We have a design with a number of TPS54623 buck converters, all running from a rail of 12.6 V. We see a noisy pulse width on the supply with the lowest output voltage, 1.2V. A demo board with the same circuit and output voltage produces the same problem. The situation is made worse by higher input or lower output voltages; probably, more specifically, the ratio between the two. Matters are also made worse by decreasing the output current or decreasing the output inductor value. If conditions are pushed far enough, I see the pulse width break into alternating wide and narrow pulses. I believe that what I'm seeing is referred to as sub-harmonic oscillation, found on current-mode devices. Is there an easy fix for this condition? There does seem to be a weak interaction with the compensation network, which may only be indirect. It would seem that the 47 pF that comes with the demo board actually worsens the situation.