RCD snubber power loss?

[thomasx]

Hi

What is a reasonable power loss in a flyback converter RCD snubber circuit in relation to output power?

E.g if output power is 75 watts, then what power loss is reasonable to expect in the RCD snubber circuit?

[Konkedout]

It all depends...for example on how much voltage margin you have in your switching transistor, and how well coupled is your transformer.

But I would try to get under 1 watt for a 75 watt flyback.

Tell us more about max and min input voltage and also the output voltage.  These things all matter.  For example, if you were making 5V @ 15A (I hope not) then a tiny bit of parasitic inductance in series with your output rectifier will put greater demands on your primary snubber.

If you do not have a wide input range, you might be able to use a clamp winding to return much of the leakage inductance energy back to the rectified input voltage.

If you are winding your own transformer, I recommend a big honking ferrite core with a round center leg, such as PQ type.  It is much easier to make a good tight winding on a round center leg.  That will have a big impact on leakage inductance.  And a big core also makes the task easier.

[thomasx]

It all depends...for example on how much voltage margin you have in your switching transistor, and how well coupled is your transformer.

But I would try to get under 1 watt for a 75 watt flyback.

Great, thanks for your response, that's precisely what I wanted to hear.

Tell us more about max and min input voltage and also the output voltage.  These things all matter.  For example, if you were making 5V @ 15A (I hope not) then a tiny bit of parasitic inductance in series with your output rectifier will put greater demands on your primary snubber.

If you do not have a wide input range, you might be able to use a clamp winding to return much of the leakage inductance energy back to the rectified input voltage.

Here are the design parameters I am working with

Input 240VAC nom (220VAC min, 265VAC max)
Output 12VDC 7A, 84W

I have an ETD34 transformer 1.55mH with 56uH leakage inductance that I am using.

Switch MOSFET is an STD16N65M5 with 710V Vds (link to datasheet below)

Rectifier diode Vf = 0,5V

When I do my calculations I end up with a peak switch current around 1,8 A and a power loss in the snubber circuit of 8 W which isn't really feasible.



Looking at the formula for the power loss, what I have to work with is Ipeak, Vsn and switching frequence. However, the switching frequency also affects Ipeak in the opposite way, so changing switching frequence really has no impact.

Left to work with is Ipeak and Vsn. To lower Ipeak I need to increas the transformer inductance to what feels like unreasonable values.

Then what is left to work with is Vsn. What I understood from reading up on it Vsn is normally around 2-2,5 times nVout.

So I guess my questions are

1. Did I miss something?

2. Can I select Vsn and if so - How do I select it and make it become what I selected?

I can of course go for a different MOSFET with higher Vds if that makes the difference.

Thanks!

https://www.st.com/resource/en/datasheet/std16n65m5.pdf

[Konkedout]

I guess I am not ready to view your snubber equations.

I think you need a bigger ferrite core.  ("You need a bigger boat." )

OK it was a plastic enclosed switcher, but I used a PQ32/30 for a production 50W flyback. This was a "critical conduction mode" design.   I do not remember for sure but my design frequency (full load, minimum Vin) was somewhere in 25 - 50 KHz.  Do not go higher than that.   I did use 60 KHz for one 48V output design but parasitic/leakage inductance is less of a problem with that higher output voltage.

The PQ32/30 has almost twice the core area of your ETD34.   75W is high for a flyback so I would use a core with really big core area.  Maybe PQ50/50 for example.

Drive the core hard so that your maximum peak flux density is in the range of 0.20 to 0.22 Tesla.  (2000 - 2200 Gauss).

These steps will help reduce your leakage inductance energy and power.

Both of your inductance numbers sound high.  But I think you should show your basic design parameters:  Frequency, turns primary and secondary, peak flux density, CCM versus DCM.

The transformer would almost definitely need to be "sandwich wound" with half of the primary applied before the secondary, and half applied after.  This is the way flyback switchers are done unless power is much lower than your situation.

Are you planning CCM or DCM (or maybe critical conduction)?  They both have their advantages but critical conduction is less useful now due to better control ICs

[Konkedout]

And...do not treat that ST MOSFET as anything more than a 650V device. I SEE NOTHING ABOUT IT BEING AVALANCHE RATED!! That means that an excessive peak drain voltage spike might destroy it easily.  When you go into output current limit with maximum Vin, the ABS MAX peak drain voltage should be under maybe 625V.

If you need more than that, you probably should consider using a silicon carbide FET.  Unfortunately, those are more difficult to drive.  I have had only a little bit of experience with them.

[thomasx]

Hi

I have been working with a 40kHz switching frequency, turns ratio 116:12, CCM.

Do you have any suggestions for a similar MOSFET that would be better from an avalange perspective?

[Konkedout]

I think your peak flux density comes out to about 0.25 Tesla (not the car!) so a little bit high.

Here is an 800V Infineon MOSFET with avalanche rating.    It has higher RDS on than your selection but my calculations say you will get less than 400 mW conduction loss (very preliminary) in the MOSFET so that ought to be OK.

The higher peak Vds max will make your snubber a bit easier.

