I need to come up with a way to daisy chain a number of sensors together and uniquely read the high/low pin from them individually. Think Digital RGB LED http://www.adafruit.com/products/306 or I2C. I like how when you power the LED strip up, it assigns a unique id. Is there already a module that will do this?
Hello Alex,
To accomplish what you want, you'd need something that is daisychained, or at least have a chain of devices where the first in the chain can 'disable' all others. I2C , without modifications or other circuitry, won't be able to do what you want. As it is a bus-topology, all devices are being 'talked to' at the same time, and all devices with the same address will respond simultaneously, thus causing bus collisions.
You'd either need a device with an 'ingoing' and 'outgoing' bus, or something that disables all other devices until it is being addressed. In both cases you could do this with a simple microcontroller. Which is of course the kind of work you're not waiting for 
Charlie,
The example you gave is interesting, but (at least the examples) only apply to output devices, not inputs. The MCP23S17 does not mention daisy chaining in its datasheet.
If Alex is still alive (haven't seen any reaction since the original post), he might do this: read through the datasheet of the WS2811 to understand how to do daisy chain addressing[pdf], and get a free LPC800 board to implement the same protocol 
Victor,
The OP mentioned "uniquely read the high/low pin from them individually" which I interpret as an input. I agree that the MCP23S17 datasheet doesn't specifically mention daisychaining, and quite possibly its not possible with that particular chip, but I was attempting to discover what sort of sensors he means, whether digital input, analogue (ADC/temperature/pressure etc) or whatever.
I've personally used 40 MCP23S17 chips in a single solution, although there I used 5 CS signals for each group of 8 addressed chips, with a common CLK/SI/SO across all 40, for a total of 8 lines. 40 chips at 16 I/O per chip=640 I/O channels 
It occurs to me that maybe a topology like that would be sufficient for Alex? I know that doesn't specifically answer his daisychaining question, but sometimes its easy to accidentally make assumptions in an original question.
I hope Alex IS still alive....
Best regards,
Charlie.
p.s. I just looked again at the MCP23S17 datasheet. I think that the SPI input timing diagram (Figure 2-5 on page 32) shows that this particular device is not suitable for daisychaining, because the data coming into SI is not copied to SO. I've used other SPI devices which do exhibit that behaviour though.
p.p.s. I may look into the free LPC800 board you mentioned. As ever its finding the time...
p.p.p.s Here's an example of an IO expander using SPI which can be daisychained : http://datasheets.maximintegrated.com/en/ds/MAX7317.pdf Mind you, the datasheet gives a limit of 10 on a single chain because of propagation delays of SI/SO versus CLK. Hmmm... Maybe SPI daisychaining isn't the way to go....
