diyAudio (
-   Power Supplies (
-   -   1kW PFC-Boost (

ChocoHolic 19th September 2005 11:46 AM

1kW PFC-Boost
2 Attachment(s)
Finally my boost converter is reaching an acceptable performance.
It gave some more headache to me than expected.
- ...tuning the regulation loop for stable operation at low power, while
providing proper THD....
- Reverse capacitance of the MosFets caused the PFC-chip to oscillate,
due to incredible turn OFF du/dt of Vds....

At 230V input it can continuosly deliver about 800W.
All temperatures a remaining below 70°C during this load.
Max. power is around 1kW.

But I am not so lucky with my snubber losses.
The CDR snubber (C:2n2 , D: BYT08P1000, R:500 Ohms) looks
nice in schematic and is working fine in reality, but the losses.
...especially at low power loads... 200kHz this snubber will make losses of
P=200kHz x 0.5 x 2.2nF x (470V)^2 = 49W.
:bawling: :bawling: :bawling: :bawling: :bawling:

I already tried some variations.
- smaller snubber cap:
Well, high du/dt causes two drawbacks. EMI issues and additional ir
directly can disturbe the PFC chip itself. In the beginning I was working with du/dt up to 30kV/us.... Don't ask ! My radio made more noise than music, laptop ignored its touch pad....
This happens only at high loads, because then the current in the MosFets goes up to 12A and these 12A are forcing the sloping speed of Vds during MosFet turn OFF.
The reverse capacity of the MosFets is demanding quite a low driving impedance to and I had to implement the five BC560 to drive and decouple these beasts from the PFC chip.
The snubber with 2.2nF now limits the sloping speed to about 5kV/us. Which is fine for proper operation and also for my radio and all other stuff I am using here...

I also tried to reduce the switching speed and save the snubber, because the MosFets could easily handle some turn OFF losses.
a) a higher driving impedance, but I got some turn OFF oscillation around 10MHz.
b) a well damped RC-feedback from drain to gate, also this
resulted in inconvinient turn OFF oscillations at high power load.

OK, I hoped that I could at least use some of the snubber energy for the chip supply... , well the snubber pulses caused some irritations to the chip and the detect signal was not detected proper anymore....

Hm, now I am thinking about some resonant "lossless" snubber, but
my brain did not come up with a proper idea up to now....
(...typically running into demag issues already in theory.)
Good ideas are welcome. Sorry for the poor quality of attached schematic.

ChocoHolic 19th September 2005 11:52 AM

2 Attachment(s)
...took this pic during building the power stage....

mzzj 19th September 2005 07:53 PM

30kV/us, sounds scary. I am sure that all Hams around you have been mad as hell during your tests :D

ChocoHolic 19th September 2005 08:25 PM


Originally posted by mzzj
30kV/us, sounds scary. I am sure that all Hams around you have been mad as hell during your tests :D

...hams: I am afraid you are right.
Also some kids with RC cars might have had some headache.
:angel: :angel: :angel: :angel:

Eva 19th September 2005 08:50 PM

30Kv/us slopes?

For a 400V output that would imply a 13ns rise time. I dont remember what MOSFETs are you using, but I think that it's quite hard to go below 50ns due to output and reverse transfer capacitances.

Also, such a rise time would generate a huge voltage spike due to leakage inductances and due to the fact that diodes take several dozens of nanoseconds to start showing a low impedance.

Furthermore, that voltage-rise spike would be coupled to the other side of the boost inductor due to winding capacitance and will be very hard to filter. At these frequencies everything is just an antenna. In comparison, a piece of 10Mhz ringing would be much easier to handle.

I don't like that "several-hundred-Khz-MOSFET" design philosophy. It requires smaller magnetics just at the expense of bigger heatsinks, more expensive switching devices and snubbers and poor efficiency. The more I experiment with >500W stuff, the more happy I am with switching frequencies in the 30Khz to 50Khz range, for I can even use $1 switches and single sided boards without EMI issues :D

Jaka Racman 19th September 2005 08:59 PM


I am just designing a new bridgeless PFC, but I have 3.5kW design in production for some time. Here are some tips:

-use center tap boost diode and symmetrical layout. Put one fet on each side of the center tap diode. Use PCB with GND plane on top and 400V plane on the bottom to connect fets, boost diode and output caps. Make layout as tight as possible.

-use only electrolitycs for output caps, I use 4 x 470u/450V in parallel, in earlier design I even used 11. Snap-in electrolytics have only 20nH inductance, so do not worry about spikes.

-I use separate inductor for each mosfet. Wind inductors single layer to minimize winding capacitance. I use Kool-Mu (Magnetics Inc.) toroids. I also mount inductors on separate PCB which is perpendicular to the main PCB, to make layout as tight as possible.

-use separate driver for each fet. If using emitter follower pair, make sure pnp is mounted as close to the fet as possible and connect it directly to the source pin.

