The Wireless Sensor Networks Blog just posted an interesting link to an article at Popular Science . Microsoft unveiled a prototype sensor module reference design at Microsoft TechFest that they are currently using to monitor the temperature from their server hardware. I can only assume that it came from Microsoft's Networked Embedded Computing division at the Microsoft Research Labs. Well, I have nothing against Microsoft, and in fact, I enjoy using Windows (XP, not Vista). However the article does claim that the sensor modules are cheap, so I thought I would give a best guess-timate of how much the reference hardware bill of materials cost.

Luckily, the article provides an excellent clue when they said that they are using a TI transceiver that costs about $3. Well, that sounds exactly like the good ol' TI (Chipcon) CC2420. Why not the CC2520 you ask? Well, the CC2520 was just released recently (see my review of that chip ) and I am assuming that for them to make the PCB, write the software, create a nice plastic enclosure, and already have it running in their server rooms, the project must be more than a few months old. Hence, I am going with my initial assumption that they are using the CC2420.

In my guess-timate, I am assuming an order quantity of about 1000 pieces. This is because they mentioned that the CC2420 could be had for about $3.00. Other than the CC2520, the CC2420 is the cheapest 2.4GHz transceiver chip that TI makes and Digi-Key lists it in single piece quantities for $6.60. 1000-piece quantities can bring the price down to about $4, but I am assuming that they are Microsoft and they got a special price break down to $3. There are many other unknowns too so I'm going for a wild, ball-park figure based on my experience building a 2420 board and some basic pricing from Digi-Key. Here are some of the other assumptions I am making:

  • I'm improvising and using a $5 price for the microcontroller, assuming something in the same class as an Atmel ATMega64 (8-bit MCU, 64 kB flash, 4kB RAM).
  • For the PCB, I'm using $2 which is a reasonable price if they are making a 4-layer PCB according to the TI reference design.
  • The enclosure is a wild card because I'm not so familiar with making professional enclosures. I normally stop at the green-board since I can start working on the software from there.
  • The misc other parts inlcude high precision resistors and capacitors required for the 2.4 GHz radio balun, the MCU crystal, decoupling capacitors, pullup/down resistors, and various other things like LEDs.
  • I'm assuming that the assembly cost per board was approximately the same cost as the PCB. I think its a reasonable assumption since there's usually an NRE to cover the cost of setting things up and organizing the parts, and then a per board cost on top of that. Im not too familiar with it though since normally, I do my own assembly.
TI CC2420 Radio
MicroController  $5.00 
SMA Coax Connector $2.00 
2.4 GHz Whip Antenna
16 MHz 10 ppm Crystal $1.00 
DC-DC Switching Boost Converter
2 AA Batteries $1.00 
PCB $2.00 
Enclosure $5.00 
Misc Other Parts $3.00 
Total BOM cost

Well, a cost of $27.00 is not too bad, but we need to figure out the final in-store cost. A BOM cost of $27.00 does not include packaging, certification (FCC, UL, etc...), marketing, and distribution. For distribution, you can have it go through a distributor to get on to store shelves, however the distributor will probably require from 15-50% depending on the distribution terms. And once its in the store, the store will want to charge a decent markup above the wholesale cost that it got from the distributor. Here we can assume another 15-50%. Electronics don't command margins as high as things like sundry goods. Lets take a middle of the road value for both markups at 30% each. Adding a little extra for the packaging, certification testing, and marketing and rounding would give us a (very rough) in-store price of $50 per sensor module. Please remember that I'm making very gratuitous assumptions and this is more of a back-of-the-Peet's-coffee-shop-napkin calculation. 

So is $50 cheap enough for the average consumer to start throwing up sensor networks in their house? Well, if it takes about five sensors to build a sensor network, then at a cost of $250, it's possible. It would cost a lot more to install the wiring for a new burglar alarm. But you'd have to be a pretty big geek to spend that kind of money on something as Star Trekky as a WSN.

However, the prices can be brought down even more. At higher volumes, the costs of the components will shrink due to volume discounts. If this is an estimate for 1000 piece volumes, then you can probably decrease the cost by about 40%+ at 1,000,000 piece quantities. However are those quantities feasible for Microsoft? If they are supporting their own protocol, then I doubt it. If Microsoft is seriously thinking about releasing something like this to the consumer market, they should join Zigbee, 6LowPAN, or ULP Bluetooth.

p.s. What if I'm completely wrong about everything? Well, it wouldn't be the first time I'm talking out of my ass Wink