Author Topic: UCC28019 based PFC blows up the mosfet  (Read 3162 times)

0 Members and 1 Guest are viewing this topic.

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
UCC28019 based PFC blows up the mosfet
« on: October 16, 2021, 12:08:39 pm »
Hi,
I am trying to design a PFC circuit (370V, 1kW) using the UCC28019. A prototype single-layer board has been made for this purpose and calculations were done as per design note SLUC069. Initially, a couple of 60W bulbs (parallel) were placed in series with the input of the board, and the Output was observed. The mains is a 240V source. The NTC thermistor was heating to quite an extent along with input bulbs glowing brightly (slow flickering).

I then tried to provide it with a variable input through an autotransformer without the series bulb and that ended up blowing the thermistor and the MOSFET (all leads were short). This process was repeated again but now monitoring the gate waveform and found out that the gate PWM becomes almost close to 100% once the input crosses the VIN enable threshold level which in our case was around 85V. Before reaching this level, the gate was low.

I also tried removing the MOSFET and replace the IC and then operate it with varying input and found out that the same thing happens as above with some of the pin voltages being

VCOMP = 7.55V, VSEN = 1.57V, VINS = 2.08V, ISEN = 0, VCC = 11.85V and ICOMP = 1V.

The Inductor waveform (wire through toroid core) shows that the PWM (65kHz) to be in kind of a burst mode followed by a slow switching waveform (~12Hz) which is visible through the flickering of the input bulb. There are also some spikes on the waveform when the PWM is active which makes me believe that perhaps the overcurrent (PCL) might be triggering. The images are of PWM and slow flickering respectively,
[ Specified attachment is not available ]

The gate waveform is as follows,
[ Specified attachment is not available ]

The PWM triggers after every 20m sec (T = 80m sec, 12.5Hz) as seen above.

Note: A 60W Incandescent bulb has been used as a load. The component values in light blue are the ones that are used in the circuit.

Here's the schematic of the PFC board
[ Specified attachment is not available ]

inductor waveform with visible bursts of PWM
[ Specified attachment is not available ]

Any kind of help or suggestion is highly welcomed and appreciated,

Regards,
Ashwin619.





« Last Edit: October 17, 2021, 06:13:33 am by Ashwin619 »
 

Offline Faringdon

  • Super Contributor
  • ***
  • !
  • Posts: 1949
  • Country: gb
Re: UCC28019 based PFC blows up the mosfet
« Reply #1 on: October 16, 2021, 02:13:01 pm »
Your C7 looks a bit too big in value.
Also, whats your switchign frequency?
Dont make it too high.
You coudl be better off using eg a very soft diode, (powerintegrations make some)  or SiC

Your MOSFET is here (so we can see it when we come back to help you more later)
https://www.onsemi.com/pdf/datasheet/fcp190n65s3r0-d.pdf

The attached is your PFC in a "quick" sim, with the control "hacked" to constant DC at the level which gives 1kW of power.
Its at 100khz, but you can change this down as you please....it gives you an idea of what goes on without taking ages to simulate like most pfc stages.
this one is in LTspice
« Last Edit: October 16, 2021, 02:36:11 pm by Faringdon »
'Perfection' is the enemy of 'perfectly satisfactory'
 

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
Re: UCC28019 based PFC blows up the mosfet
« Reply #2 on: October 17, 2021, 06:48:22 am »
Hi Faringdon,
Thanks for your reply,

The UCC28019 CCM PFC ic works at a constant PWM frequency of 65kHz and it's visible in the in-between bursts of the inductor waveform. The period between two PWM bursts is around 80msec (~12.5 Hz).
As far as the C7 value is concerned it was obtained through the design note SLU069 from TI.
Do you think that the cap might be getting fully charged (kind of like filtering) the pulsating DC which thus might not be able to provide the required input voltage reference?

I have done the simulation of the above circuit in TINA TI software and it seems to work there.

Note:- Initially I had connected the Vbridge_gnd net as the GND reference for the IC and its passive components, however someone from the TI forum suggested providing the MOSFET source referenced gnd to the IC and other components. I have made the changes however it hasn't resolved the issue.

I'll surely try your LTspice simulation. I wish TI could have provided with an unencrypted spice model of UCC28019  :(

Please let me know if there's any other possible correction in the circuit.

Regards,
Ashwin619.

