Design suggestion for a bus

Then by that measure, using a device like a 23017  http://ww1.microchip.com/downloads/en/DeviceDoc/21952b.pdf  would give 16 valves on a board and reduce the count of boards considerably and only needs a couple of OCTAL Driver chips to interface to relays / Solenoids

I suspect this is compacting the control too much though but this is not known as your description does not provide too much of the operational details and is quite vague

If you can provide more details of how this will be configured and what it is controling we can advise better

Hi Peter

I will try to explain the problem

I have pneumatic valves that can be placed side-by-side as shown in this picture. You can (currently) have up to 38 valves

I need a board that controls these valves using a serial bus. There are two requirements:

1. If (let's say) the central board does not switch on, all the remaining valves must continue to work normally

2. The board must be "plug-and-play": when a board breaks down, the maintenance person take a spare part from the box, replace it and that's all

Hi Peter

I will try to explain the problem

I have pneumatic valves that can be placed side-by-side as shown in this picture. You can (currently) have up to 38 valves

I need a board that controls these valves using a serial bus. There are two requirements:

1. If (let's say) the central board does not switch on, all the remaining valves must continue to work normally

2. The board must be "plug-and-play": when a board breaks down, the maintenance person take a spare part from the box, replace it and that's all

The obvious way to do this is to use Devicenet

https://www.google.co.uk/search?q=device+net+idc&ie=utf-8&oe=utf-8&gws_rd=cr&ei=8mO0VvPDI4fEOvCvgsAO

If actual Devicenet is too expensive you could trawl the extensive literature for ideas.

MK

Hi Michael

The obvious way to do this is to use Devicenet

Not sure why this is the obvious choice, since the devices will require a commissioning phase before installation and there are tens of industrial buses out there that are more widely used than DeviceNet

If actual Devicenet is too expensive you could trawl the extensive literature for ideas.

Yes, of course there is Google out there that can provide all the answers, but may be this community could provide suggestions based on hands-on experiences

Well because Devicenet exists, uses low cost wiring, is simple enough that you might make the nodes small enough so that you could have one per valve.

If you MUST have each node knowing its address without programming it must be plugged in to  a particular slot in the network (star connected) or be able to determine its position in  a daisy chained network.

Both methods are quite high risk - being very susceptible to mis-plugging in the field.

I'm sure that a position discovering daisy chain scheme could be devised but it's hard to imagine one that would be very reliable (all the obvious ways of doing it require nodes to actively process all messages that pass through them).

My hands on experience of building machines and control electronics says use star wiring, design a reliable hardware driver and replace the whole board if it fails.

If you can't get good enough reliability like that (and it wouldn't be good enough for some applications) then you are into a different realm alltogether and without very detailed information about the system I couldn't make any suggestions.

MK

Hi michael

Which information do you need?