Shhh! Be vewy vewy quiet - I'm hunting Power Suppwies!

Today has been a long day.  I've spent just about all of it looking into a solution for powering the Mini-ITX board from a battery.

Research

To start with, I looked for articles on SEPIC converters, and efficient alternatives that might exist.  One article on Maxim's website explains the better efficiency achieved by using a boost converter and low drop-out regulator in place of a SEPIC converter where the inlet voltage can be above and below the output voltage.  Specifically, the supply was a single cell Lithium battery, and the output was a 3.3V rail for digital ICs, as used in a lot of portable equipment.  The results were fairly promising, though the overall efficiency has to take into account the proportion of the battery life through which the inlet voltage will exceed the output.  During this time, the linear regulator will cause moderate loss.

In this application, the output needs to be 12V.  A lead-acid battery's output doesn't usually stay much above 12V for long after it's removed from charge.  This makes the boost + LDO option quite attractive, especially if the components are more easily available than for a SEPIC converter.

Off-the-Shelf Units

Though I like reinventing the wheel, it's sometimes sensible to leave the design and construction of power electronics to other companies, who spend their time designing and testing this sort of kit.  I found various solutions using off-the-shelf hardware.  Some of them aren't very affordable though:

• ATX units that would be able to run directly from a battery are available - they cost about £85 from RS, so are far too expensive to contemplate.
• DC-DC converters to supply a steady 12V running from the battery cost about £26 and £35 from RS - for 2.1A and 4.2A output.  These are a bit more affordable but are still a bit pricey.

There shouldn't be a need to get something that delivers an ATX output unless it's cheaper than buying a 12V output converter.

WEBENCH

National Semiconductor has an online simulation facility which basically does all of the circuit design for you if you put the parameters in.  I tried it, but the output isn't all the helpful - there's not much explanation of some of the traces, for example 'Vout' appeared to show the AC component of the output ripple, but no mention was made of this and so I'd expected to see the DC component too.  Since there's nothing to say what it represents, I'd be unhappy trusting the design.  If National actually provided models for their control ICs, I'd simulate them myself using LTSpice or something similar.  I don't intend to use this facility again, as it was fairly appalling to use.

Sensible Requirements

When the DC-DC converter specification was drawn up in the previous project run, it was specified for a 6A output current.  Given that the board consumes between 1 and 1.5A without the hard disk connected (and somewhere around 2A max with it connected), this seems a bit on the excessive side.  This might be worth bearing in mind, because higher current specifications make things very expensive, especially when you also need low output voltage ripple.

SEPIC ICs

Since it looks to be far cheaper to build a converter than to buy one, I looked into the controller ICs available.  Just about all of them are in some form of SOP package, which is very very tiny, and would be an absolute nightmare to work with and mount.  I did find some ICs in DIP and SOIC packages, but these tend to be the more integrated ones designed for much lower currents, such as small portable device supplies and LED drivers.  I considered adapting one of these into a suitable SEPIC circuit, but at the moment I don't really know where to start.

The Ordinary Way

It seems that there are a number of people who use Mini-ITX boards for things in their cars.  This is basically the same problem that I'm faced with, and one web page I found recommended the use of an inverter and a standard ATX supply, as it works out much cheaper than buying specific low-voltage devices.  I've considered this idea, but I think the equipment would be much too bulky - not to mention the PAT testing requirements, etc.  There is a possibility that some ATX supplies might have a low voltage side that can be somehow bodged to work from a 12V nominal inlet.

Best Course of Action

At the moment it looks like an off-the-shelf converter would be the way to go, partly because the unit could be expected to work flawlessly, and most units are protected against things like short circuits - they're quite robust.  Another option I saw was a demonstration board, for example from Linear Technology.  They are very expensive though, at about $150 a piece.  They provide a 12V, 5A regulated output across a really wide input range.

I'm still considering a fully discrete design as a viable option, though the quantity of work required might be excessive.