Forums

Active PFC and square wave supply

Started by Pimpom June 27, 2013
I started using a Tagan 700BZ PSU in my computer a couple of months 
ago - my first PSU with active PFC. The UPS is a basic one that has 
served me well for 10 years, and outputs a square wave on battery. 
While setting up this particular combination, I briefly wondered how 
the PFC circuit would react to working with non-sinusoidal power but 
then forgot about it - until it died a couple of days ago. It blew the 
fuse on the UPS and shows a short at the AC input. One of the two 
paralleled 20N60C3 MOSFETs was a dead short.

The UPS battery - a 70Ah car battery - is of the same age as the UPS 
(10 yrs) and has little backup power left. So I always shut my 
computer down quickly in the event of a power outage, of which there 
were dozens during the 2 months I've been using this PSU.

The PSU died while I was in another room. I didn't notice a power 
outage or fluctuation during that time but can't be 100% certain. Is 
it likely that there was an outage and the PFC transistor was finally 
killed by the square-wave supply?

To put it another way, if I repair the PSU, is it likely to go poof 
again unless I change my UPS to a more sophisticated one? I'd never 
done a detailed study of practical active PFCs before and could well 
miss something with a quick analysis.

(I did some online reading about active PFCs and non-sinusoidal 
supplies after my PSU died but they're mostly end-user discussions and 
are all inconclusive).


Pimpom wrote:


> To put it another way, if I repair the PSU, is it likely to go poof > again unless I change my UPS to a more sophisticated one? I'd never > done a detailed study of practical active PFCs before and could well > miss something with a quick analysis.
Hmmm, I can't imagine a PFC can work properly when fed a non-sinusoidal wave. But, the typical PFC is a boost regulator, so the transistors are ACROSS the DC bus (with an inductor in series). I can imagine that maybe the boost regulator is asking for high duty cycle = maximum boost at the zero voltage point, then the UPS switches on and the PFC creates excessive boost voltage before the PFC regulator can recover. But, that is just a guess, and I'm assuming the UPS is not really true square wave but "modified sine wave" which means there is a zero Volt perion between each output pulse. Anyway, I cannot imagine a PFC supply can properly handle a UPS with non-sine output. As these PFC supplies become more common in larger appliances like computers, the UPS makers surely are going to have to address these problems. And, it sure isn't clear how to do that very cheaply. Mostly, they'd need PWM synthesis of the sine wave, and then enough filtering to smooth out the PWM carrier. Maybe smarter PFC's could detect square-wave input and modify their PWM logic to compensate. Jon
Jon Elson <jmelson@wustl.edu> wrote:
> Pimpom wrote: > > >> To put it another way, if I repair the PSU, is it likely to go poof >> again unless I change my UPS to a more sophisticated one? I'd never >> done a detailed study of practical active PFCs before and could well >> miss something with a quick analysis. > Hmmm, I can't imagine a PFC can work properly when fed a non-sinusoidal > wave. But, the typical PFC is a boost regulator, so the transistors > are ACROSS the DC bus (with an inductor in series). I can imagine > that maybe the boost regulator is asking for high duty cycle = maximum > boost at the zero voltage point, then the UPS switches on and the > PFC creates excessive boost voltage before the PFC regulator can recover. > > But, that is just a guess, and I'm assuming the UPS is not really true > square wave but "modified sine wave" which means there is a zero Volt > perion between each output pulse. Anyway, I cannot imagine a PFC > supply can properly handle a UPS with non-sine output. As these > PFC supplies become more common in larger appliances like computers, > the UPS makers surely are going to have to address these problems. > And, it sure isn't clear how to do that very cheaply. Mostly, they'd > need PWM synthesis of the sine wave, and then enough filtering to > smooth out the PWM carrier. > > Maybe smarter PFC's could detect square-wave input and modify their > PWM logic to compensate. > > Jon
An interesting test would be to see what a PFC does when you feed it plain old DC. Has anybody exploded any switching power supplies of any sort by just feeding them DC, filtered or with 120Hz ripple? They should int theory love that, but I'm not sure about the autoranging device they use before the filter caps that either rectifies or doubles the AC input.
My advice is use a double conversion true sine UPS or none at all. With 
a double conversion UPS, you will know soon enough if everything is 
compatible when the fecal matter hits the fan. That is, line and battery 
operation are similar since the inverter is always running.

