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.

12V full bridge over heating

Status
Not open for further replies.

mrinalmani

Advanced Member level 1
Joined
Oct 7, 2011
Messages
463
Helped
60
Reputation
121
Reaction score
58
Trophy points
1,318
Location
Delhi, India
Activity points
5,285
Hi
I am working on a 12V phase shifted full bridge connected to a 1:38 step-up transformer. The output of the transformer is rectified using SiC diode bridge. Fsw = 80KHz.
The MOSFET is PSMN2R030YLDX with 2mOhm resistance and maximum 2.5mOhm at Tj = 80C.
- When driving 40A current into the transformer, the power loss per FET is 8W (Measured using temperature rise, thermistor placed directly on each FET's metal tab).
- However the calculated loss is roughly 2W per FET. Approximately 1.5W Rds_on loss and another 500mW switching loss.
- There is no cross conduction or Miller turn-ON. I have checked.
- I measured the current waveform using a Mn-Zn ferrite current transformer with 1:50 ratio. I suppose the CT bandwidth should be at least 5MHz which should be sufficient to see the current characteristic.
Here are some photos. Blue = Vds across low side FET. Yellow = Current through transformer primary measured using CT.
1596722249480.png

Figure-1: Current during FET turn-off.
1596722340849.png

Figure-2: Zoomed in view of Fet turn-off
1596722431358.png

Figure-3: FET turn-off at higher current. Notice bell-shaped voltage overshoot.

What could be the possible cause of 400% more power loss than calculated? Reverse recovery?
Also, a 22nF NP0 capacitor is in series with the transformer output (HV side) to compensate for leakege. This provides partial compensation, the net impedance seen by the bridge is still inductive.

Please share your views.
Thank you!
 

Hi,

You ask about power... but don't really show how you come to the values.
* There may be a circuit problem
* There may be a calculation error
* There may be a measurement or design error.
How can we know?

Klaus
 

Hi
Thanks for the reply.
Here's the approximate loss calculation method I used.
Rds_on at Tj = 80C = 2.5mOhm max. Rise and fall time = 25ns approx. V_ds = 13V, I_Out = 40A avg. Based on this:
I^2R loss = 40x40x2.5m = 4W. But the switch is only active 50% time in full bridge so loss = 2W.
For switching loss, the transition time is roughly 0.2% of the switch time period. so I am assuming power loss of no more than 0.2% of the delivered power. For 400W output, this should mean 800mW switching loss distributed between the switches. So per switch switching loss = 200mW. Ignoring Coss loss because its's far too low, the total power loss per switch = 2.2W . But the problem is I am getting a loss of 8W per switch. Even if it was 3W or 4W instead of 8W, I would think of measurement errors but its straight away 4x more than calculated.
Measurement procedure:
1. Input power vs output power shows 45W difference at 400W load. 10mA resolution Fluke 325 meter used with proper calibration. SiC bridge loss and transformer loss is nearly 3W and 6 to 8W respectively. I have heated the transformer using programmable DC supply and measured the temperature rise at different power dissipation with constant current. When driving transformer with the bridge the temperature of transformer suggests around 6 to 8W loss and it is a PQ30mm core and any more than 10W will destroy the winding. So it's not the transformer that's eating most of that 45W.
2. I supplied variable DC current from 1A to 12A through the MOSFETs body diode and measured the temperature rise (delta T) vs power for each MOSFET. It stands nearly constant at 3 degree C/W from 2W to 8W power dissipation. Then, when driving the transformer with the MOSFETs at 40A current, the temperature rise of MOSFETs suggest 8W loss per MOSFET.
I understand there may have been measurement errors but I think that should be limited to say 10% or max 20% but here we are talking of 400%. Probably missing out an entire physical phenomenon that's creating the loss.
3. I have also measured the RDS_on by passing 12A current and it came close to 2mOhm.

Thank you
 

Hi,

I_Out = 40A avg
For ohmic power calculations you should use RMS instead of AVG.
But if you calculate with "constant 40 A during ON", then (only then) your calculation is correct.

--> I can not verify the 40A in your scope pictures. In the upper picture I see 2A/division and a ratio of 50:1 which makes a total of 100A/div.
With 3.5 divisions(assuming center line = zero current line) this means 350A ... which makes no sense...

But the other measurement setup and even the values make sense on a first few.

did you do the input and output current measurement on the "clean" DC paths? input: before the bulk capacitor, output after bulk capacitor.

*****
To exclude reverse recovery loss: can you add schottky diodes in parallel to the FET-internal body diodes? ... for a test.
Mind short wiring.

Other ideas:
* In the transformer and other wiring there may skin effect and proximity effects causing additional loss that can´t be verified with a DC method.
--> Try to shorten the wiring and maybe use HF litz.
* What about the capacitors? Are they made for high frequency switching with low ESR? Do they get warm?
* do you have an IR camera to detect heat sources?

Klaus
 

No, the waveform title line is apparently showing correct scale. You see 40A in the third waveform.

Problem is however that we don't see transistor currents. And also don't know about gate drive pattern.
 

if your control is not pk current mode - then you can get extra current in the pri
 

Hi, thanks for the reply.
I understand that RMS is what we should be talking about instead of average, but I think the error of approx 300% to 400% is too large to depend on RMS or average especially as the duty cycle approaches 100%.
- The current magnitude is different for different snapshots. For the one with 2A/div, it actually is 2A/div after accounting for the 1:50 probe attenuation.
- The transformer is already wound with litz. And more over it is the FETs that are heating up, the transformer although hot, is not yet the major concern.
- There are no aluminium capacitors at the 12V input. There are several MLCCs.
- The output bulk capacitor is 220uF 450V and it does not heat up.
- Only the MOSFETs are heating up.
I will try to solder schottky diodes directly on top of the FETs and see what happens.
Here are a few wave forms of that depict the transformer primary voltage vs current waveform. It's a typical phase shifted full bridge patterns.
 

Attachments

  • scope_10.png
    scope_10.png
    32.5 KB · Views: 92
  • scope_11.png
    scope_11.png
    28.1 KB · Views: 85
Last edited:

what is your output circuit? have you considered and current doubler? layout of the primary is critical to low turn off losses ...
 

I found out the problem.
Majority of the loss is coming from high amount of ripple current circulating between the decoupling capacitors and the cable inductance of power cable connecting 12V source to the PCB. (Measured 1.5uH using LCR meter)
Strategic placement of capacitors on the PCB minimizes losses to a great extent.
 
Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top