 

Offline Faringdon

  • Super Contributor
  • ***
  • !
  • Posts: 1949
  • Country: gb
Re: UCC28019 based PFC blows up the mosfet
« Reply #3 on: October 17, 2021, 11:15:07 am »
What about PCB layout, did you follow the attached
'Perfection' is the enemy of 'perfectly satisfactory'
 

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
Re: UCC28019 based PFC blows up the mosfet
« Reply #4 on: October 18, 2021, 04:21:35 am »
Thanks for providing the PCB layout guidelines, Here's my design,


I am aware that I may have skipped quite some PCB layout guidelines (especially the length of the MOSFET gates track length), however, I don't think that it should be a reason for the circuit to not work at all. Let me know if you feel otherwise.

« Last Edit: October 24, 2021, 07:08:41 am by Ashwin619 »
 

Offline TimNJ

  • Super Contributor
  • ***
  • Posts: 1649
  • Country: us
Re: UCC28019 based PFC blows up the mosfet
« Reply #5 on: October 18, 2021, 09:49:13 pm »
It seems the MOSFET fails (somewhat quickly?), and places a low impedance across the input. Too much current (excessive die/package temperature) and excessive drain-source voltage are usually what kill MOSFETs..

1. Is the PFC inductor sized properly? A saturated core leads to high peak currents through the MOSFET.
2. What is the peak drain-source voltage across the MOSFET when the MOSFET switches off?

It's hard to tell what you are measuring in your oscilloscope photos. You show an "inductor waveform" with square wave shape. What are you measuring here? Inductor current should be ramping up/ramping down in a triangle wave.

I can't tell completley from your PCB drawing, but do you think you might have 'GND' and 'Vbridge_gnd' flipped or maybe mushed together somehow?

'GND' should be the common net that the IC is referenced to (Pin 1), and 'Vbridge_gnd' is where the current-sense signal should be taken from (into Pin 3). Take a look in the yellow highlighted area.

To be more useful, suggest attaching a PCB layout where you can see the component names (i.e. R2, R3, etc.)

 

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
Re: UCC28019 based PFC blows up the mosfet
« Reply #6 on: October 19, 2021, 11:29:31 am »
I agree that the mosfet is failing due to the reasons you have mentioned above, high voltage seems unlikely as I haven't even crossed 120V to date. The current mosfet (IRF840) is rated for 500V and far as the max Vds is concerned I believe that the max voltage the mosfet can face in such a boost topology would be the output voltage when the diode conducts and the mosfet is off which happens to be much lower than the rated one.

The Inductor saturation is something that I believe must be happening and it's also kind of visible in the Inductor voltage waveform which I have measured using a wire wound around the toroid and a differential probe of the DSO is used to measure it. The square wave that you have mentioned is nothing but the voltage across the inductor itself (or at least a part of it)

As far as the GND and Vbridge_gnd nets are concerned they are separated by the Rsense in between them and initially the IC ground PIN 1 was referenced to Vbridge_gnd which now I have corrected to GND, so are the other passives around it.

Note:- Thanks to your highlight I have noticed that I may have connected the ISENSE pin (PIN 3) to GND rather than Vbridge_gnd due to the changes I had made recently. I'll try to rectify that, however before the correction, PIN 3 was surely connected to Vbridge_gnd and the converter had behaved in a similar fashion.

I have attached the image's of the copper layout before and after the changes made to the GND references,
The red line I where the copper plane is separated and the blue is where new connections are made. Thick grey lines are the jumper wires.

Please let me know if there are any further corrections or suggestions regarding the design.
Thanks for your valuable inputs !!

Regards,
Ashwin619.



 
 

Offline TimNJ

  • Super Contributor
  • ***
  • Posts: 1649
  • Country: us
Re: UCC28019 based PFC blows up the mosfet
« Reply #7 on: October 19, 2021, 04:34:16 pm »
I agree that the mosfet is failing due to the reasons you have mentioned above, high voltage seems unlikely as I haven't even crossed 120V to date. The current mosfet (IRF840) is rated for 500V and far as the max Vds is concerned I believe that the max voltage the mosfet can face in such a boost topology would be the output voltage when the diode conducts and the mosfet is off which happens to be much lower than the rated one.

The Inductor saturation is something that I believe must be happening and it's also kind of visible in the Inductor voltage waveform which I have measured using a wire wound around the toroid and a differential probe of the DSO is used to measure it. The square wave that you have mentioned is nothing but the voltage across the inductor itself (or at least a part of it)

As far as the GND and Vbridge_gnd nets are concerned they are separated by the Rsense in between them and initially the IC ground PIN 1 was referenced to Vbridge_gnd which now I have corrected to GND, so are the other passives around it.

