We put an LM25017 into a product prototype for evaluation. The equivalent Webench result predicts 3.34mV of Voutp-p. We see at ~900kHz about 200-300mV for Voutp-p. Slowing the scope down, shows the output moving 200mV at 60Hz (line sync'd) for approximately 500mV Voutp-p. The horrible output looks like a "hairy sawtooth". The PCB layout is extremely close to the suggested layout in the datasheet (especially the ground layout). We have ohm-ed out, tested, and reloaded this PCB several times to ensure all components and solder joints are good.
I am happy to share Webench design with anyone interested in helping out. For completeness, I'll share here: VinMin 29V, VinMax 34V, DC, Vout 21.5V, Iout 0.5A, Ta 30C, fsw 884kHz, Low O/P Ripple Solution (type 3). On the PCB we have: Cin 1.00uF, Ruv2 100k, Ruv28.2k (for 14.2V off/16.2V on), Ron 270K, Cb 10nF, Cbyp 1.00uF, L1 100uH, Rr 100k, Cac 10nF, Cr 470nF, Rfbt 10K, Rfbb 680, Cout 1uF + 4.7uF ceramics in parallel.
The board remains unloaded except for the LM25017 supply and is being tested with a resistive load connected to test points. The input is a 60Hz half wave rectified DC with a few thousand uF of low ESR capacitance just like will be provided in the customer system (i.e. I cannot change this). Any suggestions will be appreciated.
Thanks,
Kyle