Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

[SOLVED] Flyback snubber not working!

Status
Not open for further replies.

coates

Member level 1
Joined
May 20, 2013
Messages
38
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,286
Activity points
1,660
Hi guys,

I'm trying to replicate and understand an offline flyback converter and I have some questions for those with experience. The circuit I'm trying to replicate is this one. It uses the NCP1216 controller.

I'm using a variac to soft-start my line voltage so I can identify problems before components blow. The secondary has no-load, just the capacitors+bleed resistor.

When I'm ramping the voltage up to ~120VDC (minimum mains) the controller is going flat-out, i.e. 50% duty (because the FB pin is still high), the primary current gets quite high, and the voltage across the snubber capacitor greatly exceeds what one would expect (>1kV). Once the voltage ramping has reached ~120VDC the primary current falls to nothing and the voltage across the snubber falls to normal levels (~150V). Everything is OK. I'm pretty sure the primary current only falls because the controller enters "skipping" mode because the feedback pin.

As soon as I put a load on the secondary the snubber voltage ramps up beyond normal levels again and the secondary rectifier diode fails. I think this is because the reflected drain voltage exceeds the diode voltage: Np/Ns = 6.25, 1kV/6.25 = 160V > 100V. The diode fails before the transistor even though both their voltage ratings are being abused. I think this is because it's rated better for avalanche mode.

Why is the RCD snubber not clamping properly? Does this circuit really work?

Thank you.
 

voltage across the snubber capacitor greatly exceeds what one would expect (>1kV).

The capacitor (in the DCR snubber) must quickly absorb the large spike at shut-off. If it is too small then it charges to a skyhigh volt level which is the problem you report.

Then the resistor must discharge the capacitor during one-half of the cycle.

I think it will help if you make C larger, and R smaller. Notice if R is too small, then it tends to cause continued current flow around a small loop which includes the transformer primary. You may need to bump up the resistor's W rating lest it burn up.

----------------------

Several factors need to be right, so that they interact to give the amount of power you wish. These aren't the only factors but they are adjustable parameters in the animated simulator I use.

* switching frequency
* duty cycle
* turns ratio
* primary Henry value

It's our human tendency to build a small transformer and then try to push a lot of power through it. We might adjust duty cycle, frequency, etc. But it brings out behavior that starts burning up some component or another.

- - - Updated - - -

This recent thread might be of interest:

https://www.edaboard.com/threads/350960/
 

Thanks for your reply, Brad.

I think it will help if you make C larger, and R smaller. Notice if R is too small, then it tends to cause continued current flow around a small loop which includes the transformer primary. You may need to bump up the resistor's W rating lest it burn up.

I doubled C by putting another in parallel with it (I don't have many high-voltage caps!) but it had no effect. The snubber voltage was still way too high.

I should mention that I'm using a 7W wirewound, enamelled snubber resistor and it's burning up as it is. Will making it smaller really help? Won't it just burn even more?

More importantly where is this power coming from? I short-circuited the secondary winding(s) and measured the inductance on the primary with my crappy LCR meter to get an approximate reading of the leakage inductance. It gave an approximate reading of 3.5uH -- 3.5% of Lmag -- which is what the web says it would roughly be. I measured the primary current under load and it got to a maximum peak of 4A. This suggests 28uJ of leakage energy to dissipate. Multiplied by 65kHz this suggests 1.8W. I don't think the resistor should be so hot considering it's using less than half its capacity.


Regards

- - - Updated - - -

Additionally, I spotted something that may be of concern. After the leakage spike has been dissipated the drain voltage drops below ground briefly before the Lmag/Coss ringing. Please see the images below:

1.jpg

2.png

Took me a while to realize that it was the body diode of the MOSFET conducting because it matched the forward voltage and reverse-recovery time in the datasheet.

I have never seen this before in flyback waveforms. Is this caused by the rectifier-ringing being reflected back to the primary and making the drain go negative? If so, how is it prevented?

Regards
 

I should mention that I'm using a 7W wirewound, enamelled snubber resistor and it's burning up as it is. Will making it smaller really help? Won't it just burn even more?

