CPU fan buck converter, miniature

Oops, possible problem with the DropBox link.

Note the 1-ohm resistor in series with Cout. The LM5163 needs this for a feedback signal. Another unusual aspect, the boost cap, C13, must not be larger than 2.2nF. But it's a super-handy IC design, working to 100 volts.

What happens if the bootstrap cap is bigger?

The fan speeds up, flies off the board, and heads for parts unknown. :) GH

The chip has an onboard 5V supply that run s fromthe up to 100V input. That supply has n o bypass capacitor, but runs everything on the buck converter, incuiding both MOSFETs. They point out youmusty avoid overburdening that supply, anf the boost cap is too bi gthe chip , canbe damaged. Frankly, I'm glad they avoided both bypass cap and catch diode.

how well does that work on a board full of bypass caps ?

Usually a big, moderate-ESR cap (like a tantalum) adds a lot of damping to a power pour which has a bunch of low-ESR ceramics lurking on other sheets of the schematic. In Win's case, the only load is the fan.

I Spice these sorts of things when I can, or sometimes breadboard to check stability.

Sometimes a big super-low-ESR cap, like a 56uF polymer or a 1500uF aluminum, is brute-force stable too. On a regulator (or on an opamp output) to ground.

tirsdag den 28. januar 2020 kl. 19.56.23 UTC+1 skrev John Larkin:

sure but the datasheet has formulas for calculating the required ESR and explains that the resulting ripple is required for operation

Datasheets don't usually, or ever, address the multiple-capacitor case, like one big aluminum or tantalum in parallel with a bunch of ceramics.

Ripple is required? We don't want ripple!

Some regulators are stable for any amount of low-ESR caps. That's rare. Many older data sheets don't even discuss the issue.

I invented the C3 trick here, to stabilize an LM317 with a big zero-ESR output cap.

I'm not sure how that works, but it works.

Yup. It's a super low-Z lead/lag network.

Until it dries out a bit. ;)

Cheers

Phil Hobbs

...

Thanks for sharing - looks like a great part for an update project for automotive sensor that needs to go from 12 to 24V world.

Most survive the 1-year warranty period. Actually, most survive a lot longer.

I really like aluminum polymers, or at least some of them.

Note the high value, 1 ohm, to generate appropriate voltage ripple at the feedback node. That's way higher than a typical esr, and could be spoiled by lots of bypass caps. The LM5163 datasheet, page 11, is worth reading. They allow direct Cout to ground, if instead you use a three-part network to generate a triangular ramp that's in-phase with the inductor current. Note, this isn't an "esr" situation, but for my application, it's OK. A Cff cap across Rf can reduce output ripple.