I'm in the process of designing a small, 5x5cm ATmega1284P-based microcontroller board. I would like it to regulate 6-18v input to 5v output, with a max continuous current rating of 1.5A.
At first, I was looking at an LDO. But for this much power draw, it would be really inefficient.
I then moved on to looking at a UBEC. Instead of having a linear converter on the PCB with the rest of my components, I was going to use a UBEC, which is a switch-mode power supply that sits inline between the battery and the microcontroller PCB.
However, I then thought about putting my own switch-mode power supply on the PCB. I made a circuit using WEBENCH that has an efficiency rating of over 90% and max power dissipation of just over 700mW. It is based on the TI LM20333. However, it seems a little complex and hard to solder by hand (since it uses several 0402 SMD packages). Also, I can't seem to find a power inductor that is not huge or expensive.
Would I be better using a UBEC, or making my own switch-mode power regulator circuit? Is there a certain simple and efficient circuit I should be using?
Best Answer
Switch mode power supplies (SMPS) can be difficult to make by hand, without a custom PCB. You have several issues. Even on a 2 layer custom PCB this can be difficult but is many times harder when you are hacking things together on a perf board or worse. The reason for this is that the stability of a SMPS is highly dependent on the inductance and "routing" of the signals. If a signal is not routed nicely then it can pick up too much noise and the SMPS becomes unstable. This alone is a topic worthy of several books.
Things can be much more difficult when using tools like National Semiconductors WEBENCH. While WEBENCH is useful for getting you "approximately in the right neighborhood", it falls far short of getting you a robust and reliable design. WEBENCH, and tools like it, often don't check for some basic things that become very important when designing SMPS. There is the easy stuff, like are you violating the power specs on resistors or voltage specs on diodes? Or the intermediate stuff like are you violating the duty cycle limits for the SMPS chip? Or the hard stuff like, is the tool properly simulating the MOSFET/Diode/Cap you're using or does it allow you to add/change the snubber circuit? I have seen WEBENCH and tools like it fail to properly take these things into account-- resulting in a circuit that just barely works or even fails catastrophically. WEBENCH will give you a place to start, but it is no replacement for reading and understanding the datasheet and doing the proper homework on the design. You can't just take a WEBENCH design and run with it as if it were a fully tested and robust design.
My advice: If you're building your own custom PCB then make your own SMPS. Add into your schedule a good amount of time to tinker and tweak the SMPS circuit. If you are going with a one-off build then use something like the UBEC.
Advice #2: Linear Tech's LTSpice simulator is an order of magnitude more useful than WEBENCH. It is also good for non-SMPS analog circuits. You will still need to do your homework, but LTSpice will get you much farther along. The downside is that Linear Tech's chips are a little more expensive than National Semi's and others. But, in my opinion, unless you're building thousands of these things then the better software tools is worth the extra cost.