excessive undershoot, LM2676 voltage mode controller

[zilog]

I have a circuit that uses a LM2676 voltage mode smps chip to generate a 5V logic supply voltage. The problem is that this voltage is most of the time unloaded, or loaded at only 10mA. When a certain part of the circuit goes active, it pulls current in slow pulse-trains of 100mA added, which causes excessive undershoot of this 5V line (the voltage drops ~100mV for 2ms before returning to regulation).

One solution I have found is to increase the minimum load to 100mA by use of resistance, but I cannot afford this constant power drain from both battery and thermal point of view.

Does anyone know of a way of getting rid of this behaviour short of replacing the LM circuit?

[Cristi]

what's the value of the input voltage and the inductor ? for such low current requirement the regulator works in discontinuous mode, and you might have to increase the inductor value.

[zilog]

Originally Posted by Cristi

what's the value of the input voltage and the inductor ? for such low current requirement the regulator works in discontinuous mode, and you might have to increase the inductor value.

12V in, 5V out, approximately 10-500mA load. The inductor was 33uH, but some calculations showed that I need 41mA min load at 150uH inductor. Empirical testing also shows this to becorrect if I want the inductor never to run dry.

Thanks for hinting me in the right direction.

[Cristi]

u mention before that your current draw is zero or 10mA. for this value, the inductor must be much larger than 150uH. considering 50mA, a 1mH inductor is resonable. also, why don't u use some better regulator with much lower quiescent current which will fit better for battery powered applications ?

[zilog]

Originally Posted by Cristi

u mention before that your current draw is zero or 10mA. for this value, the inductor must be much larger than 150uH. considering 50mA, a 1mH inductor is resonable. also, why don't u use some better regulator with much lower quiescent current which will fit better for battery powered applications ?

Measurements show that the problem goes away at 40mA and 150uH. The reason for not redesigning is that a huge stock of PCBs are already produced and populated, and I am called in to help out another person's design. Changing component values in a few places is all that is economically viable.