Active PFC PSU designs

[Mark R]

aPFC has been conventionally implemented as a 'boost' pre-regulator, positioned between the bridge rectifier and the reservoir capacitors in an SMPS.

The high voltage generated by the boost action, reduces I^2R losses in the DC-DC converter, allows design flexibility allowing reduction in main transformer core size and losses and allows greater energy storage in the reservoir caps (cap volume is proportional to C.V, but energy is proportional to C.V^2).

However, aPFC suffers significant losses due to switching, much attributable to the boost diode, which has led to the development of a whole series of 'active PFC' diodes made from GaAs or other special semiconductor.

The problem of passive rectification/clamping has been recognised for a long time - and the use of switches (MOSFETs) as active synchronous rectifiers is well established, as are actively clamped transformer based regulator topologies (e.g. half/full bridge and active clamp forward).

The boost topology requires a high side rectifier which makes control difficult - however, this type of switch is already used in a bridge-type circuit. So would it be possible to use this to kill two birds with one stone?

Even better - with a full bridge, you've already got a pair of switches that could short the PFC inductor to 0V - it's then simply a matter of switching off the low-side switch and turning on the low-side switch on the other side of teh transformer to catch the boosted voltage from the PFC inductor. Cool.

img128053435252880625.jpg

The full-bridge design produces nice voltage swings on the secondary so you could use self-driven active rectifiers.

So, a high-efficiency PFC SMPS with just 4 primary MOSFETs and no lossy PFC diode.

If you wanted to make a PC PSU - the multi 12V rail business gives you a nice option. Instead of using post-regulators, you could dispense with them and use several main transformers. E.g. have 1 transformer supplying co-regulated 12V and 5V - and another transformer providing co-regulated 12V and 3.3V. (I think I've seen some of the new high end PSUs do something very similar).

Full bridges are very efficient anyway - and if you could dispense with the post-regulators then the efficiency could be very impressive. A conventional full bridge can get over 95% efficiency, so I don't suppose it would be that hard to get 90-93% with PFC as well.

The problem is the control. I haven't the faintest idea how you could actually build a controller for such a hybrid thing - presumably you'd have to use a DSP.

 

[bobsmith1492]

Hm, a guy in my class made a power inverter using essentially that topology (with PWM to turn it into a sine wave). It was relatively painless to control it with a microcontroller, although he picked one with only 1 PWM channel and needed four... (one for the DC-DC boost, 12 to 2x200V, 3 for 3 phases). Then again, I think that's not related at all to your application... Good luck finding anyone on here knowledgeable enough to chat with... this is the kind of thing I want to study, but have no time with freakin' school, now.

Oh, and don't ever let school get in the way of learning. The smartest people never learn anything useful in school.

Oh, and sorry about the ranting; it's late and I'm tired.

 

[Mark R]

Been thinking about that circuit I described.

Aside from the fact that I drew it incorrectly (you can't put a capacitor in because that will defeat the PFC).

You end up with the problem that it has no reservoir capacitor - and therefore the reservoir caps have to go on the secondary side - which sucks. The other problem is that the PFC inductor will generate massive voltage swings across the transformer - this would likely sizzle the low-side switches. You could probably get around this with a switched capacitor (in much the same way as the forward converter could be actively clamped).

 

[futuristicmonkey]

I thought SMPS designs rectified the line voltage (hence the large reservoir caps) and then sent that thru a push-pull transformer topography, stepping it down at anywhere from 10 to 100KHz.

 

[bobsmith1492]

I don't think PC power supplies rectify the input first...

 

[futuristicmonkey]

Originally posted by: bobsmith1492
I don't think PC power supplies rectify the input first...

Why do you think the caps are rated for ~200 volts?

Text

From that site's excellent little article: Text

 

[Mark R]

Originally posted by: futuristicmonkey
I thought SMPS designs rectified the line voltage (hence the large reservoir caps) and then sent that thru a push-pull transformer topography, stepping it down at anywhere from 10 to 100KHz.

