I am working on a new design that requires to quickly charge up to 200V an electrolytic capacitor from a 24V voltage source. I am going to use a dedicated IC from Linear (LT3751) that is designed to control a non-isolated flyback topology.
As far as I can see, making a minor change to the flyback topolgy I get a coupled-inductor boost topology: I have tried to simulate it with LT-Spice and the IC seems to perform good control, exactly as it does with the flyback, with the advantage of a lower voltage stress on the mosfet (because there's no reflected voltage during the switch-off interval).
Does anyone can explain me which are the real differences (pros & cons) between these two non-isolated topologies ? And why one should choose one instead than the other ?
Thanks a lot.
Davide

Perhaps a schematic might be useful to everyone to help answer the questions. There is probably more than one version of a coupled inductor boost topology.

Charging capacitors....ever thought of using a discontinuous series resonant converter?

the LT3751 is expressly designed to perform high voltage capacitor charging...

The inductor has 1:1 ratio. Is it right ?
with this ratio:
- coupled inductor boost is better to deliver higher output voltage.
- coupled inductor boost should have less problem with primary snubber. Most probably it will survive without snubber at all.
- coupled inductor boost could have trouble with inrush current. Normally it is easy to solve.
- coupled inductor boost does not allow output voltage less input voltage. Can be problem with some flash gas discharge lamp.

The coupling ratio is 1:5 (why did you suppose it was 1:1 ?).
The output capacitor is 150uF, not an issue for inrush current.
The output voltage must be always higher than the input voltage.
Is the coupled-L boost topology still better ?

Hello Davide,
I think coupled-L topology make particular sense if we want to use standard one-to-one trafo.
If the ratio is not 1:1. And we are free to choice the ration like 1:5 or what else.
Then I see no reason to use coupled-L topology.
In this case we can use standard flyback or non isolated flyback which is the same except feedback.
Probably leakage-L will be little enough to ignore primary high voltage spike.
if the trafo is not so good then use the circuit shown on page 29 of data sheet, components D3, D4,C5.

Regarding reflected voltage and MOSFET voltage stress.
Voltage is reflected in any case, and MOSFET is stressed.
For flyback: Vrflctd = Vout/CR where CR coupling ratio, 5 times
For coupled-L: Vrflctd = (Vout-Vin)/(CR+1) so it is less a bit

Regardles of the value of the output and input voltages, a flyback converter processes 100% of the power, while in a tapped boost part of the power is transferred directly from the input to the output.
The lower the Vout/Vin ratio, the better the boost (or tapped boost will be: when the input and output voltages are equal, the boost converter (and its losses) essentially vanish.
As the Vout/Vin ratio increases, the benefit of the tapped boost vs flyback diminishes while the inability to offer protection against a shorted output remains.

The coupled inductor boost offers no noticeable benefit when boosting the voltage from 9V to 350V, while being susceptible to failure in case of a shorted output.
It is incumbent upon the fellow that designed the circuit to justify his topology selection, so maybe you should contact him and ask - I would love to hear the answer!

I love these coupled inductor variations, they are fun to think about.

Note the coupled boost you have drawn will no longer have a continuous input current, which is a disadvantage over the normal boost.

Other disadvantage over normal boost is that there will be spikes on the voltage due to leakage inductance. Of course, the flyback has both these disadvantages too.

Curious to know how the RHP zero is affected - not sure I have time to derive that though but has anyone done it already?

Same story with DC gain. Maybe a student online here can do that for us.

OK I couldn't help myself. I threw Vorperian's PWM switch model into the circuit to get the DC gain. My calculations show that the gain in CCM is given by

(nD/D' +1/D')

So the tapped boost helps with the step up, of course.

I'd probably concur with others that the flyback would be preferable. I'd probably do a dual output and stack them to get the 350 V.

Alternatively, two boosts in series might be a good solution as well.

But it would be interesting to hear from the original designer what his compelling reason was for choosing this.

The use of coupled inductor topology may be due to the fast response that it can provide?

I don't think it will be a fast response. There will be a RHP zero, and the crossover is limited by that.

The specs on this are a bit vague - do you charge as a current source, or try to run continuously as a regulated output?

It makes a big difference on the voltage specs too. 24 V to 200 V can be done with a single boost, but 9 to 350 V is a lot tougher.

Sometimes topologies are chosen for no sound reason, but having finished the design, they may make a good app note anyway. Always be careful when biting off a new topology, there will always be issues that you didn't expect.

My application requires to have a charged capacitor that acts as an high voltage reservoir to shot a some current pulses that will charge a piezo actuator (because the piezo is a capacitive load, the current pulse will be limited in hysteretic mode by a mosfet leg).
I guess such application could be performed charging the cap as a current surce but I'm not sure.
Anyway the LT3751 can operate in both way: charging-capacitor-mode or regulated-output-mode. What are advantages and disadvantages of the two approaches ?

I have provided a snubber made with a 1500Wpk TVS (SMCJ20A): I hope it will be enough. Anyway the application circuits from LT always show snubberless solutions in favor of a high-voltage-rated FET (look at the LT3751 data sheet).

I would personally go with the RCD, and not use a TVS. The problem that they have is that the dissipation of energy is very brief, and there is more EMI. The cap solution spreads it out over time. Most industrial solutions I see have the RCD clamp.

To keep things straight, the TVS or RCD are voltage clamps, not snubbers.
I disagree with Ray that the TVS will necessarily generate more EMI than the RCD (a detailed rebuttal is ouside the scope of this discussion).
The main difference between the two is that the RCD will dissipate significantly more power than the TVS when the converter is operated in "burst mode" in order to reduce light or no load power consumption.

Good point on the burst-mode dissipation. It's definitely lower dissipation for that.

Just to keep you informed, I come from testing the circuit in "flyback" mode and it works fine. Very fast capacitor charging (150 uF at 300V in less than 100ms), great voltage stability, no emc problems even during the charging interval. Stray and leakage inductances are so small that the snubber/clamp is absolutely useless even with a 60V rated mosfet. Thanks to everyone for supporting. And thanks to Linear for their great products....

Good to know the flyback worked. Simpler is often the right way to go.

did you do a single output? Make sure you measure the stress on the diode, especially during startup when the converter is in CCM mode.

It has single output. Which kind of measurement should I do to evaluate the stress on the diode ?

I recommend you put a scope on it. I usually float all the grounds except for the ground of the scope. Then put the scope ground on the quiet side of the diode and look at the reverse voltage. From my experience, overvoltage on this diode is a common failure mechanism. Worst case stress usually occurs during startup.