5
\$\begingroup\$

These days businesses are all talking about the Internet of Things and how Cisco predicts 25 billion devices will be connected by 2015.

All around the hobbyist community various internet shields and internet ready devices like the Raspberry Pi cost somewhere around $35. The ENC28J60 is $21. The Electric Imp is $30. The Nanode Classic is $40.

When I check out PICs and AVRs chips with built-in ethernet, they all seem to cost something like $6. Toss in an RJ-45, various other components, a PCB and a lot of time and I seem to be back to something almost as expensive as the hobby grade systems.

With all this talk of the Internet of Things, and the cheapest 32-bit PICs around $1, it just feels out of place that I haven't encountered any complete ethernet solutions for under $3. I can't imagine any designers willing to equip an alarm clock with internet connectivity for $25.

As a CS student about to graduate, I am trying to create a small volume product with the electronics costing under $10 that will send very simple analog data by ethernet to my web service. I have built the web app and prototyped with a Raspberry Pi, but $60 for an internet connected sensor is just too expensive.

Can anyone shed some light as to whether web connected sensors under $10 are possible for a guy like me to develop or use? Maybe they exist and I just haven't found them?

\$\endgroup\$
  • 2
    \$\begingroup\$ An ENC28J60 chip is $3 directly from Microchip. A PIC18F67J60 is $3.10 and has an on-chip 10BaseT interface. \$\endgroup\$ – Wouter van Ooijen Oct 29 '13 at 6:39
  • \$\begingroup\$ RN-171 from microchip microchip.com/wwwproducts/Devices.aspx?dDocName=en558370 has a on board 32-bit micro with ADC and RTC , I think is best for you.Its cost is more, $26. \$\endgroup\$ – Hemal Chevli Oct 29 '13 at 7:58
  • 3
    \$\begingroup\$ If high-volume products cannot be made and sold for under $10, I wouldn't expect to be able to make a small-volume product for $10 or less. \$\endgroup\$ – RedGrittyBrick Oct 29 '13 at 9:45
  • \$\begingroup\$ You are only looking at half the problem. Even if your device only cost $10 (which is possible in volume), you still have to pay for the ethernet cable and the pro-rated cost of the switch to plug the other end into. \$\endgroup\$ – Olin Lathrop Oct 29 '13 at 12:20
  • \$\begingroup\$ I guess I am assuming about a 5K volume. I have found this setup from Tuxgraphics: shop.tuxgraphics.org/electronic/eth-eclient.html Although the product is sold for more than I would hope, the components seem to be relatively cheap, though over $10. I thought Ethernet may have been the best bet because the sensor would be a few inches from a corporate computer and wouldn't require any installation. \$\endgroup\$ – JHAWN Oct 29 '13 at 20:51
6
\$\begingroup\$

Once you include your MCU, possibly an Ethernet MAC/PHY (PIC18F97J parts have integrated MAC/PHY, as do a few other MCU's), Ethernet magnetics & jack, power supplies, I/O & power connectors, board manufacture... everything adds up in a hurry when you're paying one-off prices. You'd struggle to make one for sub-$20, let alone sub-$10. Once you can make them by the thousand the price-per-unit comes down, but until then you're pretty much screwed.

\$\endgroup\$
3
\$\begingroup\$

Bluetooth dongles are available from $1. You won't be able to get anywhere near that price, but yes, it indicates that sensors-you-can-connect-to-your-bluetooth-enabled-web are available from that price. The Ethernet magnetics & jack are always going to keep the price up for wired devices, but a lot of the internet of things is going to be wireless.

\$\endgroup\$
3
\$\begingroup\$

As david says, most of the effort is going into wireless "things", because people don't want to run extra wires everywhere round their home, a high bandwidth connection is not essential, and the cost issue you have identified.

Zigbee, 6LoWPAN and the Nordic Semiconductor proprietary system are popular here. There are various companies with wireless home sensor networks which use one of those protocols to connect to an ethernet-enabled "base station" and thence to the Internet. It's not a very profitable market in the home.

It's worth looking at RS485 or CANBUS if you want to connect a wired sensor network on a house scale.

\$\endgroup\$
  • 1
    \$\begingroup\$ On the topic of wires - if you're using ethernet, then you need an ethernet cable for every device, and an ethernet port on a router for every device. That gets expensive. \$\endgroup\$ – Li-aung Yip Oct 29 '13 at 10:10
3
\$\begingroup\$

I don't think you can do it in small volumes (especially if you insist on wired ethernet), but there are wireless chips floating round designed for this "internet of things" which will be mass-produced, require little supporting hardware, and will likely be dirt cheap in volume.

I have a hazy idea that there's one of the MSP430 family (or a relative of) that includes wireless & MCU in one.

The bigger question is - what's it for? It's been possible to network things to a greater or lesser extent for years, these days for minimal extra cost, and the idea has been round since the year dot, but so far no killer app has emerged that makes me want to rush out and buy things which are internetted.

It's already cheap & easy enough to internet things which people decide they want internetted - if $20 on the price of your $500 appliance results in something awesome (or even moderately OK) people will buy it - and people do buy it in some cases.

However, to my perhaps overly cynical eyes, a lot of this stuff is a solution looking for a problem. Do I want all my lightbulbs internetted so I can turn them on & off from anywhere in the world? Would I pay more for my lightbulbs to be able to do that?

$60 is maybe steep for an internetted sensor for a hobby project, but to anyone (or any company) with even a half-good reason, that's a bargain. Remember the Pi is, relatively speaking, a supercomputer relative to the processing needed to read one sensor and update it over the net, and most real applications you would want to read multiple sensors, so adding more sensors to one Pi reduces the cost-per-sensor, or if you prefer, increases the quality of the data. Could you connect 100 sensors to a Pi, for under $100 total? Probably. Could you also have it take digital photos of whatever's being monitored? Yep. Is that pile of data giving you a much better picture than a $10 single-sensor - very probably.

Most of the cost for remote sensors in industry is the provision of the connection - be it a telephone wire, broadband connection, GSM/3/4G phone, radio link, sattelite, etc.

Sure it's easy to internet all your lighbulbs - but what if all your lighbulbs are in street lights running 100 miles across the countryside with no friendly WiFi router nearby?

This has turned into a bit of a ramble, sorry.

Nurse! Fetch my coffee!

\$\endgroup\$
  • 1
    \$\begingroup\$ This is an excellent answer. I work on a large instrument that is Internet connected. The total hardware cost of this connectivity is probably around $1,000 (we provide a hardware firewall with it) or so but since the instrument itself costs over half a million dollars, and the benefits of the connectivity is lower service costs to us, that's $1,000 well spent. \$\endgroup\$ – lyndon Jan 11 '14 at 16:44
0
\$\begingroup\$

Think Google Chromecast, at a selling point of $35, made in huge quantity. Production cost is probably less than $10, but I doubt you can get there in small quants.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.