Note:- Thanks to your highlight I have noticed that I may have connected the ISENSE pin (PIN 3) to GND rather than Vbridge_gnd due to the changes I had made recently. I'll try to rectify that, however before the correction, PIN 3 was surely connected to Vbridge_gnd and the converter had behaved in a similar fashion.

I have attached the image's of the copper layout before and after the changes made to the GND references,
The red line I where the copper plane is separated and the blue is where new connections are made. Thick grey lines are the jumper wires.

Please let me know if there are any further corrections or suggestions regarding the design.
Thanks for your valuable inputs !!

Regards,
Ashwin619.

I will take your word that the layout was correct with respect to GND vs VBridge_gnd for the prototype you built.

I suggest to measure the voltage across the two sense resistors with an oscilloscope. This shows you the inductor current, of course. The current should be triangular. It should not have any "hooks" or look exponential. This indicates saturation. If there is no load on the PFC, I am a little skeptical this is the issue, although there may be some problem at startup, I suppose.

The waveforms you show do not appear to show inductor current. Where was your "inductor waveform" measured? Voltage across the inductor?

500V rating is possibly okay for a PFC, but if there is lots of ringing (I see quite a lot on some of your waveforms), the peak ringing voltage may exceed the Vds rating. It would be smart to confirm that Vds is okay. Poor layout with lots of inductance will give poorer results. Even a good layout may give 50V+ ringing. If you see a problem, you can slow down the turn-off time of the MOSFET to reduce the dv/dt on the turn-off. On your schematic, there is no "turn-off resistor", just a diode, so the gate is pulled down very hard and oscillation is likely. You can just remove the diode D3 at first and increase the value of R6.

 

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
Re: UCC28019 based PFC blows up the mosfet
« Reply #8 on: October 21, 2021, 07:44:21 am »
Thanks, TimNJ for your suggestions,

Currently, I have replaced the aux supply with a 12V battery source to avoid messing with the GND reference any further. So now all the components including IC gnd (PIN 1), passive components, and 12V battery source are sourced to the GND net, whereas the R2 resistor is connected to the Vbridge_gnd for current sense. Other than these no further changes have been made to the circuit.

The waveforms in the Siglent DSO are that of the inductor voltage (measured via a couple of turns through toroid) whereas the waveform in the red DSO kit is that of the mosfet gate waveform. I'll try to measure the current waveform via differential probes across the Rsense as per your suggestion, perhaps using the red DSO.

I agree with your point that parasitic inductances via long traces cause quite some ringing and other issues however as I have mentioned before the PWM signals are triggered after every 80msec which is visible in the Inductor waveform image, also I haven't even crossed the input of 120V as the mosfet blows up before we reach this point. The probability of having 500V+ spikes with such input seems highly unlikely. Although ill try to measure Vds to ensure this. I'll also try to add a 10ohm resistor for the gate sink to avoid hard switching.

I'll try to post a video with some proper explanations regarding this.
Thanks again,

Regards,
Ashwin619.


 
 

Offline TimNJ

  • Super Contributor
  • ***
  • Posts: 1649
  • Country: us
Re: UCC28019 based PFC blows up the mosfet
« Reply #9 on: October 21, 2021, 06:15:46 pm »
However you decide to measure the inductor current, it should look something like this, if zoomed in on one half of an AC cycle.



The square waves I see don't make too much sense for inductor current.

As far as Vds voltage stress goes, the MOSFET sees the PFC set-point voltage + the boost diode drop. So...like 370V + 0.7V, or so. You can say that the 'voltage boost is developed across the inductor'. Therefore, doesn't matter if the input voltage is 80V or 240V, the off-state Vds voltage will be the same. The turn-off spike, however, may change with input voltage, more specifically it may change with input current. All PFCs I've designed use 600V transistors minimum.
 

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
Re: UCC28019 based PFC blows up the mosfet
« Reply #10 on: October 22, 2021, 06:46:39 am »
Its true that the max Vds the mosfet faces will be something close to Vout + any spike due to ringing but i think this might be unlikely as the output hasn't crossed 150V yet.

I have attached a video link to show what exactly is happening, please have a look at it

https://drive.google.com/file/d/1mzWMNR4ex1TjvduxO-GF3vQVQpR1nfsp/view?usp=sharing

One more thing that I have noticed is as soon as I crossed Vout of 150V the snubber resistor (220Ohm) started to release magic smoke.
Do you think this might be due to ringing ?
   
