For a number of reasons, Jon and I are not happy with the powerline modem options. Chief amoung those reasons are that they are expensive and proprietary. Building them into our product seems to be too restrictive.

So, we decided to investigate the wireless options which seem to be improving in price recently. Bluetooth, Zigbee and Wifi are all comparably priced these days (<$10). However, Wifi seems to still require a somewhat expensive CPU to work and Bluetooth doesn't seem quite right for meshing networks. A closer examination of the Zigbee standard led us to believe that it is the cheapest and easiest low data-rate command network to deploy.

So, we decided to experiment with it by creating our own light switch dimmer. Since we might be abandoning Insteon, we will have to make our own dimmer/control panel for our product. Here is a picture of the layout:

You can find the light switch dimmer schematics here.
You can find the BOM here (OpenOffice Format).
Here are some things to note:

1. CPU. We are using the Freescale MC13224. This is a third generation 2.4GHz 801.15.4 SOC. In addition to an integrated antenna transmit/receive it also has two ADCs and many GPIO lines. O... and it is very cheap. This makes it an excellent choice for a low cost dimmer.
2. Antenna. We are using the Planar Inverted F Antenna recommended in the application notes of the MC13224. This antenna has a very good omni-directional pattern and is low cost due to it being constructed from copper on the PCB. The antenna is designed to be a quarter of a 2.4GHz wavelength (124.0mm) and have a characteristic impedance of 50 ohms.
3. Power. The entire dimmer design runs 'hot'. That is, it is not issolated from the main AC power lines. This is typical for light switches but does pose a safty risk for the developer (I have a power limited AC source in my lab for safty). The DC voltage rail is formed by clipping the 170V AC signal with a 5.1V rectifier. To limit the current (so the rectifier doesn't burn up), a 1uF capacitor is put in series with the mains. This acts like a 2.6K ohm resistor which limits the current entering the circuit to about 50mA. A 100mA 3.3V regulator is then used to stabalize the board power.
4. Triac. We are using a 600V, 4A triac to control the power to the load. The MC13224 controls the duty cycle of the triac so that it is only on during part of the AC sinewave. This has the effect of varying the power to the load (thus, dimming it).
5. Power Monitor. We have put a 0.050 ohm resistor in series with the load in an attempt to measure the current going to the load. The MC13224 has an integrated 12bit ADC so by measureing the voltage drop across the shunt resistor, we should be able to estimate the current. Since the curcuit also includes a zero-crossing detector, we should be able to calculate the power (including power factor) going to the load.
6. Temperature Monitor. We decided to include another circuit to measure the voltage drop across a varistor that is sensitive to temperature.