PFC Switched-Mode Power Supply

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
To Eva: this is too dogmatic approach.
I used to attend school somehow and I still remember how to calculate elementary things like energy storage :xeye:
Still, there was 0,5W resistor in snubber prototype (produced in thousands quantities) and I used the same in my variant. The resistor was barely hot during operation.
There should be another explanation; obviously energy stored in small 5uH inductor is not fully dissipated in that resistor, instead maybe transferred into main output capacitor.
Or, stored energy is of smaller order :confused:

Forgot to add, small diode there is SF26, ultrafast type too.
 
I'm sorry that I cannot find that PFC schematic propotype now, and I also don't have those pcbs near me so that to make picture.
I made a mistake in my above description of 2nd snubber (diode is the opposite connected) so put its schematic here as I recall it. I'm not sure about capacitor value but it's not hard to define since resistor used was 5W cemented wirewound type. This snubber unlike first one operates at full voltage pulses so energy waste is higher. Looks like they include it to decrease switching losses and normalize voltage rise on FET channel after it goes into off-state.
This circuit was also present in schematic by STM engineers, so prototype PFC contains 2 snubbers.
 

Attachments

  • snubber_2.gif
    snubber_2.gif
    2.2 KB · Views: 820
answers to more PFC questions

Steve

1. The PFC I posted was designed for 85-260V rms mains input. It was tested OK for continuous operation at max load at 110V input and 240V input. Measured efficiency was about 92% at 110V input and 93% at 240V input.

2. The 235uH boost inductor was designed for 260uH at I=0, and this drops to 183uH at I=Imax=13A peak. It was designed to
have two half windings of 22 turns of 1.5mm ecw each, wound in opposite directions, following the recommendations of the ST appnote I mentioned in a previous post. Calculated losses were 3.8W and temp rise 28 deg C, which was ballpark confirmed in practice insofar as the coil barely got warm, even under full load. To fit in the windings in the real world with the Law of Conservation of Misery acting, I had to take 1 turn off each winding. For continuous full power operation at 85V the inductor is a bit on the edge and in hindsight I would have traded some of the low copper losses for more turns of thinner wire. Another thing you can do, if you have height available a, is to use two of the Magnetics 55439 cores glued on top of each other in a
stack. This trick increases the AL significantly, while the
permeability with Idc characteristic remains the same. Note: cost is 2x and these cores are expensive

3. As someone has observed, the bulk of the losses will be in the hard switching FET. At 110V, eff = 92%, 70W are lost for 760W out. I can account for 3W in the sense resistor, 4W in the inductor, approx 15W due to Rds_on losses in the FET and a guesstimate of 5-8W in the boost diode. This leaves around 40W for switching losses in the FET.

Alme and Eva are discussing a snubber circuit in this thread and this would be a useful addition to reduce switching losses if it can be worked out. The configuration in question is for a magnetic snubber that looks to be the same one I saw in an ST appnote. It came with a number of formulae and my notes show I calculated some values as follows, based on the appnote:
L = 4uH, based on Vout/(di/dt = 100A/us)
C = 200nF based on (Energy(C) = Energy(L) = 250uJ) and Vc = 50V
If this energy has all to be disipated in the resistor then the resistor would need to be 36W!
R was not calculated and I appear to have run aground here and abandoned the snubber. The high dissipation I calc in the R is consistent with Eva's explanation, but I'm not at all confident enough on snubbers to be pedantic about it. (I'm usually just grateful if my stuff doesn't explode!)

But this dissipation would render the snubber useless, which surely can't be the case. Perhaps, as Alme has suggested, there's something we're not seeing here, or our initial assumptions about how this thing works may be wrong. . .

Come guys, thinking caps on . . .


Cheers


John H
 
Hi there....Those wanting to use melf surface mount resistors be mighty careful in pfc circuits. In telecom many boards fail due to board flexing after the soldering process ...then mounted......the serious and explosive condition that can happen if the OVP and voltage feedback dividers beome o/c. With no reference the duty cycle goes full in one cycle to reach breakdown voltages.
It happened to me on a 1 kW 500V pfc and with so many joules stored up in the o/p caps the switching components exploded.

