Heating in LLC HB mosfets ?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi. Yes, 60v for safe testning. The basic waveforms are similar to 230vac operation. Yes, the transition is very fast and I think that is what is triggering the huge current-spikes when the parasitic capacitors are charged. The currents should not be xtreme, but maybe I have a hidden error some where.
 
Switchnode-transition at 20v/div and current in lowside source@15mOhm shunt 1v/dv. 50nS/div!
 

Attachments

  • DSC_1343.jpg
    DSC_1343.jpg
    946.7 KB · Views: 81
Just remeasured the inductances in xformer and the external coil. 17+75uH in total. Its low, but the peak output is 120v@8A so peak currents Will be high.

Xformer is approx 1cm2 cross with 0.8mm gap and 27turns primary.

Capacitance is 44+44nF
 
Last edited:
Disabled Account
Joined 2010
Keep in mind that magnetizing current is triangular shaped with peak values at the switching transitions. Operating at primary series resonant frequency load current crosses zero at the same time. Following this, the current that charges the bridge leg capacitace is independent of actual load and thus ZVS can be tested with no load as well.
The current spike you measure across your shunt might be ok or not, this kind of measurement is a bit treacherous and might show wrong results. You better measure the current returning through the resonant cap(s) so you do not see any "cross-conduction" of the half-bridge but the primary current you are looking for. I use a toroidal magnetic current sensor with 50 turns that delivers 1V/1A loaded with 50 Ohms burden resistor. If you do not have access to such a current probe, a resistor with 100mOhms in the ground return path of the lo-side cap will do the trick.
 
Last edited:
Disabled Account
Joined 2010
It is a good idea to measure primary series resonant frequency. Measuring primary stray inductance with some LCR-meter (with test frequencies of 10kHz or less) and calculating based on resonant cap values did not prove very reliable to me. So I power the controller with its fixed voltage, short the secondary winding, and supply the half-bridge with +5V only to stay in a non-destructive range. Now the controller frequency is adjusted either to max primary current, max ac-voltage across resonant cap or to sinuoidal primary current.
These points are all close together.
btw these measurements will give you some insight of the resonant tank behaviour.
 
Last edited:
Disabled Account
Joined 2010
Just remeasured the inductances in xformer and the external coil. 17+75uH in total. Its low, but the peak output is 120v@8A so peak currents Will be high.

Xformer is approx 1cm2 cross with 0.8mm gap and 27turns primary.

Capacitance is 44+44nF
The cross-section of 1cm2 and 1kWac power are at the bleeding edge imho. I use a ETD39 to produce 400W at a resonance below 100kHz.
 
I know that the peak-power is way higher than normally for this size core. I'm not sure how to design LLC's for audio where peak power is allways much higher than average power?

I'm beginning to suspect that what I see is "Cdv/dt shoot-through": the switch-node has such a high dv/dt that the mosfets are briefly turned on during transition? - that would explain the current-peaks, but on the other hand: the current-peaks could also just be the internal parasitic mosfet-caps being charged/discharged.

I tried to connect schottkys (MUR340) over D-S to short out the internal diode, but it did not change anything.
 
I tried the same setup with a smaller gap in the xformer. Like 0.5mm and before it was maybe 1.1mm. Can't remember the exact number, but it made a huge difference.

Mosfets are cold at No load and 10degC above ambient at 100w load. Not sure if the prototype can deliver the same voltage under load, and Im still dont fully understand why the gap made that difference.

Imductance changed from around 75uH to 110uH. Will re-measure.

Kind regards TroelsM
 
Disabled Account
Joined 2010
I'm beginning to suspect that what I see is "Cdv/dt shoot-through": the switch-node has such a high dv/dt that the mosfets are briefly turned on during transition? - that would explain the current-peaks, but on the other hand: the current-peaks could also just be the internal parasitic mosfet-caps being charged/discharged.

I tried to connect schottkys (MUR340) over D-S to short out the internal diode, but it did not change anything.
that is why I recommend measuring primary current, not bridge current. You cannot distinguish capacitve charge pulses from dangerous cross-conduction.
Parallelled Schottky will not change anything in the hi-speed domain because wiring inductance is too high. btw MUR340 is rated 40V, not suitable for off-line application.
 
Last edited:
Your test indicates that magnetizing current was way too high, so the MOSFETs heated up according to conduction loss Rdson*Irms².
You should learn to measure primary current and simulate the primary tank with LTspice.