If you go through
http://www.xbitlabs.com
website, they test power supplies run from a crappy UPS. just to see how 
much power they can deliver before the UPS protection circuitry cuts in. 
Some supplies really piss off the UPS!

The drawback to double conversion is the thing is running 24 and 7. And 
it makes noise. Oh, and they aren't cheap.

The advantage to double conversion is you have no brownout issues, line 
to battery switchover is fast since the inverter is already running, and 
the surge suppression is as good as it gets since they are filtering DC 
prior to the inversion.

Cydrome Leader wrote:


> > Has anybody exploded any switching power supplies of any sort by just > feeding them DC, filtered or with 120Hz ripple? They should int theory > love that, but I'm not sure about the autoranging device they use before > the filter caps that either rectifies or doubles the AC input.
Years ago I worked on an outboard PFC that supplied PCs with pure DC at 340 V, worked fine as LONG as the PC had a classic voltage doubler/switch PWM supply. But, it wouldn't fly, as the PC's power switch and fuse were not rated for DC. And, of course, woe be to anyone that plugged in, say, a monitor with a 60 Hz transformer or AC degaussing coil! I think many PFC front ends might work fine with DC input within the right range, they'd just become a nice boost converter. But, of course, it depends on the control loop of the PFC chip. Jon
Jon Elson wrote:
> Pimpom wrote: > > >> To put it another way, if I repair the PSU, is it likely to go >> poof >> again unless I change my UPS to a more sophisticated one? I'd >> never >> done a detailed study of practical active PFCs before and >> could well >> miss something with a quick analysis. > Hmmm, I can't imagine a PFC can work properly when fed a > non-sinusoidal wave. But, the typical PFC is a boost > regulator, so > the transistors > are ACROSS the DC bus (with an inductor in series). I can > imagine > that maybe the boost regulator is asking for high duty cycle = > maximum > boost at the zero voltage point, then the UPS switches on and > the > PFC creates excessive boost voltage before the PFC regulator > can > recover. > > But, that is just a guess, and I'm assuming the UPS is not > really true > square wave but "modified sine wave" which means there is a > zero Volt > perion between each output pulse. Anyway, I cannot imagine a > PFC > supply can properly handle a UPS with non-sine output. As > these > PFC supplies become more common in larger appliances like > computers, > the UPS makers surely are going to have to address these > problems. > And, it sure isn't clear how to do that very cheaply. Mostly, > they'd > need PWM synthesis of the sine wave, and then enough filtering > to > smooth out the PWM carrier. > > Maybe smarter PFC's could detect square-wave input and modify > their > PWM logic to compensate. > > Jon
You're right in that my UPS is not true square wave but has a dead time in between transitions. My son has been using a PSU with active PFC for more than a year and his 800VA UPS ($45 with a 9Ah internal battery) outputs the same type of waveform except that mine has noticeably steeper edges. Maybe that's what makes the difference. Some of my friends are also using PSUs with PFC and, judging from what they paid for their UPSes, I'm sure none of them is a sinusoidal type. The focus here is on how the PFC components bear up to a non-sinusoidal input rather than on the actual PF correction. Similarly, although my computer needs less than half of the PSU's 700W rating, I used it because it returned good performance figures in reviews and I got a used unit cheap.
Jon Elson <elson@pico-systems.com> wrote:
> Cydrome Leader wrote: > > >> >> Has anybody exploded any switching power supplies of any sort by just >> feeding them DC, filtered or with 120Hz ripple? They should int theory >> love that, but I'm not sure about the autoranging device they use before >> the filter caps that either rectifies or doubles the AC input. > Years ago I worked on an outboard PFC that supplied PCs > with pure DC at 340 V, worked fine as LONG as the PC > had a classic voltage doubler/switch PWM supply. > But, it wouldn't fly, as the PC's power switch and > fuse were not rated for DC. And, of course, woe be > to anyone that plugged in, say, a monitor with a > 60 Hz transformer or AC degaussing coil! > > I think many PFC front ends might work fine with DC input > within the right range, they'd just become a nice > boost converter. But, of course, it depends on the > control loop of the PFC chip. > > Jon
