Switching Loss: High Side MOSFET Overheating in Half Bridge Converter

Thread Starter

Randriad

Joined May 10, 2022
8
Hi everyone,
It's my first post here, I hope I'm in the correct forum.

New in power electronics, I'm currently trying to design Half Bridge DC to DC Converter with 10A output on 20kHz frequency (50us period).
As of right now, here's what my circuit looks like:
1652265648899.png
High Side

1652265673633.png
Low Side

Shown NMOSFET are true, but I'm using KST42 for NPN and KST92 for PNP transistors. I'm using 50nF ceramic capacitor and 1N4148 to achieve 1.8us turn on time, without it the turn on time would be about 10us.

The circuit works with 90% efficiency, but the high side MOSFET gets really hot, around 85C on 50% on 10A output. I don't think this is normal since I tried with P-channel with 20mohm Rds(on), achieving similar temperature rise.

The oscilloscope showed huge ringing (in High Side Gate, High Side Source, and output) when turning on the High Side MOSFET, lasting for about 6us (my pulse width is 25us). Smaller ringing in turn off also exists in high side gate, source, and Vgs. I see this as the main cause why the high side MOSFET is heating up.

The goal is to decrease the temperature rise by looking at the circuit design, am I correct to address the main problem as switching loss? I'm open to any comments and suggestion to improve the design! Oh! I'm trying to design my circuits without the gate driver IC, nevertheless, I'm ordering a few to test if the ringing and MOSFET temperature would decrease using the IC.
Thank you in advance!
 

DickCappels

Joined Aug 21, 2008
10,138
There will always be a large voltage drop because you high side switch is a voltage follower -it be better if you drive it with a transformer or have a floating power source for the gate drive.

Alternatively change it to a P-channel MOSFET with the source connected to +24 volts.

You managed to get pretty good turn-on time for that relatively slow KSP42. You might consider TS 2N5550 or 2N5551.
 

DickCappels

Joined Aug 21, 2008
10,138
Good catch @AlbertHall!

One other thing - for the ringing on the gate put a 20 Ω resistor in series with each gate, near the MOSFETs. That should be enough to significantly damp the ringing and thereby further reduce power dissipation.
 

Thread Starter

Randriad

Joined May 10, 2022
8
Thanks for the quick reply everyone!
I tried several things yesterday to make sure of things:
1. I used a combination of series RC Snubber values parallel to the high-side switch. Not resulting in any decrease of high-side MOSFET temperature rise. Also, the snubber was hot, around 160C and I can't reach 10A with my usual settings.
2. I tried lowering my switching frequency, and it does decreasing the temperature rise. Soooo, I'm inclined to assume that the problem is switching loss.


There will always be a large voltage drop because you high side switch is a voltage follower -it be better if you drive it with a transformer or have a floating power source for the gate drive.
I'm sorry, can you provide any keyword to what you meant so I can look it up on google? Also, is the large voltage drop one of the losses I should be aware of, other than conduction and switching loss? Thanks!

Alternatively change it to a P-channel MOSFET with the source connected to +24 volts.
So, a PMOSFET as high side switch and a NMOSFET as low side switch, correct? I haven't tried it before, but with a higher Rds(on), wouldn't it have a higher conduction loss? Or would it solve the ringing?

You managed to get pretty good turn-on time for that relatively slow KSP42. You might consider TS 2N5550 or 2N5551.
Thanks! An old-school trick I found on the internet, could have used a MOSFET but I'm happy with the results. I might try the transistors you just mentioned.

Your diagram shows the gate of Q1 connected to the bases of the driver transistors rather than their emitters.
Is this a drafting error or are they actually conncted like that?
Oops! It's a drafting mistake! Thanks for pointing that out. Will edit the schematic soon!

This is fine for a learning exercise,
but no discreet-parts Circuit can beat a proper Gate-Driver-Chip in simplicity, cost, or performance.
Thank you for the comments! I'm currently ordering IR2104 right now, will try it soon. My seniors got me to try with discreet parts, reflecting on the past IC shortages.
 
Last edited:

BobTPH

Joined Jun 5, 2013
8,765
Imagine your high side switch is fully on. What is the voltage on the source and on the gate? Do you see the problem?

