3 phase PFC - why so hard?

[MagicSmoker]

So I'm pretty familiar with PFC of single phase AC which has been full wave rectified and (naively?) expected that a similar approach could be employed with 3 phase AC (ie - a boost converter after the 3 ph. bridge rectifier) with the only real difference being that the duty cycle of the PFC switch wouldn't vary so much because the DC output of a 3ph. bridge has much less ripple than a 1ph. bridge.

After digging into various papers on IEEE Xplore and elsewhere, it seems that doing PFC of 3ph. AC is worthy of masters' theses and doctoral dissertations, so I suspect I am missing something here...

And the explanation in this eetimes article bounces between making perfect sense to me and sounding totally wrong:

http://www.eetimes.com/document.asp?doc_id=1272370

More specifically, and referring to figs 2 & 3, why would the duty cycle of the boost switch go to zero for half of each cycle if the PFC controller?

Like I implied above, I suspect I am missing something important here, but damned if I know what. Anyone want to take on the thankless task of setting me straight?

[wraper]

More specifically, and referring to figs 2 & 3, why would the duty cycle of the boost switch go to zero for half of each cycle if the PFC controller?

Because it is obvious. First of all those figures are not about boost converter itself. Voltage after 3 phase rectifier never drops down to 0. No current flows from the phase when it goes below a certain voltage level because another phase catches the load once it's voltage level goes above another phase.

[mikerj]

The issue is that the PFC will be trying to simultaneously correct two phases at any particular time, and having no effect at all on the third phase.  This is why each phase needs to be treated separately.

Figure 3 is just showing what the current in one phase looks like, and therefore the parts of the cycle where a single PFC circuit would be active on that phase. As soon as one of the rectifier diodes on a particular phase becomes reverse biased the PFC can no longer have any effect on that phase.

[MagicSmoker]

The issue is that the PFC will be trying to simultaneously correct two phases at any particular time, and having no effect at all on the third phase.  This is why each phase needs to be treated separately.

Figure 3 is just showing what the current in one phase looks like, and therefore the parts of the cycle where a single PFC circuit would be active on that phase. As soon as one of the rectifier diodes on a particular phase becomes reverse biased the PFC can no longer have any effect on that phase.

Yeah, your explanation is the one I came up with when the eetimes article makes sense; the alternate explanation I come up with that throws everything into question is that when the boost switch turns on it effectively shorts all 3 input inductors to the 0V rail, so as long as the voltage on any phase is above 0V (absolute magnitude, because the bridge diodes will steer the polarity as needed) then current should build in the inductor. When the boost switch turns off then each inductor will act as a current source and deliver the energy stored within to the output capacitor.

The only real problem I can see with doing PFC on full wave rectified 3 ph. AC is that the reduced variation in duty cycle for the boost switch will result in higher harmonic distortion and/or incomplete pf correction, but probably good enough to drastically improve usable power on a given branch (assuming the load being fed is not a simple leading or lagging one).

[T3sl4co1l]

The conduction angle of a 3ph FWB (aka 6-pulse rectifier) is less than 360 degrees for each phase.

Even if the load current is made as sinusoidal as possible (or some funky wave shape to tweak some harmonics, at others' expense), there will still be "dead time" where the other pair of phases is powering the load, and one free-wheels.

So you need at least two independent channels (it's a degrees-of-freedom thing: suppose there were two pairs of wires, with a 90 degree phase shift between them*), or three for symmetry.

*Tesla's original configuration.  But, it uses one extra wire, and some percent more copper, so the equilateral three phase system was pretty quickly adopted.

Or you can use an altogether different method, like an active compensator.

Take in power through a triplet of current transformers.  Connect a trio of inductors, and a three leg inverter.

The inverters are switched in such a way as to act as an active rectifier, so you generate a controlled DC bus.  Put a nice beefy filter cap on there.

The magic comes here: control the switching of each leg, so that the instantaneous current through each inductor (as sensed by the CTs) is in phase with that phase's voltage.  Now the inverter draws unity power factor, by definition.  Fancy!

Over a longer time constant, control the magnitude of the three current channels, so as to regulate the DC bus.  Otherwise, drawing more current than needed would simply supercharge the capacitor, which would be bad.

So, as an active rectifier: if you draw a load from the DC bus, it supplies output power.  Or if you pump current into it, it absorbs that power back into the mains.  Sweet, we can make a combination battery charger and grid-tie inverter this way!

Here's the really magical part.

Attach additional loads after the CTs.  They act in parallel with the inverter, but the inverter still controls the total line current so that it remains sinusoidal.

Which means, if the loads aren't sinusoidal, the inverter will deliver the difference.  Bam, active inline PFC!

If you need only PFC, then the DC bus simply stores the reactive energy needed to restore power factor.  If you need to draw or dump power into the DC bus, you can do that, no problem.

The single phase equivalent is the two switch, direct-AC-to-boost circuit.  (Or the Cuk version, which uses a "single" switch, but it's bidirectional..)  Which if you use an H-bridge instead of two switches and two diodes, you can do grid-tie inverter duty just the same.

Tim

[Wolfram]

Even though the waveforms look quite bad, a three-phase rectifier with a resistive load presents a theoretical power factor of 0.95. This is good enough in many cases, and indeed this solution is widely used in industry. If significant smoothing capacitance is added across the output of the rectifier, the power factor degrades rapidly however. In cases where significant DC bus energy storage is needed, or where low DC bus ripple is required, it often makes sense to use a three-phase rectifier followed by a single PFC stage.

