Your schematic is so beautiful that I have missed several "bugs" but in this case, I missed D1, which makes what I thought was a problem -not a problem. I would worry about the repetitive peak current through D1, being a small signal diode, but I still worry more that your source can only go about 11 volts while the drain is at 24 volts, meaning that your source follower switch will be less than 50% efficient. Much less.

To repeat in paraphrase, it is ok to use a source follower for the high side switch if you can provide gate drive that is a sufficiently higher voltage than the drain.

Do you have an oscilloscope?

 

View attachment 267089
Your schematic is so beautiful that I have missed several "bugs" but in this case, I missed D1, which makes what I thought was a problem -not a problem. I would worry about the repetitive peak current through D1, being a small signal diode, but I still worry more that your source can only go about 11 volts while the drain is at 24 volts, meaning that your source follower switch will be less than 50% efficient. Much less.

To repeat in paraphrase, it is ok to use a source follower for the high side switch if you can provide gate drive that is a sufficiently higher voltage than the drain.

Do you have an oscilloscope?

It's a boostrap circuit! That's what C1 does (That's why I was checking to see if it really was connected as per the diagram). The source will get to 24V because C1 charges to 12V when the lowside switch is on, then pulls the gate to +36V when Q3 turns off.

 

Yes, you are right. I noticed that when I finally spotted the diode. If the diode weren't there the source of the high-side switch would be driving a 1 uf capacitor.

I think the maximum voltage on the gate is about +12V - the two diode drops because the capacitor charges from +12 - 2 diode drops to ground (not +24V to -12V as a casual glance might suggest. The low side switch switches to ground.

If @Randriad wants to continue to use a bootstraped source follower circuit he should modify it to pull higher than +24 volts so the MOSFET can saturate but not put too high of a Gate-Source voltage into the MOSFET. Or he can buy a floating isolated gate driver already made.

 

I think the maximum voltage on the gate is about +12V - the two diode drops because the capacitor charges from +12 - 2 diode drops to ground (not +24V to -12V as a casual glance might suggest. The low side switch switches to ground.

If @Randriad wants to continue to use a bootstraped source follower circuit he should modify it to pull higher than +24 volts so the MOSFET can saturate but not put too high of a Gate-Source voltage into the MOSFET. Or he can buy a floating isolated gate driver already made.

Sorry, but from the oscilloscope, I can see that the voltage on the gate (Vg) is exactly +36V, which is 24V from source + 12V from bootstrap, so the high side MOSFET is already saturated with +12V Vgs. Or is there something else that you meant?

Yesterday, I removed D2 and D3 and I got faster gate voltage rise and fall time (but ringing still exists and the temperature rise is still large), I could have sworn that I got faster rise and fall time on Q3 and Q4 when I added D2 and D3, but I guess it's because I haven't added the MOSFET yet by then. Sorry! I'm ordering schottky diodes to check if I can get faster rise and fall.

I also tried using an IR2104, unfortunately I don't get enough dead time using my IRF3205, the IR2104 also have slower gate voltage rise & fall time compared to my last discreet part circuit. I'm trying to add more dead time using the shut down pin, but I'm struggling with it. From what I've collected, if the IR2104 (with longer rise & fall time) gives lower temperature rise, then the ringing is what causing my MOSFETs heating up in my discreet circuit, or is there any other factor I should look into?

 

Do you have a load connected? If so what is it?
Try it at a lower frequency and see what happens. If the temperature rise reduces, then it is switching-related. If it stays the same, then it is conduction-related.

 

Do you have a load connected? If so what is it?
Try it at a lower frequency and see what happens. If the temperature rise reduces, then it is switching-related. If it stays the same, then it is conduction-related.

I have a dummy resistance load to test the circuit. I tried to run it using lower frequency and it does lower the temperature rise, therefore I'm inclined to say that it's switching loss, my guess is from the gate ringing.

And also, my Vgs rises in about 3us (0V until 12V) and falls in about 1.2us (12V to 0V). The rise from 0V to 4V is about 700ns with a plateu about 1us, then rise again to 12V for the rest. The fall is similar but a bit faster, 12V to 4V in about 200ns, plateu in 4V for 200ns, and then 0V in 350ns.
Is it regarded slow for driving 10A in 20kHz? Since it's switching loss, I think it's either the ringing or the slow Vgs rise and fall.