Low voltage PFC

he

't

ld

The idea should work. The PFC chip shouldn't care what the real voltage is so long as everything is scaled correctly.

You have a couple of ways you can go. One would be to use the lower voltage transformer and a booster. Another would be to go to a higher voltage transformer and make a bucker.

It works indeed. I've seen the designs with the PFC behind the transformer. It is convenient to have the control part on the secondary side. Also, the transformer acts as the additional filter.

For that purpose, boost topology has advantage.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

High-frequency switching PFC does not significantly reduce low frequency ripple voltage amplitude on the output filter. It only makes it look sinusoidal, by removing the higher harmonics.

The inductive storage element is intentionally ineffective at low frequencies, to provide dynamic control of the 120Hz modulation, so the low frequency energy storage is all in Cout, as in a simple capacitive rectifier filter. This energy storage is proportional to CV^2, which is why low voltage low frequency capacitive filters use large C values.

Large inductors can reduce ripple, but I doubt this is an option here. You seem to be imposing some pretty arbitrary component limitations, as it is.

RL

Actually, intentionally modifying the low-frequency rectified current waveform at the line frequency will aggravate the need for large output capacitance, at low voltage.

The only way to reduce Cout is to maximize the source conduction angle with current equal to or greater than the load - there will still be droop that depends on the Cout value at any time that Iout is less than the load value. For pulsating or sinusoidal sources, this will occur regularly at the source or double-source frequencies.

If the input voltage was applied to a simple boost converter, the source current waveform would be initially peaky, level off at the load value, then peak again before reducing to zero, as the source voltage fell below the modulator's drop-out. Outside of the modulator's operation, simple capacitive droop in the output cap would dominate.

The OP wants to optimize for low output C.

'Low' C values are acheived only if an intermediate energy storage is provided at high voltage.

RL

You need capacitance to store energy for some time. While the AC is near the zero crossing, which state lasts for milliseconds, there's no power coming in, so the caps have to power the load. So there's no particular advantage to using a PFC chip.

Conservation of energy again.

Some sort of reverse-PFC circuit could help a little... it would draw

*more* input current as the AC got lower, unlike a PFC chip that commands less. That is actually just a regular boost converter.

PFC chips need big filter caps. One advantage they have is that high-voltage electrolytic caps have higher energy storage per volume and per dollar than low voltage ones, and PFCs usually boost to around

400 volts.

Why are you constrained to use only X7R? Temperature? Have you considered polymer aluminums?

John

No real issues at all. I have a working PFC running off a 16VAC secondary for 34v@3A. But I think you'll need a electrolytic to filter out the ripple.

Cheers

You guessed it. 350F operating temperature. I have a PFC IC that works at this temperatures for our mission life (~2weeks). Wet tantalums are an option but they are huge for the space I have avaialbe. trying to find something small that will work.

Define 'work'. What are you actually stuck with at 180C?

If you can increase the voltage before regulation, without fancy electronics, you can apply the fancy electronics to the regulation function, which I assume is the ~aim, not just reducing the ripple of a low voltage filter.

RL

On Apr 1, 8:10=A0am, legg wrote: [... PFC to reduce the need for capacitance ...]

I disagree with this statement.

With a simple rectifier feeding a capacitor, the conduction happens only from the time that the input voltage passed the remaining voltage on the capacitor.

The conduction will stop just after the peak in the waveform.

With either a booster or a higher voltage transformer and a bucker, the conduction of the rectifier starts earlier and ends later.

he

't

ld

I just had a different idea that won't be PFC but also won't be a linear.

If you use a transformer with a higher output voltage and a MOSFET switch to disconnect the rectifier at the peaks, you can reduce the ripple. Your circuit will conduct before and after the peak but not actually at the peak.

You're right, anything that increases the conduction angle will reduce ripple.

Ideally the pfc modulation will reduce ppk ripple by a factor of about

3, for 1A on 1000uF, if the average output voltage is held constant between the two situations.

Using pfc just to reduce ripple is waste of parts, if the aim is actually regulation of the output voltage.

RL

This is a bit of a fallacy - this method of switched rectification provokes the same switching energy loss relationship as switching any voltage source into a capacitive load. Linear-like losses are absorbed by the switch and the lf magnetic coupler or distribution wiring as the peak to average current is aggravated. Attempts at pushing commercial applications of this idea have been attempted, in spite of this effect and in spite of cooking isolation transformers present in demonstrations of same.

On top of this, though the conduction frequency is increased, the conduction period is actually reduced, producing similar low frequency ripple filtering problems - all of which is fairly easily simulated.

If you are allowed to increase the source voltage, a buck regulator would be more productive and more easily achieved than the pfc circuit that was previously entertained.

You might also be allowed to switch the source lines to carry pre-regulated DC, which might reduce stress and complication at the high-temp load site, if the source location does not suffer the same temperature stress.

RL

e the

he

don't

could

I see no fallacy I what I wrote. I was pointing out this idea as a method to reduce the ripple voltage for a given filter capacitance. It draws an ugly looking current waveform but the switching device, has a much lower loss than the linear part doing the same job would.

It only cooks the transformer if the transformer is undersized for the ugly wave form.

For a given capacitor, the peak to peak ripple is reduced. The OP is in need of a method to do that. He may be willing to pay the price in the transformer.

I suggested exactly this idea earlier in another sub-thread. It does raise the parts count but it makes very good use of the capacitor.

If isolation is needed, the frequency could be raised and perhaps a squarewave used.

Are you not the OP in this thread?

RL

Sorry about the confusion between you and the OP. It's just that when you started talking about 'another idea', I'd assumed this was the OP offering a sequitur.

I think it's best to let the OP start suggestions of altering sources as a route to a solution - the issue of simple voltage and frequency relationship to stored energy and droop is already there. Finding out more about how extensively the temperature restrictions apply was just my attempt to get weaseling room on the basic issue of filter caps, real estate and application environment.

RL