USB Lab Power Supply

[Maxlor]

In some older episode, dave mentioned the idea of a USB-powered lab power supply. The idea struck a chord with me, even though Dave seems to have abandoned it. I like the idea because it would be somewhat of a challenge to implement. I could take parts of Dave's PSU suggestion and modify it to make it work from USB. At first glance, it seem like a nice idea, because USB is ubiquitous, you'll find cheap chargers or a nearby laptop everywhere, and most projects don't need all that much power. I mean, even the arm-based GSM-and-OLED-equipped thingie I write software for at work stays under 1.2 Watts most of the time.

So I've been mulling over how to design such a PSU. I'm thinking the power supply must be isolated from the USB port, in order to protect a laptop if things go wrong. So an isolated transformer-based DC/DC module would be the first part. I found some parts from Analog Devices (their isoPower devices) that do just that. Then a step up converter to get, say, a 15V from which the 0-12V working voltage could be derived from, maybe using buck/LDO combination for efficiency like Dave uses in his current lab power supply design.

Now, the problem I have is that just sketching this in my head results in an overall efficiency of just about 50%. The isolated dc/dc modules seem to hover around the 70% mark, and the following voltage transformations eats into the budget some more. Now, a standard 500mA USB port doesn't provide all that much power, and a mere 1.25W (2.5W times 50% efficiency) isn't all that much, I'm thinking the whole thing would be pointless then. 2W of output power off a standard USB port would be nice, I hear you can fly to the moon with that (and ouf course, if some charger delivers more, all the better.)

So what I'm asking you is: do you have any clever ideas to improve the efficiency of a circuit that transforms an input voltage of 3.6-5.75V (which seems to be what USB ports output under varying load conditions) to an output voltage of 0-12V? Eventually, I'm hoping to figure out how to achieve 80% overall efficiency.

This is really my first forray into analog electronics, I've only done digital stuff so far, mostly ignoring the analog side apart from some simple filters do get rid of undesired ringing and so on. If you think this is too advanced, I'd like to hear your thoughts too. Cheers!

[Zbig]

Seems like a good solution for powering world's tiniest (electric) violin amp...

[mariush]

There's nothing physically limiting the usb current to 500mA, in most computers.  In general, up to 4 usb connectors share a power source and there's a polyfuse or some protection at about 2A.  So you could use a single cable to draw more than 500mA - I would consider 1-1.2A safe.
The alternative would be to design something that would use two usb jacks just like some old scanners and some external hard drive racks used to be sold, giving you 1A+

However, keep in mind that USB isn't required to give you exactly 5v, the pc can give +/-5% I think.. or maybe it's +/- 10%... I don't know right now, I think you can officially have about 4.75 - 5.3v on the USB. Good power supplies will do 5v or more.
After that, you have to keep in mind the resistance of the wires in the usb cable itself - sometimes the cable is so crap that at 0.5A there's about 0.2-0.3v drop on the cable itself, so even with 5v at  connector, you could have 4.7v at the other end.

If you want to have the power supply isolated (remember, the usb on a computer is not floating), you'll have a lousy efficiency due to converting dc to ac then going through transformer losses, then back to dc .. after that, yeah, there are some boost/sepic ICs out there that can do about 92% under some parameters (depends on vin-vout difference, current, etc).

If I were forced to design something I would probably consider using some lithium battery, charging it through usb port and decoupling the battery from usb (using maybe a relay) when the power supply is in use (to have it isolated from computer). The battery can provide the high instant current and it can charge even when you turn off the computer, if you plug the power supply in a usb port that's powered from Vsb (stand by power of power supply).

[Maxlor]

Actually, USB Voltage can vary a lot. From my research, the limits are 4.25 and 5.75 (so, +- 15%). Current limited chargers that don't do any data transfers are usually designed so the linearly drop to 3.6 at maximum current, then drop very rapidly if you draw more.

Using a battery was an alternative I've considered too. A couple thousand Ah should fit into a small enclosure. Let's be conservative and say it has 2000mAh, which would be around 7Wh-ish. The advantage would be that you can draw 7W from it (assuming 1C capability, again being conservative), but of course it'll be empty after a mere hour, and we all know time passes very quickly if you're tinkering

The isolated USB connection could then be used to just charge the battery. That works, but it won't increase efficiency much, i.e. after losses you'll probably still only put 1.2W into the battery. But maybe it's enough? Would that be enough for your tinkering? How'd you feel about a PSU that can output 7W short term (up to an hour) and 1W constantly? Or maybe you'd just like a bigger battery?

[mariush]

No, USB voltages should NOT vary that much.  The USB standard allows for a maximum of +/-5% or +/-10% ...I can't be bothered to read and link the specs now and point you to the page, but it is there.

If you have something that outputs 4.25v then it's a shitty power supply or something, either way it's out of spec.
If you have something that outputs more than 5.3v or so, again it's shitty... but there's plenty of phone chargers out there that use 5.6v zener diodes as a crude mechanism to block overvoltage, so I'm more lenient here.

Now of course, it should be 5v +/- something AT THE CONNECTOR. If you pull 500mA or thereabouts through a flimsy cable and measure the voltage at the end of the cable, I wouldn't be surprised to see you measuring 4.25v - doesn't mean it's right.

ps. I didn't check personally the specs, but I've heard USB 3 is allowed to go as low as 4.4v @ 900mA I don't think you thought of that when you said the numbers.

[alm]

I'm pretty sure the 4.25 V figure is about right for worst case. The 5% or 10% figure might be at the USB port, but the USB spec also allows for some voltage drop over the (max. 5m) USB cable and a bus-powered hub.

The ~1A figure might work for USB 2.0 ports, sometimes. Until you plug in four of those devices (or other high-power USB devices like external hard drives) in all four free USB ports. Or plug them into a laptop. That's the difference between a hack and a properly engineered device.

I would probably also go for the battery idea, wasn't that also Dave's plan before he abandoned it?