High efficiency DIY buck converter

[drummerdimitri]

I'd like to build a buck converter to step down the 9-12.6V of a lithium ion pack to a constant 5V and 8A max as the adjustable one I bought from Aliexpress is terribly inefficient and generates too much wasted heat.

Is there a simple circuit diagram that achieves above 80% eff. and is capable of delivering the required power or is that too much to ask?

[NiHaoMike]

Look for one based on a controller with synchronous rectification and low quiescent current. One I have used is the LTC1775, can do over 95% efficiency and not too difficult to build a complete regulator based on it from scratch.

[ogden]

Quick search revealed "DC-DC Buck 9-35V to 5V 5AStep Down Synchronous Rectification Power Supply Module" that may meet efficiency numbers, yet may not be powerful enough. In case it is one-off project, then you may consider "branded" DC-DC module such as Meanwell SKA60A-05 (Input 9-18Vdc; Output 5Vdc at 12A).

[SiliconWizard]

Define inefficient.
For this kind of ready-made modules, 80% would be about what to expect, and could hardly be qualified inefficient. Finding something as cheap (which I suppose it is) with better efficiency may take a while.
Designing it yourself  - yeah, doing better than 80% if you have no preliminary experience with that, good luck!

Now if your module has significantly worse than 80% efficiency (are you sure? how did you measure it?), you probably either got ripped off, or you're not using it properly.

Lastly, just in case you haven't really measured the efficiency, but are basing your impression on the fact that the converter gets "hot": well, if it has say 80% efficiency in your conditions, it's still 20% loss. For the worst case, going from 12.6 to 5V  @8A, that's a dissipation of 12.16W. Certainly enough to get a small converter very hot if it's not properly heatsinked.

[mariush]

There's ready made dc-dc converter power supplies, here's some prefilter (anything 7.5A or higher, but make sure 5v is within its output range) : https://www.digikey.com/short/pfnq27

Basically you only need to add a couple capacitors and a resistor to set the output voltage if it's not fixed.
Datasheets contain the instructions on what's needed

There's also these dc-dc power supplies: https://www.digikey.com/short/pfnqn0


There's modules like this one.. expensive though: LTM4626 https://www.analog.com/media/en/technical-documentation/data-sheets/LTM4626.pdf

Something cheaper ( look up examples in datasheet)

UP to 96% efficient: https://www.digikey.com/product-detail/en/vishay-siliconix/SIC437DED-T1-GE3/SIC437DED-T1-GE3CT-ND/10244417

MPQ8634BGLE-P  https://www.digikey.com/product-detail/en/monolithic-power-systems-inc/MPQ8634BGLE-P/1589-2005-1-ND/9817899

AOZ2264 https://www.digikey.com/product-detail/en/alpha-omega-semiconductor-inc/AOZ2264QI-18/785-1898-1-ND/10258107

[splin]

This might do it:

https://www.ebay.com/sch/i.html?_from=R40&_trksid=m570.l1313&_nkw=dc-dc+300w++xl4016&_sacat=0&LH_TitleDesc=0&LH_PrefLoc=2&_osacat=0&_odkw=dc-dc+%28300w%2C280w%29++xl4016

This review shows it achieving 89.2% efficiency with 12V in, 5V out @ 7.5A. (8:27):



It failed his 5V, 8A test at 7:44 however. It managed an impressive 92.3% at 12V in, 9V 8A out. Dirt cheap too at $5 shipped.

The efficiency will improve with lower input voltages so it will probably deliver 8A or more with less than 12V input. How long for might be another matter so more heatsinking would likely be a good idea. There are plenty of other reviews around.

My main concern though would be the consequences of the module failing if you were using them to charge big lithium batteries. Many of these cheap chinese modules are not exactly reliable when pushed to the limits. I bet those input smoothing capacitors will have far more ripple current than they are rated for in the worst case scenario with output = half the input voltage at max current.

You could parallel two modules and set the current limits and voltages appropriately to ensure they share nicely.

