I am working on a small solar power system. The critical problem right now for me is how to do the DC-to-DC conversion. Essentially, the problem is that you have an array of large 12V lead-acid batteries and need to power a wide variety of different DC devices which typically require voltages like 3, 5, -5 and 9. Also, you may have 12V devices but need a regulated supply so even for hooking up the batteries to a 12V device is not trivial.
Problem #1: confusing multiplicity of circuit possibilities
The main off-the-shelf options seem to be either buck converters or DC switchers. One problem is that there are many different ways to design such devices and what you can buy is often a black box, so you often don't know what kind of circuit you are buying. The manufacturers claimed efficiency is often much different than the real efficiency you actually get when you plug it in and measure it. Therefore, it is hard to know without a lot of trial and error which products might be good or bad.
Problem #2: compromise circuits
A lot of switcher products accept a range of input voltages, thus they may be "compromise designs" that sacrifice efficiency for flexibility. In my case I have just ONE input voltage: 12V, so I don't need a chip that accepts a range of input voltages. So, if I a buy switcher chips I may be unnecessarily losing efficiency to get flexibility I don't need.
Problem #3: single-input single-output designs
Virtually all commercial products are not designed to output a range of different output voltages; they take one input and generate one output. Therefore, to use off-the-shelf components, you have to use a whole bunch of different independent converters, which is bulky and probably a lot more inefficient than an integrated design.
To avoid these problems, the alternative is to make a custom design: a board that takes a large 12V input, divides and regulates it to a range of different smaller voltages according to what is wanted. So, for example, the board could have 6x 3V outputs, 8x 5V outputs, 5x 9V outputs, etc, whatever I want. The challenge is that making such a board, especially one that would have good efficiency is potentially extremely complicated. For example, I have been reading the book "GaN Transistors for Efficient Power Conversion" by Alex Lidow and "Switch-Mode Power Supplies" by Christophe Basso and it is evident that making a such a board could be extremely complicated and take a long time, at least for me. Although one alternative I guess would be to pay an expert to design it.
There is a halfway solution here and that is to make a custom board, but to use switchers, which come in SMD packages. You just plaster a bunch of switchers on the board, so you solve the bulky multiplicity of devices problem, but the design is not really integrated--it's just a large number of switchers in the same box.
So, what would be a good solution here? If I am trying to get >95% efficiency, is cobbling together commercial products "good enough" (Option 1) or should I look at a multi-switcher box (Option 2) or go whole hog and try to make a custom board (Option 3)?