Hi guys,
I am currently working on a buck converter using GaN FETs (EPC2103) as power switches. With my current setup I can archieve a switchnode rise time of about 1ns without parasitic ringing.
So far so good but there is one issue with GaN FETs: They don't have a body diode and therefore the switchnode voltage goes down do -3V during the delay time, where both the high side and the low side switch are turned off. The buck controller chip (MIC2127A) does not like this at all, which is why I installed a Schottky diode in parallel to the low side switch do limit the switchnode voltage to -1V. This diode (PMEG10010) is not fast enough and at around 10A output current, the voltage got so low that it destroyed the controller. So my quention is the following: What is the fastest power diode you know? Minimum requirements: If=1A, Ur=75V
I attached two pictures, one shows the switchnode voltage
and the other picture shows a closeup of the switchnode voltage where you can see the negative voltage (orange and yellow) as well as the low side gate voltage (blue).
Not sure if it may help, but a good source of diodes are the body diodes in other mosfets
Not sure if it may help, but a good source of diodes are the body diodes in other mosfets
I believe that mosfet body diodes can experience current crowding during reverse recovery, sometimes leading to destruction. That is to say, different parts of the die recover at different rates, and the slowest areas are eventually left with the entire reverse recovery current. This is especially so if the gate was turned on after the body has begun to conduct fully. That was 20 years ago, might not be an issue nowadays.
Why did you change the design to use Gan FETS ?
I cannot see the point in your risetimes, they just cause trouble as you have discovered.
Why did you change the design to use Gan FETS ?
These FETs have lower switching losses because they are faster, lower losses due to a lower output capacity, lower driving losses due to lower gate capacity, no reverse recovery losses, lower parasitic inductances...
I cannot see the point in your risetimes, they just cause trouble as you have discovered.
The high rise times with the low switching losses enable a higher switching frequency and therefore a smaller PCB design. Only a few percent points higher efficiency can make the difference between having to use a heat sink or not. High rise times in itself do not cause problems for me, as you can see there is no ringing.
Not sure if it may help, but a good source of diodes are the body diodes in other mosfets
in that case I could replace the low side GaN FET with a Si FET but as explained above I want to avoid that.
I am looking for a fast Schottky diode, maybe on the basis of GaAs, InP, do you know any such component?
in that case I could replace the low side GaN FET with a Si FET but as explained above I want to avoid that.
I am looking for a fast Schottky diode, maybe on the basis of GaAs, InP, do you know any such component?
Not the same, by using the body diode of a mosfet whose Vgs=0 you are using just the diode and maintaning the switching characteristics of a GaN.
In a certain application I did the same because body diodes can have a favorable forward recovery characteristic.
Not the same, by using the body diode of a mosfet whose Vgs=0 you are using just the diode and maintaning the switching characteristics of a GaN.
In a certain application I did the same because body diodes can have a favorable forward recovery characteristic.
That's interesting, can you explain the application you used a body diode and do you have any resources on body diode behaviour? Where did you learn that and what FETs would you recommend?
The quated speed of diodes is normally the time for turning off. The time to turn on is more determined by parasitic inductance and thus the layout and phsical size. There is some forward recovery, but mainly is noticable with PIN diodes made for high voltage, as they have additional votlage drop from the low doped zone. This can lead to a higher forward voltage for the initial time until an equilibrium carrier concentration has build up.
With a high current the voltage drop at the diode may just get to high. So it may need a larger diode or more of the small ones in parallel.
There are fast SiC based diodes, but I am afraid there forward voltage drop may be too high.
There is no faster diode than schottky. The GaN itself behaves as a schottky with Vf ~ Vgs(th) + Vgs(off). Both are minority carrier devices, and suffer no recovery or time delay other than that given by their dimensions, RLC equivalent circuit, whatever.
These waveforms are on a time scale of single nanoseconds. It would seem the better question is: what diode can you even wire in parallel with the low side, that will provide the required properties -- namely, stray inductance in the diode-FET loop, and junction capacitance.
In other words, speed is limited by small-signal bandwidth, which is wholly dependent on layout geometry and switching impedance.
Have you considered:
- Using a GaN-specific controller?
- External gate driver?
- Those GaN-driver-and-inverter-in-a-chip parts?
You may not be completely screwed, with this combination (the controller plus GaN). A possibility is putting the series gate resistor in the negative (PGND?) path between controller and bridge, so that the negative swing can be clamped by a (smaller) diode, without upsetting the controller. Whether this is feasible given other limitations (e.g. permissible AGND/PGND voltage, whether the controller itself needs to be referenced to a global GND e.g. because of a programmable voltage setpoint or something, etc.), you'll have to figure out.
