control system choice

[pb]

I'm trying to decide on an automation system to use for my new (well, renovated) house.

My main requirements are that:

- it should have a dedicated, hard-wired control bus, ideally but not necessarily something that I can run over standard cat 5/6 wiring.  X10-style data-over-powerline comms are not acceptable; I can't convince myself that this will be reliable enough.  Any kind of radio link is also undesirable for the same reason.

- there should be a decent range of off-the-shelf modules available.  My primary interest is lighting control and, although I am happy to roll my own bits to some extent, I think that life is probably too short for me to build and debug my own light switches and dimmers.  In particular the switch modules themselves need to look good enough to satisfy my wife, which probably rules out anything I can make myself at reasonable cost.

- there should be at least some level of openness, so that I can interface my own hardware to the bus.

So far the best plan that I've come up with is to use the Dallas 1-wire bus, plus a home-grown controller of some kind, for simple centralised monitoring tasks (temperature sensors, heating control, etc) and Clipsal C-BUS for control of lighting and other more distributed things.  C-BUS is not quite as open as I would like but it seems to fulfil all the other requirements.  Both of these will run over standard 4-pair UTP wiring. 

But, the C-BUS equipment seems to be awfully expensive: the input modules (light switch plates) are over £100 each, and a four-channel dimmer costs the best part of £500.  Based on those prices, the idea of doing something for myself suddenly seems more appealing again.

Any other suggestions or recommendations?

[guydewdney]

If you think thats expensive - dont look at Lutron (I have a 6 channel, grafik eye, wall plate controller, occupation detection, PC interface and calendar controller)

[dhaslam]

A way to make the wireless switches a bit more reliable would be to wire  your  control system  to the remote  control for each switch.  This way the remote control  can be near the switch it controls.     

[wookey]

pb, as you have found everything is very expensive except 1-wire. The choices I've found are modbus, knx, idranet. Zigbee is quite good but if you really don't want wireless then it's no good. Idranet is relatively cheap but windows-only and not open. KNX proudly claims to be open but that actually the sort of open where you need to have over 1000 euro just to read the spec. Siemens sell loads of kit and it's all expensive.

The whole field is desperately in need a good 'openness' kicking IMHO, which is starting to happen with projects like opnode and the jokomajo institute http://www.jokamajo.org/ . Even things like 'opentherm' the open thermostat protocol won't tell you what the damn protocol is without a load of money: fortunately some guy in holland has reverse-engineered it for us.

Take a look at the Webbrick stuff - that's relatively sensible (they'll control all sorts of stuff of various types). Not very open though.

Opnode: 1-wire and xAP (i.e. open bus and control mechanisms). http://sites.google.com/a/usapiens.com/opnode/xap They use ethernet for data bus, and interface to 1-wire for sensing and x10 for actual switching, which is nice and flexible (but not how I'd do it).

If you go with 1-wire (and that's what I'm doing - everything else is too closed _and_ too expensive), then cat5/6 is ideal, quite a lot of kit is available (e.g from hobby-boards.com, opnode and welserver). Sadly there isn't that much connector compatibility, so plugging things together can be tiresome. I'm quite keen on filling the '1-wire mains switching' hole myself as it's such an obvious gap in the market, using the designs discussed on this board. But of course I have to find the tuits.

[pb]

Thanks.  The webbrick site seems a bit short on detail, but I see they will be exhibiting at ecobuild next week and I will try to have a chat with them then. 

From what I could see on their website it looks like their offering is fairly similar to the Clipsal C-BUS stuff but it isn't clear whether it is any more open.  Allegedly Clipsal do publish at least some of their specs for free although you have to sign up to their developer programme in order to get hold of them.

It is quite tempting to have a go at designing a dimmer system to work with 1-wire.  I wonder whether the performance would be good enough though in terms of response time to switch input on a large network.  If I remember right 1-wire is single master so all the switches would need to be polled.  Hm.

[pb]

Thinking about this some more I wonder whether CANbus would be a better choice for my kind of application than 1-wire.  You can still run it over cat5, it'll go a reasonable distance, and it's multi-master so each control point can just emit its own messages onto the bus.

[wookey]

1-wire is probably not suited for things like light-switch control, because I'mnot sure that an end device can say 'I have amessage I need to send now' - they wait till they are asked. I've wondered about CAN bus but not looked into it much. A lot of small microcontrollers have the hardware now. I have no idea how easy it is to create end-devices. The protocol seems to be open (but subject to a patent licence for hardware implementations). I couldn't find any cheap switches (mains or otherwise) in an initial search.

As I say Idranet was the only off-the-shelf thing I found that wasn't extortionate, because it didn't try to be _too_ clever. Useless to me though due to lack of Linux support.