More importantly where is this power coming from? I short-circuited the secondary winding(s) and measured the inductance on the primary with my crappy LCR meter to get an approximate reading of the leakage inductance. It gave an approximate reading of 3.5uH -- 3.5% of Lmag -- which is what the web says it would roughly be. I measured the primary current under load and it got to a maximum peak of 4A. This suggests 28uJ of leakage energy to dissipate. Multiplied by 65kHz this suggests 1.8W. I don't think the resistor should be so hot considering it's using less than half its capacity.

My knowledge is insufficient to disagree with your formula, but seeing the schematic it appears you have upwards of 110W going through the transformer (output 24v x 4.5A). This indicates the amount of energy that must be absorbed briefly by the capacitor at turn-off. It may not be 110W continuous, but it is momentary. It still charges the capacitor to a few hundred volts. The goal is to keep the capacitor voltage from getting so high that it exceeds the diode rating, or mosfet rating, etc.

You must discharge the capacitor sufficiently during each cycle. The spikes of current from the transformer are invincible. No matter how high the capacitor voltage is, another spike arriving automatically generates sufficient voltage to push it even higher. More than 1000V appeared at terminal 1 of your transformer. That's how it charged your capacitor to more than 1kV. The solution is that you have to keep that node voltage below your mosfet rating (650V). Therefore your resistor needs to be low enough ohms so that it discharges the capacitor almost completely during each cycle.

Several W is the power rating of your resistor, and agrees with the schematic. That is probably adequate. But I think the schematic's value of 20k is too high ohm value.

This DCR snubber appears to be the optimum configuration of snubber, to do the job of dissipating single polarity spikes emerging from the transformer. There is a tradeoff of RC values which ought to work best, and it requires testing with several combinations of values, to find the proper tradeoff.

- - - Updated - - -

There is another way to look at RC values. C can be a large value, so that its voltage goes up and down only a few V during a cycle. This means it hovers around a certain voltage. It does this because R discharges the capacitor at a more or less constant current. In effect, R sets the 'hovering' voltage level. Picture how R does this. High ohm, high V. Low ohm, low V. You can try to get away with a high ohm value, but the voltage eventually rises to dangerous levels. This is where the proper tradeoff must be found.
 

there appears to be A LOT of leakage inductance in your transformer, causing the massive turn off voltage and subsequent ring to gnd...!
 

here is a flyback simulation in the free ltspice, you can play with it to experiment with RCD snubbers
 

Attachments

  • Flyback bias winding and transformer leakage.txt
    16.8 KB · Views: 102

My knowledge is insufficient to disagree with your formula, but seeing the schematic it appears you have upwards of 110W going through the transformer (output 24v x 4.5A). This indicates the amount of energy that must be absorbed briefly by the capacitor at turn-off. It may not be 110W continuous, but it is momentary. It still charges the capacitor to a few hundred volts. The goal is to keep the capacitor voltage from getting so high that it exceeds the diode rating, or mosfet rating, etc.

You must discharge the capacitor sufficiently during each cycle. The spikes of current from the transformer are invincible. No matter how high the capacitor voltage is, another spike arriving automatically generates sufficient voltage to push it even higher. More than 1000V appeared at terminal 1 of your transformer. That's how it charged your capacitor to more than 1kV. The solution is that you have to keep that node voltage below your mosfet rating (650V). Therefore your resistor needs to be low enough ohms so that it discharges the capacitor almost completely during each cycle.

Several W is the power rating of your resistor, and agrees with the schematic. That is probably adequate. But I think the schematic's value of 20k is too high ohm value.

This DCR snubber appears to be the optimum configuration of snubber, to do the job of dissipating single polarity spikes emerging from the transformer. There is a tradeoff of RC values which ought to work best, and it requires testing with several combinations of values, to find the proper tradeoff.

- - - Updated - - -

There is another way to look at RC values. C can be a large value, so that its voltage goes up and down only a few V during a cycle. This means it hovers around a certain voltage. It does this because R discharges the capacitor at a more or less constant current. In effect, R sets the 'hovering' voltage level. Picture how R does this. High ohm, high V. Low ohm, low V. You can try to get away with a high ohm value, but the voltage eventually rises to dangerous levels. This is where the proper tradeoff must be found.

I tried lots of different combinations but could not get the snubber resistor to a reasonable size! :bang:



you'd better post your circuit as else all is guess work...

Click me to see the circuit.

