TPS62120 Efficiency too low


Efficiency ~14v in, 3v and 10mA out expecting 80%, getting 72-73% efficiency

The Zener diodes, D20 & D22 on the input and D3 & D4 on the output, are voltage limiters.   I was thinking I could remove those to see if they have any effect on efficiency.  They should not be conducting during my tests.    

My load requirements are about 36 mW,    3V and  12 mA.

I have observed efficiencies of 70 to 74% in my testing.   

I have increased the input capacitor to 10uF, with little change. 

For the inductor, L2, I’m presently using  SUMIDA      CDRH6D28NP-220    

I have tried values in the same SUMIDA series  from 10uH up to 47uH  and C63 values from 4.7uF up to 22uF, with very little change.  

I am testing this on my PC board, which has four layers.  There is a GROUND plane layer adjacent to these components and the VDD, 3.0 Volt output, on a POWER plane, the next  layer. 


Based on the datasheet curves, he should be getting higher efficiency.  But his difference is only ~5mW, which is not much power to lose.  There are several differences between his circuit and the EVM that the datasheet graphs were taken with that could cause this.

Yes, I would recommend he use a 10uF, 25V, X5R input cap as recommended by the datasheet.  I would also recommend that he change his output cap, C63, to a ceramic.  The ceramic would be cheaper, smaller, and more efficient.  The ESR of the tantalum causes extra output voltage ripple (especially in power save mode) and can reduce the efficiency.

I would also remove the zener's as they can be leaky.  They are partially on even below their clamp voltage.  This can definitely cause a few points of efficiency loss.

At this light load, the current through his FB divider will also be significant.  He could increase his resistances to increase his efficiency.  Limit R82 to 360k.

This app note will help him accurately measure light load efficiency:

His inductor is very different from the one used to take the datasheet data.  Though it has lower DCR, this is of less importance than its core losses at such a light load.  Being bigger, my guess is that its core losses are higher.

He can also order the EVM and measure its efficiency to see if his different components or layout have an impact.


Thanks for your help!  

I made use of a demo board to test performance and then used the inductor from that demo in my circuit.  Turns out the inductor is a big part of the inefficiency of my unit.   Also, I found that there was a significant impact from the Zener diodes, D3 and D4, on the output.   Between those two things I was able to move the efficiency up to about 83%, almost 10% improvement.