[mariush]

From the links I posted in my previous message:

91% eff.. 6$ :
product: https://www.digikey.com/product-detail/en/bel-fuse-inc/VRAE-10E1A0G/507-1384-ND/1754625
datasheet: https://www.belfuse.com/resources/PowerSolutions/VRAE10E1A0Series/VRAE-10E1A0_DS_J_122815.pdf



95$ eff.. 7$:
product: https://www.digikey.com/product-detail/en/bel-power-solutions/SLDN-12D1ALG/179-3029-ND/8028727
datasheet: https://belfuse.com/resources/datasheets/powersolutions/ds-bps-sldn-12d1ax-series.pdf

[Siwastaja]

Designing it yourself  - yeah, doing better than 80% if you have no preliminary experience with that, good luck!

Market is literally full of completely integrated converter ICs with synchronous rectification and efficiency around 90-95%, with layout examples in datasheets. It's about placing an inductor and a few capacitors properly. Luck doesn't need to be a big part of the equation, nor thousands of hours of experience is needed.

An example: AOZ2261QI satisfies the spec.

Of course some experience required to be able to follow the recommended footprint and layout and solder the beast down, but minor beginner errors shouldn't degrade the efficiency to <80%, as you seem to suggest.

Chinese Ebay DC/DC modules (not all) tend to use counterfeit copies of 30-year-old non-synchronous ICs (or functionally equivalent Chinese versions), with very crappy layouts, and unsuitable inductor types and too much capacitor ESR as well. It's very easy to do better: modern fully integrated ICs almost "force" a tight layout, higher fsw allows all-ceramic solution minimizing ESR losses.

Most Ebay DC/DC converter modules look like they are getting rid of excess part inventory.

[drummerdimitri]

Sorry for my initial ballpark guess. I thought the energy wasted was a lot higher than actual since the heatsink of the power transistors reaches 60C after some time and stays at that temperature.

I did some efficiency calculations and it is actually around 93.5% which is really good!

Only issue is the heat as it will have to be inserted in a 3D printed case and I'm not too keen on having it run that hot inside and adding a fan to blow over the heatsink will result in noisy operation which is not acceptable so Should I try mounting those transistors onto larger heatsinks?

[schmitt trigger]

Thermal management is as important as electrical efficiency.

Larger heatsinks work if you have proper convection. Or since you are 3D printing the case, perhaps leave some space for a small fan.

[mariush]

If you're 3D printing the case, maybe you could make a sort of hybrid case?

For example, maybe look for a rectangular copper or aluminum piece that would be a part of the case walls, then you have your 3d printed case made of two parts with the metal ring sandwitched between the plastic case
Then you could further drill some very fine holes in the aluminum or copper for ventilation or natural convection or you make some striations on the metal (cut some lines in the metal) to increase the surface area.
If not the side walls, maybe have the bottom as a copper/aluminum sheet?

[drummerdimitri]

I will include slots for natural convection at the sides of the case but definitely will not be including a fan as it is not a quiet solution.

How far can I relocate the transistors from the main PCB and what wire gauge should I use?

[KaneTW]

A low RPM fan will be basically inaudible and increase heat transfer by _a lot_. http://www.crydom.com/en/tech/hs_wp_fa.pdf for some ballpark math.

[floobydust]

I'm happy using the Ali/eBay boards from QSKJ like QS-1205CBUM-8A.
Most use TPS40057 synchronous rectification and two or three external mosfets of varying spec. The DPAK are better rated than the laptop MSOP mosfets.
Surprisingly, the inductor gets hot and limits power output so they recommend using a fan above 6A or so. Many reviews of these boards out there.
For cold temperatures, I swap in solid polymer caps.

[schmitt trigger]

In addition About the fan: it doesn’t need to run continuously. Only when the components exceed a certain preset temperature.
Which may be only reached if you are pulling a full load continuously, and/or have a high ambient temperature.

