Buck/Boost design, iterative calculation or just name a figure?

[Simon]

I am looking at this IC: https://www.analog.com/media/en/technical-documentation/data-sheets/lt8392.pdf

Page 17 gives the inductor calculations. It asks for a figure that is not a percent as suggested but a ratio  I think. OK, so say I just wave a figure in there like 2, this will give me an inductance. But this figure is dependant on that inductance....

Is it just the case that the ripple current over average current in reality will always be less than what I guestimate with the inductor that comes out?

I am just slightly tetchy as I want this to work from 2 totally different supplies which is why I need to nail my component choices and then see what happens.

[Simon]

The question really is if I need a 8.3µH inductor to buck and a 20µH to boost and I use 20µH what happens in buck? that includes running max frequency in boost and min frequency in buck. I don't need to seamlessly switch as it's either 36V or 12V in.

I suppose, what is there to stop me using 2 inductors and separating out the transistors. At this point I am saving less and less parts but the thought just struck me before I go read the datasheet with that in mind so don't hit me just yet 

[T3sl4co1l]

Hmmmm, they give all three conditions (buck, boost and subharmonic oscillation) as minimums.  Evidently you can use whatever the hell you want, go nuts.  Thing is, subharmonic oscillation is worse at high inductance (low ripple fraction, typically below 30% is impractical to use; the amount depends on the slope compensation factor and how much spread in current limit is acceptable*).  I don't see anything unusual in the block diagram that would suggest they have a solution for this.

*Slope compensation lowers the ripple fraction for stability, at expense to making peak switch current dependent on pulse width, which is to say, supply and load voltages.

This may be an example where LT wasn't thinking at 100%.  Whether because the subharmonic condition is a typo and meant to be lesser-than, or something else.

Tim

[Terry Bites]

https://www.analog.com/en/design-center/ltpowercad.html
lets you quickly optimise performance.

[sandalcandal]

I had a quick gander at the datasheet.

"For any given ripple allowance set by customers" I think the idea is you come up with your desired maximum ripple as a fraction of average current and plug it in to get your inductor value lower bound. You don't need to know what the average current is. Plug \$ \Delta I_L \% = 0.2 \$ into the equation and you'll find values you need for 20% ripple. Obviously values outside \$ 0 < \Delta I_L \%<1 \$ will be invalid or cause some discontinuous current modes which might not work with the IC (would have to read in more detail).

The equations in the R_sense selection look like the same thing rearranged and for boost subbing \$ I_{L(AVG)} = \frac{V_{OUT}I_{OUT}}{V_{IN}} \$ which is what you would expect from \$ I_{L(AVG)} = I_{IN(AVG)} \$ and \$P_{IN} = P_{OUT} = V_{IN} I_{IN(AVG)} = V_{OUT} I_{OUT(AVG)} \$ and for buck subbing \$ I_{L(AVG)} = I_{OUT}\$ which matches my understanding of how buck and boost both work.

I guess they should add somewhere:
\$ I_{L(AVG)(BOOST)} = \frac{V_{OUT}I_{OUT(AVG)}}{V_{IN}} \$
 \$ I_{L(AVG)(BUCK)} = I_{OUT(AVG)}\$
Maybe an oversight on expected knowledge.

From what's said in the datasheet, the buck inductance minimum bound appears to be for maximum inductor ripple current and the boost inductance minimum bound appears to be for minimum inductor ripple current so I don't really get why the boost inductance lower bound equation is there because I can't really think of why you need to set a minimum bound on inductor ripple current? Looking closer at the equations identifying boost as minimum ripple might be a misstep, depending on the values input e.g. V_out close to V_in(max), ripple in the boost mode could be higher than the buck mode depending on values. Maybe should try bothering the support engineer for this IC/datasheet if you can.

Picking a larger inductance than the given lower bounding equations should just result in adequate ripple and stability at the cost of a more expensive/larger inductor. Setting a lower ripple (and larger inductor) should reduce magnetic losses. Obviously you also need to also be aware of inductor saturation/max current and other limits when picking your inductor.

Not familiar with the subharmonic oscillation condition so deffering to what Tim has said there. Edit: I'm having a read on slope compensation right now here https://www.eetimes.com/understanding-nonlinear-slope-compensation-a-graphical-analysis-part-1/?page_number=2

As other have suggested, I'd just run an LTSPICE sim anyway since they tend to be pretty good with LT ICs anyway and there should be ready made LTSPICE circuits for the IC too.

[SiliconWizard]

https://www.analog.com/en/design-center/ltpowercad.html
lets you quickly optimise performance.

You can of course also use LTSpice, their models are pretty close to the real deal IME.

[Simon]

Well the chip is unavailable anyway. Maybe I should use separate buck and boost converter. Maybe 12-48V input is a bit too much of a range. With the availability at the moment I may end up considering rolling my own till everyone buys up the comparators

[poorchava]

To be honest, I usually use Excel for SMPS design. I find it that most tools make many assumptions which are often invalid for particular situation.

Just out the formulas into a sheet and manipulate values until it makes sense. I sometimes use some VBA for automated bruteforce approach.

It turns out that designing a dumb offline flyback can be a really complex optimization task if you have some constraints (like total transformer size, space constraints, forced to use particular core type, etc).

[Terry Bites]

Roll your own is not so bad if you have the time. I used to do it in the dark days before RS even stocked an inductor. Well, maybe some 4H job for your power amp.
You might have to build your own comparator from plywood as things stand- even that's in short supply.

[Simon]

Oh dear, that's what I was wondering, how long before everyone has to do it. I know soon all the cheap 8 bit micro's will disappear as we start using those just for the comparators in our SMPS designs.