Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Phase shifted full bridge, Ringing, EMI

Status
Not open for further replies.

skn96

Junior Member level 3
Joined
May 28, 2009
Messages
26
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,281
Activity points
1,667
Hello All
I’m working on a phase shifted full bridge dc/dc power supply prototype right now. The converter has been designed for 800W continuous power and must regulate for max 2 seconds of 2400W over load. The input is 400V coming from external power supply for now (later from PFC) and the output is 170V in complementary fashion i.e. +85V/-85V. Schematic has been attached. I anticipated ringing due to the use of 400V secondary diodes and I have added enough snubbers as you can see in attached schematic to solve it. I do not have a major problem with that . I have used very fast 650V Infineon mosfets from CFD2 series to make sure I have no problem with mosfet diode reverse recovery. My problem right now is the ringing on the input side of the converter due to the discontinuous nature of input current in full bridge convetrer. The current sensor CS1 is on the secondary side close to controller so I had to use a 16 Gage wire 20 cm length as loop so that current can be sensed by current sensor(CS1) which is located on the secondary. Wire loop starts form TP211 goes to secondary and come back and ends at TP201. I have twisted this 20cm wire but still it has some parasitic inductance contribution on the input side which is added to the parasitic already there from PCB traces. At power level above 200 W I do not have that much problem with ringing which is caused by input side parasitic inductance and mosfet capacitors. Picture 1 shows the input current waveform which is conducted through the wire loop. The voltage of drain to source for switches Q201 and Q200 has been shown also with no major ringing on top o them. But at very low load since the mosfet capacitors are not discharged before switching and there is no ZVS , enough energy for the ringing around 10Mhz is available as you can see in second picture. Obviously the effect of this ringing on the drain to source voltage of Q201 and Q200 is bad and very close to Mosfet voltage raining which is 650V. Now the question is how to eliminate this 10Mhz ringing on the input? I’ve heard people use saturable reactors and ferrite rings. I have no experience with suppressing these kind of ringing. Can you please help me with that?
 

Attachments

  • PSFBschematic.pdf
    814.4 KB · Views: 259
  • Higherload.png
    Higherload.png
    198.5 KB · Views: 544
  • lighterload.png
    lighterload.png
    161.6 KB · Views: 344

The high ringing level in a bridge converter suggests insufficient DC bus bypassing or bad circuit layout, probably both.
 

I have 3 big 390 uF cap connected on the input which has not been shown in schematic along with several 600V, 0.1uF ceramic caps. Layout is good but even if the layout is improved more there would be still some nH inductance due to current sense primary wire loop which I can not eliminate. That is why I'm looking for a solution to attack ringing caused by this.
 

I'm also designing a phase shifted full bridge at the moment and also chose CFD2 mosfets for the primary (probably, or C7).


Are you sure the secondary ringing is under control (remember the turns ratio)? How about primary gate drives, they can ring and cause those problems? Can you show the primary waveforms with the excessive fet voltage? In theory the intrinsic diodes clamping to the well bypassed input bus shouldn't allow that type of overshoot.

EDIT Nevermind you provided the primary waveforms. To FVMs point, how about the voltage of the bus as measured directly across Q200.D/Q202.S

Does it ring just as badly on the opposite transitions?
 
Last edited:
  • Like
Reactions: skn96

    skn96

    Points: 2
    Helpful Answer Positive Rating
Hi skn96,
Try an RC snubber across CS1 at node C204 and node <Q200+Q201 drains> in such a way that it is also across the 16 Gauge wire. Snubber capacitor value should be 2 to 10 times (2*Cds) and the snubber resistor value should be √(Leakage inductance/(2*Cds)).

I hope this helps.

--Akanimo.
 

Dear asdf44
Thank for your response. The Dc bus voltage measured directly across Q200.D/Q202.S as you can see in the picture 1 below is as bad as the drain to source voltages(What does it prove?). I just added an EMI toroid core around the 16 gage wire as you can see in picture 2. I was able to bring down the ringing drastically as you can see in picture 3 and 4. But I'm just not sure if this the best thing to do or If I'm doing something wrong? I should I have the current sensor on primary so the I do not have to use 20 cm , 16 hage wire loop and vias TP211 and TP201? Do you any idea how much inductance the 20 cm 16 gage wire and two vias can add? if it is in the order of nH then it is comparable to other PCb traces and Mosfet leads and not the main contributor but if it is in the order of tens of nH then it will be the dominant parasitic here.

- - - Updated - - -

.1.png2.JPG3.png4.png

- - - Updated - - -

Thanks for your response Akanimo
The reason I was reluctant to add scrubber across the 16 gage wire was I though it will steal some of the current that is sensed by CS1 and may affect precision of current control at the fundamental frequency which is 100kHz. . If I want the snubber to damp 10mHz ringing it will have some minor impact on the 100kHz current also. But I don't know if it is ok or not. Do you any experience with that?. It is my firs kW watt experience with full bridge topology and wasn't sure adding snubber in parallel with current sensing path how will impact the current control performance. By the way in the formula you presented by leakage inductance do you mean the resonance inductance which is in the order of 20uH or the parasitic inductance of input traces and 20cm wire which is in the order of nano hanries?
 

