Wired sensors around the house

[JJalling]

Hi all,

I plan to place a handful of various sensors around the house (both inside and outside). Some of them is just simple temperature sensors, but others will be counting pulses (utility meters), light intensity level, air pressure and possibly other things.
My idea is to have a central device for handling all the measurement data, communicating with the sensors via. RS-485. Each sensor will have a small uC to handle the communication between the sensor and the central device.
But what about power? My initial plan was to have 4 wires to each sensor node. 2 for power and 2 for communication.
Is 12VDC, with a 5V regulator at each sensor a bad idea? Should I use 5V instead? Or maybe AC?
How would you do it?

BR Jonas

[IanB]

It seems to me that a higher voltage power distribution with local on-board regulation makes most sense. You could distribute power at 12 V DC as you suggest and that should give plenty of headroom for stable 5 V supplies where needed. I don't see any particular gain in distributing AC instead of DC.

Local regulation gives you the advantage of a stable 5 V without any voltage drops from other loads on the system. You can also use thinner wiring (you could tolerate a 1 V drop in the 12 V supply at the point of use, but you could not tolerate a 1 V drop in a 5 V supply).

[JJalling]

Thanks Ian,
I'm glad my idea wasn't that stupid:) I will feed the sensors with 12vdc and have a local 5 or 3.3v regulator.

Br Jonas

[David_AVD]

Just be aware that voltage drop in your DC 0V supply wire can also affect the signals that are referenced to it.  If you're using RS485 there's quite a reasonable common mode range, but something to keep in mind if you did go for something else.

[magetoo]

If you're installing 4-pair CAT-5/CAT-3/... then you have two pairs for half duplex data and another pair for power and still a spare pair.

IIRC, both the Power-over-Ethernet standard and the consumer grade passive PoE common practice is to use two pairs for power.  Using just one of course works too, but in that case it would be a good idea to use a custom connector so there's no confusion with Ethernet cabling and no random plugging in of equipment that expects something more standard.  (and no accidental shorting out 12V to ground)

It's a cool idea for a project by the way!  I've been thinking of doing the same thing, since there's already Cat-5 cables installed in my apartment.


Related to that (hope I'm not hijacking here); is there an obvious and cheap way to couple a differential signal onto the power pairs?  I've thought about just copying what Ethernet does, but I'm not sure how that would work when the signal going into it has nothing to do with Ethernet. (different data rates, expects termination, etc)

Edit:

evb149, you mentioned capacitively coupling; the part that is making me think twice is when the conductors of a pair is tied together for power - I'm not clear on how to use a pair for power in that manner without shorting out the signal.  (unless one goes the full PoE route)

But it's of course possible that is a stupid idea to begin with.  It seems that PoE is mostly useful for point-to-point connections, and a multi-device bus should probably just use different wires.

[dfmischler]

I would try my best to use 1-wire.

[JJalling]

I'd look at using low power RF like 802.11.15 / Zigbee, bluetooth low energy, something like that.

...
The nRF24L01 and similar are popular for wireless.. TI, NORDIC have bunches of options, lots of off the shelf modem modules / shields.

Thanks for the input.
Currently all the rooms have a small wireless humidity/temperature sensor, based on nrf24l01+ radio modules. This eases the placement of the sensors, as I don't have to get a wire to every on of them. They run on a cr2032 battery.

This works fine for sensors that can be shut down, but it's not very good for sensors that have a constant power draw, for example the ones used to monitor utility meters. That's why I want to put some of them on cable.

In my old apartment I had 1-wire sensors to monitor the temperature in all the rooms, but Maxim/Dalas have discontinued the 1-wire pulse counter device (DS2423) so I needed to find another solution for that.
Again, that's why I'm building my own wired sensor nodes.

Br Jonas

[dfmischler]

What kind of pulse rate do you need to track?  If the pulse rate is low enough you could count pulses with an external binary counter that is read by a DS2408 (8-bit 1-wire PIO).

It is pretty easy to interface various analog sensors using the DS2438, like this humidity sensor based on the Honeywell HIH series parts.  This approach should work for many other sensor types, as well.

Of course, other approaches are fine, too, but it is still hard to beat the cost of 1-wire for low data rate sensors.

[JJalling]

What kind of pulse rate do you need to track?  If the pulse rate is low enough you could count pulses with an external binary counter that is read by a DS2408 (8-bit 1-wire PIO).

It is pretty easy to interface various analog sensors using the DS2438, like this humidity sensor based on the Honeywell HIH series parts.  This approach should work for many other sensor types, as well.

Of course, other approaches are fine, too, but it is still hard to beat the cost of 1-wire for low data rate sensors.

Why do it simple, when you can do it complicated?
I hadn't thought of that approach. That could be a viable solution indeed.
I will consider it, and maybe go down that path - thanks.

BR Jonas