Power Supply Design - Multiple Voltage Regulators

I'm trying design a benchtop DC power supply for powering small projects, breadboards, arduino, etc. This is a list of what I want from it:

• Fixed voltage rails for 24V, 12V, 9V, 5V & 3.3V
• Edit: Current requirements: 1A will be alright, 2A I'll be happy with and anywhere close to the full 3A on offer from from the laptop power brick I'm using (see below) will be great.
• Variable voltage rail from ~0V to ~24V
• Pulse width modulation on the variable rail
• Voltmeter panel on the variable rail
• Power on indicator LED
• USB ports for charging USB devices
• (Other stuff that's not related to this question such as selectable methods for dealing with transient voltage spikes from using PWM to power inductive loads, adjustable PWM frequency ranges, etc)

I will use a 240V AC to 24V DC laptop power brick capable of outputting 3A.

To produce the desired voltages, I intend to use a linear voltage regulator for each rail, each taking 24v input from the power brick. An overview of my intended setup is shown below.

The PWM circuit I'll be using will be powered from the 12V rail. The USB ports will come off the 5V rail. The power on indicator LED (not shown) will most likely come off the 3.3V rail.

Here are my questions:

1. I'm assuming the voltage regulators should be in parallel, each taking the full 24V input, even though there will be a large difference between input and output for a few of them (eg 24V to 3.3V). I thought about putting them in series so that the 12V output of the first regulator would feed the 9V regulator which would feed the 5V regulator and so on, but the circuitry of each regulator block in my diagram above would divide away some current, leaving me with very little of the 3A max input for the lower voltage rails. Is this correct? I'm also assuming that the larger reductions in voltage necessary for the parallel arrangement as shown will produce more heat than in series, but each regulator will have a nice fat heatsink and the whole thing will be enclosed in something with at least a fan or two and plenty of holes for ventilation.
2. The variable voltage rail will be controlled with a potentiometer into a variable voltage regulator (LM350). Should I use different fixed voltage regulators for my fixed voltage rails, or use an LM350 for each, and set their outputs with either fixed resistors or trim pots?
3. As the laptop power brick I'm using outputs 24v already, shall I wire that directly to an output terminal for my 24V rail, or should I also put a voltage regulator in there? Ideally, there'd be no point regulating 24v to 24v, but I'm not sure how reliable the voltage is from a laptop power brick. I'm not even sure you can get out of a regulator exactly what you put in - there must be some voltage dropped along the way. If necessary, is there another way I could ensure a constant 24v output?
4. I saw a tutorial somewhere a while ago where someone hooked up a voltmeter to a psu, but the instructions mentioned using a 9v battery and a relay to power the voltmeter rather than drawing power directly from the internals of the PSU. Should I do this? There was no reason given in the tutorial I read. As mentioned above, I was going to draw power from the 12V rail.
5. Is there anything missing that could be considered essential to the design of a power supply? Any kind of safety features? The LM350 appears to be rated for 3A and has inbuilt overcurrent protection. I'll also be adding fuses to each rail. Will this be adequate in the event of a short circuit within any external load (eg poorly wired breadboard, etc)
-
I'm confused by '0-24V DC PWM' - is this an adjustable peak voltage rail with a variable frequency square wave (chopped) output? – Adam Lawrence May 16 '13 at 12:36
Many of the self-contained DVM modules require that their power supply is isolated from the voltage being measured - your item 4 would refer to these meters. However, there are some DVM modules that do permit their power ground to be connected to the ground of the voltage being measured - read the specs carefully for this feature when selecting a DVM module. – Peter Bennett May 16 '13 at 16:26
Madmanguruman - Yes, that's the plan. Peter Bennett - +1 for clearing that up. – John May 19 '13 at 13:51
Could you give us ballpark numbers for how much current you want on each voltage rail? That could possibly make a big difference in linear vs switching regulators, parallel vs serial, etc. – Joe Baker May 19 '13 at 17:01

I agree with others that switchers are a better choice in terms of efficiency, but they can be somewhat complicated to deal with if you're inexperienced, and there can be lots of weird effects that aren't immediately obvious (precharge sinking, beat frequencies, etc.) that can make life difficult. Assuming you've figured out your power dissipation and know how much current each rail can deliver, if the linears will work for you, stick with them (at least for the first pass).

If you're trying to achieve a variable-amplitude square wave output on your adjustable rail, the chopping may introduce noise into the main 24V rail, which could show up on the other rails. You may want to have an LC filter between the main 24V rail and the regulator input to provide high-frequency isolation, and will probably need extra capacitance on the adjustable regulator output (bulk electrolytic as well as low-impedance ceramic) if you expect the square wave edges to be sharp.

1, 5) There are some dangers with your scheme.

Power dissipation in the linear regulators will be

$(V_{out} - V_{in}) \cdot I_{out}$

which is significant, especially for the lower output rails. 78xx-type regulators have built-in thermal protection around 125°C, and (without heatsinking) a junction-to-air thermal resistance of 65°C/W. Your thermal management will be challenging.

Another potential problem - if the series-pass element in any of your low-voltage regulators fails or gets bypassed (shorted), you'll present the full 24V input to the output. This could be catastrophic to low-voltage logic. You should protect your low-voltage rails with SCR crowbars that can sink enough current to put the DC/DC brick into current limit and collapse the 24V rail (they'll need big heatsinks too). Fuses are unlikely to be good protection since the 24V brick likely isn't stiff enough to generate the $I^2 \cdot t$ needed to blow a fuse.

2) Whatever floats your boat.

4) Meters aren't huge loads. Just use one of your rails.

3) Correct - all regulators have headroom requirements. If you want the maximum 24V out, you'll need a direct connection, and will have to rely on whatever intrinsic protections the brick will provide you.

-
+1 for the crowbar suggestion – Andy aka May 16 '13 at 13:38
This is my first real project so it looks like easier linear regs are the way to go. I think regulating 24v to 3.3v @ 3A will kick out about 62W from a single chip. Less so for the higher voltage rails, but still significant. So it seems the only sensible option is to link them up in series with at least one intermediate between 24v and 12v. Something like 24->18->12->9->5//5(USB)->3.3V Hopefully this will cut the power dissipation for individual regulators to a more manageable level. I shall look into what you suggest for square wave output and find out what a crowbar is. Thank you! – John May 19 '13 at 14:10

You haven't mentioned current output requirements but, the biggest problem you'll have is power dissipation and I'd urge you to consider using switching regulators for 12V, 9V, 5V and 3V3 and if necessary use them to generate 13V, 10V, 6V etc. and have Low-Drop-Out (LDO) linear regulators to take the drop to the final voltage you need.

The benefits are a significantly lower heat dissipation on the design and higher current outputs especially on 3V3 and 5V.

1) Regulators can be in series or parallel but it will be the 12V regulator that generates most of the heat if using linear regs in series. If using switchers then providing they can operate with an input at 24V+ then parallel is probably better BUT if you find a really decent 12V switcher and tee-off the other lower voltage switchers/LDOs from its output then this will work too.

2) I'd use switchers and linear LDO (optional)

3) I'd wire it in directly

4) I don't understand this point BUT neither do I see hanging a meter on the output as a problem.

5) Add fuses before the regulators if you must but it's likely that most regulators will have over-current protection easily added if not intrinsic to the device. I might be tempted to run the USB from its own 5V reg circuit. I might also be tempted to provide a -9V (or -12V or -5V) rail too.

EDIT - I'd probably put a LED on each output

-
I've been reading up on switching regulators after reading your answer, and it seems they would be the way to go in an ideal world. I assume you recommend regulating to Vout+1 on each rail, then using the LDO regulators to take it down a volt to reduce noise caused by the switching regulators? However, I have quite a few LM350 linear regulators available, and as I'm on a budget and can't seem to find any suitable switching regulators, I may press on with a linear regulator-based solution. Thank you! – John May 19 '13 at 13:55
Drop-out voltage is 2.8V on the full 3A for an LM350 so they could be used with a switcher but the switcher would need to supply 14.8V for the 12V regulator etc.(see fig 10 on LM350 spec). It might be worth considering making provision for a switcher and bypassing it for initial setup and test. The LM2679 is dead easy to use and comes in 3.3, 5 and 12v fixed versions or an adjustable version ti.com/lit/ds/symlink/lm2679.pdf – Andy aka May 19 '13 at 15:47