I have used the NCP1216A instead of the NCP1216, to limit the duty cycle to 50% because I wanted it to be a DCM design since most posts suggest this is the easiest. I believe the linked circuit only needs CCM because of the overload requirement.

I tried my best to replicate the transformer with what I have. The transformer is made of two EC52 halves made of 3C90 material with an air gap to give an inductance factor of about 126nH/T². Primary is a single layer of 28T of enamelled 22AWG. Secondary is 5T of enamelled 22AWG. Separated by layers of PET tape. This approximately gives the required 100uH primary inductance. I did some flux calculations to ensure the transformer didn't saturate. Looking at the primary current it isn't saturating.

The circuit is nailed together on stripboard! I know this will cause noise problems but I'm only trying to get to grips with the basics first.



there appears to be A LOT of leakage inductance in your transformer, causing the massive turn off voltage and subsequent ring to gnd...!

I understand that my awful hand-wound transformer won't have the best performance but I measured the leakage and it's still only 3.5% of the primary, which is not too bad according to web sources.



I put an RC snubber across the rectifier diode and the ringing improved massively. I then put a further RC snubber across the primary switch and improved the ringing even further. However I believe it's only diverted the power to the new snubbers because the output is regulating more.

Do you guys think I'm barking up the wrong tree trying to make a 4A supply with a DCM design, as opposed to CCM? Won't the current spikes of ~9A cause noise problems later on? The CCM design seems "softer".



Many thanks.
 

so your spec is
vin = 120-240vac
vout = 24vdc
iout = 4.5A

...THIS WOULD ONLY BE POSSIBLE WITH A SINGLE SWITCH FLYBACK (sorry for caps) if you can get the leakage inductance low enough.
You need to do sandwich (interleaved) winding to get Lleak down

Remember whatever is the peak current in the leakage inductance, then the energy is 0.5.L.I^2, and power is 0.5.L.I^2.f…….the actual dissipated power in the clamp is often up to three times that, due to the primary power current getting diverted into the clamp for a short period.
So you need to tell us you lleak value. And also the peak pri current, and f(sw)

- - - Updated - - -

also, you really need pfc for your smps because its over 75w...unless its just at 24*4.5 Watts for very short intervals?
 

so your spec is
vin = 120-240vac
vout = 24vdc
iout = 4.5A

Yes, close enough!

...THIS WOULD ONLY BE POSSIBLE WITH A SINGLE SWITCH FLYBACK

I am aiming to use a single switch flyback for simplicity but why couldn't any other switched-mode power supply be used? What's wrong with a half-bridge, for example?

So you need to tell us you lleak value. And also the peak pri current, and f(sw)

Short-circuiting the secondary winding and measuring the primary inductance gave approximately 3uH. The Fsw is between 65-85kHz. The maximum primary current would be about 9A because of the low inductance but it won't be this in practice because of feedback.

also, you really need pfc for your smps because its over 75w...unless its just at 24*4.5 Watts for very short intervals?

I tried to find a standard controller IC with built-in PFC but couldn't find one that's popular (I looked at the stock on supplier websites). No point choosing an IC that will be discontinued if it becomes unpopular. Can you suggest a sustainable way of implementing PFC?



Many thanks.
 

for PFC you could use a "boundary conduction mode" boost PFC chip.
I think ti.com do them.
Infineon might do some

- - - Updated - - -

if you do use a pfc stage then that eases your flyback design, because your primary voltage will be around 400v and your primary current will be very low.

- - - Updated - - -

here is an ltspce sim of your flyback, also the excel design file.
Its quite a few watts in the rcd clamp at min vin. ..but not impossible
You can go ALT+LEFT CLICK to see dissipations, which of course, are just guideline.
 

Attachments

  • Flyback 24V 4.5A.txt
    6.5 KB · Views: 60
  • 24v 4.5a flyback.zip
    4.6 KB · Views: 65

for PFC you could use a "boundary conduction mode" boost PFC chip.
I think ti.com do them.
Infineon might do some

- - - Updated - - -

if you do use a pfc stage then that eases your flyback design, because your primary voltage will be around 400v and your primary current will be very low.

- - - Updated - - -

here is an ltspce sim of your flyback, also the excel design file.
Its quite a few watts in the rcd clamp at min vin. ..but not impossible
You can go ALT+LEFT CLICK to see dissipations, which of course, are just guideline.

