Combined PFC AC/DC & DC/DC SMPS

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi Eva
In order to understand your snubber correctly:
Is the attached schematic showing your snubber in principle?

@Steve:
...looking forward to your new project...
I hope that I will be able to send some pics and measurements from
mine during summer. Sorry, can't do this earlier, because I will be
sometimes on business trips and have lots of other topics in the next two months....
My digicam is broken anyway :mad: . It didn' survive my bike tour in the sahara.... :whazzat: :whazzat:
 

Attachments

  • eva_snubber.jpg
    eva_snubber.jpg
    40.6 KB · Views: 1,301
Hi,

originally posted by Eva

buffered gate drive in order to avoid blowing the control IC everytime I blow an IGBT

maybe you will find attached scematic useful. It is a simple desaturation protection circuit which works fine in my PFC. I have not lost a single Mosfet in the last 5 years.

Best regards,

Jaka Racman
 

Attachments

  • clipboard02.gif
    clipboard02.gif
    10.8 KB · Views: 1,349
...hm, this could also be a magnetic snubber, Eva.
...but in contradiction to your explanation this one would not store energy when the switch is turned on...
Your comments made me curious!
Here the parasitic winding capacitance of the 390uH choke will
help to reduce the inductive overshoot. Or do you have further
snubbering for your IGBT?
I guess the snubber which estimated first will not only add the reflected output voltage (450V : 3 = additional 150V ) to the IGBT, but also some inductive peak due to the leakage inductance of the magnetic snubber....

...bye & good night 2 all
Markus
 

Attachments

  • eva_snub2.jpg
    eva_snub2.jpg
    28.6 KB · Views: 1,155
Chocoholic :

Looking at the layout of the PCB my snubber topology is easily figured out. The primary is connected between the diodes and the storage capacitors

So I made a big mistake when explaining the energy storage behavior of my magnetic snubber [I explained the first version I tried, that looked like your first schematic]

For the current version :

- During off period the inductor is fully energized and at turn-on it starts de-energizing as the switch current rises and diode current falls. The current across the inductor drops to zero and starts rising in the opposite direction during diode reverse recovery time. Finally the diode stops conducting and the reverse-current energy stored in the inductor is dumped back to the storage capacitors [three times slower than it was stored]

- At turn-off the coupled inductor works as an autotransformer, it keeps the Vce of the IGBT below 600V while the primary inductor is energized and until the required current is reached [this also happens three times slower, at 15A/us]. Both the parasitistic capacitance of the boost inductor and an additional RC network placed in paralell with the inductor help to reduce the likelihood of avalanche. The leakage inductance may still cause the IGBT to enter avalanche mode but during a very small period of time <20ns
 
This schematic shows the switching section of my PFC :

An externally hosted image should be here but it was not working when we last tested it.


Note that I made another mistake in the previous explanations since I'm actually using SKP10N60 IGBTs [rated 20A at 25ºC]

Edit :

I've discovered one more mistake : Initially the coupled inductor had 10uH primary inductance but I forgot that I then tried 5uH without significant performance degradation and I decided to adopt that new value [so diode current downslope is actually 90A/us]

Excuse me for the mistakes, this project has been inactive for the past eight months and I'm quite lazy taking notes about what I do
 
Hi Eva!

That's great. Thanks a lot for your schematic !
Working principle matches basically to my second snubber picture.
But your arrangement is more fortunate regarding isolation requirements in the magnetic. I really like your snubber!

In the mean time, I considered about the advantages/disadvantages of transition mode vs. well snubbered continous mode.
Just to get some numbers I compared a boost with 200Vdc input and 400Vdc output. Operating frequency: 50kHz.

