Slashdot Mirror
Google Offers a Million Bucks For a Better Inverter
An anonymous reader writes: With the Little Box Challenge, Google (and IEEE, and a few other sponsors like Cree and Rohm) is offering a $1 million prize to the team which can "design and build a kW-scale power inverter with the highest power density (at least 50 Watts per cubic inch)." Going from cooler-sized to tablet sized, they say, would make a whole lot of things better, and the prize is reserved for the best performing entrant. "Our testing philosophy is to not look inside the box. You provide us with a box that has 5 wires coming out of it: two DC inputs, two AC outputs and grounding connection and we only monitor what goes into and comes out of those wires, along with the temperature of the outside of your box, over the course of 100 hours of testing. The inverter will be operating in an islanded more—that is, not tied or synced to an external grid. The loads will be dynamically changing throughout the course of the testing, similar to what you may expect to see in a residential setting." The application must be filled out in English, but any serious applicants can sign up "regardless of approach suggested or team background." Registration runs through September.
Word is, there was no spec for lifecycle so the devices met the contract as stated, and the government couldn't return the devices.
So I'd recommend to Google: At some point, look in the box.
Oliver's law of assumed responsibility: If you're seen fixing it, you will be blamed for breaking it.
Must be able to handle up to 2 kVA loads
Must achieve a power density of equal to or greater than 50 W/in3
Must be able to handle loads with power factors from 0.7–1, leading and lagging in an islanded mode
Must be in a rectangular metal enclosure of no more than 40 in3
Will be taking in 450 V DC power in series with a 10 resistor
Must output 240 V, 60 Hz AC single phase power
Must have a total harmonic distortion + noise on both voltage and current of 5%
Must have an input ripple current of 20%
Must have an input ripple voltage of 3%
Must have a DC-AC efficiency of greater than 95%
Must maintain a temperature of no more than 60C during operation everywhere on the outside of the device that can be touched.
Must conform to Electromagnetic Compliance standards as set out in FCC Part 15 B
Can not use any external source of cooling (e.g. water) other than air
Does not require galvanic isolation
In the USA/Canada typical residential setups use two conductors at 120V to ground, but the conductors are out-of-phase so there is 240V between them.
There really isn't any such thing as 110V or 230V in the USA/Canada, both of which you'll sometimes see referenced. 208V does exist, it's the difference between two legs of a 3-phase setup where each leg is at 120V to ground.
Except it requires more wires. 220/240V split phase requires 3 wires.
3-phase generally requires 4.
And unless you really need 3-phase, split phase is easier to deal with - with 3-phase you need to monitor all three phases to ensure they are working (failure of one phase is a common failure mode that requires immediate shutdown of the other two phases lest any dangerous currents develop).
Though, one thing I don't get about this challenge - they're using they want 2kVA output, but then demanding 50W/in^3 with a max size of 40in^3, meaning you have to provide 2000W.
And 2000W can mean providing way more than 2000VA. (The reason we use VA for inverters instead of watts is VA captures virtual power. 2000VA requires just as much power handling components (transformers, transistors, etc) as supplying 2000W at a 1.0PF (i.e., all resistive). Even if you have a really bad power factor and your real power draw is only 1000W - the hardware has to be able to instanteously supply the current and voltage for 2000W at periods in the cycle. The virtual power is virtual, because it's "given back" during another part of the cycle, but that means all the equipment has to handle it.
A lot of electric companies will have a power factor surcharge because of it - if your power factor can't be corrected to within limits, they charge more because they have to install bigger equipment.
The only real saving grace is that the input voltage is 450VDC, so you're really just doing a buck converter.
Stupid objection the first: "This is worth a lot more than a million dollars."
Response:
Stupid objection the second: (something stupid about 12 volts)
Response:
I know that slashdotters don't RTFA, but seriously, all of you jaw-jacking about 12 volts or about how a million is chump change are a bunch of Useless McToolbags. STFU already.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
and this is the reason it is 48V.
Copper is expensive. You don't want your 42U rack, which pulls 8kW powering it's 42 1U servers, drawing upwards of 700A.
For a 1% power loss over 1 metre at 700A, you need 0000AWG cable. It's about 1kg of copper (that's a single conductor, you'd need one for the return path, another kg and another 1% power loss)
But 0000AWG can't actually handle that amount of current without active cooling. for 90C rating, it's only 260A.
700A for 1M distance has cost you 160W of power and 2kg of copper and you need some fancy cable cooling technology.
If you upped the voltage to 48V, you only need 175A.
You'll be fine with 2AWG cable if you can keep it cool and only lose 30W. You could use 0AWG, which would cost you 1kg of copper and only have 19W of wasted power.