[Baz]

The cheapest system is still to hardwire it all back to a central point. RS232 is the next best if you want to distribute a bit more data to local i/o points as there are dozens of small boards for it. There are lots of small development boards Basicx, pickaxe, stamp, etc for your own i/o boards too.
The only fully open control protocol is Crestron but it is the most expensive.
It is worth remembering that electronic just don't last - 10 to 15 years tops, (yes I've got a 40 yr old radio too but also 10 broken ones) whereas a house at 100yrs is 'new'.

[pb]

Wookey,

Yeah, CANbus is a fairly open standard.  As you say there are patents on it but the cost of those is generally absorbed into the price of the controller chips; you'd only need to worry about the patent licensing if you were rolling your own interface out of raw materials.

A bit of googling reveals:

http://www.ilight.co.uk - lighting control products based on "iCANNet", which seems to be basically CANbus under a vanity brand name.

http://caraca.sourceforge.net/ - AVR based CAN interface platform including open source firmware

I've no idea how much the iLight stuff costs.  I will send them an email and try to find out.  If it's not too extortionate then it might be a good option: even if they are using their own proprietary messaging at the application layer, it probably would not be very hard to reverse engineer if the transport layer is standard CAN.

In terms of homebrew hardware, the Farnell price for an AT90CAN32 microcontroller is £5.24 at 10 off.  The physical interface chip PCA82C250 is another 85p.  Add a pair of RJ45 sockets at say 40p each and that seems to be pretty much all you need for a non-isolated CANbus peripheral.

If the iLight dimmers do turn out to be expensive then it doesn't seem like it would be very hard to design a little board which would interface CANbus to a standard DIN-rail dimmer (using either DMX or 0-10V analogue).  That would avoid the need to grapple with the power electronics part of the problem.

The controller end seems more tricky, mostly because I can't think of any easy way of making a homebrew light switch controller with an appearance that my wife would tolerate.  Best plan I have come up with so far is to buy a cheap touch dimmer, throw away the electronics and just keep the front panel, but that still isn't a very appealing option.

[wookey]

I've just realised who I'm talking to. Hi there, wlecome to Navitron-world :-) (We should narg further about electric cars/batts sometime.) Which days are you planning to go to ecobuild? I'll probably visit too.

I was rather hoping that someone else had done the simple/cheap CAN device board, as you describe, but I guess if they haven't then it is perfectly do-able to DIY some.

If the actual dimmer is DMX, why not just use that directly and avoid the complexity of CAN? It's an addressable scheme too (so you don;t need individual wiring) isn't it?

I reckon a good way of having aesthetically pleasing user control for HA is to use something like a phone or webpad device. It does mean a whole pile of UI code but there are various people writing this, as you well know.

Obviously having a physical light-switch is farily necessary too, but that could just do simple on/off and the fancy stuff could all be via the more versatile interface? Otherwise I think you are right that you have to get a suitable face-plate from something else and change the internals. I looked for cheap LED dimmers a couple of weeks back and found none that would go in a standard patress without spending a fortune.

[pb]

Heh.  Hello wookey!

I did consider DMX briefly as the main bus, but it doesn't work very well in a distributed control kind of environment: it's designed for a single master to control multiple slaves and it would be hard/impossible to use any other way.  In particular, there is no target addressing in the conventional sense - every packet contains the complete set of settings for every slave device, so there is no way to change the output of just one dimmer unless you also know the current state of all the others.  The DMX "address" that you configure on the slaves is just the byte offset into the packet that the device will use to find its own settings.

A bit more time with the Farnell site suggests that a 2-channel touch sensor IC costs something in the region of £1.30, or you can trade up to the 4-channel version for another 30p.  So, total bill of materials cost for a CAN touchswitch control seems like it should come in at under £10.  You'd then need to add the faceplace from something like...

http://www.tlc-direct.co.uk/Products/VLIVS002.html

... which, annoyingly, seems to cost anywhere from about £10 to £20 depending on what finish you want.   So I guess the total cost for the whole thing would come in at somewhere around £40-£50 including the PCB and assembly.  Still not exactly a low budget solution, but at least you get something open for your money rather than being locked in to a proprietary system.

Looks like I will probably be at ecobuild on the Wednesday.  I'll send you an email about that separately.

[pb]

Having pondered the light switch thing a bit more, I now think I would be better off forgetting about touchsense and just designing something to use standard push-switches.  The AT90CAN32 has plenty of I/O lines so it would be straightforward to interface it to a good number of switches and still have a few left over to drive LEDs, sense thermistors/LDRs or whatever else.  No faffing around with stray capacitances, and the switches can be chosen to suit budget/aesthetic concerns.

Anybody else interested in such a thing if I was to get some boards made?