Hi All.
I have been playing with a small LLC for a while.
The first prototypes had very little heating in the primary HB mosfets.
Now I´ve assembled a new prototype and the mosfets heat up more than before. It works and output is stable, but the mosfets get to 80degC with an output of approx 75W.
Mosfets are IRFP450 and heatsink is approx 40x60mm alu plate. Same setup as the old prototype. It is pure ZVS-switcing according to the voltage-dip at the gate before turn-on.
The gate-waveforms look fine to me. Deadtime is OK. around 100Khz and thats a little above resonance.
A potential problem is that the highside-gate is not "lifted" more than 10V above Vcc. Although IRFP450 should be fully on with Vgs at 10V it looks odd. Vcc to L6599 is around 15V.
The HB switchnode it dos not get a all the way to Vss or Vcc and that look strange to me. Although the circulating currrents in the LLC-network is pretty high, I dont think they are high enough to justify that big a "on-resistance-loss".
The IRFP450 are NOT fakea, but the L6599 is from ebay.
Technical details:
Input: 230Vac
Output: +/-60Vdc
Res-cap 88nF (split-cap-design with 2x44nF)
Inductance: 105uH + 17uH
I know its hard for you guys to troble this remotely but any hints and ideas are welcome.
I've ordered a 100x100mm PCB for the next prototypes. If it works I'l be happy to share the files.
Kind regards TroelsM
I have been playing with a small LLC for a while.
The first prototypes had very little heating in the primary HB mosfets.
Now I´ve assembled a new prototype and the mosfets heat up more than before. It works and output is stable, but the mosfets get to 80degC with an output of approx 75W.
Mosfets are IRFP450 and heatsink is approx 40x60mm alu plate. Same setup as the old prototype. It is pure ZVS-switcing according to the voltage-dip at the gate before turn-on.
The gate-waveforms look fine to me. Deadtime is OK. around 100Khz and thats a little above resonance.
A potential problem is that the highside-gate is not "lifted" more than 10V above Vcc. Although IRFP450 should be fully on with Vgs at 10V it looks odd. Vcc to L6599 is around 15V.
The HB switchnode it dos not get a all the way to Vss or Vcc and that look strange to me. Although the circulating currrents in the LLC-network is pretty high, I dont think they are high enough to justify that big a "on-resistance-loss".
The IRFP450 are NOT fakea, but the L6599 is from ebay.
Technical details:
Input: 230Vac
Output: +/-60Vdc
Res-cap 88nF (split-cap-design with 2x44nF)
Inductance: 105uH + 17uH
I know its hard for you guys to troble this remotely but any hints and ideas are welcome.
I've ordered a 100x100mm PCB for the next prototypes. If it works I'l be happy to share the files.
Kind regards TroelsM
Last edited:
To my understanding there are two things that can cause FET heating in a situation like this.
The first is useful work being done in which case the current through the transformer and then output from it should be proportional to what you'd expect should be flowing through the FETs.
You should be able to back calculate ball park expected losses within the FETs and see if matches the temperatures you're seeing.
The other is cross conduction through the FETs either through too little deadtime compensation or some other issue with the switching causing what you don't want to happen. Ie the heating is from small pulses of short circuits through the half bridge. Not long enough to destroy anything but nonetheless unwanted.
Have you scoped the waveforms of the high and low side FETs together to check for any cross conduction?
The first is useful work being done in which case the current through the transformer and then output from it should be proportional to what you'd expect should be flowing through the FETs.
You should be able to back calculate ball park expected losses within the FETs and see if matches the temperatures you're seeing.
The other is cross conduction through the FETs either through too little deadtime compensation or some other issue with the switching causing what you don't want to happen. Ie the heating is from small pulses of short circuits through the half bridge. Not long enough to destroy anything but nonetheless unwanted.
Have you scoped the waveforms of the high and low side FETs together to check for any cross conduction?