 

Offline Wolfram

  • Frequent Contributor
  • **
  • Posts: 378
  • Country: no
Re: UCC28019 based PFC blows up the mosfet
« Reply #11 on: October 22, 2021, 12:52:35 pm »
You have VBridge_gnd and GND switched in some places, as TimNJ points out. The schematic shows the controller GND (pin 1) connected to GND (MOSFET source), this is correct. Your layout shows controller pin 1 connected to VBridge_gnd, this is incorrect.

The loop between Q1/D4/C19||C20 needs to be kept small, it does not look too bad in your layout but it could be improved by moving the SMD caps closer to the MOSFET and diode.

The loop between the UCC and the MOSFET gate needs to be kept small as well. Your gate lead is not too long, but the return path is very convoluted. Due to the incorrect connection of the controller ground, the gate return current has to pass through R5||R15, and these will probably have significant inductance in addition to their resistance. These will not only drop voltage due to the gate charging current, but also the drain current of the MOSFET. Make sure you ground the control IC in the right place, and make the loop for the gate drive signal and the return current as small as practical. This is easier on a two-layer board but many successful APFCs have been made on single layer boards.
 

Offline TimNJ

  • Super Contributor
  • ***
  • Posts: 1649
  • Country: us
Re: UCC28019 based PFC blows up the mosfet
« Reply #12 on: October 22, 2021, 09:31:16 pm »
I watched the video. Some questions:

1. Does the power supply fail with no load (or very light load) on the PFC output?
2. Does the rail regulate at 370V at no load (or very light load)?

Traditionally, the "series light bulb test" shows you if you have some sort of unexpected short circuit or low-impedance. To me, it almost points towards a wiring error or PCB layout error. I would expect the output to regulate at 370V at no load/light load, as long as the wiring/layout is correct (or at least matches the datasheet/app note design).

When a load is added, there is the possibility that the boost inductor saturates and that the over-current protection cannot adequately protect the MOSFET. In this case, the inductor stops acting like an inductor and instead just looks like a piece of wire with a small amount of resistance. If L2 is saturated and Q1 is on, then the AC input is effectively shorted through Q1. Again, if you look at the inductor current, not voltage, you can see quite easily if it is entering saturation.

If you purchased the inductor, what is the manufacturer part number? And if you wound the inductor yourself, what core size, core material, and number of turns did you use?
 

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
Re: UCC28019 based PFC blows up the mosfet
« Reply #13 on: October 23, 2021, 06:37:37 am »
You have VBridge_gnd and GND switched in some places, as TimNJ points out. The schematic shows the controller GND (pin 1) connected to GND (MOSFET source), this is correct. Your layout shows controller pin 1 connected to VBridge_gnd, this is incorrect.

The loop between Q1/D4/C19||C20 needs to be kept small, it does not look too bad in your layout but it could be improved by moving the SMD caps closer to the MOSFET and diode.

The loop between the UCC and the MOSFET gate needs to be kept small as well. Your gate lead is not too long, but the return path is very convoluted. Due to the incorrect connection of the controller ground, the gate return current has to pass through R5||R15, and these will probably have significant inductance in addition to their resistance. These will not only drop voltage due to the gate charging current, but also the drain current of the MOSFET. Make sure you ground the control IC in the right place, and make the loop for the gate drive signal and the return current as small as practical. This is easier on a two-layer board but many successful APFCs have been made on single layer boards.

Hi Wolfram,
Thanks for your suggestions,
As I have mentioned before the grounding issue has been corrected in the layout by referencing the PIN 1 gnd of the IC to GND net along with the other passive components. Removing the R5 and R15 resistor shows clear isolation between the two ground nets i.e. GND and Vbridge_gnd.

The layout optimization for better return paths/loop is something that I agree with, however this seems to be somewhat a different issue. The problem is with the input series bulb there's flickering and removing the bulb causes quite an inrush probably due to the sudden application of almost close to 100% PWM.

I recently tried this setup again with the bulb removed and then gradually increasing the input with an autotransformer. After reaching an input at around 95-100 Vac the autotransformer started to vibrate (this had happened previously too and I assume that it was due to high current perhaps due to a short which may have been caused by the PWM being very high ~100% for long duration), however this time the mosfet didn't blow (shorted pins). I believe the current limit may have been kicked in due to the correction in the ground plane.

