2003A&A...399..223V

Both nulling and subpulse drifting are poorly understood phenomena. We probe their mechanisms by investigating how they interact in PSR B0809+74. We find that the subpulse drift is not aliased but directly reflects the actual motion of the subbeams. The carousel-rotation time must then be over 200s, which is much longer than theoretically predicted. The drift pattern after nulls differs from the normal one, and using the absence of aliasing we determine the underlying changes in the subbeam-carousel geometry. We show that after nulls, the subbeam carousel is smaller, suggesting that we look deeper in the pulsar magnetosphere than we do normally. The many striking similarities with emission at higher frequencies, thought to be emitted lower too, confirm this. The emission-height change as well as the striking increase in carousel-rotation time can be explained by a post-null decrease in the polar gap height. This offers a glimpse of the circumstances needed to make the pulsar turn off so dramatically.