LM3433 circuit failure, MOSFETs blown away

It's hard to say without more information. But I can say that driving 28A with a single phase synchronous converter requires very careful layout. Most often if devices are damaged it is due to layout and some node(s) either get over-voltage or negative voltage events that cause them to fail. I recommend to anybody to look closely at the layout of the EVM and follow it as closely as possible. We took multiple precautions to contain noise and generated fields. That being said, if you would like to attach layout pictures along with a BOM (at least what inductor and FETs you are using) and a schematic I would be happy to look them over for you. If it looks like it may be layout related I could pass along suggestions.

Hi
We are using both Q3 and Q4 for the LS Side and Q1 , Q2 for HS side. The MOSFET is common i.e. SiE808DF which has a Vds of 20V and Rds of 1.6mOhm. Most of the times, Q3 blows off although we have put all the components as per AN1937 design and the PCB is also an exact copy of the TI design. Only difference is that the layers are 3oz instead of 4oz as mentioned in the design.

Hopefully you can help us understand what is the reason for this behaviour and how can we rectify it

Thanks in Advance

The only other times I have seen a low side FET blow up is either the load is disconnected during operation at high currents and no discharge diode is used (D3 on the schematic) or more commonly the ADJ is floating. I have seen the second one a couple of times and in both cases the ADJ pin was open. Once the shunt used on the EVM to connect ADJ to POT (J2, pins 2 and 3) was missing, the other had the potentiometer go bad. In either case if ADJ is floating it can start up with much higher current than expected and it can damage the FETs. Check to be sure ADJ is not floating and it is a known voltage.

The FET blows off during operation. We have the ADJ pin tied to POT, have the LED attached to the output and D3 in place. After the shorting of the MOSFET, the output current becomes 10A approx. The LED is on but nothing happens on adjusting the Pot.

Hmmm. Have you tried putting a thermal camera on it to check if it is a thermal issue? The EVM was designed basically for evaluation at room temp due to the high power dissipation of the circuit. That is why we used many vias and thick copper, but it's still borderline at higher currents. That is part of the reason we used the FETs we did and added the fan footprint. The FETs can have heatsinks glued to them and the fan can help prevent the inductor from heating up to much (which could cause loss of inductance and higher peak currents possibly causing runaway until it saturates).
That is the first thing I would check, especially with the thinner copper planes you are using. Otherwise I would need to know more information such as:
Input voltageOutput voltageInductor manufacturer/part number
Are you using the resistor/ferrite bead combo?

We haven't checked the temperature of the MOSFET yet. The issue is that it doesn't happen all the time. A circuit could be working and at some point of time the whole thing pops off. I can stick heatsinks on the MOSFET and check how many from a lot fail. The inductor is rated for 48A while we run it at 17A or 27A. The issues are the same for both the circuits

The details of the circuit are as follows:
Input Voltage: -12V
Output Voltage: 3.6V
All components as per AN1937 including Inductor (Coilcraft SER2915H-332KL/ SER2915L-332KL) and resistor-ferrite bead combo

Another issue is that the SiE808 are a little difficult to come by these days. Would it be OK to put SiE868 or SiE882 in its place. Both of them have higher Vds but the Rds & Rg are also slightly higher. Would you also know how one can calculate for a MOSFET for such a circuit?

Either of the FETs you mentioned would be fine for substitutes. But using the 868 you would have to be even more careful with thermals since the Rdson is higher.

In either case we should rule out thermals as the cause. Do you have access to a thermal camera? The EVM really was designed for basic demonstration purposes. It is built without the fan and heatsinks simply because that would make them cost prohibitive, so they get pretty hot. It is possible one circuit is more prone to thermal runaway than another. Also check the diode D3. I've seen where that can get very hot being directly next to the sense resistors and as a schottky it can fail short if it gets too hot which can damage other components.