Thanks for this information, treez. However, the turns ratio is incorrect in the simulation. When I change the inductances to the correct ratio I can see subharmonic instability. Is there an easy way to modify the simulation to get rid of this?


Many thanks.
 

the turns ratio is 1:1 and I made it like that to get the right flyback CCM waveforms.
You need a low primary conduction time, and longer secondary conduction time.
If you have changed the turns ratio and are getting subharmonic oscillations, then that is because you have not changed it correctly, why did you want to change it?......subharmonic instability occurs at duties of 0.5 plus and this flyback should never be operating with 0.5 pus duty cycle.
 

the turns ratio is 1:1 and I made it like that to get the right flyback CCM waveforms.
You need a low primary conduction time, and longer secondary conduction time.
If you have changed the turns ratio and are getting subharmonic oscillations, then that is because you have not changed it correctly, why did you want to change it?......subharmonic instability occurs at duties of 0.5 plus and this flyback should never be operating with 0.5 pus duty cycle.

I changed it to represent my transformer, operating in DCM. Otherwise the snubber power dissipation would be incorrect.


Regards.
 

ok i'll look into a DCM design if you wish it.
But here attached is another CCM version, this time with ns/np reduced down to 0.53...this means less reverse voltage on the secondary diode.

- - - Updated - - -

Oh and by the way, if you want a DCM design over the entire input voltage range, and you are not using pfc, then you may want to be doing it with a BCM flyback chip, or else otherwise you may find yourself too deeply in DCM at high line.

- - - Updated - - -

Also attached now here is the DCM Flyback version of your design, where you can note that high secondary current as you already discussed.
(LTspice sim of DCM flyback attached)
 

Attachments

  • 24v 4.5a flyback _1.zip
    4.7 KB · Views: 58
  • Flyback 24V 4.5A _1.txt
    6.5 KB · Views: 63
  • Flyback 24V 4.5A _DCM.TXT
    6.5 KB · Views: 55

I tried lots of different combinations but could not get the snubber resistor to a reasonable size! :bang:

Here is my simulation. It focusses on snubber action. (I do not know what answer to give regarding other issues being discussed.)



Notice the waveforms on the snubbing components. The resistor needs to be rated for tens of W. This is just a theoretical simulation, of course.
 

When I'm ramping the voltage up to ~120VDC (minimum mains) the controller is going flat-out, i.e. 50% duty (because the FB pin is still high), the primary current gets quite high, and the voltage across the snubber capacitor greatly exceeds what one would expect (>1kV).

It really does sound as though the transformer is not designed for the power thru-put you want, i.e. too small...!

- - - Updated - - -

Exactly what is the leakage inductance referred to the primary? if it is 10uH say and the peak current is typically 4.5A say then the snubber has to remove approx 0.5 x 0.5 L I^2 freq = 10 watts at 100kHz, see why the leakage is so important...!
 
  • Like
Reactions: coates

    coates

    Points: 2
    Helpful Answer Positive Rating
coates claimed 3.5 µH leakage throughout this thread, unfortunately the number doesn't fit the observations of burned snubber resistor. But we didn't yet see an actual snubber circuit with component values. Of course it's possible to get much more than leakage inductance energy by choosing unsuitable values. Also, can we be sure about correct secondary winding polarity?
 

coates claimed 3.5 µH leakage throughout this thread, unfortunately the number doesn't fit the observations of burned snubber resistor.

Another possibility (perhaps remote but good to remember) is resonance; the LC frequency (snubber) should be far away from the PWM frequency. Resonance can get you in serious trouble.
 
  • Like
Reactions: coates

    coates

    Points: 2
    Helpful Answer Positive Rating
also, avoid operating the flyback at low input voltages, as that will mean your error amp is saturated and youll get really highest peak current, ...if you look at the sims I sent you above, youll see I meant for you to use enough bus capacitance to have at least 140vdc minimum at the input to the flyback...when at 120vac.
You need to avoid operating with vin too low, -that will overstress the snubber, and really you need under voltage lockout so you cnat operate below a certain voltage.
 
  • Like
Reactions: coates

    coates

    Points: 2
    Helpful Answer Positive Rating
Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top