![IMGDEAD]](/community/proxy.php?image=http%3A%2F%2F%5BIMGDEAD%5Dhttp%3A%2F%2Fi67.tinypic.com%2Ffdact2.png%5B%2FIMGDEAD%5D&hash=f952f669c0415515177d3e11811db893)
IRFP is 0.40 ohm on resistance, maybe the reason, and you need really to drive the mosfets so it is in on completely.
The LLC is quite critical, you need experiment with it. I do not now it is for audio amp purposes, because LLC is maybe
to slow reacting on peaks of power.
regards
I borrowed a better scope at work and made a few measurments.
This is a test-setup with a DC-source instead of the mains.
The two gate-signals look fine to me: ( shot taken without anything connected to the HB-midpoint, to get a "clean" reading).
Yellow is highside Gate. Blue is low side gate. Vcc 30v
Blue is low side Gate. Yellow is HB mid-point.
No load in the secondary and freq is locked around 100Khz.
With 22Vdc i get this:
Gate swings nicely to approx 15V. HB mid comes close to GND. Looks good
But raising the voltage above that, reveals something odd: HB-mid does fall quickly any more.
This is with Vcc at approx 30V. Gate still looks good.
To my understanding this is a sign of the lower mosfet not turning on fast enough.
IF it was cross-conductiong for that long is would heat up very fast even with no load.
Power draw from psu is around 40mA at 30Vdc
Without load fets are still cool- but with load they heat up. I makes sense that the I^2/R rises if the mosfet does not turn fully on fast enough, but WHY is that slow, when the gate is fast?
This is a test-setup with a DC-source instead of the mains.
The two gate-signals look fine to me: ( shot taken without anything connected to the HB-midpoint, to get a "clean" reading).
Yellow is highside Gate. Blue is low side gate. Vcc 30v
An externally hosted image should be here but it was not working when we last tested it.
Blue is low side Gate. Yellow is HB mid-point.
No load in the secondary and freq is locked around 100Khz.
With 22Vdc i get this:
An externally hosted image should be here but it was not working when we last tested it.
Gate swings nicely to approx 15V. HB mid comes close to GND. Looks good
But raising the voltage above that, reveals something odd: HB-mid does fall quickly any more.
This is with Vcc at approx 30V. Gate still looks good.
To my understanding this is a sign of the lower mosfet not turning on fast enough.
IF it was cross-conductiong for that long is would heat up very fast even with no load.
Power draw from psu is around 40mA at 30Vdc
An externally hosted image should be here but it was not working when we last tested it.
Without load fets are still cool- but with load they heat up. I makes sense that the I^2/R rises if the mosfet does not turn fully on fast enough, but WHY is that slow, when the gate is fast?
Last edited:
Can you also post the gate drive circuit? What value are the gate drive resistors?
There seems to be a problem in the device switching. If it switches slowly,
the conduction loss can be high with a load. Are you using floating differential
voltage probes? The upper gate-source voltage waveform would also be useful.
Last edited:
Drive is text-book ( I think, beginning to suspect some rookie-mistake in my prototype)
The signals change very little with different values for the drive-resistors.
If the gate-drive was totally off it would show up on the gate-drive-waveforms above?
No floating probes for this as the voltage is only 30Vdc for this test
An externally hosted image should be here but it was not working when we last tested it.
The signals change very little with different values for the drive-resistors.
If the gate-drive was totally off it would show up on the gate-drive-waveforms above?
No floating probes for this as the voltage is only 30Vdc for this test
Not familiar with this IC, but perhaps it needs a larger HV for the charge pump to work properly.
It is an unusual arrangement without an external diode. Instead it uses an internal synchronous diode.
Use a floating differential voltage probe to check across the bootstrap capacitor to see if its voltage
stays high enough under operating conditions.
Last edited:
I also designed a LLC SMPS.
I found for some reason it was takin guite a lot of current even without a load.
In the end I traced it to a bad secondary winding on the transformer.
I hadn't got the bi-wired output windings exactly the same length and It was basically shorting out the two windings.
Other than that hot mosfets are due to shoot through or not turning on and off fast enough.