I am pretty curious about this, but not enough to test it on my computer. Modern computers do lack AC blowers, real power switches and LCDs lack degaussing coils, so I decided to see what I switching power supplies I can explode with DC. Then things got "interesting". The laziest place to get 160or 340VDC is from a switching power supply. I grabbed some genericy open fram 110 watt unit from Artesyn and noticed it only had one, not two filter caps on the HV side- a single 400V fed by just a 4 pin bridge rectifier and the usualy noise filtering stuff. There was no autoranging chip, that that would do anything too useful with a single filter cap. The label on the power supply clearly states 100-240VAC 50/60Hz. I plugged it in and sure enough, there was only 164V on the filter cap. I'm going to have to guess you 330-ish if you feed it 240. I've never seen that before. So at least for this power supply, it would seem feeding it any type of DC would probably be OK. The real test is to run this power supply off the filtered 164V from it's twin. Then it's off to see what happens to stuff like power adapter bricks and maybe a CFL bulb. Anybody want to vote on if a low end CFL bulb with light up (flames don't count in this case) on DC?
Pimpom <pimpom@invalid.invalid> wrote:
> Jon Elson wrote: >> Pimpom wrote: >> >> >>> To put it another way, if I repair the PSU, is it likely to go >>> poof >>> again unless I change my UPS to a more sophisticated one? I'd >>> never >>> done a detailed study of practical active PFCs before and >>> could well >>> miss something with a quick analysis. >> Hmmm, I can't imagine a PFC can work properly when fed a >> non-sinusoidal wave. But, the typical PFC is a boost >> regulator, so >> the transistors >> are ACROSS the DC bus (with an inductor in series). I can >> imagine >> that maybe the boost regulator is asking for high duty cycle = >> maximum >> boost at the zero voltage point, then the UPS switches on and >> the >> PFC creates excessive boost voltage before the PFC regulator >> can >> recover. >> >> But, that is just a guess, and I'm assuming the UPS is not >> really true >> square wave but "modified sine wave" which means there is a >> zero Volt >> perion between each output pulse. Anyway, I cannot imagine a >> PFC >> supply can properly handle a UPS with non-sine output. As >> these >> PFC supplies become more common in larger appliances like >> computers, >> the UPS makers surely are going to have to address these >> problems. >> And, it sure isn't clear how to do that very cheaply. Mostly, >> they'd >> need PWM synthesis of the sine wave, and then enough filtering >> to >> smooth out the PWM carrier. >> >> Maybe smarter PFC's could detect square-wave input and modify >> their >> PWM logic to compensate. >> >> Jon > > You're right in that my UPS is not true square wave but has a > dead time in between transitions. My son has been using a PSU > with active PFC for more than a year and his 800VA UPS ($45 with > a 9Ah internal battery) outputs the same type of waveform except > that mine has noticeably steeper edges. Maybe that's what makes > the difference. Some of my friends are also using PSUs with PFC > and, judging from what they paid for their UPSes, I'm sure none > of them is a sinusoidal type. > > The focus here is on how the PFC components bear up to a > non-sinusoidal input rather than on the actual PF correction. > Similarly, although my computer needs less than half of the PSU's > 700W rating, I used it because it returned good performance > figures in reviews and I got a used unit cheap.
I don't have the answer, but the middle of this page shows some types of PFC methods and how they work: http://www.fairchildsemi.com/applications/motor-control/solutions/power-factor-correction/ My guess is whatever is cheapest to make and just passes interference standards is what you're going to see in a PC power supply. The boost one IS just a plain switching power supply anyways, so I'm not sure why you'd string yet another switching power supply after that.
On Thu, 27 Jun 2013 17:36:30 -0500, Jon Elson wrote:

> Hmmm, I can't imagine a PFC can work properly when fed a non-sinusoidal > wave. But, the typical PFC is a boost regulator, so the transistors are > ACROSS the DC bus (with an inductor in series). I can imagine that maybe > the boost regulator is asking for high duty cycle = maximum boost at the > zero voltage point, then the UPS switches on and the PFC creates excessive > boost voltage before the PFC regulator can recover.
Most, if not all PFC circuits use feedforward from the incoming waveform. -- "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." (Richard Feynman)
I ruined a notebook power supply with a modified sine inverter. Use that 
shit at your own risk. Like I said, get a true sine double conversion 
UPS or use nothing. You are far better off losing power than feeding a 
PC with a modified sine.

They should call it a modified square wave! That is closer to the truth.