Again, you are operating on 200 - 250 VAC mains only; not on 90 - 130 VAC; right?  Lower Vin would require higher core flux density.

https://www.infineon.com/dgdl/Infineon-IPD80R900P7-DS-v02_00-EN.pdf?fileId=5546d4625b3ca4ec015b4287da753ee9

This is just a preliminary list at Digikey; not selected for Avalanche:

https://www.digikey.com/en/products/filter/transistors/fets-mosfets/single-fets-mosfets/278?s=N4IgjCBcpgbFoDGUBmBDANgZwKYBoQB7KAbRAA4AGAJjAGZ4BdAgBwBcoQBlNgJwEsAdgHMQAXwKwAnAhDJI6bPiKkQAFkpS6WkMxDtOPASPEEArAHZKs%2BYtwFikMnQtW1MvQcjc%2BQ0RJBYSnhoOVRMexUnEBdyKXjdVg5vIz9TQLBqG3ClB1VMygs6MxACNTpi6hKCAot3UsDXVzoGq1cpeAI6SjVMrK6aanJqmMHyNQbu6iGIAemLGTnqC3JJwdc1%2BZaBuDjWtTNqWBHu%2BjqG6niq7ZBLqSrZ26u1VYI76gOL54m355v3tT9J73OhA950R6XShxH63OJ0NSwobkIKdOGwOhUC4olxo5ZSJoXVxmQpEuoJN4WWAWOAXWC9chg%2BlgciPOqHal0hiIrkHPEYg7-WAfMCLW7C3p4szxXoXMzkFYjKoK8j-MzSqRK9UyuUayHavZvA0WXVSFGm4ataT3fXSrFG6ULU0bB0dJHajEXNTUzRen1470k163REaSGI7Qmt7laiYr1gdzWaOZdUXBhSQGJfTJHzGfxvYKrUK2CLKRxkIKUAAEADUGsLq3XJBpa-WSa2CFZG61uh2QBZ202KJRu54cwBVQT8NgAeRQAFkcGgsABXXg4dIAWiyxagfBXZdUJUYYgCJVCLCgYFYl8gYBHp6AA

[thomasx]

Hi

Thanks!
However, RDSon 0.9 ohm seems a bit high, especially compared to the 0.28 of the STD65N15.

I checked the datasheet of the STD65N15 and it does state the following avalanche parameters

IAR: Avalanche current, repetitive or not-repetitive (pulse width limited by Tj Max) = 4 A
EAS: Single pulse avalanche energy (starting Tj = 25 °C, ID = IAR, VDD = 50 V) = 200 mJ

And when doing some comparisons that seem to be pretty good.

Yes, I'm operating on 220-265VAC only.

My big concern is the snubber power loss. 

If I understand correctly, with

V in max = 265VAC = 375VDC, Vout 12V and Vf 0,5V
Vds = 710V --> 90% = 640V

This gives a maximum snubber voltage of 640 - (375+12,5) = 252,5 volts?

And then I need to calculate the R and C to match that maximum voltage.

And then the power dissipation is what it is based on the other design parameters?

Have I understood things correctly?

And then the only practical way to lower the power loss is with an increase in inductance.

[Terry Bites]

Here's another design guide or two.

[jkostb]

I think you made a mistake with power loss formula. In my opinion Ploss=0.5*Leakage*Ipeak^2*fsw
With 40Khz, 1.8A and 56uH the loss is 3.6W. You can reduce this by using an active clamp circuit

[Konkedout]

It is possible that I missed something on the ST datasheet.  But:

Comparing the Rds of two FETs is like comparing the fuel economy of a motorcycle to that of a bus, without including the fact that the bus can carry 30 times as many people.

I have my own flyback calculator spreadsheet with which I calculated that the conduction loss in the Infineon FET is less than 500 mW so is small compared to that in the snubber. 

The snubber is maybe #3 or #4 on the list of items to calculate after designing a good transformer and choosing the MOSFET.  Both increasing the Vds rating of the MOSFET and improving the transformer can reduce the power dissipation needed in your snubber.

I did design production flyback switchers (90 - 265 VAC) using 600V MOSFETs, but with the wide input range, the peak input current is higher, and I had a better transformer.

First things first!!  You need a bigger transformer center leg together with "sandwich windings".

https://www.google.com/search?q=transformer+sandwich+winding&rlz=1C1RXMK_enUS1029US1029&oq=transformer+sandwich&gs_lcrp=EgZjaHJvbWUqBwgAEAAYgAQyBwgAEAAYgAQyBggBEEUYOTIICAIQABgWGB7SAQoxMDA3OGowajE1qAIIsAIB&sourceid=chrome&ie=UTF-8

[thomasx]

Hi

I swapped for a 1700V SiC MOSFET and recalculated the snubber. Now the situation is much better even with this transformer. But I will recalculate the transformer as well.

Now however, I have a problem with a large voltage drop already at 30W load, from 13,6 volts down to just above 10 volts, I didn't see this before. Strange.

I also noticed I did not get any higher spikes in the switching with this new MOSFET and snubber circuit.

[thomasx]

I think you made a mistake with power loss formula. In my opinion Ploss=0.5*Leakage*Ipeak^2*fsw
With 40Khz, 1.8A and 56uH the loss is 3.6W. You can reduce this by using an active clamp circuit

You are not including the Vsn/(Vsn-n*Vo) in your calculation, should this not be included?
I think it should, otherwise it is totally independent of the snubber voltage, and increasing the snubber voltage seems to be one of the "common" ways to reduce the snubber power loss.

I have to look into an active clamp circuit, is this a better option than the RCD snubber?