-I use no snubbers, but rather rely on tight layout, which is much easier without snubbers. I have no overshots or ringing on the waveform, turn on gate resistor is 5R6, turn off gate resistor is 0.5R.

-put control circuitry on small daughter PCB which you can mount directly on fet gate and source connections. I have no experience with Infineon controller, but if they are something like TDA4700 from Siemens, I would rather change chip for UC3854 or L4981. TDA4700 tended to reset itself with pulses introduced from Weller soldering station.

Best regards,

Jaka Racman

powerbecker 20th September 2005 06:37 AM

:cannotbe: Do you use Infineon Cool-Mos...?

ChocoHolic 20th September 2005 07:09 AM

Hi Eva: Yes, that fast slopes were crazy. I did not expect such a speed. But look, an inductor that forces ongoing 12A ...
...just some pF of parasitic capacitance cannot slow the resulting
sloping speed...
Now with the snubber it takes approx. 100ns to shoot from ground to 600V . At home with my cheap USB scope at home this are just some single dots.... and some further precision drawbacks of that 200EUR toy.
But I don't want to borrow all the things from work.
I have some overshoot due to additional noise suppressor
ferrite beads in series to the PFC diode. These ferrite beads
are completely counter-intuitive at this position, but have prooved
about 10 db less EMI issues in the RF range. The do not act
like a simple inductor. This would just cause ringing. Their losses
in the range above 1MHz are smoothening RF oscillation quite eefective, but do not generate noticable heat. Nevertheless the
overshoot is an additional stress for the MosFet. That's why I am
using the 800V CoolMos 17N80C3. The gate charge of this beasts
is just about 100nC, but if drive them slow with a impedance
they tend to ring in this 100% inductive loaded application. So
I drive them with some single Ohms for turn OFF. And this really
seems to turn OFF the drain-source path within some 10ns...

At home with my cheap USB scope even the snubbered 100ns slopings are just some single dots.... and some further precision drawbacks of that 200EUR toy.
But I don't want to borrow all the things from work and if you know some principal behaviour, then even a toy with a sampling rate of 20MHz and scrabby internal amplifiers (only 5MHz) and
not perfectly adjustable probes... well, still such cheap home equipment is sufficient to get most things running.

Please note, I am not doing a several hundret kHz design.
At full power the frequency will be just above the audiable range.
200kHz is only happening at low load. This large frequency range is the tribute I have to pay for the critical conduction mode.
But I am also feeling that a fixed frequency design and a magnetic snubber like yours might be smarter for high power applications. My approach simply to enforce a 100W-topology for a
1kW application is working, but clearly showing its limitations.

Hi Jaka:
My proto is a folded foil design. Ground plane also for the control circuit. If my time schedule works out today, I can post some pics
of my ultra ugly "3D-P2P-origami-proto".
My PNP driver is situated directly in the 20mm between the PFC chip and MosFet gates. Also the gate drive loop area (PNP, emitter resistors, gates, source resistors, ground connection source resitors and collectors of the PNP) is small.
I also did not observe heavy spikes on the DC output rail. This seems to be fine.
Changing the chip, but keeping the basic design would lead me to the L6562 of STM. Do you think that would be more rugged vs
external disturbances than the infineon type?


Eva 20th September 2005 08:58 AM

Sorry, I had forgotten that you were using critical conduction mode. This is nice as to what concerns to switching losses since frequency goes down as current increases thus keeping losses under control.

Have you tried to place a 2.2nF (or maybe 1.5nF) capacitor directly between D and S pins of the switching device? The basic idea under that is to have an intentionally slow gate turn-on so that the capacitor charge peak current is kept low, and a relatively fast gate turn-off (but not fast enough to require a dozen of PNP buffers!). Such a capacitor in conjunction with MOSFET turn-off time and internal drain-source capacitance will limit rise time to approx 100ns even with 12A load, but in a less dissipative way, and shall ringing be present it would happen at fairly low frequencies (so little will be radiated). Note that your particular inductor design nicely allows for slow turn-on since the inductance is high for low currents, so inductor current rise in the first 500ns should be quite small, thus producing low losses.

The boost inductor will tend to resonate with that capacitor but I don't consider that as a problem.

Also, I would try these bead ferrites placed in the inductor side instead of the diode side. You shouldn't slow down diode commutation.

ChocoHolic 20th September 2005 01:50 PM

2 Attachment(s)
This picture shows the heart of the power stage. The power devices
are mount on a aluminium base plate. The base plate is isolated but
HF-connected to GND by 1nF (please refer to the schematic, where
“shielding” is mentioned.) Unfortunately the capton isolation looks quite similar
as copper and makes it harder to understand the construction from the pictures.
MosFet drains directly connected with 0.3mm copper foil
to the PFC diode. The cathode is again directly connected to foil
of the positive rail. The foil is splitted and carries the ferrite beads.
Below the foil of positive rail you can see the film cap.

All times are GMT. The time now is 05:58 AM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 18.75%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright ©1999-2017 diyAudio