In transition mode:
The values above require for a 200µH choke. Current is ramping up and down between zero and 10A.
I_eff= 5.78A

In continous mode:
In order to bring down the peak currents, I had chosen a 800µH choke.
The current will be DC with ripple, ramping between 3.75A and 6.25A. I_eff=5.05A

At first glance for the choke both requirements do look not so bad.
In transition mode you would need much less inductance, but higher
saturation.
In fact you need 25% more turns for that 800µH design for
continous mode, but have less I_eff and only one quarter of
the inconvinient HF components. As the HF effects in the winding cause horrible losses, I guess that a choke for the continuos mode design can be slightly smaller than for the transition mode.
In the end this advantage may be canibalized by the magnetic snubber.
But for the switches it is definitely a big advantage to deal with 6.25A instead with 10A....

Further I found during designing my 150µH/24A choke, that the
winding might fit nicely to my ETD44.... but the air gap.... :hot: :hot:
A gapped ferrite design does definitely not match to that.
I would have to give 10mm gap each every leg !!!!
...did not look like a choke, more like a high power pirate antenna. ...not my favorite....
Also in this regard the continuos mode design is better. I could have
seen this earlier by comparing the aount of stored energy.
The stored energy in the continuos mode design would be less than
half compared to the transition mode design. And in the same portion the gap would become smaller.

But I am stubburn.
:D
As long as you have cool semiconductors (..and just by good luck: I have them.) and find a fortunate choke design. Then the simpler transition mode design is attractive also at 1kW.
OK, I have an idea for the choke.
Combining several torroids (MPP or cool_mu) and make a double
aperture core. The attached pic shows such design.
Of course this is not the final PFC choke. I will have to take care for the isolation and get propper cores....
I am thinking about 6 pieces of MPP cores from magnetics, or similar.
http://www.mag-inc.com/powder/powder_cores.asp
6 pieces of core number 55930 will end up in a small
PFC core of about 54mm x 27mm x 34mm and nearly all the
winding would be place inside that core.....
And these MMP cores have a much lower permeability than
ferrite and allow an ungapped design for energy storage.
No gaps, less leakage fields.

BTW: From geomtrie such double aperture designs give quite a good coupling. I think putting together several ungapped E- ferrites ( should also be possible with U- ferrites)
to such an double aperture design, could be interesting for the push pull transformer!

Bye
Markus
 

Attachments

  • pfc_chok.jpg
    pfc_chok.jpg
    53.4 KB · Views: 845
Remember that fast switching is required in order to reduce dissipation and that the boost inductor has to withstand high dV/dt transients both at turn-on and at turn-off

Iron powder cores tend to suffer from too high loses in these circumstances, you will experience this by yourself when you try the first prototype

Also winding capacitance is very critical since the boost inductor works as an antenna and radiates EMI like hell [it resonates quite easily]. The best approach I've found is to use a gapped ferrite core with a single-layer winding [several layers of identical windings connected in paralell to reduce Rdc also appears to work]. This requires a big E core

I don't like ETD cores, they are designed to reduce the amount of ferrite used and this is undesirable for gapped inductors


About the second stage :

It's not practical to wind mains and 12V primaries over the same core, winding requirements are absolutely different and winding area is limited. For 12V the best approach is to use several 1:1 smaller transformers with secondaries connected in paralell, so primary turn count can be kept above 4 turns [getting good coupling with less turns is very hard]. For mains voltages a single bigger transformer is the best solution


About the transformer coupled PFC :

QSC uses it and it works fine but the output is not tightly AC regulated, it has only DC regulation and it will show ripple at 100Hz and its harmonics [essentially the waveform you obtain when you feed a capacitor with a rectified sine current wave]. I've not tried this approach but I think it will require also a coupled-inductor snubber to prevent excessive voltage build up across the switches during the energizing phase of the leakage inductance of the transformer [the snubber should transfer this energy to the secondary side quicker]

Also, note that the transformer will be working at a wildly variable duty cycle so copper and diode losses are higher and the output has to be pi-filtered

Anyway, this kind of PFC makes a great substitute for the classic bulky toroid transformer [keeping the usually huge storage capacitors] and it features DC regulation and lower 50Hz ripple for the same storage capacitance as a bonus
 
Hi Eva!

