### Calculating the AIO card!

My last post seems to be getting some "buzz" in the "gadget" communities. Listening to some of the comments, it makes me think that not many know about thin clients.

The concept of "carrying a computer in a card" is already there with SunRays. For example I can insert my Sun's employee ID card into a LCD screen (below - notice a slot to insert the card), do some work. Take out the card, go home insert the card into similar LCD screen. I get my display back where I left it. Say I prepared a presentation, left it open in full screen mode, then fly to London. I could buy food with the same card. I could walk into a presentation room full of people, and put the card in a Sun ray device connected to a projector. Viola! my presentation is ready to be shown. I didn't waste a second trying to bring up the slides. Then I can fly to Bangalore, insert the card into another Sun Ray LCD screen and still get the same display I left in London! There is absolutely no need to carry a laptop, if only I could insert the card into the LCD screen on the plane seat!

Now lets make some calculations about the feasibility of the AIO Card. My physics is a bit rusty. Last time when I calculated how big a biggest computer can be, I made no use of physics! We are now finding the answer to the question how thin can the thinnest thin-client be?. If you have attended a class in physics, the very first thing a good physics teacher would say is that all your calculations will be based on assumptions and approximations. He might ask you to learn the art of approximate arithmetic. Even rocket scientists make assumptions and approximations, (just that the results will be more accurate)!

These are my approximation and assumptions based on input from Googling the latest in today's technology. If you find a component that performs better, or if you see a mistake, please feel free to point it.

Bandwidth:
In an update to last post, I said that the fact that Sun Rays work was a proof that remote display works great over wireless with today's bandwidth conditions.

I did this experiment: run a few GUI applications securely and remotely using ssh -X while capturing the network traffic. I ran calculator, a clock and gthumb(image browser) and mozilla. Browsed a few pictures, clicked on some websites. This was for about 2min 44sec.

The graph by Ethereal summarizes the traffic. Blue line is the bytes (totally 10,870,488) coming from the server, red is the bytes (totally 802,022) sent from the client's end. For such workload the network bandwidth requirements are as much as ~ 70 Kbyte/sec = average 600 kbit/sec and peak ~1 Mbits/sec.

For real time video and voice refer to www.videohelp.com/dvd. Being a bit conservative, take real media for example, only requires 350kbits/sec for video + 64kbits/sec for audio.

Size of the wireless devices
Bluetooth and/or wifi needed to make it work with other peripherals like speakers, headsets, mouse and keyboard. Searching google for smallest bluetooth device gives: www.3g.co.uk/PR/Feb2003/4961.htm (8mm x 6.5mm x 1.6mm) (made in 2003) which easily fits into our AIO card. Or these antennas for bluetooth and WLAN 802.11 which are < 1mm in thickness. A good 2.4Gz wireless transmitter requires at least 6cm long antenna, which also fits into our credit card size(8.56 cm).

Power requirements
Looking at transmission power of small Linksys USB Wifi device gives ~10 dBms = ~ 0.01watt = 100 milliwatts.

For the display, E-Ink website says that a 90.6x122.4 mm2 area requires 750 milliwatts for changing the display (1mW for standby) For our credit card size (85.6mm x 53.98mm) this is about 300 milliwatts. The reason these consume very less power, is because they do not emit light.

Searching for some of the small, low power processors, gives us www.via.com. We get some of their full featured low power processors which work at power less than 1Watt. So we can make a guesstimate that we need about 1Watt of power to run the card, if our processor does not need all the processing features, and it operate at lesser clock speeds enough to keep persistence of vision.

It should also be able to do encryption and decryption. In Feb this year Sun demonstrated a coin sized secure web server!

You may argue that if all the wireless, processing and display circuit was just one integrated circuit, it might be more energy efficient and compact.

Power supply:
Now the problem is to supply 1 Watt of power. Recently NEC announced that they have a flexible gel battery, made for things like Smart Cards. (what a coincidence!) Looking at their site, they say it provides 1mWh per cm2. It is 300 microns thick. One credit card sized layer provides 40milliwatts for one hour. Another good thing is that these take only 30 sec to charge. but we would actually need 88 of such sheets to supply power for 3 hours = (20mm in thickness) not what we wanted, but we may still use it as the battery for Smart Cards.

