I am experiencing some unusual regulation issue in the LM3102 and am seeking ideas or if anyone else has experienced a similar effect or has any explanations.
At light load the output voltage is high and has high ripple. Now this is partly expected due to the constant on time topology. However what is strange is once switching commences it continues even if the FB pin is above 0.8V. It continues to switch until FB hits 0.92V. Then it stops switching and of course the output voltage will slowly drop. When the FB drops below 0.8 it starts up again. This last point appear correct. I cannot work out why once running it continues with FB>0.8V. So the bursts are longer and the ripple much larger than expected.
The other unusual effect is under load. At approximately 200mA the output voltage drops abruptly to slightly below correct set point. From 200-800m the output voltage is very flat with a very slight down slope – it appears like correct regulation but a little low in voltage. Above 800mA a gradual increase in output voltage is noted. Maximum occurs at around 1.8A where the voltage is about 2% high. Beyond this current it decreases to the set point at about 2.5A. Roughly follows a half sin curve.
Vin=24V
Vout=10V
Fsw= approx 500-600k
Ron=180k
L=22uH
Layout should be good. Feedback path is extremely short – maybe 10mm including resistor lengths. Solid ground plane under design. No signs of instability - switching waveform is periodic and stable for all loads.
I have tried numerous circuit mods with no major changes in this behavior: changes to L value, increases to input/output caps, change of set output voltage, Ron modified to 100k, feedforward capacitor value changes. Even tried small caps from FB to GND and no feedforward capacitor to remove any possible switching transients on the FB pin.