SMPS feedback loop phase and gain margin

[Jamie Morken]

Hi,

What is a good way to increase the stability of a SMPS feedback loop?
I have a SMPS that uses this algorithm:

if inductor current > current_limit, pulse off

Duty cycle = ((desired voltage - current voltage) * gain) - (Inductor
current * gain2)

I am not sure how to put in the analog or digital adjustments for making
sure the feedback loop is stable.

I am reading up on these pages but am not sure how to implement this on
a PCB, either analog or in an FPGA:

http://en.wikipedia.org/wiki/Phase_margin
http://en.wikipedia.org/wiki/Bode_plot#Gain_margin_and_phase_margin

cheers,
Jamie

 

[MooseFET]

Hi,

What is a good way to increase the stability of a SMPS feedback loop?
I have a SMPS that uses this algorithm:

if inductor current > current_limit, pulse off

Duty cycle = ((desired voltage - current voltage) * gain) - (Inductor
current * gain2)

You didn't say what sort of DC-DC this is. What you are using as a
control equation makes me think it is a simple bucker topology. The
equation you have shown won't yeld a very good load regulation.

If this is a design with a continuous inductor current, think of the
PWM's signal as having been magically low pass filtered so that you
cant see the switching and only see (Duty * Vin)


(Duty * Vin) ----)))))-----+------- Load
!
---
---C1
!
GND

As you can see, from (Duty * Vin) to the load we have a LC filter. If
you assume that your equation for doing the control is feeding into a
linear system, you have a fairly simple way to make a Bode plot of the
system. The issue of the delay in the modulator is not taken into
consideration in this plot.

To deal with the modulator, you will usually find that assuming a
simple linear delay of 1/2 of a switching cycle will get you into the
right ball park.

Basically, you want the duty cycle to be controlled by a PID
controller running from the error signal. Your term where you brought
the inductor current into the equation is trying to do the "D" part by
saying that the current in the inductor tells you the rate of change
in the capacitor (C1) voltage.

Using the inductor current in this way doesn't lead to very good
performance because you don't get to bring the high frequencies of the
load current into the "D" term. This greatly reduces your ability to
track a sudden changes in the load.