Buzzing SMPS under load

[cthree]

I built up a 5V SMPS using a LM22676-5.0 and other components. The design is from WEBBENCH. I made a PCB and assembled it but it's behaving badly and I would appreciate a second look. I've attached schematics and scope traces.

I'm using this to power a BeagleBone SBC from a 24VAC source. The output (X4) connects to the BBB's 5V DC jack. Unloaded and with a load of up to about 200mA it appears to work great but as soon as the load increases to above 200mA it starts to buzz (120Hz I presume, it is quite audible) and it looses regulation. LED1 provides a nominal minimum load of 18mA.

The three scope traces: first is no load. Second is < 200mA load. Third is > 200mA load. CH2 (blue) is VIN, CH1 (yellow) is VOUT. The dip in VOUT seems to grow and shrink with load. This SMPS is supposed to be able to output 5V@3A but clearly thats not happening. Does anyone have any sage advice as to why this thing might be dropping out of regulation like it is? Or likely suspects?

[owiecc]

You have 10V input voltage ripple at no load. You have no input caps!

[jmaja]

Having just ~10 uF at the input would likely be OK, IF you had a DC source. But you have a rectified AC source at 60 Hz, which means it will have zero output at 120 Hz. Thus your input capacitor needs to supply the ~10 ms in between the peaks. Your current capacitance only somewhat works up to 200 mA and even at lower currents your huge input ripple probably causes quite much output ripple.

You need at least an order of magnitude more input capacitance. The actual value depends on the maximum output current and how much input and output ripple you can tolerate.

[ResR]

I would use at least 47uF or 100uF 400V capacitor for this purpose, yeah the supply may oscillate little bit longer after disconnecting, but it's much more stable DC that way. Little bit redo the PCB to fit the much larger capacitor is inevitable, so why not to go for keeps.
Cheers.

[owiecc]

This  circuit is using a 24Vac input voltage. A 1000uF/50V cap would be better.

[cthree]

 

Thanks everyone!

I should have seen it but it got lost in the cracks as things do. I added an additional 100uF and that lowered the input ripple down to ~200mV at 450mA. It's just a first prototype right now and I will need to respin the PCB due to a physical interference problem anyway. Just describing the problem to post it here caused me to see it. Thanks for indulging.