12V to 5V Buck converter design

[nour]

I want to make 12v to 5v step-down converter with the following specs.

Input voltage: 12v
Output voltage: 5v
Output current: 2.5~3A
Efficiency >=90@2.5A
High switching frequency >1Mhz ~ 3Mhz
Small footprint for both the buck converter and the inductor
low heat dissipation @ 2.5A output
I want to add overvoltage\overcuurent\short circuit protection
I am going to make a 2 or 3 units (maybe 5 max)

After doing some search on Linear Tech.  website I have chosen this one(only the converter):
LT8614 (not a final choice)

Can you recommend something for those requirements
Also is there a solution for over voltage\current short circuit protection in one package?

[nour]

Any suggestions!

[amb101]

To get to know buck converters I played around a bit with the readily available and cheap "MP1584 Adjustable 3A DC-DC Converter Step Down Voltage Regulator".
Turning them into LED driver and such.
You can learn a lot from understanding the MP1584 datasheet.
Since then the MP23070N seems to have superseded it.

[tecman]

Your requirements are relatively easy to hit.  My only comment is switching frequencies mentioned are high and will present most of the problems in building.  I would suggest trying to stay under 100 khz, or even 50 khz.  Life will be much easier.

paul

[nour]

Your requirements are relatively easy to hit.  My only comment is switching frequencies mentioned are high and will present most of the problems in building.  I would suggest trying to stay under 100 khz, or even 50 khz.  Life will be much easier.

paul

I want to have the inductor as small as possible!

[james_s]

Buck regulators are fairly simple these days, you can pretty much just copy the reference design from the datasheet. Pay close attention to PCB layout as it's critical at higher frequencies. I've found that classic "dead bug" construction on blank copper clad works remarkably well for this sort of thing. I built a buck converter that way just for fun a few weeks ago and it worked perfectly on the first try.

[T3sl4co1l]

LT stuff, and TPS54xxx something or other, comes to mind.  Single chip solution, plus the usual inductor, caps and resistors, and maybe a diode.  I'd have to go shopping to see what parts exactly, but it's definitely doable.

Tim

[james_s]

I've used some of the LT parts and they're good, but quite expensive. I'm also familiar with the TI SimpleSwitcher series and they're also quite good, although efficiency is only around 80% on many of those. The integrated switch devices are nice because you only need an inductor, diode and a couple of capacitors for a minimal setup. An additional LC filter can be added if a cleaner output is needed.

[monqy]

Check out the TPS82130 from TI.  I believe it might meet all of your requirements, and it also includes the inductor mounted on top of the package so it takes up very little space.

[T3sl4co1l]

An additional LC filter can be added if a cleaner output is needed.

This goes for the input moreso than the output side, by the way! 

Tim

[DutchGert]

I have used the LT8614 in some automotive project and it preforms really nice. It's a bit expensive but works a treat: high efficiency at a 'high' Fsw and low noise/ripple.


If you want to go cheap with very good specs have a loop at MPS : https://www.monolithicpower.com/Products/Parametric-Search/DC-DC-Power-Conversion/Switching-Regulators/Step-down(Buck)/Converters?categoryID=50#filters=

Or save yourself the trouble and use a module: https://www.monolithicpower.com/Products/Product-Detail/1146/DC-DCPowerConversion/PowerModules/MPM3630
These puppies are rated to operate at there rated current full time and its very hard to make something that performs better for the same price and footprint......
I love doing a DC-DC design and lay-out but these modules are getting very attractive...

With something like the LT8614 you can get lower noise, lower EMI and a bit higher eff but it comes at a price. 

[DutchGert]

LMZ31704 for the lazy engineers. All in one, including coil, just add caps and resistors and you get a working 94% 12V->5V DC/DC in minutes.
I've used its bigger brother, LMZ31710, which is 100% compatible, just with higher (10A) capacity, in a CC-CV laser diode driver application.
Keep in mind that LMZ317xx family doesn't have precision current limiting (it has over current protection), so if you need CC-CV, you need an external opamp to do the LM2596 CC-CV trick.

A MPM3630 is smaller, cheaper and performs better above the 1MHz the opp wants...

[Gibson486]

LT stuff, and TPS54xxx something or other, comes to mind.  Single chip solution, plus the usual inductor, caps and resistors, and maybe a diode.  I'd have to go shopping to see what parts exactly, but it's definitely doable.

Tim

Agreed...they both also have those lines where the inductor and switcher is built into the die. Those are awesome. 

[Zero999]

LT stuff, and TPS54xxx something or other, comes to mind.  Single chip solution, plus the usual inductor, caps and resistors, and maybe a diode.  I'd have to go shopping to see what parts exactly, but it's definitely doable.