By the way, do you think it's the right way to gradually increase the input voltage via autotransformer or should I directly apply mains to it via an existing fuse and NTC?

« Last Edit: October 23, 2021, 07:09:27 am by Ashwin619 »
 

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
Re: UCC28019 based PFC blows up the mosfet
« Reply #14 on: October 23, 2021, 06:48:32 am »
I watched the video. Some questions:

1. Does the power supply fail with no load (or very light load) on the PFC output?
2. Does the rail regulate at 370V at no load (or very light load)?

Traditionally, the "series light bulb test" shows you if you have some sort of unexpected short circuit or low-impedance. To me, it almost points towards a wiring error or PCB layout error. I would expect the output to regulate at 370V at no load/light load, as long as the wiring/layout is correct (or at least matches the datasheet/app note design).

When a load is added, there is the possibility that the boost inductor saturates and that the over-current protection cannot adequately protect the MOSFET. In this case, the inductor stops acting like an inductor and instead just looks like a piece of wire with a small amount of resistance. If L2 is saturated and Q1 is on, then the AC input is effectively shorted through Q1. Again, if you look at the inductor current, not voltage, you can see quite easily if it is entering saturation.

If you purchased the inductor, what is the manufacturer part number? And if you wound the inductor yourself, what core size, core material, and number of turns did you use?

1. Currently I have applied a load of 128 Ohms (two incandescent bulbs in series) at the output, it is connected continuously and I guess the resistance will further increase once the filament starts heating.
2. I haven't even reached 370V as the transformer starts vibrating (a possible over current/short) condition as I reach an input of around 95-100V. The output certainly gets boosted to around 180-195V but overall it feels that something is wrong.

Here is the link to the inductor
https://in.element14.com/abracon/atca-08-471m-v/inductor-470uh-7a-20-radial/dp/2665051?ost=atca-08-471m-v

I am planning to connect the circuit directly to the mains (240V) via a B10 curve MCB, and in case that doesn't work I'll come up with a new board layout.
« Last Edit: October 23, 2021, 07:11:04 am by Ashwin619 »
 

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
Re: UCC28019 based PFC blows up the mosfet
« Reply #15 on: October 23, 2021, 03:35:38 pm »
I have just tested the circuit again, this time without any input series bulb and with input MCB just for safety and it has suddenly started working ;D. Here's the video of its working,
https://drive.google.com/file/d/1I9rDiz45V3ZBMv3ARzVjqG3qoo4sPfiA/view?usp=sharing

The output voltage in the video is low due to some error in the output reference, after adjusting it I am getting the desired voltage of around 370V. The mosfet (currently without any heatsink) and the inductor seem to heat to quite some extent. I believe the high Rds on of the mosfet i.e. IRF840 (~0.8 Ohm) and a possible case of inductor saturating might be the culprits behind it. Probing the inductor current waveform via Rsense as suggested by TimNJ may help to shed some light on it.

Nevertheless, I appreciate the help and suggestions provided by you guys. I Will let you know once I apply some different loads across it.

Thanks and Regards,
Ashwin619.
 

Offline Ashwin619Topic starter

  • Contributor
  • Posts: 48
  • Country: in
Re: UCC28019 based PFC blows up the mosfet
« Reply #16 on: October 25, 2021, 07:04:37 am »
Hello everyone,

So today I tried testing the circuit again along with the measuring of the inductor current waveform (via DSO probes across the Rsense). I have also changed the mosfet (with a heatsink) from IRF840 to FCP190N65 which is rated for almost 600V and has a better Rdson value.

The inductor still heat's to quite an extent (if kept operating for more than a couple of minutes). There are quite some spikes on the inductor voltage waveform (at an input close to 120Vac), which decreases as we approach higher voltage (~ 180 - 220 Vac). This perhaps indicates some momentary saturation during turn ON and turn OFF of the mosfet. The current waveform doesn't clearly show any sign of saturation but maybe that's dues to the bigger timescale and poor resolution of the red DSO kit.

The power levels I have measured are as follows,
Vin = 160 Vac, Iin ~ 0.6 A
Vout = 371.6 Vdc, Iout ~ 0.25 A
Efficiency ~ 96.7%

The currents were measured using a clamp meter and all the values are in RMS. I hope this calculation makes any sense.

The heating of the inductor is something that I am worried about as it seems somewhat abnormal.

Please let me know if there's any solution to this other than beefing up the inductor.
 


« Last Edit: October 27, 2021, 09:54:42 am by Ashwin619 »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf