Switched Mode Power Supply Overshoot problem

A copy of the schematic would be very helpful. But increasing the capacitance of the output filter may help. The best cure is probably in the feedback divider and filter section of the circuit.

Good luck with that one. It is a highly complex, read I've never really looked very hard at that stuff much before, controller. Working out what it thinks its transfer function might be could be hard. It's probably some wierd thing that is not suited to 'normal' linear analysis.

The dirty sheet does mention possible instabilities due to the capacitor ESR zero being too high. Voltage feedback is a simple divider and there is no feedback available to tailor the response of the internal error amplifier if there is one in there.

That might suggest that the overall voltage loop including the capacitor is second order up to the ESR zero where it becomes first order and crosses over at some higher frequency. If the ESR zero is too high then is crosses over second order and is unstable...... BTW this is speculation.

Anyway, one possibility is that with such a large VIN/VOUT differential you can rapidly ramp up your filter inductor current to some excessive level and when the output voltage reaches the set level the controller switches things off BUT you have still put energy into the inductor and that gets delivered to the output causing the overvoltage.

Cures for that would be reducing the absolute current limit level and/or increasing the inductor value so it doesn't store too much excess energy.

If I'm right about the loop behaviour (pigs might fly) then you might split the feedback resistor from the output node into two in series with a ten to one ratio and add a capacitor across the larger resistor..... and then experiment with the value of the capacitor. That has two effects.

The first is it introduces a low frequency zero at the input making the loop first order up to the frequency where the smaller resistor cancels that zero. This is good because is makes the loop less conditionally stable.

The second is that such a response results in a more damped behaviour and, in particular, a less snappy start up. Rather than starting up balls to the wall it should bring the output voltage up 'gradually'.

Further than that I wouldn't like to guess. All of the above is speculation based on 'normal' controllers and should but might not apply to this one.

Good Luck

DNA

Certainly is a convoluted architecture.

This thing claims to have a built-in slow start. With a constant on-time topology, this is frequency modulated. Are you starting up by using the enable line, or just whanging the input with 12V?

Without a well established 5V bias supply, you can't count on the slow start working. This is also the supply that is needed to switch from

At lower output voltage settings, the app circuit has the output terminal directly returned to the FB input, preventing the conventional use of output feed forward.

Are you using the current sensing terminals?

Are you using the application circuit layout?

RL

Hi all, Thanks for all the help! Just wanted to inform you that the problem was solved, with help from MAXIM support: The current-limit setting in my design was wrong: I supply 4A max, and my limit was set to 14A, so the initial current was too much for the controller. the fix was to set the current limit threshold via divider resisors, and change the low side mosfet to one with higher Rds resistance, as per the datasheet. Thats all, thanks again for all the help! Harp.