Hi All,

I am struggling to understand how can I stabilize a hysteretic buck regulator when I add an extra loop to improve its steady state accuracy. The loop simply senses the feedback voltage, compares it against a reference and sets up the control signal for the comparator. I have seen this architecture being used successfully but I cannot find any stability analysis of the whole architecture. I can post link/attachments to show what I mean. But I guess people who has worked on this already got it.

The problem is a classical hysteretic buck is actually not stable in terms of small-signal linear control system point of view. So the way I see it as the output actually oscillates at its frequency of oscillation, but if the attenuation is large enough to reduce the ripple size -- it is stable for practical purpose as long it is oscillating around the target output voltage with small enough ripple. Issue is it sometimes deviates from its target output value. To improve the DC performance if I attempt to add an extra feedback loop it looks ugly! I have managed to make the final output voltage look nice enough i.e. the ripple is small but it is not good enough I believe, because I see the output of the error amp (the high gain amp in the extra loop that compares the feedback voltage against a reference) is swinging rail to rail!! I will list 23 observations related to this:

1. if I make the bandwidth of the error-amp infinite (using a model) I get a nice output voltage with small ripple, but the output of the error-amp is still swinging rail to rail.

2. if I use an error amp with a small bandwidth (i.e. a low freq pole) it just doesn't work!

3. If I use an error amp with larger bandwidth (a larger pole) I see the performance improving but still the error amp output swinging rail-to-rail. The output voltage ripple is also not negligible unlike 1, thereby exposing the oscillatory behaviour of the hysteretic buck!

So the question is what I need to with this additional loop? How can I make this loop stable so that the output of the error amp is stable, and let the final output voltage "essentially oscillate" with a small ripple around the target DC voltage?

Thanks,
Rajdeep