Bob
 

Thread Starter

Randriad

Joined May 10, 2022
8
Imagine your high side switch is fully on. What is the voltage on the source and on the gate? Do you see the problem?

Bob
If it's fully on, the gate would have 36V, and the source would have 24V, assuming the duty cycle is small enough to let it reach 36V (50% duty cycle is more than enough by my trials). Is there any problem there?

I think your drafting error will help greatly. I don't know if you built it to the wrong schematic.
I actually built the circuit with the high-side mosfet gate connected to the emitters. Sorry for posting the wrong schematic :(
 

Ian0

Joined Aug 7, 2020
9,618
I don't see why that shouldn't work. Do you have the bottom of C1 connected to the right place?
When the low-side transistor is on C1 will charge to 12V because the lower end is connected to the negative supply via the Low-side MOSFET
When the low-side transistor is off, the top of C1 will rise to 24V+12V = 36V because the bottom of C1 is connected to the bridge output - enough to switch on the top MOSFET.

By the way, your Baker Clamp (D2) should be a schottky diode for it to work.

[EDIT] swap Q3 for something that will stand a bit more voltage - at 36V, you're a bit close to the 40V Vceo limit for a 2N3904.
[EDIT] You need a resistor in series with D1, because otherwise, when the low-side MOSFET switches on it connects C1 directly (and suddenly) across the 12V supply and a large amount of current will flow, disturbing the 12V supply. That (and the lack of gate resistors) is probably the cause of the ringing.
[Yet another EDIT] Does your drive waveform have sufficient dead-time?
 
Last edited:

Thread Starter

Randriad

Joined May 10, 2022
8
One other thing - for the ringing on the gate put a 20 Ω resistor in series with each gate, near the MOSFETs. That should be enough to significantly damp the ringing and thereby further reduce power dissipation.
Thanks for the suggestion @DickCappels! I admit that I removed them from my circuit to have faster turn on & turn off time. I tried to use 22 Ω resistor on both gate just as you suggested, and it does not damp the ringing at all. I tried larger values, and I see that the ringing is decreasing by adding 100 Ω on Q6 gate (70V peak to 52V peak, 6us to 4us), also reducing the temperature rise to 75C. Adding more gate resistor on Q1 however, does not reduce the ringing, but increasing the turn on/off time and the temperature rise.
Why is it that adding gate resistor on Q6, resulting in ringing dampening on Q1?

I don't see why that shouldn't work.
Hi @Ian0! The circuit works, but with the high side MOSFET heating to 85C due to switching loss. I'm trying to find ways to reduce the ringing and the temperature rise.

[EDIT] You need a resistor in series with D1, because otherwise, when the low-side MOSFET switches on it connects C1 directly (and suddenly) across the 12V supply and a large amount of current will flow, disturbing the 12V supply. That (and the lack of gate resistors) is probably the cause of the ringing.
Oh! I haven't thought about this. I'll try several values for it. Thank you!

By the way, your Baker Clamp (D2) should be a schottky diode for it to work.
I'm aware of that. I don't have a schottky around right now so I used 1N4148 and it did reduce the switching time. I'll try with schottky when I have one available.

[EDIT] swap Q3 for something that will stand a bit more voltage - at 36V, you're a bit close to the 40V Vceo limit for a 2N3904.
I'm using KST42 for NPN and KST92 for PNP in my actual circuit, I wrote this on the post. Again, sorry for not posting the correct schematic :(.

Does your drive waveform have sufficient dead-time?
I do. I made sure that the Vg of Q1 and Q6 never meet halfway.
 

Ian0

Joined Aug 7, 2020
9,618
I do. I made sure that the Vg of Q1 and Q6 never meet halfway.
[/QUOTE]
Try rather more dead-time, at half-way (6V) they will both be almost full on. Try it so that one doesn't set off until the other has arrived (if you see what I mean!)
 

Thread Starter

Randriad

Joined May 10, 2022
8
Try rather more dead-time, at half-way (6V) they will both be almost full on. Try it so that one doesn't set off until the other has arrived (if you see what I mean!)
Truly sorry for not being clear, I get what you mean. I meant that I set enough dead time (3us, is it too large?) so that the rising and falling of Q1 and Q6 never meet.
 