Anyoner using a magnetic snubber would be hard to convince me of the worthwhile conditions at high line Vin. The odd % + in efficiency doesn't warrant the extra cost of a fast diode /components and extra space. What everyone should be looking at is reducing the wretched bridge diode loss at low V in with high VA.....this in my circuits is by far the greater losses than either IGBT or fast diode.

I'm using IGBT's for all my pfc designs even at 100Khz op......why not at 500V ? I'm seeing so many circuits basically using the wrong type of fast output diode and switching semi in average current designs........Don't all shout at once........

richj
 
Bridgeless PFC

Richj

Quite right. There's about 20W lost there in my design, which I forgot about completely in the calc in my last post! This would make my FET switching losses then about 24W.

I've seen some bridgeless designs in an appnote (think it was ST) where they use 2 FETs instead of a bridge. But I've never scrutinised these designs to see how the losses from the second FET switch compare with the loss of the bridge. It may be a false saving, it may be worthwhile.

I once designed a 240W PFC/converter PSU for a company, and it had rectangular 200V 1206 resistors in the divider chains for both feedback and OVP. As luck would have it, these both failed at the same time while the CE examiner was testing the unit in his lab. The +400V bus shot up to 720V and 2 x 100uF electrolytics exploded! The examiner was VERY SHAKEN, but had to concede that this was a low probability case of a double fault.

Regards

John H
 
Hi there John.....I think anyone new with the idea of toiling with real world PFC would soon realise that two of us who've had the same experience get the same results when things start to go wrong would soon be put in the picture .......those fresh to this technology game be well aware of the dangers. By all means start with other toplogies.....switchmode is brilliant stuff.......RF background is important.

I've also studied the notion of improving the input rectifier end. However I came up with a compliance problem which conflicted the mos rectifier idea. I made a unit up which was far more bulkier than the modern slim KBUxx series wafer bridges, but the bog standard AC rectifier bridge is actually quite lossy at RF frequencies. I found to get the same results with a mos bridge, alot more Y caps had to be strung in the circuit. More anon.
So apart from transformer input and unless something really reliable comes along I'm having to stay with ye faithful KBU series.

A worthwhile improvement area is using white alumina insulators (pair of them greased ) on TO220/247 tabs to reduce device charge capacitance to heatsink. The thermal resistance is excellent and I also insert a thin copper strip which is then grounded to the switching circuit ret/ground. Anyone can work out the actual charge current v.s frequency & capacitance to realise how high the impulse current really is.....
Trends wise....I'm steering away from snubbing by careful tight layout and generous magnetics.......There's lots of nS diodes about...I will work 500V on a 600V rated version but no more. The high speed Philips BYC10-600 ranks high on my menu.

The cherry on the icing is post smps with ZVT to keep efficiency up....The UCC3895 ic .....have you used this devil ?

richj
 
UC3895

richj

I do sometimes use the alumina washers (3.0mm) on PowerDAC output stages to reduce the amount of switching pollution on the GND line.

On the SMPS I posted I did the usual trick - as you mentioned - of an insulator with a sandwiched screen to COM.

Of the various Unitrode chips, the UC3854 has no independent OVP, so I don't use it. I've used the UCC3817 in a production product but I wouldn't use it again because it is terribly sensitive to static, spikes on the power rails, and EMI. I had major difficulty getting a design based on this chip through EMC susceptibility. And in production, high percentage (5-10%)of the parts arriving from the supplier are DOA. I've spoken to the TI bods about this and they sort of 'half acknowledge' it's not a terribly robust chip. I suspect it's the BiCMOS process they use, for I find similar problems with other chips of the same process, like UCC3804.

I toyed with the idea of using the UC3855 (??) device, which has a soft switching mechanism, but the small efficiency improvement they quoted (relative to UC3854-based circuit) didn't justify all the extra parts, some of which were a bit esoteric (saturable inductors etc). The overall impression I got from reading the app notes on this chip was that the writers didn't feel very impressed about their own chip. Just a hunch.

The UCC3895 - I was planning to use them in my next ZVS dc:dc converter instead of the power hungry ML4818, but I'd be concerned if they were BiCMOS types.

Cheers


John H
 
Hi there......the UCC3895 is one of those sorts of Bicmos ic's......requires a sep driver ic as o/ps are lowish current. I've used this chip in many converters.....but there's a serious bug in using it in phase shifted current mode.
The condition in very light load or no load at min duty cycle is a taxing condition...(open circuit operation is one of those standard market conditions) ......the current reference ramp amplitude becomes so low that the peak current mode ramp internals waiver.....the result that the sep AB/CDoutputs can phase conflict.......result is instability leading to switch destruction. The trick is to tickle the current ramp with a bit of Vref....at the moment I've mislaid the Unitrode information update which highlighted this problem.

