How can I design a robust and professional power supply circuit?

Question

I'm planning on doing a large project that would include an ARM microcontroller @ 50 MHz interfacing with a graphic LCD, user buttons, some sensors and some actuators (DC motors, valves) controlled through relays. I want to design a robust and durable power supply for this project that must have a 10+ years lifetime. I have a 24V power supply and I have to design the circuit that will give me 3.3V/5V and at least 1A. What protection components should I include in the circuit? Where does all this components come into place: PTCs, ferrite beads, filter capacitors, EMI filters, fuses etc.? Could you give me a block diagram of such a circuit?

Answer 1

Sounds like a homework question :-).

If you need to ask at the level of detail you are asking you will need more input and learning than you will get in a single answer here.

If this is a real project that is worth money then that is the sort of task that is worth spending money on the design of for professional input.

But:

Assume design life is 100,000 hours = 11.4 years x 8765 hrs pa.

• Capacitors can be aluminum electrolytic IF you design properly. A 2000 hour capacitor will need to be run at 60C under rated temperature . So 105C caps need 105C - 60 = 45C ambient or less and preferably much less. That sounds easy but must account for local temperature + enclosure rise + self heating + radiation from other components + any local hot spotting from other air flow. Care will be needed. 3000 or 4000 hour caps will help. Be aware that aluminium wet electrolytic caps die FASTER at a given temperature when unpowered than when powered (due to dryout).

• Solid Aluminum caps will be easier to use where size and price allows.

• Tantalum caps are tempting and work well if design is immaculate and if reality follows theory. If this is in a game it's worth the risk, perhaps. If it's in a sub or spacecraft then send tantalum packing now.

• Understand temperature derating, ripple current derating.

• Buy components of known trustable brands AND ensure that what you buy is what it claims to be. Incoming inspect as much as needed to maintain certainty.

• Properly manage ESD issues (electrostatic discharge), if it says don't bend closer to xxx from seal then don't, if it says clamp lead to prevent shock damage while cutting or bending then do. Similarly take proper note of manufacturers advice re max storage time at xx% RH, retreatment required for packages open too long, reflow soldering temperature profiles, advice that part may not be solder by means of xxx, ultrasonic cleaning warnings, solvent cleaning warnings, do not stack xxx way, do not apply force to xxx, ... manufacturers advice.

• Design properly. Use worst case parameters, pore data sheet for exceptions and special requirements. err on conservative side.

• For any of the following that you wish to be protect from take due note: Assume worst worst case mains transients, sags, brownout, lightning strike, acts of God, acts of children, acts of drunks and people of low IQ, acts of mice rats ants and cockroaches (gnawing, urinating, defecating, nesting, dying, ...), 100% condensing atmospheres, low humidity, air conditioning failure, coffee spills, Coke spills ... .

• If you care, assume that 110 VAC equipment will be plugged into 230 VAC mains. Assume that 60 Hz equipment will be plugged into 50 Hz (iron cored transformers care, other things may) and vice versa.

• Understand longitudinal and transverse mains filters. Understand X & Y filter capacitor ratings. Realise that mains to output ground Y caps can produce destructive output voltages(typically half mans voltage).

• Allow component degradations and changes of characteristics with time - capacitor dryout, LED degradation (especially including opto couplers), iron cored coil binder thermal degradation, ... .

• Be aware of why components have rated values - voltage ratings for resistors, surge (not fusing) ratings for fuses, temperature rise for tracks, resistor current as opposed to power ratings, power semiconductor peak vs max operating ratings, dV/dT opto ratings ... .

That's a once over lightly out of my head start. There can be much more. Skimp on or ignore almost any of these and your 10 year lifetime is suspect. 100,000 hours is a long time. Survival is usually via gross dumb overkill or skilled design but seldom due to luck. If you are feeling lucky you probably wont be.

Produce a total picture of what you wish to do, known hazards, know mitigations, sensible solutions. Worry it to dearth if you don't want it to die.

More ...

___________________________________

This superb reference, supplied by @davidcarey, is essential readung.

Underestimating Complexity of Power Supply Design - Underestimating the complexity of power supply design can lead to schedule slips, cost overruns, and excessive field failures.

Useful:

Best manufacturing practices site - Reliability
and their download list - some relevant.

MORE POWER FOR THE DOLLAR 149 pages
Price vs Value - A Technical Guide
NAVSO P-3641A (Replaces NAVMAT P-4855-1A) October 1999

Power Supply Standards

Answer 2

The most vulnerable components as far as lifetime is concerned are electrolytic capacitors. Standard lifetime is 2000 hours, long-life parts are usually 5000 hours. That's not much, but there' a rule-of-thumb which says that for every 10°C you stay below specified you can double lifetime. So a 105°C cap working at 55°C will have a lifetime of 2000 * 32 = 64000 hours. That's 7.4 years of continuous operation. Long-life types will get you over 20 years.
Here we touched an important factor: temperature. High temperatures decrease components' lifetime. If you want to go from 24V to 3.3V at 1A in a linear regulator you'll have to dissipate 20W. That's a lot. A switching regulator may be a better solution.

Concerning special protection components I would use a fuse to protect the product as a whole, and a varistor on the power supply input to protect against spikes specifically.

Answer 3

Designing for such requirements goes far beyond hobbyist / hacker competency.

First, you need to have a lot of knowledge about power. You need to fully understand whatever topology you intend to implement to be able to build an accurate mathematical model of it before you start building (to make sure that what you're going to build will work). You need to know where the stresses are, how to measure them, and how to improve upon them if necessary.

The laundry list of components you mentioned in your question leads me to believe that you're not an expert in power supply design. If this is the case, I strongly recommend either finding an expert to take on the design, or contract the design out to a company specializing in power conversion.

Beyond knowledge - unless you have a well-equipped laboratory (well-suited for power development, that is) you won't even have the ability to properly measure most critical parameters. Without this data, there's no way to know if what you're doing will have any chance of long-term reliability (or short-term explodability, to coin a term.)

Do you have an oscilloscope with good input bandwidth? High-bandwidth probes? Accurate current shunts? Accurate high-bandwidth current probes? A gain-phase analyzer? A spectrum analyzer? An EMI analyzer? A power analyzer? A data acquisition unit? Thermocouples? An IR camera? This is just a sample of what you need to properly evaluate a power supply.

Yes, I said evaluate. Even if you don't do the design yourself, without most of this equipment you cannot reliably test a power supply. You're at the mercy of the supplier.