Low Load on a DC/DC buck controller

So I'm building a DC/DC power rail based on the LM27403 with the major driving factor being efficiency. Im looking to achieve over 95% with a 5v power source and .8-1v core voltage output with total load being under 15 amps. Typical load will be between 3-10 amps. To achieve this I'm running at 250khz switching frequency.

Im trying to pick an inductor for this, and most Specs for LM27403 + CSD87353Q5D say a 300nH inductor in the IC examples...which im pretty confused about, do these example circuits just assume there will never be any low loads? If I used a 300nH inductor like it suggests the inductor ripple current would be at 11A at 1v output....this is no small load and the circuit can't be expected to operate at only >11A.

I would need at least 1uH inductor to keep ripple under 100% @ 3A load. Am I missing something with this buck that allows it to operate at high inductor ripple currents during low load?

The example IC is in section 9.2.5 here: https://www.ti.com/lit/ds/symlink/lm27403.pdf

Parameters are exactly what I'm running except I don't need 25A of load.

jstefanop said:

Do these example circuits just assume there will never be any low loads? .

Click to expand...

Yes, pretty much so.
A ten amp sample design will usually be optimised for ten amps.

If you want a 3 amp supply that can also work at 15 amps then you calculate the inductance for 3 amp minimum load, maybe for one amp peak to peak ripple current.
But the inductor will also see 15 amps, so it needs to be physically much larger.
A higher inductance, with 15 amp wire size and a core that will not saturate with 15 amps.

One solution is to use a powdered iron core that will have sufficient inductance at only 3 amps, but will be driven well into the magnetisation curve at 15 amps resulting in a much lower inductance at 15 amps, which is not be a problem.

*edit*

Are you planning to buy an inductor, or design one yourself ?

With that kind of step-down ratio, you'll need some really
sweet power switches and perfect input filter caps to come
close to your goal. Presuming your 95% is "wallplug" and not
"marketing efficiency".

Last integrated POL buck I worked on, hit 95% for 5 -> 3.3V
and half load but was down to low 80%s at 5V -> 1V, at 10A
out. We didn't try to get peak efficiency at full (reliability limit)
load since none of our particular customer base would derate
less than 50%, ever. Based on this we tried to optimize the
sweet spot at half load and 1MHz. But we couldn't get any
more out of the high-step-down corner.

Warpspeed said:

Yes, pretty much so.
A ten amp sample design will usually be optimised for ten amps.

If you want a 3 amp supply that can also work at 15 amps then you calculate the inductance for 3 amp minimum load, maybe for one amp peak to peak ripple current.
But the inductor will also see 15 amps, so it needs to be physically much larger.
A higher inductance, with 15 amp wire size and a core that will not saturate with 15 amps.

One solution is to use a powdered iron core that will have sufficient inductance at only 3 amps, but will be driven well into the magnetisation curve at 15 amps resulting in a much lower inductance at 15 amps, which is not be a problem.

*edit*

Are you planning to buy an inductor, or design one yourself ?

Click to expand...

Im just buying one. There are plenty of 1uH inductors that will work at 15A. One I have picked out is 1.3 with 20A rating. That gives me a 2.5A ripple. Typical load will be around 5A so this should work fine. I just want to cover my bases for really low end. Should I expect anything horrible to happen if ripple approaches 80% of load?

dick_freebird said:

With that kind of step-down ratio, you'll need some really
sweet power switches and perfect input filter caps to come
close to your goal. Presuming your 95% is "wallplug" and not
"marketing efficiency".

Last integrated POL buck I worked on, hit 95% for 5 -> 3.3V
and half load but was down to low 80%s at 5V -> 1V, at 10A
out. We didn't try to get peak efficiency at full (reliability limit)
load since none of our particular customer base would derate
less than 50%, ever. Based on this we tried to optimize the
sweet spot at half load and 1MHz. But we couldn't get any
more out of the high-step-down corner.

Click to expand...

Im simulating 94% right now with my setup. But this is with ultra low ESR caps, low DCR inductor, and a single integrated dual mosfet power block from TI.

It's usually impossible to keep continuous conduction mode (CCM) of a buck converter at low loads. Percent ripple current specifications are for rated load.

As the absolute ripple current amount doesn't increase at low loads, there shouldn't be a problem of related output voltage ripple with usual output filters. But changing from CCM to DCM affects the controller loop gain and must be considered when designing a stable control.

jstefanop said:

Im just buying one. There are plenty of 1uH inductors that will work at 15A. One I have picked out is 1.3 with 20A rating. That gives me a 2.5A ripple. Typical load will be around 5A so this should work fine. I just want to cover my bases for really low end. Should I expect anything horrible to happen if ripple approaches 80% of load?

Click to expand...

Eighty percent ripple will be fine, just so the load does not fall so far that it goes discontinuous on you.
If it starts getting hot it will probably be more core loss, but that is unlikely because the voltages are so very low.

As your output voltage is on the lower side, nothing horrible will happen in your circuit.

One way controllers achieve higher efficiencies at low loads is by employing burst mode, frequency foldback or similar techniques.
Does the controller you chose have these features?