That is the conventional design. However, with active PFC there is an extra switching stage before the reservoir caps. An inductor limits the current flow into the reservoir caps, with a switch used to control the current flow - a control circuit matches the current so that it is proportional to the line voltage.

I've drawn a simplified schematic here:
IMG_6448.jpg

The idea I had was to combine the PFC and main DC-DC switching stages into one. This avoids the need for a PFC switch, while also avoiding the serious efficiency problems of using a passive diode for the PFC circuit. [Diodes don't shut off instantly - this means that when the PFC switch closes, the diode will (very briefly) allow the switch to short out the reservoir cap].

 

[NeoPTLD]

How would such device cope with modified square-wave input? Non-PFC CFLs, computers, ballasts runs fine on UPS, but but I have some PF >.99 active PFC electronic ballast that would malfunction and flicker when powered on quasi-square wave inverters and UPSs.

Requiring sinewave UPS would be an unacceptable demand, because the consumer do not directly benefit from power factor corrected PSUs.

The only direct beneficiaries are power companies and major account utility customers through reduction in KVA demand&poor power factor penalty and through reduction in KVA demand on customer owned distribution infrastructure.

So server rooms and computer labs with hundreds of KVA in SMPS loads have a valid need for active PFC, but for consumers it might actually be nothing but trouble.

 

[Mark R]

How well a high power factor SMPS handles modified-square wave input really depends on the PFC control system. There are a few known problems - mainly when using 'ferroresonant' circuits. It used to be common for high-end modified square wave inverters to use a ferroresonant transformer to smooth the square wave and also hold up the output while the UPS switched in. Under some circumstances, the combination of active PFC and ferroresonant transformers can result in malfunction of the PFC system.

In general, modern PFC controllers directly measure the AC line voltage and the AC line current, and force the current so that it is proportional to the line voltage. In the case of a modified square wave input, the 'dead time' between cycles will mean no power available to keep the reservoir caps topped up - so the output voltgae of the PFC stage will develop more ripple than it would normally do - however, the output of the PFC stage is only going to be regulated again - so this isn't usually much of a problem. Like any SMPS, there is a lot of finesse to designing a good PFC system, and it's very easy to make one that works badly.

Harmonic reduction (PFC) has been mandatory for Computers, TVs and monitors since 2001 in Europe, and is also mandatory in several other regions of the world. While most manufacturers initially chose to use passive harmonic reduction techniques, the reduction in the cost of aPFC, and impracticality of passive PFC as power levels approach 600-1000 W, has led many manufacturers to revaluate aPFC for their products. Depsite this, there are not really many issues with compatability of aPFC and inverters - most low-cost inverters are modified square wave, and aPFC has been used for a number of years without anyone really documenting any systemic problems.

Having said that, I'm not surprised that lighting ballasts don't work properly. Again, this is driven by the need for mandatory harmonic reduction for industrial/commerical lighting. However, in lighting systems, because of the low unit cost, and need for small size - there is immense pressure on the PSU designers to shave costs to the very bone, and to put as much as possible onto a single chip. This means that lighting aPFC is extremely crude - often the controller has no way to measure line voltage or current, and just assumes that the input voltage is sinusoidal. The other issue is that active PFC allows a reduction in size of the reservoir cap - and it may not be big enough for a modified-square wave. You're practically guaranteed for the system to malfunction with a non-sinusoidal input waveform.

And yes, harmonic reduction doesn't do much directly for the domestic end-user. There are some, theoretical, marginal reduction in energy costs due to lower voltage drop in the domestic wiring - but savings are of the order of $.20 per month for a domestic media server - so hardly worth bothering about, and potentially overwhelmed if the addition of the active PFC reduces the overall PSU efficiency (which in the case of PC PSUs is typical, mainly because cost is the number 1 factor in designing a PSU).

You can argue over the indirect effects - but there would not be an international standard for harmonic limits if they didn't cause significant problems. Unlike reactive low power factor (common in industrial applications, and a major factor in power grid management), harmonics are not easily managed at grid level, and the preferable approach is to address the issue at source.