Ian0

Joined Aug 7, 2020
9,618
Truly sorry for not being clear, I get what you mean. I meant that I set enough dead time (3us, is it too large?) so that the rising and falling of Q1 and Q6 never meet.
If you achieve 1.8us risetime, it should be just enough. See if 5us makes any difference, or see if you can improve the risetime.
(Fall time is more important than risetime)
 

ronsimpson

Joined Oct 7, 2019
2,954
IRF3205 8mohm, which I will call 10mohm to simplify the math. current = 10A 100% dutycycle. watts=resistance * current^2
watts=1.
I know this is not how you are using the MOSFET but ..... just saying the part needs a heat sink, even if there was no switching losses.
Are you certain there is no point in time when both transistors are on?
 

DickCappels

Joined Aug 21, 2008
10,138
I don't see why that shouldn't work. Do you have the bottom of C1 connected to the right place?
When the low-side transistor is on C1 will charge to 12V because the lower end is connected to the negative supply via the Low-side MOSFET
When the low-side transistor is off, the top of C1 will rise to 24V+12V = 36V because the bottom of C1 is connected to the bridge output - enough to switch on the top MOSFET.

By the way, your Baker Clamp (D2) should be a schottky diode for it to work.
1652362715882.png
I should have noticed C1. Definitely a problem if it is really connected that way.

If you use a Schottky in your Baker clamp it might be faster than with a silicon diode. But if you wanted it faster you might want to use a faster transistor or an integrated circuit driver.

You can search for isolated high side switch and floating high side switch driver
1652363306608.png
Direct Coupled from ground-referenced gate drive for a high side switch.
1652363428666.png
Capacitively coupled from ground-referenced gate drive for a high side switch.

1652363657363.png
Transformer from ground-referenced gate drive for a high side switch.


1652363762321.png

Using a dual driver with a floating gate drive circuit for the high side switch (actually I think this one uses two floating gate drivers and one happens to be grounded.

With a P-channel switch or an N-channel switch and some extra voltage the switch will saturate and if you pick the right MOSFET the voltage drop will be very low. You would lose the several volts (and the corresponding watts) by using an N-channel MOSFET as a voltage follower.

Here is one place in which to look.
https://www.analog.com/en/parametricsearch/11395#/
 

Attachments

BobTPH

Joined Jun 5, 2013
8,765
If it's fully on, the gate would have 36V, and the source would have 24V, assuming the duty cycle is small enough to let it reach 36V (50% duty cycle is more than enough by my trials). Is there any problem there?
Sorry, I totally missed that the high side driver was a bootstrap circuit.

Bob
 

Thread Starter

Randriad

Joined May 10, 2022
8
If you achieve 1.8us risetime, it should be just enough. See if 5us makes any difference, or see if you can improve the risetime.
(Fall time is more important than risetime)
I'm having a hard time improving rise and fall time, right now. I'll try using a schottky diode and using a faster transistor as @DickCappels suggested. Thank you!

I know this is not how you are using the MOSFET but ..... just saying the part needs a heat sink, even if there was no switching losses.
Are you certain there is no point in time when both transistors are on?
I'm certain. Well the low side transistor is floating(?) at 1V when it should be off, and touched 0V after the high side transistor is totally on. 1V is under the Vgs(th), so I thought that it won't make it conduct and make it hot. And the Low Side Switch is fine, around 40 C, only the High Side Switch is heating up.

I should have noticed C1. Definitely a problem if it is really connected that way.
If it was a propper bootstrap circuit the transistor would not be over heating.
Sorry, I might missed something in my C1 connections (frequently do). Do you mind pointing out what the problem might be?

You can search for isolated high side switch and floating high side switch driver
View attachment 267005
Direct Coupled from ground-referenced gate drive for a high side switch.
View attachment 267006
Capacitively coupled from ground-referenced gate drive for a high side switch.
View attachment 267008
Transformer from ground-referenced gate drive for a high side switch.
Thanks for this information! I'll learn more about this.

BTW @DickCappels, I added gate resistors in high side and low side switch as you suggested. And I found that adding gate resistors in low side switch DOES reduce the ringing in the high side switch, and adding resistor in high side switch does not. Any idea how it might occur? Thanks!

My gate driver IR2104 just arrived today and I'll try it out. I want to check if it will reduce the ringing and the temperature rise.
 
Top