Improving on the cheap MP1584EN converters

[ezalys]

Hey All,

I need to buck 24 volts down to 5V 3A regulated, and 12 volts unregulated. I like the idea of the MP1584EN converters available online, such as https://www.amazon.com/eBoot-MP1584EN-Converter-Adjustable-Module/dp/B01MQGMOKI/ref=sr_1_3?keywords=MP1584EN&qid=1559531634&s=gateway&sr=8-3 . My worry is the very dubious reviews. I'd like to try improving on the design, but I'm a little worried about where the flaws are. The main thing that sticks out to me is the lack of the via array which the data sheet calls for. This and what I might guess is 1/2 oz copper might be responsible for these guys popping at full load. I also have heard these converters are super noisy. Why might that be? Ultimately I'd appreciate the insight of someone who both knows what they're doing and/or has poked at this design a bit. I guess I wanna know why this thing doesn't meet spec and how to make it meet spec.

My main concern with redesigning this is the thermal engineering. Part of me really wants to just go with an older converter with an external through hole FET and stick a huge heatsink on it and just take a silly safety margin to avoid wasting much time on a low volume project. What's the best way to just overengineer this a bit so it's just completely bulletproof and reliable? Is there a sort of... surface mount big-ass heat sink that I might be able to use with this part? I've seen there are heat sinks that will just mount over parts which don't contact the part itself and just cool through radiation. Would that be appropriate here? I've also seen heat sinks that you can stick on the opposite side of the board and just use conduction through the vias. Does solder mask interfere with these heat sinks? Which of these strategies is better?

Third, I want to return to the 12 volts unregulated rail. I'm tempted to just use two buck regulators to get the 12, but what might be more fun is sticking a second winding on the inductor. I just don't know where to get a 5:12=1:2.4 turns ratio coupled inductor with the proper characteristics for the project. What are peoples' thoughts on this approach and potential appropriate inductors?

Thanks all!

[floobydust]

What is your ambient temperature spec?

Another way to look at it, is the heat generated. With 85% efficiency 2.7W to get rid of, although mostly in the buck rectifier diode. That's a lot of heat IMHO.

I'm using TPS40057 with two external MOSFETS, one is the synchronous rectifier. You get lower losses.
You can also buy cheap modules KIM-055L or "car 5A usb charger" and some 8A versions.
If your product could need more than 3A, do a design with some extra room. I'd never run MP stuff at max. current.

[ezalys]

> I'd never run MP stuff at max. current.

Is their stuff not particularly trustworthy?

[floobydust]

Many semi's are not realistic when operated at datasheet headlines. Datasheet hype varies with every semi manufacturer. You have to read between the lines.

Does it output 3A in higher ambient temperatures, on 1oz copper?
Read the fine print: SOIC8E 50°C/W on a JESD51-7, 4-layer board.
And why are they suggesting B340A Schottky rectifier diode, when that little SMA is going to cook and die, after it burns your finger. Do the math on how hot it will get, it can't do steady 3A either.

I think MP has close ties to Asia and I know to be very careful perusing such a datasheet which is not trustworthy.

[ezalys]

Gotcha. Okay so now I’m curious what I can get away with respect to power dissipation into the PCB with SMT components. Suppose I use a buck with external switches as you’ve suggested. Can I dump 3 watts successfully into a 2-layer PCB with a PowerSO package? Most guides I’ve seen seem to want a 4 or more layer PCB, but it’d be nice to keep costs down and use two if possible. Are there any documents that talk about thermal modeling of two layer PCBs with powerso packages, or is it just not done?

[floobydust]

Can I dump 3 watts successfully into a 2-layer PCB with a PowerSO package?

No. You would need much thicker copper and a big area, and the PowerSO (according to MP1584) is max. 2.5W into an infinite 25C heatsink. TI and Analog Devices claim about half as much.
You won't get 3A out of an MP1584 for very long.

External MOSFET's can have lower RDS on so they generate less heat. Then you could do a two-layer PCB. An XL4015 has 1/2 the on resistance of the MP1584 but it's not tiny.

There's tons of thermal design resources on the web, app notes and more.
Most important is to avoid datasheet bullshit and deception. Otherwise, just build a prototype and learn the hard way.