Forums

More methods to salvage leakage energy in a flyback?

Started by Joerg February 2, 2015
Folks,

Have to do a flyback-PFC but this time with a really good efficiency, 
 >90%. The output voltage is high enough that sync rectification isn't 
needed. AFAICT the only other way to boost efficiency is to not flare 
off the leakage inductance spike via zener or snubber but to recycle it 
instead. Unfortunately a recuperating winding only works if the duty 
cycle stays under 50% at all times. Or if it's a double winding 66%, 
similar in a tapped design, et cetera. Long story short that doesn't 
work in a PFC because the duty cycle is much higher around the zero 
crossings of the AC going into the unit.

A non-dissipative snubber such as in Weinberg converters can be iffy 
because the results are highly dependent on component tolerances in the 
FET and in the flyback transformer.

Are there any other tricks to sponge off the leakage spike and dump it 
back into the pot? Maybe a fast and low capacitance "siphoning buck" 
that only takes off energy when it sees 2x input voltage or something? 
Must limit kind of hard though.

-- 
Regards, Joerg

http://www.analogconsultants.com/
In my opinion, leakage is best dealt with first, in and of itself.  What 
winding options do you have on this job?  Custom OK?  Current stock only?

You want LL/(stray C) ~= Vsw / Isw.  Too much interleaving is just as bad 
as too little.

Tim

-- 
Seven Transistor Labs
Electrical Engineering Consultation
Website: http://seventransistorlabs.com

