The input is DC (450V), so there is no need for bulky capacitors.

The relative low output current ( 8A at 240V) allows the use of MOSFETs at high frequency, so ouput filter inductors can be small. And no transformer is needed as galvanic insulation is not required.

Yes, I am perhaps optimistic too, but I think that it is possible to put these 2 kW into 655 cm3. I think that the main problem is to get rid of the heat, so a small fan can be necessary.

Hi Francesc,

thanks for sharing your opinion.

I had similar thoughts especially regarding the semiconductors.

Using the latest MOSFETS, I envision Rdson of

How hard can it be? Pretty darn hard or Google would not be offering a large cash prize. I really hope someone will win!

Achieving the design specifications is not extremely difficult. I am sure I could do it for much less than 1M$.

An advantage is that there is no limit in the cost, so you can use very low Ron MOSFETs (even the new ones not made of silicon), powerful drivers, expensive ceramic capacitors, etc. And I would use resonant transitions to reduce switching losses and EMI. 

Reaching a 98% efficiency, which seems feasible, putting the power electronics on aluminium substrate and using the box as a heat sink, a box of 10x60x1 cm (0,6liter) with the appropriate finish, seems to be able to dissipate the heat with no fan. Another option would be to make the unit more cubic shape and use a small fan.

However the problem is that you do not get the1M$ just designing something that meets the specifications, your design has to be the best of all. Even if your desing is one of the 18 selected as the best, you do not get nothing.

And if you leave outside the USA you cannot send the prototype, you have to carry it personally.

So the risk of doing a lot of work and expenses for nothing is extremely high, too high for independent designers or small companies. So after having used the chalenge as brain teaser for some minutes, I leave it.

Dear Claire,

Francecs has a valid point - its an all or nothing situation and I don't see small  companies or even freelancers participating in this challenge.

But my point is, that today it is a stand-alone industry that builds inverters, doing so based on off-the-shelf components. The key to win this challenge in my opinion is not about components but about integration, thus I think it will be a semiconductor manufacturer to win this challenge.

Who else could integrate all possible items into a single, small package and make it a mass product?

Go ahead and count the parts in an inverter and compare to what is integrated in a microcontroller.

I don't hope for "someone to win" this challenge because for sure someone will :-)

My prediction today is, it will not be a well-situated inverter building company.

What do you think?

They are offering funding of about $20k to $40 to university researchers.

https://research.google.com/university/relations/littlebox.html

This could be a year of funding for a graduate student... to work on the project while completing a PhD. From my point of view, a graduate student could be pretty happy with $20K US even if there is no hope of winning the prize. My the little box challenge is a) seeing future cost reductions in inverters and b) seeing improvements in thermal conductivity technology.

Dear Martin,

As you note this should be doable and being with Infineon you (already) should have a good idea. I think it would be best for any project to possibly work with a manufacturer to do this. I see Google does provide such a list.

As Claire notes the prime focus will be cost and volumes as this is what drives the power semiconductor market. It would be best to try and integrate all power electronics, DSP, sensors etc all onto a single substrate to reduce the packaging costs and allow for easier thermal dissipation.

But if 20 kW/l is available then then 4 kW/l should be doable.

I think the issue is not so much power density but as Claire notes heat dissipation as this has always been and remains the bugbear as at some point the reduction in packaging costs get swamped by rising cooling costs. As I understand matters greater power densities may be achieved as this depends upon the line width technology and the power electronic industry is still several generations behind in this regard. Together these form the (a) and (b) of Claire's post. These are the classic "ideal switch" holy grail.

Finally, if as you note that this is achievable then Google are using this as marketing to generate interest.

I think if Google made the prize 100,000 units as a guarantee you might see a different interest. It is one thing to do a prototype and another to produce in volumes and assure quality, which is beyond the realm of any student.

Dear Gregory,

do you see a contradiction between cost, volume and efficiency?

My thinking is, that if you spend the money on cutting edge semiconductors to reduce the losses, there will be less cooling effort, smaller heat sink and eventually not even a fan.

The integration you name, putting most of the silicon parts into one package, will lead to more reliable designs using well-known technologies, so I don't expect the price to massively grow.

On the contrary, I suppose it will become even cheaper in producing highly efficient, small units in mass production.

This leads to the question, why inverter manufacuters as of today don't do this on their own and I imagine it is due to history.

They are simply used to work on components they can buy from the shelf, taking care of not infringing patents and build their designs as it is described in the books.

It is like the first radios that were not build on tubes but on transistors - as big as they ever were. Take a look how small a FM radio got due to integration. Buy now it is part of your smartphone :-)

In my opinion, inverters are as big as they are today because of similiar processes.

However, a semiconductor manufacturer - not building inverters but components - will allways buil devices for the inverter market.

We start turning in circles - it remains something like a hen-and-egg problem.

Thus, I consider this littlebox challenge a good idea to trigger some people thinking about the overall system instead of componetns only.

Maybe just like some designer thought about unleashing the power of transistors to FM radios.

If I understand matters that current semiconductors are still a few line width generations behind processors then there is scope for higher power density. But bonding wires and heat dissipation from the IC core becomes problematic. I have not delved into this for some time but have seen very little change in 20 or so years.

Some time back resonant link converters were all the rage as they allowed zero voltage switching but these too seem to have all vanished into the ether. The unified integrated system has been a long desired objective and some have already done this for custom / specialized designs. The problem with a system is designing a platform that may accommodate many different applications, controller and sensor designs, a sort of universal interface. 

That is primarily why many semiconductor manufactures remain at the topology or component level as they prefer to stay in their comfort zone. Others then build on this.

Twenty or so year ago I saw a 50kW water cooled servo-drive module that could be produced for DM3000 for an electric vehicle. It was the size of a Galaxy Tab albeit thicker.

Today we have very powerful FPGA and s/w with which to develop new web server based controllers (some even Google OS based) so the ability to create a platform similar to the smartphone is there and does exist. The difference is PE systems can inflict serious damage if used incorrectly.

ABB for example have the PEC800 platform which allows one to simulate and design in Matlab / Simulink and then download on to their controller platform and control power up to 10 MW and Siemens have same / similar as do others.

In Europe there is the DRIVECOM user group who have well defined standards and even the Infineon VECON processor so all the bits and pieces are there and like you say why has it not happened. It is because of the diversity of application and the huge amount of application IP as well as reliability, conformance and safety.

For Google this all very left field and as we saw with Google Power Meter it soon lost interest when it realised it could not play in the niche utility market.

I agree it is a good idea but only time will tell as to  its usefulness.

My understanding is there are some university teams pursuing this. Here is the timeline:

"July 2015

Registered teams must submit a technical approach and testing application by July 22, 2015.

October 2015

Up to 18 finalists will be notified of their selection for final testing at the testing facility. They are required to bring their inverters in person to a testing facility in United States by October 21, 2015.

January 2016

The grand prize winner will be announced sometime in January, 2016."

I am very interested in this as my employer manufactures a thermal spreader with 1500W/m.K in plane thermal conductivity.

On the other hand, there is plenty of "volume real estate" under the solar modules on roof mount solar. Most micro-inverters (actually all I have ever heard of) are mounted directly under the module. A large (65"x39") thermal spreader could be mounded on offsets under the module, to dump heat toward the roof via radiation. (Though the module is currently also radiating heat from the back side, and module temperature increases will impact efficiency.)

Hi David,

being involved in this challenge, I no longer believe it will be a cube of 7cm length.

Too big :-D.

I suppose, we get to see designs in the low 300cm³ while the 7cm cube would be 343cm³.

Deadline July is closing in...