SEPIC Inductor losses

[ConnorGames]

I just finished assimbling a SEPIC regulator based off of the lt1615 to find that it had poor efficiency, around 65-70%. It is the appnote "2-cell Lion to 3.3v SEPIC", which is 72% efficient in the LT-provided ltspice sim.I have found that changing the inductor in the simulation makes a big difference, and seems to depend mostly on "parallel resistance" as long as DCR is kept in-check, with efficiencies ranging form 70% to 83% depending on the inductor. I cannot find "paralell resistance" on any inductor datasheets, and am currently using the LBC3225T100KR. Do you think this is the source of the efficiency issues? What is the best way to go about diagnosing losses in a switching regulator that operates at currents too low to make anything get warm?

Thanks for the help!

[ConnorGames]

This is the switch node waveform. 5v in, 3.4v out, 17mA in, 16mA out. Sorry for the horrible image quality!
5v/div, time/div 1st Image : .2 uS/div, 2nd Image 2uS/div

[HackedFridgeMagnet]

What was the app note for this?
Is the circuit in there?
Not sure what "paralell resistance" is either in regards to an inductor. where did you get the term? Might be magnetising losses.

Hopefully you were searching for parallel not paralell.


.

[ConnorGames]

Sorry, the circuit is in the 1615 datasheet, not an appnote. My bad.

"parallel resistance" is a parameter for the inductors in the LTSPICE sim.

Inductors that work well seem to have this value as high as possible.
Here is a screenshot. the inductor in the screenshot seems to work well in SPICE and fit the existing board, so I will try purchasing a few soon.

[BravoV]

Just a suggestion, post a photo of your working circuit and the schematic as picture here, most of the experienced or experts in this forum here can rule out majorities of the switching power problem made by beginner in just few miliseconds. 

[HackedFridgeMagnet]

http://www.intusoft.com/articles/inductor.pdf This document explains the parallel resistance in an inductor model.

Probably you can avoid most of the parallel resistance effects by avoiding magnetic saturation and keeping well below the resonant frequency.

I think at such low power your overhead and switching losses are always going to be significant.
Not much you can do except get a good inductor and caps.
D1 should be shottky too.

AFAIK The data sheet hasn't actually got  a  "2-cell Lion to 3.3v SEPIC" circuit.

If you have 2-cell Lion and you want 3.3v then a buck topology would be better.

[ConnorGames]

I meant 1-cell, sorry. it is the "1-Cell Li-Ion to 3.3V SEPIC Converter" on page 7. I am actually using it with a 5v supercap, though.

The diode is a shottky, all the caps are all X7R or X5R and rated at least 16v, but the inductors are crappy unshielded 15 cent SMT inductors, with an unspecified core material. I think this might be the problem, especially when you look at how much more the recommended parts cost! If a 15 cent part worked as well as a $2 one, they wouldn't recommend the $2 part!

[Niklas]

The inductor selection might be the key to higher efficiency. There is a difference in the material properties in the ferrite core between energy storage inductors and filtering chokes. Filter chokes are made with high losses in mind.

[qno]

A supercap is not very good as a power source.
The internal rsistance is to high and the discharge voltage curve is very bad.

[ConnorGames]

The ESR is 130mOhm. That is not "Very bad" in my opinion . I do dislike the discharge curve, that is why I had to choose a SEPIC. The supercap is a 5v device to help with that,too. If only lithium coin cells could be charged at high C-rates, I would LOVE the greater energy density and higher charge-rates. LiPo is too dangerous for this device, and also does not charge as fast as supercaps. see the "OSHW acceleration Logger" Thread for info on the project. Please note the github files and first-post images are out of date, I have been busy lately and have not uploaded the newest changes.