Harmonics waste energy within the grid, can unbalance distribution systems (in a 3 phase system, normally the phase currents cancel in the neutral wire - harmonic currents sum in the neutral wire), and distort the mains voltage waveform. This distorted waveform causes EMI to radiate from power lines, it decreases the efficiency of electrical appliances powered from that supply, and causes other problems (e.g. motors and transformers run hotter and experience more vibration) when powered from a power grid polluted by SMPSs, and harmonics (due to the skin effect) cause greater power loss in power lines/transformers than you would expect from power factor alone.

 

[NeoPTLD]

Originally posted by: Mark R
How well a high power factor SMPS handles modified-square wave input really depends on the PFC control system. There are a few known problems - mainly when using 'ferroresonant' circuits. It used to be common for high-end modified square wave inverters to use a ferroresonant transformer to smooth the square wave and also hold up the output while the UPS switched in. Under some circumstances, the combination of active PFC and ferroresonant transformers can result in malfunction of the PFC system.

In general, modern PFC controllers directly measure the AC line voltage and the AC line current, and force the current so that it is proportional to the line voltage. In the case of a modified square wave input, the 'dead time' between cycles will mean no power available to keep the reservoir caps topped up - so the output voltgae of the PFC stage will develop more ripple than it would normally do - however, the output of the PFC stage is only going to be regulated again - so this isn't usually much of a problem. Like any SMPS, there is a lot of finesse to designing a good PFC system, and it's very easy to make one that works badly.

Harmonic reduction (PFC) has been mandatory for Computers, TVs and monitors since 2001 in Europe, and is also mandatory in several other regions of the world. While most manufacturers initially chose to use passive harmonic reduction techniques, the reduction in the cost of aPFC, and impracticality of passive PFC as power levels approach 600-1000 W, has led many manufacturers to revaluate aPFC for their products. Depsite this, there are not really many issues with compatability of aPFC and inverters - most low-cost inverters are modified square wave, and aPFC has been used for a number of years without anyone really documenting any systemic problems.

Having said that, I'm not surprised that lighting ballasts don't work properly. Again, this is driven by the need for mandatory harmonic reduction for industrial/commerical lighting. However, in lighting systems, because of the low unit cost, and need for small size - there is immense pressure on the PSU designers to shave costs to the very bone, and to put as much as possible onto a single chip. This means that lighting aPFC is extremely crude - often the controller has no way to measure line voltage or current, and just assumes that the input voltage is sinusoidal. The other issue is that active PFC allows a reduction in size of the reservoir cap - and it may not be big enough for a modified-square wave. You're practically guaranteed for the system to malfunction with a non-sinusoidal input waveform.

And yes, harmonic reduction doesn't do much directly for the domestic end-user. There are some, theoretical, marginal reduction in energy costs due to lower voltage drop in the domestic wiring - but savings are of the order of $.20 per month for a domestic media server - so hardly worth bothering about, and potentially overwhelmed if the addition of the active PFC reduces the overall PSU efficiency (which in the case of PC PSUs is typical, mainly because cost is the number 1 factor in designing a PSU).

You can argue over the indirect effects - but there would not be an international standard for harmonic limits if they didn't cause significant problems. Unlike reactive low power factor (common in industrial applications, and a major factor in power grid management), harmonics are not easily managed at grid level, and the preferable approach is to address the issue at source.

Harmonics waste energy within the grid, can unbalance distribution systems (in a 3 phase system, normally the phase currents cancel in the neutral wire - harmonic currents sum in the neutral wire), and distort the mains voltage waveform. This distorted waveform causes EMI to radiate from power lines, it decreases the efficiency of electrical appliances powered from that supply, and causes other problems (e.g. motors and transformers run hotter and experience more vibration) when powered from a power grid polluted by SMPSs, and harmonics (due to the skin effect) cause greater power loss in power lines/transformers than you would expect from power factor alone.

Most UPSs still use 60Hz switcher with using the same steel core tranny used for charging the battery while most "ac power inverters" use KHz range switcher and ferrite core trannies.