True 3-phase PFC, as described by Tim, is ideal for cases where a very low level of harmonic line currents are required, and for cases where you have bi-directional energy transfer. True 3-phase PFC can also be made somewhat more efficient in theory, as the rectifier diode losses are eliminated.

[MagicSmoker]

Thanks for the replies so far*, though it seems I have not framed my query correctly...

I get that a given diode in a 3ph. FWB feeding a purely resistive load only conducts for part of a cycle (120 degrees, to be exact), and that the current waveform tends towards a square wave of the same conduction angle as load side inductance is introduced and goes towards infinity. Or to put it another way, without PFC 2 phases out of 3 are supplying current to the load.

What I am getting hung up on here is that, intuitively, I would think the conduction angle for any one diode would increase to nearly 180 degrees when (high frequency) inductors are inserted in series with each phase input of the FWB and a boost switch immediately follows the FWB because when the boost switch turns on it effectively clamps the output of the FWB at 0V, so current should build in all 3 boost inductors any time the instantaneous phase voltage is not at zero.

Ergo, the conduction angle should increase to 180 degrees, though I can't say for sure via intuition alone that the power factor will meet IEC or EN harmonic standards, for example.

Now based on the replies from both T3sl4co1l and Wolfram I suspect the highly complex schemes for 3ph. PFC that I have read about on IEEE Xplore are an attempt to improve the PF from, say, 0.95 to 0.995, or to reduce the THD below 30% or whatever.

At some point I will likely work this out in LTSpice, but for now I am basically checking to make sure I understand what is going on correctly.


* - except for wraper, who, judging by his post history, seems incapable of behaving in a civil manner.

[Wolfram]

I thought a bit about the idea you're describing, and I couldn't immediately spot any problems with it, so I did a quick simulation in LTSpice.

The method seems to effectively reduce input current harmonics compared to the PFC with the inductor after the rectifier, but the overall power factor is reduced due to the inductors in series with the input. The following plot shows the output power and voltage, along with the line currents and one phase voltage when running from 230 V phase-phase, with 5 mH line inductors, 10 kHz switching frequency and 14 kW output power. The total power factor is 0.75 and the dominant harmonic is the fifth at 5 % of the fundamental current.

Reducing the line inductors to 1 mH (which is a more practical value anyways) results in the input currents appearing closer to the post-rectifier PFC case. The power factor is 0.91 and the fifth harmonic dominates at 24 % of the fundamental.

[T3sl4co1l]

IOW, it's no better than passive PFC.

Which is also a valid method, as long as you don't need terrific regulation and squeaky clean harmonics.

Another common way is to move to a 12-pulse rectifier.  This requires a power transformer, but power electronics customers are usually comfortable with the size and expense of passive and extremely reliable hunks of metal, like transformers.

Anyway, the method is, you use a delta-wye transformation, which happens to have a 60 degree phase shift.  A 12-pulse rectifier simply has three more legs of diodes, and thus the total input current flows with two levels of steps (instead of being 'on' or zero) that are interleaved.

Tim

[jahonen]

There exists a special 3-phase rectifier or a PFC, Vienna Rectifier. Anybody tried that out yet?

Regards,
Janne

[oldway]

Perhaps using an active power filter associated with 3 phases bridge rectifier ?

See also:
https://docs.google.com/file/d/0B5vXY4-Kg5GeTTNNLWJfU2tnSnM/edit?pref=2&pli=1

And:
http://intranet.ctism.ufsm.br/gsec/livros/eletronica.pdf
17.4 Active filters for power conditioning

[MagicSmoker]

I thought a bit about the idea you're describing, and I couldn't immediately spot any problems with it, so I did a quick simulation in LTSpice.

Close! The one thing that needs to change - and which I feel will improve PF & reduce harmonics in the current waveform noticeably - is to control the boost switch with a constant on-time, boundary conduction mode strategy (popularized in, e.g., "one cycle" control ICs like IR1155 [which Infineon - nee IRF - seems to be discontinuing... grumble]). Granted, BCM is not the control scheme of choice for high power due to the high peak current in the switch (~2x the average), but I'm only considering the relative simplicity of one scheme over others in the academic literature (including - nay, especially! - the modulation scheme for a Vienna Rectifier). I also have to point out that much of what is published in, e.g., IEEE Xplore, is totally impractical from a commercial standpoint, even if it does work in anything besides MATLAB or PSIM (which, sadly, is also relatively uncommon).

@T3sl4co1l & @jahonen - see above.

@oldway - I have the second book in my library, but not the first one - thanks for the link! I am going to assume you are suggesting a shunt active filter (because what has been discussed above is, effectively, series active filtering). The big problem with shunt active filters is that you rarely know what other harmonic-producing/reactive loads are present and therefore it is difficult to adequately AND economically specify the power rating of the components in the filter. Also, the modulation scheme for the switches is no less difficult to implement (frequently needing vector control techniques with rotational transformations - much like a high performance AC motor drive). It is this control difficulty (and, frequently, circuit complexity) which drives so many industrial power conversion equipment designers to choose line-frequency transformers and passive PFC techniques (e.g. - 5th harmonic traps) when isolation and PFC is needed.