Tim
As per Tim's advice, you can't really get faster than a Schottky or the reverse recovery characteristic of the GaN HEMT.
If I am reading correctly, the problem caused by the negative voltage swing here is the damage to your controller, not excess reverse conduction loss. My suggestion is that instead of trying to reduce the reverse voltage across the power switch using a power diode, you try to reduce the reverse voltage across the controller using a smaller (faster) signal diode. You can likely find some small signal diode (maybe an RF diode worst case) that you could put in reverse parallel (or something else) to your controller in order to clamp the reverse voltage there.
In other words, speed is limited by small-signal bandwidth, which is wholly dependent on layout geometry and switching impedance.
The Schottky diode is placed on the bottom side of the PCB, exactly below the EPC2103 with 5 vias each connecting it. I am aware that the inductance is critical but there is no better way. I simulated the whole circuit and adjusted the diode inductance in the simulation until the result matched the measured curve, when the low side FET turns off and the high side FET turns on. You can see, that the voltage dips into the negative and returns to about -1V which is the Vf of the Schottky at load current. In the simulation I get the same dip when I insert an inductance of 100pH which is already quite low.
The strange thing is that when the high side FET turns off, the diode does not seem to do its job because the voltage drops way too low. The only difference to the previous case is that the diode has a reverse bias now before turning on. I don't know why it does that. It seems like it doesn't even turn on at all. Do you know why that could be?
Have you considered:
- Using a GaN-specific controller?
Yes, I would really like to and if you know one pleas tell me but to my knowledge there is no GaN specific buck controller going up to 75V. The MIC2127A is listed on EPC's website as a suitable controller which is ironic because it clearly doesn't work and I already had a long video call with the chip developers from Microchip discussing this issue. They also have no better suggestion.
- External gate driver?
- Those GaN-driver-and-inverter-in-a-chip parts?
maybe possible but buck controllers of this kind have a current sense feature for which the switchnode voltage has to be connected to the controller independent of the gate driver.
A possibility is putting the series gate resistor in the negative (PGND?) path between controller and bridge, so that the negative swing can be clamped by a (smaller) diode, without upsetting the controller. Whether this is feasible given other limitations (e.g. permissible AGND/PGND voltage, whether the controller itself needs to be referenced to a global GND e.g. because of a programmable voltage setpoint or something, etc.), you'll have to figure out.
Something simular was suggested by the Microchip developers but after thinking about that for a long time and simulating it I am pretty sure this would destroy the IC because it messes with the voltage reference for the gate path. If you are really interested I can show you a detailed analysis of this...
I'm going to guess that if these don't help, you're shit out of luck
made my day
looks interesting, I will look into it tomorrow
Thanks for all the helpful responses
Not sure if it may help, but a good source of diodes are the body diodes in other mosfets
The MOSFET body diode is formed from the base-collector junction of the parasitic bipolar transistor and not particularly fast.
I believe that mosfet body diodes can experience current crowding during reverse recovery, sometimes leading to destruction. That is to say, different parts of the die recover at different rates, and the slowest areas are eventually left with the entire reverse recovery current. This is especially so if the gate was turned on after the body has begun to conduct fully. That was 20 years ago, might not be an issue nowadays.
It can be current crowding, but not of the body diode. The parasitic bipolar transistor can operate in inverse mode if the body diode snaps off and dV/dT is high enough, resulting in avalanche breakdown or secondary breakdown of the parasitic bipolar transistor. This may appear as a second recovery time waveform after the body diode's recovery time waveform. Newer power MOSFETs are more resistant to this.
With a high current the voltage drop at the diode may just get to high. So it may need a larger diode or more of the small ones in parallel.
That is my thought as well. I do not see how it could matter, but at that high a current density, the parasitic PN junction in the Schottky diode is being forward biased normal PN reverse recovery applies.
Multiple diodes in parallel will also lower the total parasitic inductance.
The MOSFET body diode is formed from the base-collector junction of the parasitic bipolar transistor and not particularly fast.
Jep, had a look at a few MOSFETS datasheets and the body diodes are not particularly fast
That is my thought as well. I do not see how it could matter, but at that high a current density, the parasitic PN junction in the Schottky diode is being forward biased normal PN reverse recovery applies.
I am now pretty confident this might be the issue, I am going to try simular diodes without guard ring structure and let you know...
At turn off of the high side MOSFET, the free wheeling diode goes from reverse bias to forward bias, how exactly can reverse recovery be relevant there?