Hi. I think I have a decent knowledge of simulating this stuff, but I'm probably overlooking something, or maybe I have a "hidden" parasitic in the prototype that does not show up in the sims i have done.

The supprising part is that the sims dont show drastic changes from 75uH to 110uH. I don´t think I'll have the time to full analyze the difference right now, but hopefully I can get back to the issue later.

Kind regards TroelsM
 
I have use this core a E71\33\32.

But is for a welder inverter.

Have someone a good simulation schematic for LLC with models? I have the ncp1395 mod and osc only it is in zip with schematic.
 

Attachments

  • llc inverter.zip
    15.1 KB · Views: 50
  • IMG_20190424_121108.jpg
    IMG_20190424_121108.jpg
    400.2 KB · Views: 94
Disabled Account
Joined 2010
Last edited:
pls complete your zip archive, I got "could not find symbol VC02"
you may find this link useful: https://archive.eetasia.com/www.eetasia.com/ART_8800468574_499501_AN_6f1e90c8.HTM

zip is complete, it does work here.

unpack it in your ltspice folder under LLC design or such, find application.lib and change path for models if needed.

Thanks for the link, these I had already.

Have you idea of a usable feedback for a constant voltage source inverter for mig welding? Using a comparator into this feedback can speedup the regulation, maybe a good idea for use with a amplifier, however I think a unregulated smps for audio do work as good or maybe better as a feedbacked one, but the impedance is higher, for low impedance what is good for impuls behavior feedback is better.

regards
 
Last edited:
Disabled Account
Joined 2010
zip is complete, it does work here.

unpack it in your ltspice folder under LLC design or such, find application.lib and change path for models if needed.

Thanks for the link, these I had already.

Have you idea of a usable feedback for a constant voltage source inverter for mig welding? Using a comparator into this feedback can speedup the regulation, maybe a good idea for use with a amplifier, however I think a unregulated smps for audio do work as good or maybe better as a feedbacked one, but the impedance is higher, for low impedance what is good for impuls behavior feedback is better.

regards
My LLCs operate unregulated, i.e. at fixed frequency adjusted to primary series resonant frequency. With this operating point the output impedance is at absolute minimum, I would call it better than toroidal xformer of same power rating.
On the other hand this implies the urgent need of some fast current limiting. Otherwise the hole thing blows at the first output short.
I prefer clamp Diodes that limit voltage excursion at the primary tank - a bullet-proof cycle-by-cycle current limiter.
For welding applications I would recommend secondary synchronous rectification. That saves power loss and delivers stiffer voltage than diode rectifiers.



The constant voltage regulation of LLC is a special beast due to the nonlinear transfer function - so I avoided this approach.
 
Last edited:
My LLCs operate unregulated, i.e. at fixed frequency adjusted to primary series resonant frequency. With this operating point the output impedance is at absolute minimum, I would call it better than toroidal xformer of same power rating.
On the other hand this implies the urgent need of some fast current limiting. Otherwise the hole thing blows at the first output short.
I prefer clamp Diodes that limit voltage excursion at the primary tank - a bullet-proof cycle-by-cycle current limiter.
For welding applications I would recommend secondary synchronous rectification. That saves power loss and delivers stiffer voltage than diode rectifiers.



The constant voltage regulation of LLC is a special beast due to the nonlinear transfer function - so I avoided this approach.

Thanks

For the mig welding I need a constant voltage and need then feedback, I can do that with a PID feedback or such, caluclating that is not so difficult.

Synchronous rectification is a nice approach but the mosfets need to be uge, and I do not now it is suitable for welding, I do not see it somewhere used in a welder inverter.

I have some resonace designs who are used as a mma welder, also open loop and these have a current limiter who works fine, a antistick, maybe usable for a high current amplifier supply. wil send if you like, it is from russia..


rgards
 
A short test tonight, - and Im even more confused. Started with the smallest gap in the xformer and inductance was 400uH +20uH external. It works and regulates ok, but does not look like, ZVS. Mosfets cool under load.

Adjusted the gap to get the design-goal of 120uH +20uH external. It still works an we have ZVS, but Mosfets heat up?

My Best guess is that the lower inductances raise the ciculating current enough to tigger Cdv/dt false-gate-trig.

Kind regards TroelsM
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.