Designing step-down power system

[PurpleAmaranth]

I'm looking for advice about designing a power supply for a board, stepping down an input voltage. I've been reading a lot of application notes and my current understanding is to use an LDO/linear regulator for a very clean output and for low input-to-output voltage difference (such that it dissipates low power), and to use a buck in all other situations. Is that the correct way to think about it?

Furthermore, when designing the buck, what is an intelligent way to go about the inductor design? TI app notes suggest sizing the inductor for 40% ripple at max load. But this rule of thumb doesn't really explain what to do in situations such as peak load that could be high but won't happen often (100s of mAs) and very low average load (maybe in the tens of mAs). Designing for max load in this case would imply DCM operation the vast majority of the time. Is that a problem? It seems that it isn't, but then why not just always use a very small inductor and always operate in DCM?

I know it's a lot of questions, but I'm basically looking for pointers on good design practices overall. I've learned the theory in school, but I'm not sure how well that translates to practice... I know I haven't been very specific but that's because I'm hoping to get more general-purpose guidelines.

[Berni]

Just choose a switchmode IC for the range of amps and volts that you need, then copy the design from the datasheet

For some common examples they not only give you all the component values or even exact part numbers, but there is often a PCB layout examples. No actual knowledge is required of how these things even work.

Then as you get to know these things better you can start tweaking the designs from datasheet to your liking. If i need to calculate some things about the switcher because my operating voltages are far from the example then i will typicaly just use a generic switchmode calculator like Weurth Electronics RedExpert:
https://redexpert.we-online.com/we-redexpert
(Yes they try to sell you their passives since they are a manufacturer of those, but the calculation results are valid none the less, and they do actually make really nice passive components)

[Solder_Junkie]

If you look on eBay, there are ready to use Buck converter boards that are typically rated up to 5 Amps. Many use the Texas LM2596 IC. For the money, it's not worth the effort of making your own, just incorporate the ready made board on yours.

I replaced a 1.2V linear regulator in an SDR receiver (QS1R) with a Buck board and it saved the day. The original design used a 3 leg linear regulator that overheats and eventually fails. While these Buck converters aren't as efficient as the eBay claims (TI state 73% for a variable one), they run a lot cooler than a linear regulator. In the case of the QS1R, there is no noticeable hash/noise produced, although I cannot claim that all such boards will be electrically clean.

Linear regulators are generally considered cleaner in terms of electrical noise, at the expense of efficiency. We "all" tend to use 7805 and similar regulators.

In all cases, consider what happens if the regulator fails and outputs excessive Voltage... In many years of using power supplies I have only had 2 fail, one blew an audio amplifier (supplied with it and under warranty) and the other was a fixed Voltage bench power supply that had been fitted with a crowbar protection circuit by myself and just blew a fuse on the crowbar board. These days I use an LTC4368 protection circuit with a pair of back to back MOSFETs, the circuit is in the datasheet for that IC.

SJ

[EPAIII]

OK, Yes a linear regulator will provide the lowest noise - as a general statement. No switching so no switching artifacts and that means lower noise.

LDO? Well, IF, and that is a BIG IF, IF you can provide an unregulated supply that is just big enough to provide the (Voltage) head room that the LDO regulator needs, then yes, it will waste less power. Linear regulators drop the Voltage down to the regulated value by controlling the effective resistance of the series element (pass transistor or FET). In other words, they turn that excess Voltage * Current = Power, into heat. So the less the needed Voltage drop the less power is wasted. The problem here is if you are designing a ONE-OFF, LINEAR power supply, then you are probably going to have to buy a commercially available transformer. And those commercially available transformers are only available in a few standard secondary Voltages. So, you are stuck with the transformer output Voltage and then the output Voltage of the rectifier and then the output Voltage of the filter capacitor and that is what you are going to have to send to the the input of the linear regulator. So, if the Voltage difference from the filtered, unregulated Voltage to your output Voltage is larger than the head room of the LDO regulator, that LDO regulator still must drop that full Voltage at your load current. In other words, in most cases, using a LDO regulator does NOT save any power at all.

Now, you can add some other element in the path, like a resistor or additional diodes or whatever you want that provides a Voltage drop, but then you are just moving the place where the extra power is turned into head and wasted. Such additions do not provide any greater efficiency. So "LDO" is not a magic device. It is only a tool that must be used properly. I have designed a number of ONE-OFF regulated supplies and have never saw any advantage for a LDO design. But they were one-offs, not mass produced items.

Designers who are working on power supplies for commercial equipment that will sell in the tens of thousands or hundreds of thousands or millions of units will be able to order transformers made for that exact item, at a advantageous price. They can use a LDO regulator to it's proper advantage by having a special transformer made for their circuit. Unless you want to pay a LOT of money for one or make it yourself, you can't.

The above is why switching power supplies are so popular. If your design criteria calls for high efficiency (low wasted power) then the easy route is to use a switching power supply and deal with the switching noise.

Linear power regulator ICs will have data sheets with design examples. You can follow them - I do.

Switching regulators are available in a wide variety of designs and doing the full design work for each and every one of them is a real PITA. Fortunately, the major makers of those switching regulators have on-line design tools that make the design work simple. Here's a link or two:

https://www.ti.com/tool/WEBENCH-CIRCUIT-DESIGNER

https://www.analog.com/en/design-center/glossary/switching_regulator.html

https://www.analog.com/en/app-notes/buck-power-stage-design-equations.html

There's a lot more available. If you go to a major supply house and pick an IC, they will have links to the data sheet and other design aids.