I found for some reason it was takin guite a lot of current even without a load.
In the end I traced it to a bad secondary winding on the transformer.
I hadn't got the bi-wired output windings exactly the same length and It was basically shorting out the two windings.
Other than that hot mosfets are due to shoot through or not turning on and off fast enough.
And also the on resistance can play a role, the IRFP450, the capacitances of the gate and the driver capable to drive enough current to charge discharge the gate it is 2.6 nF, use a good fast mosfet when have such high switching frequency.
https://www.infineon.com/dgdl/mosfet.pdf?fileId=5546d462533600a4015357444e913f4f
Maybe this calculation sheet is of use?.
Last edited:
A quick calculation with this online-tool leads me to believe that maybe, my memory is just wrong, and that the heating in the mosfets are ballpark right.
Free Flat Plate Heat Sink Calculator
If the circulating currents in the LLC+mosfets are about 3Arms, the the loss in the mosfests are about I^2/R = 9/0,5 = 4,5W and with 40x60mm alu plate that gives 140degC in the die (as calculated by the tool).
Free Flat Plate Heat Sink Calculator
If the circulating currents in the LLC+mosfets are about 3Arms, the the loss in the mosfests are about I^2/R = 9/0,5 = 4,5W and with 40x60mm alu plate that gives 140degC in the die (as calculated by the tool).
Or mosfets to slow, on resistance is quite high seek other types maybe trend mosfets.
I measured some more and I'm pretty sure its is not "classic" crossconduction.
My understanding is as follows: (please correct me)
This is an resonant LLC network and operating near the resonance means that the HB-switchnode swings high and low before the appropriate mosfet switches.
This is (as I understand it) the definition of ZVS (zero voltage switching).
This is normally detected by a short negative puls on the gate before turnon ( because the source is actually at a lower voltage than the drain).
I see that negative puls very clear and thus we have ZVS.
I dont think we can have cross-conduction and ZVS at the same time. For ZVS to work there has to be some dead-time.
So what is then causing the heat?
I connected and 15mOhm shunt between the low-side source and gnd.
The voltage across the shunt is equal to the current in the lower mosfet and I see some (very, very) short pulses of 2V across the shunt. Thats a very, very high current
.
I tried googling the subject (H-brigde, crossconduction, shoottrough and such) and found descriptions of similiar pulses simply caused by the internal capacitances of the mosfet being discharged. As the HB-midpoint swings very fast in an LLC, I sort of understand that these discharge-current can get pretty big, but i dont understand why they would heat up the mosfet?
Hope my ramblings make sense. Greatfull for the help so far. keep it comming
Kind regards TroelsM
My understanding is as follows: (please correct me)
This is an resonant LLC network and operating near the resonance means that the HB-switchnode swings high and low before the appropriate mosfet switches.
This is (as I understand it) the definition of ZVS (zero voltage switching).
This is normally detected by a short negative puls on the gate before turnon ( because the source is actually at a lower voltage than the drain).
I see that negative puls very clear and thus we have ZVS.
I dont think we can have cross-conduction and ZVS at the same time. For ZVS to work there has to be some dead-time.
So what is then causing the heat?
I connected and 15mOhm shunt between the low-side source and gnd.
The voltage across the shunt is equal to the current in the lower mosfet and I see some (very, very) short pulses of 2V across the shunt. Thats a very, very high current
.I tried googling the subject (H-brigde, crossconduction, shoottrough and such) and found descriptions of similiar pulses simply caused by the internal capacitances of the mosfet being discharged. As the HB-midpoint swings very fast in an LLC, I sort of understand that these discharge-current can get pretty big, but i dont understand why they would heat up the mosfet?
Hope my ramblings make sense. Greatfull for the help so far. keep it comming
Kind regards TroelsM
If the frequency is too low you will get a high current.
The LLC would appear as LC and at resonant frequency it is basically a short.
I blew a few mosfets up when I first started on LLC until I actually measured the inductance of the transformer. From that I set the high and low frequencies (I just used two) to be outside the resonant frequency.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.