"Joerg" <news@analogconsultants.com> wrote in message 
news:cja9aaFqc2bU1@mid.individual.net...
> Folks, > > Have to do a flyback-PFC but this time with a really good efficiency, > >90%. The output voltage is high enough that sync rectification isn't > needed. AFAICT the only other way to boost efficiency is to not flare > off the leakage inductance spike via zener or snubber but to recycle it > instead. Unfortunately a recuperating winding only works if the duty > cycle stays under 50% at all times. Or if it's a double winding 66%, > similar in a tapped design, et cetera. Long story short that doesn't > work in a PFC because the duty cycle is much higher around the zero > crossings of the AC going into the unit. > > A non-dissipative snubber such as in Weinberg converters can be iffy > because the results are highly dependent on component tolerances in the > FET and in the flyback transformer. > > Are there any other tricks to sponge off the leakage spike and dump it > back into the pot? Maybe a fast and low capacitance "siphoning buck" > that only takes off energy when it sees 2x input voltage or something? > Must limit kind of hard though. > > -- > Regards, Joerg > > http://www.analogconsultants.com/
Joerg <news@analogconsultants.com> Wrote in message:
> Folks, > > Have to do a flyback-PFC but this time with a really good efficiency, > >90%. The output voltage is high enough that sync rectification isn't > needed. AFAICT the only other way to boost efficiency is to not flare > off the leakage inductance spike via zener or snubber but to recycle it > instead. Unfortunately a recuperating winding only works if the duty > cycle stays under 50% at all times. Or if it's a double winding 66%, > similar in a tapped design, et cetera. Long story short that doesn't > work in a PFC because the duty cycle is much higher around the zero > crossings of the AC going into the unit. > > A non-dissipative snubber such as in Weinberg converters can be iffy > because the results are highly dependent on component tolerances in the > FET and in the flyback transformer. > > Are there any other tricks to sponge off the leakage spike and dump it > back into the pot? Maybe a fast and low capacitance "siphoning buck" > that only takes off energy when it sees 2x input voltage or something? > Must limit kind of hard though. > > -- > Regards, Joerg > > http://www.analogconsultants.com/ >
Vladimir Vassilevsky posted this a while back: http://www.abvolt.com/misc/antiseptic.jpg -- ----Android NewsGroup Reader---- http://usenet.sinaapp.com/
On 2015-02-02 2:16 PM, Tim Williams wrote:
> In my opinion, leakage is best dealt with first, in and of itself. What > winding options do you have on this job? Custom OK? Current stock only? > > You want LL/(stray C) ~= Vsw / Isw. Too much interleaving is just as bad > as too little. >
It's a flyback, needs an airgap. The transformer will be custom so we can decide what we want. We can't interleave much because safety isolation almost requires a two-chamber bobbin or at least a thick separation layer. It is the usual rock <-> hard spot situation. -- Regards, Joerg http://www.analogconsultants.com/
On 2015-02-02 2:43 PM, bitrex wrote:
> Joerg <news@analogconsultants.com> Wrote in message: >> Folks, >> >> Have to do a flyback-PFC but this time with a really good efficiency, >> >90%. The output voltage is high enough that sync rectification isn't >> needed. AFAICT the only other way to boost efficiency is to not flare >> off the leakage inductance spike via zener or snubber but to recycle it >> instead. Unfortunately a recuperating winding only works if the duty >> cycle stays under 50% at all times. Or if it's a double winding 66%, >> similar in a tapped design, et cetera. Long story short that doesn't >> work in a PFC because the duty cycle is much higher around the zero >> crossings of the AC going into the unit. >> >> A non-dissipative snubber such as in Weinberg converters can be iffy >> because the results are highly dependent on component tolerances in the >> FET and in the flyback transformer. >> >> Are there any other tricks to sponge off the leakage spike and dump it >> back into the pot? Maybe a fast and low capacitance "siphoning buck" >> that only takes off energy when it sees 2x input voltage or something? >> Must limit kind of hard though. >> >> -- >> Regards, Joerg >> >> http://www.analogconsultants.com/ >> > > Vladimir Vassilevsky posted this a while back: > > http://www.abvolt.com/misc/antiseptic.jpg >
That's a good name :-) A friend of mine always gets it wrong when we talk about SEPIC converters. He calls the septic converters. Unfortunately we need isolation so I can't have a cap of tens of nanofarads across the barrier. If I heave the leak energy across the barrier it would have to be via a small secondary transformer. Best would be to dump it into the source. -- Regards, Joerg http://www.analogconsultants.com/
Den mandag den 2. februar 2015 kl. 22.42.39 UTC+1 skrev Joerg:
> Folks, > > Have to do a flyback-PFC but this time with a really good efficiency, > >90%. The output voltage is high enough that sync rectification isn't > needed. AFAICT the only other way to boost efficiency is to not flare > off the leakage inductance spike via zener or snubber but to recycle it > instead. Unfortunately a recuperating winding only works if the duty > cycle stays under 50% at all times. Or if it's a double winding 66%, > similar in a tapped design, et cetera. Long story short that doesn't > work in a PFC because the duty cycle is much higher around the zero > crossings of the AC going into the unit. > > A non-dissipative snubber such as in Weinberg converters can be iffy > because the results are highly dependent on component tolerances in the > FET and in the flyback transformer. > > Are there any other tricks to sponge off the leakage spike and dump it > back into the pot? Maybe a fast and low capacitance "siphoning buck" > that only takes off energy when it sees 2x input voltage or something? > Must limit kind of hard though. >
I know nothing of PFCs but this is one of the first hits on google http://www.google.com/patents/US7649757 -Lasse
On 2015-02-02 3:25 PM, Lasse Langwadt Christensen wrote:
> Den mandag den 2. februar 2015 kl. 22.42.39 UTC+1 skrev Joerg: >> Folks, >> >> Have to do a flyback-PFC but this time with a really good efficiency, >> >90%. The output voltage is high enough that sync rectification isn't >> needed. AFAICT the only other way to boost efficiency is to not flare >> off the leakage inductance spike via zener or snubber but to recycle it >> instead. Unfortunately a recuperating winding only works if the duty >> cycle stays under 50% at all times. Or if it's a double winding 66%, >> similar in a tapped design, et cetera. Long story short that doesn't >> work in a PFC because the duty cycle is much higher around the zero >> crossings of the AC going into the unit. >> >> A non-dissipative snubber such as in Weinberg converters can be iffy >> because the results are highly dependent on component tolerances in the >> FET and in the flyback transformer. >> >> Are there any other tricks to sponge off the leakage spike and dump it >> back into the pot? Maybe a fast and low capacitance "siphoning buck" >> that only takes off energy when it sees 2x input voltage or something? >> Must limit kind of hard though. >> > > I know nothing of PFCs but this is one of the first hits on google > > http://www.google.com/patents/US7649757 >
Thanks, interesting. With a PFC this could be tough because the clapm level either needs to be controlled and persistently high or the energy in the cap C2 would have to "waft around" with the incoming AC phase. Which might be doable but requires an elaborate switcher to employ the stored energy for something useful. I am currently simulating an active clamp method, similar to this but not on the high side: https://duckduckgo.com/l/?kh=-1&uddg=http%3A%2F%2Fscholar.lib.vt.edu%2Ftheses%2Fpublic%2Fetd-71398-22552%2Fmaterials%2Fch3.pdf It's a pain. Even with an Intel i7. -- Regards, Joerg http://www.analogconsultants.com/