Well the inductance of the wire is easy to look up but regardless of what the value is it seems to be the entire problem.

Traces may have inductance too but what matters is where the inductance is. That wire inductance is right at your switches which is the worst case. That's where your bulk C should ideally be with minimized L between the bulk C and the mosfets.

I'm specifying a small CT which goes right on the primary and is very easy to lay out:
https://uk.rs-online.com/web/p/current-transformers/7157030/

If you really needed, you could then take the secondary of this CT and run it through a wire to get it across the board to your secondary. That would be much preferable to routing your actual primary current that way.

It's good the choke helped but you may want to try relocating your CT (let it float in the air if need be for testing purposes) and minimize that 16awg wire length to see how much it helps.
 

Thanks for your response Akanimo
The reason I was reluctant to add scrubber across the 16 gage wire was I though it will steal some of the current that is sensed by CS1 and may affect precision of current control at the fundamental frequency which is 100kHz. . If I want the snubber to damp 10mHz ringing it will have some minor impact on the 100kHz current also. But I don't know if it is ok or not. Do you any experience with that?. It is my firs kW watt experience with full bridge topology and wasn't sure adding snubber in parallel with current sensing path how will impact the current control performance. By the way in the formula you presented by leakage inductance do you mean the resonance inductance which is in the order of 20uH or the parasitic inductance of input traces and 20cm wire which is in the order of nano hanries?

- - - Updated - - -

Dear asdf44
The CT that you introduced is very similar to what I'm using for my pfc. I used them for my PFc and I had no problem. The reason I didn't use this kind of CT for phase shifted full bridge was that it doesn't provide 6mm distance between primary and secondary needed for reinforced insulation. That s why I'm using pulse PE51678NL for phase shifted full bridge. When I was planing on layout I had to chose between two methods: one similar to what I already did, The other one similar to what you proposed i.e. having the CT on primary very close to mosfet and local cap and then sensing the signal through the barrier. I didn't pick this second method because I was afraid the current signal may get noisy routing under the fets and transformer. Though I have 4 layer board and could rout on fourth layer but still didn't know how it will play out. But looks like I may have picked the non optimized method. I actually cut the 20cm 16 gage wire and made it only 3cm to see how it affects the ringing. It only improved slightly. That mans the 16 gage wire probably is not the dominant factor and the contribution of the two vias probably is more.
 

Thanks for your response Akanimo
The reason I was reluctant to add scrubber across the 16 gage wire was I though it will steal some of the current that is sensed by CS1 and may affect precision of current control at the fundamental frequency which is 100kHz. . If I want the snubber to damp 10mHz ringing it will have some minor impact on the 100kHz current also. But I don't know if it is ok or not. Do you any experience with that?. It is my firs kW watt experience with full bridge topology and wasn't sure adding snubber in parallel with current sensing path how will impact the current control performance. By the way in the formula you presented by leakage inductance do you mean the resonance inductance which is in the order of 20uH or the parasitic inductance of input traces and 20cm wire which is in the order of nano hanries?

Since you have already figured that the ringing is not due to the 16 gauge wire, then there's no need for the calculation.

I looked at the picture of your board layout and the positions of your connectors are not impressive.
For EMC purposes, connectors should be to one edge of the board or on two adjacent edges towards one corner.

--Akanimo.
 
  • Like
Reactions: skn96

    skn96

    Points: 2
    Helpful Answer Positive Rating
Yeah I'm wondering about the rest of the layout. Assuming voltage is solid at the bulk caps (double check), there needs to be significant impedance between that point and the point where the voltage is ringing. I assumed the wire explained that, but if that's ruled out then there must be layout issues creating impedance elsewhere.
 
  • Like
Reactions: skn96

    skn96

    Points: 2
    Helpful Answer Positive Rating
Thanks Akanimo and asdf44
It looks like I have been able to fix the issue using both of you guys comment. As you guys predicted the parasitic inductance was the issue but the point I was missing was that it was mostly the loose connection of this wire to the vias. The via were dirty and only few strand of litzs from the 16 gauge wire was loosely connected. I remove the 20cm wire and made it 2 cm and put the current sensor on the primary . Then I added a snubber across the 2cmm wire loop. I used 1 nF capacitor and 10ohm resistor. Then all the ringing issue went away.
Best
 

The MUR1540's are snappy diodes, and will need snubbing, all PSFB converters need real diode snubbing at load... if you are going to go to 2400W even for 2 sec, better to use the best diodes you can get, this will reduce the snubber power required.
 
  • Like
Reactions: skn96

    skn96

    Points: 2
    Helpful Answer Positive Rating
The MUR1540's are snappy diodes, and will need snubbing, all PSFB converters need real diode snubbing at load... if you are going to go to 2400W even for 2 sec, better to use the best diodes you can get, this will reduce the snubber power required.

Thanks For your recommendation . I understand that these didoes are snappy and I actually noticed the D201 and D211 in my converter get very hot since they have do absorb lots of ripple caused by reverse recovery of MUR1540. But the thing is that I havn't been able to find any alternative for MUR1540 which are 15 amp 400 volt ultra fast diodes. Only On semiconductor and vishay make them to my knowledge. Are you aware of any good alternative?
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top