2
\$\begingroup\$

This is something that is troubling me in many of my projects and I would like to know if there is any standard approach to it.

The question is how should I correctly design / dimension my power supplies to be sure that enough current will be available for my CPU.

I am using various packet processors from Marvell and at every one of them they provide, like CPUs from other vendors, two different set of current/power specification.

  1. One set is the TPD (thermal power dissipation), where they say that this numbers should be used for the thermal design. The numbers are measured under nominal voltage and some selected use case.

  2. The second set is the maximum continuous current that can be drawn from the power supply and is given to support the power supply design and selection of power module. These numbers are given using maximum recommended voltage specification for the power rails, at max recommended operation temperature and some extreme use cases.

One could say that the answer to my question is obvious: use the second set of data for correctly selecting the power supply, since this is why they are given for.

But the thing is, these numbers are at maximum recommended voltage (usually +5% of the nominal voltage) and at maximum operation temperature. Both of these conditions are not met at my designs.

Is there any standard approach to this? Should I just use this maximum continuous current specifications when defining the maximum current of my power supplies or should I somehow (how?) at least scale them down to the typical voltage supply?

My concern is not to over-design the power supply, since the numbers between the two sets differ -usually- a lot.

\$\endgroup\$

3 Answers 3

3
\$\begingroup\$

Design for worst case and add some derating to keep on the safe side depending how hot the ambient can get etc. Never ever design something with the mindset "hey, we're using 5A current so I'll use 5A parts, m'kay"!

There's IPC-9592B standard which is squarely aimed at codifying how much derating you actually need to be "safe" or at least it gives you something to refer to when the house burns down and investigators come calling.

Douglas Alexander's website has a derating table but these are NOT IPC standard, it's just something he's come up with experience. I don't obviously endorse them for any commercial use but it gives you an idea what a proper design margin would look like.

http://www.componentsengineering.com/procedures-guidelines-2/component-derating-guidelines/

http://www.ebnonline.com/author.asp?section_id=1541&doc_id=235329

\$\endgroup\$
2
\$\begingroup\$

Yes, the numbers differ by quite a bit, and yes, your PSU should be able to provide the max continuous current to guarantee a reliable operation. Otherwise you risk to spend your time hunting down rare transient glitches resulting from brown-outs.

As a real-life example, check out the Raspberry Pi, which requires a 2.5A PSU but actually boots with as little as 500 mA. Problem is, it's not reliable with such supplies: there are countless questions on raspi.SE (and everywhere else on the net) which end up being a PSU problem (example from 2 days ago).

IMO you'll do yourself a big favor if you over-design the PSU, at least in the prototype phase. Once you have working prototypes, you can run benchmarks to determine what current is requited in production to meet your reliability goals.

\$\endgroup\$
1
  • 2
    \$\begingroup\$ I use RPi for internal testing and video source regularly with SMPS circuit not even close to a 2.5A supply. Then again we are NOT watching videos or playing games or anything like that. So, yes, you can get away with 1.4A peak SMPS circuit but it depends on your use case. \$\endgroup\$
    – Barleyman
    Commented Jul 28, 2017 at 13:20
1
\$\begingroup\$

The supply has ouput transistors for regulation, so these transistors should always be inside their safe operating area (SOA). This is de-rated by current and temperature. Find proper transistors by examining the data sheet

The current capability should be exactly the maximum the whole system can draw, or slightly more. Because switching converters are usually the most efficient in this area of operation, close to maximum output current.

If the cpu gets too little current while operating, the voltage drops and it will detect a "brown-out", automatically shutting down. This is why the power supply should be capable to handle the maximum continuous current. If not the system will glitch under heavy load.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ I'm sorry, but how is this relevant to my question? Never said that the power supplies should exceed the rated voltage of the CPU. I was only talking about current. And what about the "brown-out"? This is irrelevant. \$\endgroup\$
    – nickagian
    Commented Jul 28, 2017 at 6:43
  • \$\begingroup\$ @nickagian I'm editing my answer. I've interpreted your question wrongly in regards the maximum voltage. \$\endgroup\$
    – user55924
    Commented Jul 28, 2017 at 12:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.