We are only presenting you the options. But in the end only you know your budget, the performance criteria and the environmental restrictions.

[T3sl4co1l]

If you want something probably-better than Ali junk, and don't have the time/money/tools to produce a board yourself: consider getting a dev board for one of the regulators.  This will be under $100 for that price range I think, and is ready to go, give or take swapping some resistor values to set the desired output voltage and such.

Tim

[splin]

If you want something probably-better than Ali junk, and don't have the time/money/tools to produce a board yourself: consider getting a dev board for one of the regulators.  This will be under $100 for that price range I think, and is ready to go, give or take swapping some resistor values to set the desired output voltage and such.

Tim

Nooooo! Don't consider a dev/evaluation board - far too expensive! Just go to digikey, or your favourite distributer and plug in your requirements. Here's one @ $7, one off:

https://www.digikey.com/product-detail/en/bel-power-solutions/SLDN-12D1ALG/179-3029-ND/8028727

95% efficiency (according to spec) with 12V input, 5V out @ 8A. Don't like that one? There are plenty more for < $10.

[T3sl4co1l]

That is, if you weren't going for a drop in module, indeed.

Tim

[Warhawk]

Ti has been publishing reference and power designs online. You can try to have a look http://www.ti.com/reference-designs/index.html

e.g. http://www.ti.com/tool/PMP30101

[drummerdimitri]

I ended up adding a fan on top of the buck module and it is almost silent due to it's small size even though I was reluctant to do so, it seemed like the correct way forward.

Now I'm trying to chose a switch with the lowest resistance to keep efficiency as high as possible without going for a ridiculously large one but I'm not sure how to calculate the power loss if I know its resistance (4 wire method)? 

I can do ohms law but not sure about voltage drop/power loss calculations.

[Siwastaja]

P = I^2 * R.

Assuming MOSFET switch, do note that R is specified at die temperature of 25 degC. At, say, 100 degC die temperature (case temperature can be much less), it's about 1.5 times higher.

Also remember to provide high enough gate voltage to minimize Rds(on).

For a buck, a separate switch usually won't make sense (the buck itself has the switch); use a buck chip/module with enable input instead.

[drummerdimitri]

P = I^2 * R.

Assuming MOSFET switch, do note that R is specified at die temperature of 25 degC. At, say, 100 degC die temperature (case temperature can be much less), it's about 1.5 times higher.

Also remember to provide high enough gate voltage to minimize Rds(on).

For a buck, a separate switch usually won't make sense (the buck itself has the switch); use a buck chip/module with enable input instead.

You misunderstood.

Sorry for the ambiguity, I was referring to a physical ON/OFF switch not a transistor.

[Siwastaja]

P = I^2 * R.

Assuming MOSFET switch, do note that R is specified at die temperature of 25 degC. At, say, 100 degC die temperature (case temperature can be much less), it's about 1.5 times higher.

Also remember to provide high enough gate voltage to minimize Rds(on).

For a buck, a separate switch usually won't make sense (the buck itself has the switch); use a buck chip/module with enable input instead.

You misunderstood.

Sorry for the ambiguity, I was referring to a physical ON/OFF switch not a transistor.

P = I^2 * R still applies. Although, for mechanical switches, the ratings are all over the place, and the actual R is more unpredictable.

Some quote a "cold resistance" and then "voltage drop at rated current", which gives you a different R when you calculate R = U/I.

[NiHaoMike]

If the converter has a low enough shutdown current, you can just use a signal level switch to disable it.

[SiliconWizard]

Sorry for my initial ballpark guess. I thought the energy wasted was a lot higher than actual since the heatsink of the power transistors reaches 60C after some time and stays at that temperature.

I did some efficiency calculations and it is actually around 93.5% which is really good!

(...) so Should I try mounting those transistors onto larger heatsinks?

So, exactly what I suspected!

And yes, you need to improve heat dissipation. And if your enclosures have no vents (or too small ones), it's going to get pretty hot in them no matter how large the heatsinks.