A Noise-Free DIY Switching Power Supply - How Hard Can It Be? -- Episode 542

That is what I had to do on the Mega IIe project’s power supply! The noise went down significantly by going to four layers.

Going forward, I will probably always use 4 layers on (standalone) SMPS designs. But as you said, pre-made modules are often a better option. 

I also just remembered an experiment of a teaching assistant on EE university. Instead of using a 2 layer PCB, using 4 layers with ground layer and a power layer. Even a sloppy designed PCB that produced a high amount of EMI got under control using this thick.

Great episode. I got to it a year after releasing, but glad I found it.

I guess one solution for less skilled HW designers might be that there will be ready made modules that would be connected to a target PCB via castellated edges or just pin headers, just like use processing power (e.g. Raspberry Pi Pico) / communication modules (many GSM / BT / GPS modules).

I'm with you. Even with good simulation tools, predicting how the passives actually work is a huge challenge.

Pre-2020 supply shortage, I would stick to a couple of switch controllers.

Any time I use a new one, I always build a separate power supply module (similar to the boards I did in this video) so I can test them a bit before incorporating them into a final design.

The upside is you get lots of design practice and a handful of modules. The downside to the shortage situation is that I am not building experience with any particular parts.

The biggest issue I have with designing switching power supplies is that I am unable to predict performance. Noise performance, thermal performance, efficiency and even current capacity. I need my prototypes to work well on the first spin. Cloning a reference design might be a way to reduce risk, but I usually find it very hard to procure the exact BoM used in the reference design, and the PCB layout is often wasteful of space. It is also common that the reference design doesn't meet all the requirements of the application. Maybe if I invested the time and money in perfecting my own reference designs I could get to a comfortable place, but I'm not there yet.

Yeah. There's been a couple of other variations I've considered. These were four I did a single night based on the common ones I see around the interwebs.

EMI differences is another angle I'd like to approach. This is where I think fat trace vs. pour might be more interesting.

I gotta say... I appreciate the illustration.  I do have a bad tendency to say "Aw, it'll be fine."     And it's not.     Being able to see the consequences of the shortcuts that I can see me taking... a least some of them...very valuable.  Still learning. 

I agree, a follow up episode showing how to impedance match traces and compare them to filled zones would be interesting!

Interesting episode, although designing for failure rubs me the wrong way. I would like to see to see if there is a difference between realistic attempts - I doubt anyone would design with a 6 mil trace meant to carry 4 amps, even though I know it was done just for illustration. It would have been interesting to see differences between heavy traces and poured interconnects. Heavy traces are easier to edit and their impedance can actually be calculated.

I had an instructor that used the same phrase - I think in connection to report writing objectives.

It seems to me the more I know the less obvious it becomes.