Now consider Solar cells. Taking the numbers from Wikipedia, a credit card sized 100% efficient solar cell should generate ~ 5Watts power. Recent cheap and thin film solar cells are only 7-8% efficient (efficiency is the fraction of light energy they convert into electrical energy). These would from a credit card sized film would generate about 400milliwatts.

Assuming that we will have a folded film, which covers the front display (opens up like a book's front cover) we can having it charging while we are using the display. Assume if we succeed in making cheap solar cells that are 20% efficient (the silicon cells give about 15% efficiency) and perhaps make our components run at 500 to 700 milliwatts, we would have actually succeeded in creating a self sustaining thin client all-in-one card, with out requiring external chargers!

While being on the future prediction theme, you can read my post last year about a movie made by Sun back in 1994, that envisioned the technology in 2004.

For solar cells, check out http://pesn.com/2005/06/10/9600107_NanoHorizons_Solar_Cell_Breakthrough/

Posted by Leen Toelen on January 02, 2006 at 07:02 PM PST #

Of course this will happen at some point soon. Until it does we have to settle for somewhat thicker handheld touchscreen wireless X terminals that open remote displays to remote mature vertical market solutions like the one shown in my URL. The software is working now - it has been for a decade. The device is in proproduction and is about to go into production. It's quite similar what you're talking about except that it's about 3/4" thick. X terminals are, of course, 'thinner' than thin clients.

Posted by Gene Mosher on January 03, 2006 at 09:42 PM PST #

for the gum-pack-sized keyboard, an uprev of something like the VKB Bluetooth virtual keyboard, at CNET reviews might prove useful ...

--retiarius

p.s. greetings chandan, from ex-sun employee #31094, i see you are stateside now...

Posted by retiarius laboratories on January 04, 2006 at 04:43 AM PST #

Do you think it possible to decode RealVideo real time on a 1watt cpu? I still think there will be major bandwidth problems besides anything thats not relativly simple or single screen stuff (ie a calculator). Any type of compression requires upwards of twice the work to do anything, and a weak slow cpu won't be able to handle that real time.

Posted by DnSn on January 04, 2006 at 04:57 AM PST #

The decoding part of the job will probably be done by the server, and not by the client..

Posted by RvE on January 04, 2006 at 07:01 PM PST #

moomins

Posted by guest on January 05, 2006 at 01:57 AM PST #

This is a test comment

Posted by John Doe on January 05, 2006 at 03:19 AM PST #

Nice calculation exercise. But i'm afraid you are far too optimistic about the power that can be drawn from the solar cell. The 5W is based on outdoor daytime sunlight. Most of the time the card will be in your wallet, or inside buildings. I'd say that interior lighting gives at most 10% of the 5W, so 0.5W x 0.20 = 100mW at most. At present 100 mW is by far not enough for complicated multimedia decoding / encoding. I would bet on the next generations low-power ICs, powered by fuel cells will be the real enabler for this kind of devices, with watts of power from milliliters of ethanol.

Posted by marco on January 05, 2006 at 05:50 AM PST #

This sounds cool :) I really like your Blog .. it is clean and easy to read ... neat :)

Posted by Hasan Bazerbashi on January 06, 2006 at 09:30 PM PST #

Posted by guest on February 01, 2006 at 06:24 PM PST #

For power, it seems solid-state lithium-ion batteries are coming soon: witness progress at Infinite Power Solutions. It already produces a 0.11mm lite\*star producing 0.7mAh http://www.infinitepowersolutions.com/index.php?txtMode=pagelevel&id=4 On 6 Sept. 2006 the NY Times reported that it will start mass production soon: http://www.nytimes.com/2006/09/06/technology/06battery.html?_r=1&oref=slogin Another in this field, with a great overview of battery technology, is Oak Ridge Micro-Energy: http://www.oakridgemicro.com/tech/tfb.htm Cymbet also has incredibly think Li-ion batteries delivering 300 Watt hours per Kg or 1 milliamp per sq. cm. http://www.cymbet.com/features.php Excellatron talks of a 10mAh battery 0.35mm thick: http://www.excellatron.com/advantage.htm

Posted by ScooterJP on September 07, 2006 at 11:36 AM PDT #

Comments are closed for this entry.