Tim

Agreed...they both also have those lines where the inductor and switcher is built into the die. Those are awesome.

Does anyone know how they do that? I've had a quick Google for integrated circuit inductors but I only get results for RF circuits.

[T3sl4co1l]

They're not die integrated*.  They'd be more accurately called co-pack, or hybrid or module.

*The ADI (and now others) monolithic transformers (ADUM series) are actually good for a little power, up to a watt I think.  These are still silicon substrate and photolithography process, but the insulating layers are polyimide and the conductor is copper (IIRC), and the scale (layer thickness and feature size) is much larger than your usual semiconductor fabrication.  (Not necessarily that they're using an intentionally sloppy process, just that it's big fat coils, because that's what you need to handle power.)

IIRC, the power isolator ones run around 200MHz and are in the range of 50% efficient.  Crummy, but possible!

Tim

[Zero999]

They're not die integrated*.  They'd be more accurately called co-pack, or hybrid or module.

*The ADI (and now others) monolithic transformers (ADUM series) are actually good for a little power, up to a watt I think.  These are still silicon substrate and photolithography process, but the insulating layers are polyimide and the conductor is copper (IIRC), and the scale (layer thickness and feature size) is much larger than your usual semiconductor fabrication.  (Not necessarily that they're using an intentionally sloppy process, just that it's big fat coils, because that's what you need to handle power.)

IIRC, the power isolator ones run around 200MHz and are in the range of 50% efficient.  Crummy, but possible!

Tim

All right that makes sense but do you have any idea of how they make the inductor so small, given the frequency isn't that high? The LMZ31704 only works at 200kHz to 1.2MHz and at 4A, a buck converter would require a larger inductor than would fit into the tiny 10x10x4.3mm QFN package.

http://www.ti.com/lit/ds/symlink/lmz31704.pdf

[mariush]

If you want  all in one, Linear has a lot of those .. see the uModule section : http://www.linear.com/products/umodule_buck_regulators

But they're expensive, for example LTM4623 is 13$ if you buy one , but goes down to around 9$ if you buy 25 (7.35 for 100) : https://www.digikey.com/product-detail/en/linear-technology/LTM4623EV-PBF/LTM4623EV-PBF-ND/5080540
It can do 12v in , 5v out at around 93% efficiency and it's small (6.25mm × 6.25mm × 2.42mm)

I know you want high frequency to get smaller inductors, but it may be worth trading frequency to get better efficiency .. high frequency regulators may not be as efficient as you want.

See for example AP3513 for a 500kHz 3A regulator that's very cheap ($0.65),

There's also AP65450 at 650kHz and 4A current and over 90% efficient across all output range.

Note though that in such tiny chips, you'd have to be careful about how you dissipate the heat (you'd need a few vias under the chip to dissipate heat on both sides of pcb etc)

On the other extreme you have TPS62135 which does 2.5mhz but it's only about 85% efficient with such high voltage.

[DutchGert]

If you want  all in one, Linear has a lot of those .. see the uModule section : http://www.linear.com/products/umodule_buck_regulators

But they're expensive, for example LTM4623 is 13$ if you buy one , but goes down to around 9$ if you buy 25 (7.35 for 100) : https://www.digikey.com/product-detail/en/linear-technology/LTM4623EV-PBF/LTM4623EV-PBF-ND/5080540
It can do 12v in , 5v out at around 93% efficiency and it's small (6.25mm × 6.25mm × 2.42mm)

I know you want high frequency to get smaller inductors, but it may be worth trading frequency to get better efficiency .. high frequency regulators may not be as efficient as you want.

See for example AP3513 for a 500kHz 3A regulator that's very cheap ($0.65),

There's also AP65450 at 650kHz and 4A current and over 90% efficient across all output range.

Note though that in such tiny chips, you'd have to be careful about how you dissipate the heat (you'd need a few vias under the chip to dissipate heat on both sides of pcb etc)

On the other extreme you have TPS62135 which does 2.5mhz but it's only about 85% efficient with such high voltage.

]

I am a Linear fanboy but the MPM3630 (http://www.monolithicpower.com/Products/Product-Detail/1147/PowerModules/MPM3630) costs 30% of the LTM4623, performs just as good and is 3mm x 5mm x 1,6mm compared to the 6.25mm × 6.25mm × 2.42mm of the LT.......

MPS does not seem to be widely known on this forum but they are some ex-Lineair engineers who thought they could do better if they just focussed on power products and so far a lot of there products perform similar or better than the LT or TI counterparts for a fraction of the price.......
They even have full automotive qualified parts

[DutchGert]

Not everyone does it that way: