher value resistors and look out for the voltage spikes and just scope and play again?

 

[Futterama]

Just a final word. I used the following links to calculate new values. The maxim paper uses a ζ = 0.5 which is not critical damping, so I got better results with these links using ζ = 1:

https://www.ti.com/ww/en/analog/power_management/snubber_circuit_design.html
**broken link removed**

Note that the document from NXP saves a lot of work with trying a lot of capacitors for reducing the ring frequency to half since they simply calculate the parasitic capacitance from the change in ring frequency with a given added capacitance value.

 

[Flyback]

the notes from ray ridley are good for snubber design.
Remember you are having to deal with the magnetising current too....to that is going to get dissipated in your snubber, which is why I don't like the pushpull topology.

So your snubber is not just a snubber, its a clamp to dissipate the magnetising energy aswell.

 

[Futterama]

Ray Ridley, what, where?

I don't have inner understandings of where all the current is coming from, all I can see is that no load input current goes from ~10mA to ~50mA when adding the snubbers I have now, so I guess they dissipate the 40mA which is acceptable for my application.

 

[Flyback]

Its attached, its about flyback snubber but some of the principles hold through. you culd try rayridleyengineering.com for a pushpull snubber doc from Dr Ridley.

Presumably the drain voltage peak diminishes when you add your snubber?
Also, a TVS or zener in there could be good, because you don't really want your snubber conducting from the referred secondary voltage...you only want to snub what goes above Vin+V(secondary referred)