Need help with a switched-mode CCCV module

[Plecto]

Hello. I'm working on a variable constant-current-constant-voltage regulator. Two PWMs are supposed to set the maximum allowable current and maximum allowable voltage and the regulator will output whatever voltage needed to meet those criterias. The schematic below is my first attempt (please disregard the two fets in the bottom right as well as C88). The output voltage of the switching regulator is manipulated by adjusting the voltage in it's feedback path, either directly from 'Output_Voltage_PWM' or through U1D which will raise it's output voltage if the output current goes above the set current threshold. This circuit works as intended as long as the input voltage is kept below ~20V, but at 48V input the regulator get's unstable in all sorts of fashions at different output voltages and more often than not it will cause my power supply to go into current limit by allowing shoot-through (the fets gets really hot). I've read the datasheet of the LM5117 and tried following the examples, but my knowledge when it comes to transfer functions and feedback compensation is limited so I'm kind of in the dark when selecting component values. I can't really follow the formulas in the datasheet completely either as I want a variable output voltage (the formulas often require a specified output voltage). Could someone help me out selecting component values that would lead to a stable regulator regardless of set output voltage? Or perhaps a different approach entirely would be better?

[prasimix]

Take into account that even with properly selected compensation network if you didn't succeed to layout PCB correctly you can expect all types of issues as you mentioned (usually that starts to be visible with increase of input or output voltage or power).

[T3sl4co1l]

Noooo!  Why do it voltage-mode only, when you have access to the internal current control node?

http://www.ti.com/lit/ds/symlink/lm5117.pdf page 12
V(COMP) - 1.2V is compared to the measured/estimated output current.  The pin can sink/source at least 3mA, implying the error amp is a conventional voltage-mode op-amp, and probably shouldn't be overdriven*.  So, we should use a feedback clamping circuit instead, like this:



The error amp and feedback network is the internal error amp, so the discrete circuit only needs to attach to the FB and COMP pins, with everything else as normal in the application circuit.

*This is handy in transconductance type error amps, where the COMP node is loaded with an R+C to ground to set compensation (rather than an R+C feedback to -in, as a voltage-mode amp uses), and V(COMP) min/max can be set with simple diode clamps.  TPS54233 for example can be used this way.

Tim