Tempus Fugit...Some 10 yrs ago I put this on the shelf at the time....before more suitable lower diode loss mosfets specially sited for ZVS operation came on the market. Perhaps I should rekindle it.

The mosfet IRFB17N50L was a typ candidate.....but way too deare at turn of 2001.
Unfortunately not all the knowledge bowels of Unitrode have moved to Texas Inst. Laslo was I believe the principal architect behind UCC3895.
At the time I was doing R&D in a leading telecoms giant and designed a reliable telcoms ZVS using current doubler o/p rectifier topology from 3 phase juice.

Your efficency seems to fall in line. 93% was about as high in the low line condition.

More anon to the pitfalls of this tecknoledgy

richj
 
davy: output signal duty cycle of L4981A generally depends on output voltage, output current, and also input momentary and average voltage, if to activate all inherent PFC functions. Check its datasheet and you'll see that it contains analog multiplier for those signals. So duty cycle for every given moment of PFC operation is a complex function of the above arguments.
If your purpose is to use L4981A in PFC followed by next stage power converter, then you may need to synchronize them together. Use pin 16 of L4981A for either input synchro (slave operation) or output synchro signal (master operation). Much information on those things is available on STM website, especially, their application notes are useful.
 
Do you really need power-factor correction for your charger?

If not, then for your 180W supply I suggest you have a look at the two-transistor forward converter design help at the following site.

http://schmidt-walter.fbe.fh-darmstadt.de/smps_e/smps_e.html

This will help you design the transformer and the output inductor.

If you really do want PFC, then you can use the L4981 to give you a 400V dc supply, and then use this +400V to feed the two-transistor forward converter. The isolation is done in the forward converter transformer. You can use a control chip such as the UCC3801 for the forward converter, and something like a IRF2110 (or a suitable transformer) to drive the gates of the power FETs. You will also need to learn about using a TL431 and an opto-coupler for the output-sensing feedback, as this has to cross the isolation barrier.

If you want to charge lead-acid batteries you may need to add temperature-compensation to the float chage voltage, to ensure long battery life.
 
well ,Ouroboros,
thanks for your explanation. it's very useful for me.

our tutor want to design a robust, and he ask me to design a battery charger for the robust.

what i want to design is a half bridge circuit, with 220vac 50hz input, and 36vdc 5A output for battery storage charger,

and i want to use a IC (such as L4981) to drive the two MOSFETs, i just post the circuit in a word document, 220vac---transformer---half bridge---battery storage , could it be done?

may be SG3525 IS suitable cause it has two output pins, but I havn't made itvery clear.
i want to pose the circuit here , but it's too big.
 
Buy a electric bicycle charger for 36V lead-acid battery.
davy_shi said:
well ,Ouroboros,
thanks for your explanation. it's very useful for me.

our tutor want to design a robust, and he ask me to design a battery charger for the robust.

what i want to design is a half bridge circuit, with 220vac 50hz input, and 36vdc 5A output for battery storage charger,

and i want to use a IC (such as L4981) to drive the two MOSFETs, i just post the circuit in a word document, 220vac---transformer---half bridge---battery storage , could it be done?

may be SG3525 IS suitable cause it has two output pins, but I havn't made itvery clear.
i want to pose the circuit here , but it's too big.
 
ML4826 PWM-PFC Combo chip

DavidF-

I am so braindead! :xeye: I just found the datasheet for the Fairchild Semi ML4826 in my stack o' stuff last week. I forgot I had downloaded it a few months back. I am planning on getting a few samples from Fairchild and giving them a test, but with new child and family, it will be a challenge on getting some time to do this.

Steve
 
Anode/Cathode on an IXYS diode?

I´m sorry, but I can not find this info anywhere;

On an IXYS 12A 1000V Ultrafast, Soft Recovery (FRED)
Epitaxial Diode (trr = 50 ns; TO-220AC package)...

... I can´t seem to find out wich is the "anod" and wich is the "cathod"?

Please advise.
Regards,
Mike
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.