Do you mean normal PN forward recovery will apply in the high current regime?
The forward recovery voltage of PN junction rectifiers is often incorrectly assumed to be an "inductive" effect, and it is a conductivity modulation effect.
It is a mistake to assume that if reverse recovery is large then forward is going to be also large and viceversa.
My application was with voltages >100V and if the lowest possible forward recovery voltage was required then the body-drain diode of a MOSFET was the best available. The peak forward recovery voltage with near instantaneous current application is limited by the resistance of the drift region, which is the ON resistance of a HV MOSFET, and this is carefully minimized by design.
Well, it can't be any less than the drop due to Rs, that's certainly true. What about peak, though?
Curiously (or perhaps not, I'm not real clear on the mechanisms of forward recovery myself), all my testing has shown little if any forward recovery (transient peak voltage above Vf) in MOSFETs. And by that I mean, less than Vgs(th) would allow, because that is an ultimate limit (at Vgs = 0, Vds << 0 will simply enhance the channel and conduct majority carriers instead, same effect as the GaN FETs above).
I suspect my relative ignorance is no accident, as forward recovery can vary wildly from part to part, at least when it's poorly or uncontrolled. Suggesting it's not well understood by those in the business even...
Tim
The peak forward recovery voltage with near instantaneous current application is limited by the resistance of the drift region, which is the ON resistance of a HV MOSFET, and this is carefully minimized by design.
What MOSFET package has <100 pH source inductance though? These edges are ridiculously fast.
On second thought ... what single package could ever get to <100 pH except something similar to the EPC MOSFETs in the first place? Wide, short, dozens of bumps.
I suspect my relative ignorance is no accident, as forward recovery can vary wildly from part to part, at least when it's poorly or uncontrolled. Suggesting it's not well understood by those in the business even...
Yes. Eventually whenever you need a spec optimized you need to test, test and test a lot of parts. Either the conventional wisdom is wrong, or the test conditions of the datasheet do not apply, or no one knows.
My application was with voltages >100V and if the lowest possible forward recovery voltage was required then the body-drain diode of a MOSFET was the best available.
Back in 2004 I was part of the design team of a 3kW SMPS for telecoms usage, in particular the PFC boost converter at I think 100kHz. It had a 450VDC buss and we were having all kinds of fun and games keeping the PFC mosfet drain voltage from overshooting too much at full load and low AC in, 160V. No SiC schottkys big enough in those days so we had a normal ultrafast diode and all sorts of inductors and caps and smaller diodes between the main diode and the big electros to look after reverse recovery, so the drain voltage wasn't clamped to buss voltage as well as it could be. Anyway, while looking at the specs of various diodes that we might be able to use in the PFC stage I do remember seeing the datasheet of some particular mosfet at the time that actually specified the forward recovery time of the body diode and it was astonishingly low. I can't actually remember what the device or the figure was. I can't actually add anything to the discussion here, just that I too have seen a body diode forward recovery spec that was outstanding.
-----
Having flashbacks of winding the big variac down and down while watching the peak drain voltage as the isolation tranny and variac hum more and more and the CRT screen of the HP DSO got more and more wavy from the magnetic field of the tranny and variac, hoping the whole thing wouldn't blow up.
doesi, have you been able to modify the FET gate voltage waveform to slow the Vds rise time of the FET turning off, just enough for the buck inductor not to cause grief? I note you have discussed gate drive resistance and where that is located. Is that a balancing act you have characterised (destruction versus additional switching loss)? I recall some controllers/drivers had a specialised gate drive characteristic that suppressed the EMI outcome from the switching transition - have you checked what other controllers have similar capabilities to the MIC2127A ?
I would try adding a Schottky diode right at the controller itself where it is electrically and physically farther from the EPC GaN FET itself and can clamp directly to the controller's Vss or Vdd, whichever you are having trouble with. Or both
I would try adding a Schottky diode right at the controller itself where it is electrically and physically farther from the EPC GaN FET itself and can clamp directly to the controller's Vss or Vdd, whichever you are having trouble with.
Unfortunately there are two reasons why this would not work:
- The whole load current switches between the high side, the diode and the low side FET and therefore these three components should be close. Any added parasitic inductance leads to massive ringing.
- If the diode only clamps the controller, there is a voltage gradient of ~2V between the switchnode and the controller for a short time. Because this connection is also the current path for the high side gate and the threshold voltage is 1,3V, this leads to an unwanted turn on of the high side -> boom
I already thought these scenareos through and did some simulations because the guys from Microchip suggested something simular.