Up to now, I also did not use iron powder cores. Especially
due to their high losses. But the MPP are not that bad. Especially facing the fact that I do not need 1kW continuously, they seem
to be the perfect match for a small size design....
I will keep you updated about my design flaws.
:p
Please be patient. I am not able to set up the things that fast
as you usually do. I would expect PFC results in June....


The approach with several small 1:1 transformers is a good idea.
Just from simple calculation and simulation.... two days back, I saw that coupling is one of the most important things for the push pull transformer.
When you simulate a unregulated push pull you can find a dramatic
drop of the outputput voltage just be decreasing the coupling from 1 to 0.995 !

I simulated a half bridge push pull coming from 480V DC +/30V-100Hz-modulation.
A single output voltage, full wave rectified, 1000µF output cap:
Output load 10 Ohm, desired output voltage around 100V.
Transformer primary 16 turns ==> 4.8 mH
Secondary: 4 turns ==> 1.2mH
Derived output voltage, min-max due to modulation of input : 108V - 123V
Not so bad. :cheerful:
Now I changed the coupling to 0.995:
Output dropped to 52V - 53V !

The leakage inductance already increased the transformer's impedance to an unacceptable value.
OK. The coupling alone is a little bit a misleading number to figure out the suitability of a push pul transformer. The absolute values of
leakage inductances seem to be more helpful. And from this it becomes obvious that high perm materials do not really help.
They bring up the coupling, yes. But the absolute values of the leakage inductances do not drop by this.....

Hm,.... now you say several 1:1 transformers for 12V input and an additional one for the supply coming from the PFC.....
Sounds reasonable. But I don't like it. To many space consuming
components!..... I would like to manage this in a single
small transformer.
Let's see how long reality will need to cure my dreams.
:clown: :clown:

I don't like ETD to much either. But their geometry is fine if you have to wind margin tapes for large creepage distances to meet safety standards....

In fact for my push pull converter I am thinking about an E shape the offers giant Ae. ....the double aperture proposal from above....

Looking forward to your support, when fighting with the real
set up in summer!

Bye Markus
 
Let me join this interesting discussion and share my modest PFC and converter design experience here :)
Chocoholic, I have tried fixed-frequency continuous mode PFC based on L4981A a few times and this is what I recommend for higher power application, especially that you will desire to synchronize it with next stage converter.
Here is additional info on it by STM.
http://www.st.com/stonline/books/toc/an/1212.htm
I made my 500W PFC just like they suggest in their application note, with smaller additional inductor (this seems to be similar way as Eva suggests here).
What do you think about this chip?
 
L4891a

Alme - Eva,

WOW! This chip looks pretty cool! :cool: I will try and get some samples to play with. I very much like the idea of synchronization.

While on ST's website, I noticed an application note about an L4891A-based "Bridgeless" PFC rated at 800W. The two lower bridge diodes are replaced with two N-Channel MOSFETs and driven synchronously by the PFC chip. The boost inductor is placed ahead of the MOSFET-diode "bridge", and according to ST, is very useful for Hi-power PFCs, where power disspiation in the diode bridge becomes a concern. Here is the link to that PDF file:

http://www.st.com/stonline/books/pdf/docs/9119.pdf


Just goes to show ya': spmeone is always thinking outside the box!

Steve
 
Eva said:

The input diode bridge is placed in an external heatsink and the EMI filter is not shown. I've successfully tested it up to 1.7KW continuous with proper cooling at 450V output and 160V AC input, however the board has space for additional components in order to get up to 4KW theoretical maximum output. Currently I'm using SKP20N60 IGBTs and MUR860 diodes combined with a coupled inductor that works as a magnetic snubber and contols the dI/dt at turn-on [and also at turn-off as a side effect]
UP we go.
two IGBT's and diodes parallei or how?

Do you have you any fiqures for effiency at 1.7kW?

I have been wondering similar project, ie 4kw PFC stage for a while now. Availlability of MPP cores just bugs me off :mad:
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.