"Joerg"  wrote in message news:cja9aaFqc2bU1@mid.individual.net...

Folks,

Have to do a flyback-PFC but this time with a really good efficiency,
>90%. The output voltage is high enough that sync rectification isn't
needed. AFAICT the only other way to boost efficiency is to not flare off the leakage inductance spike via zener or snubber but to recycle it instead. Unfortunately a recuperating winding only works if the duty cycle stays under 50% at all times. Or if it's a double winding 66%, similar in a tapped design, et cetera. Long story short that doesn't work in a PFC because the duty cycle is much higher around the zero crossings of the AC going into the unit. A non-dissipative snubber such as in Weinberg converters can be iffy because the results are highly dependent on component tolerances in the FET and in the flyback transformer. Are there any other tricks to sponge off the leakage spike and dump it back into the pot? Maybe a fast and low capacitance "siphoning buck" that only takes off energy when it sees 2x input voltage or something? Must limit kind of hard though. -- Regards, Joerg http://www.analogconsultants.com/ Joerg, you need help. What is your input AC voltage range and output Volts and Amps? Sounds like you require I/O isolation. A Flyback at >50W or >0.93PF requires a professional. Cheers, Harry
On Mon, 02 Feb 2015 15:14:33 -0800, Joerg wrote:

> On 2015-02-02 2:16 PM, Tim Williams wrote: >> In my opinion, leakage is best dealt with first, in and of itself. >> What winding options do you have on this job? Custom OK? Current >> stock only? >> >> You want LL/(stray C) ~= Vsw / Isw. Too much interleaving is just as >> bad as too little. >> >> > It's a flyback, needs an airgap. The transformer will be custom so we > can decide what we want. We can't interleave much because safety > isolation almost requires a two-chamber bobbin or at least a thick > separation layer. > > It is the usual rock <-> hard spot situation.
Bullcrap. HV transformers are easy, including flyback types. You just have to understand that a high winding count means that the latter winds need to be isolated from earlier turns. Not from the other winding. That gets done by the bobbin. Each layer of turns in the secondary gets taped, and subsequent turns get wound on top of that. The margin at the bobbin faces should be such that there is not an upper layer turn near the face and other layer turns. I had secondary designs where the bobbin hub got 75 turns of say #43, and there was about a half mm space between the winding start and the finish, and the bobbin faces. That way, the tape which goes on top of that first layer tapes to the bobbin hub .5mm at each face and the next layer is inboard too. So there is always a half mm space at the edges of each layer where the tape provides layer to layer creepage, and the turns number keeps the differential low. Volts per turn and all that. So I could lay down 1500 turns on my secondary, and place a 15 turn, single layer say #26 bifilar (or single) on top of that. But you can bet none of those higher layer conductors ever go anywhere near the lead-in line or lower layers. The lead in gets a piece of Teflon tubing over it, as does the output lead. So I could pump 4kV from just a couple CW multiplier stages, instead of counting on more stages to get there, which is less efficient.
The search term would be non dissipative snubber
Which yields  this

http://www.ti.com/lit/an/slup100/slup100.pdf

The old unitrode app notes (which is now ti) are gold.
Mark