How to debug low-efficiency of a switching converter?

I am using analog devices' ADP1613 to boost a lipo cell ~4v to 12V. I have it layed out on a PCB with mostly 0603 components

According to the datasheet, this should give me around 90% efficiency @ 200mA output. Unfortunately I am seeing something more like 30% efficiency. I am pretty certain that the ESR values of my passives arelow enough to theoretically give me close to 90%.


How do I start to troubleshoot this setup to figure out why I am getting such low efficiency?

Low efficiency can be caused by several things. Which switching frequency did you use?, this is very critical since the losses from the drivers are proportional to this frequency. The datasheet states that it can go from 650KHz up to 1.3MHz. Check if your chosen frequency is within this range.

The part is current mode controlled, meaning that it senses the output current. Then the resistor value matters since you will have power dissipation in those sensing resistors. This can degrade the efficiency severely. This also is affected by the traces of your PCB, long traces means more power dissipated on those traces. Avoid long and narrow lines for the VDD, VSS, VIN, VOUT.

Also you need to make sure that your circuit is working properly under normal operation before attempting to measure the efficiency. In other words, check if your controller is stable. try to replicate the waveforms provided by the datasheet, using a load change from Iload=0 to 200mA and check if your Vout has ripple or if it is behaving as expected. Also double check if you are measuring the efficiency in the same way the datasheet reports (n=POUT/PIN).

Without knowing any other info, I would put the issue as the inductor selection. Look at the inductance versus current spec on the inductor you are using as well as inductance versus frequency. You may also have too much Rs (wire resistance) within inductor. All of this relates to the physical size of the inductor.

If your inductor is a 0603 sized component I can guaranty that is the issue.