My LM3478-based SMPS isn't working so well. Thoughts?

[robertbaruch]

Vital statistics (with datasheet links):

Input voltage: 8-12VDC
Output voltage: 170VDC
Output current: Up to 150mA

SMPS controller: LM3478
Switch: IRF630
Diode: LQA10T300
Inductor: 22u 7A toroid

My detailed design procedure is here.

Schematic:



I built the circuit on perfboard.

So, I didn't bother with adjusting the pot, and my no-load output was 176.5VDC. Loaded down with 100k (1.8mA) the voltage remained stable. However, at 7.2k load (22.5mA) the voltage dropped to 162VDC, and the circuit was audible. Going all the way to a 1.2k load (about 150mA) caused the output to drop to about 100VDC.

Here's a shot of the gate waveform (top) and the feedback pin (bottom) with the 7.2k load. Feedback seems to be right where it should be, around 1.26V with transients. Gate waveform has a period of 82usec (12.2kHz, explaining the whine). (Sorry, I haven't bothered yet to learn how to get screenshots of the Rigol).



With low loads, the gate waveform seems to be where it's supposed to be (625kHz).



I determined that around 43k load (4.1mA) the voltage dropped very slightly, but the circuit seemed to switch between high frequency and low frequency operation.

What could be causing the voltage drop and incorrect gate frequency at higher loads?