May I suggest you restrict this one to a networked thermostat.

After this one is done a more complex family member with keypad, better display, etc is always an option. After they get the first one working they will be hungry for the second.

@arhi
What would you think of Bill selling the kits with a USB based bootloader? That way people without programing skills or equipment could download programs supplied by other users.

EDIT: If the USB port is there some clever body may setup a PC based UI to configure the thing, log data, or integrate with a larger system.

The sales potential for this kit, if it does not require programming skills, is thousands of times greater then that of the junebug.

There's the rub, I believe a thermostat especially an open source software design (I don't give away PCB artwork but the rest will be published) has the potential to go far beyond the scope of a typical setback thermostat. It would be possible to include a serial bootloader such as tiny bootloader in the device.
I've wanted to offer a thermostat kit as the first kit, it was among my first posts here when I joined as MyBlueRoom including an schematic (shown below in its last version from 2006).
One of the problems with that design were the LED displays did affect the temperature by about 2c after they were running for awhile. Also trying to squeeze so much out of a tiny PIC made for many compromises and I wanted the kit to be our flagship kit and a kit I would put to use controlling our homes Furnace & AC.
A zone based design seems the most logical and energy efficient but zoned thermostats are not nearly as cheap as a typical home thermostat and I love to tinker but your typical Honeywell or White Rogers are custom single chip designs and not exactly hack friendly.
Hopefully this new concept Cricket will attract firmware engineers that know something about temperature control (not as simple as it sounds) and will be willing to share.

On a side note I'll leave the design as simple and elegant as possible. No keypad (but I'll have to hunt down a low power encoder, or 3 pushbutton (UP,DN,SET) or gasp a joystick, or maybe IR. After all it can be connected to your PC... (Did someone mention XBee?)

Bill I agree with with your thinking on the design. It is possible to expand the potential market and sales without changing it.

As the software base for the device grows it will attract increasingly less capiable people.

In time the vast majority of users will buy a preassembled unit and flash it with user contributed code using the boot loader.

It would be easier for the user to have it preprogrammed, but then you may be looking at liability for the code. I understand not wanting to go there.

3v0

DUDE, you are having too much fun programming your little processor.
There are a lot of dynamics in trying to control the temperature of a dwelling with a heater that can only run ON or OFF. The heater has to pre-warm before turning on the blower, during which time the home continues to cool. Then when the fan turns on, at first it has to blow all the cold air through the vents before the hot air comes. On the reverse end of the cycle, once the target temperature has been reached and the heater shuts off, the fan still has to blow for a while to clear out the vents, cool down the heater, and make use of the stored energy.
How will you control hysteresis and rate of change reponse?
I am sick of squinting at LCDs. Light that sucker up!
I put a "nice" honeywell digital control up to replace a rusty old mechanical device. It has so much trouble trying to do it's job I constantly imagine putting the old thing back up.
BTW, very cool schematic. What software did you use, and what is that "Crickett" font?

That font is so old I've long forgotten its name I use ORCAD 10.3 to draw the schematic. It's old but I'm used to it.

The programming of most digital thermostats is mystery meat, who knows what's it's doing. Almost all under $100 thermostats only knowledge of the outside world is temperature (usually a cheap thermistor) and if it's a setback thermostat then time too.
I would think a smart thermostat with open source software could exceed most thermostat efficiency especially when connected to the internet.
So many options come to mind.
As for the hardware simple and extremely low power (unfortunately not battery as the communications circuitry would gobble up the battery relatively fast)
XBee is certainly on the table and might be a better choice than RS485, after all running wiring is more expensive than a $20 RF adapter.
The tough choice for me is the best use of the few pins remaining on the PIC.
Also KISS (keep it simple stupid) should apply, the more straight forward the design the less to go wrong.

PS I forgot how easy it would be to cheat on a jumper config till I looked at the old Cricket 88
Example: the DS1721 & LM75A (and most I2C temperature sensors) have a programmable address. But it doesn't have to be used that way.
Switch
1 = gas / elec
2 = C / F
3 = setup / run

and it costs zero IO on the PIC.

PS I may throw the PCB pads down for a couple or more I2C EEPROMs, seems some like data loggers.

Here's an example of a 14 segment font.

Yes I'll stick with PICs for the design.
I'm trying to avoid using TQFP just for the sake of kit building and yes I can solder TQFP I wouldn't expect everyone else to.
I was thinking of splitting the design into two parts much like the RCS-TR16. This has the head unit (2 digit display, temperature sensor and pushbuttons) in one part and uses the same 4 wire cable your existing thermostat uses but reassigns it to +,-,A,B
Where the relay and control unit are located near the furnace.
This would allow me to use the a PIC16F917 DIP with a simple 2.5digit LCD (as in my Cricket 2 design) in the head unit and a more powerful PIC18F on the control unit. It also would allow for mutiple thermostat head units to work with a single control. I could also mount a single relay in the head unit for aux control ie curtains or registers.

RCS-TR16 RS-485 thermostat.

Very nice, I'd love to design a kit that looked like the ZigBee SuperStat. I couldn't find a price on it though.

XBee is only $19US doesn't seem to pricey. Another problem with wireless is where do you get the power? ZigBee is not going to last long on a pair of AA cells. As for horsepower in the sensor; why? IMO it should be just enough to get the sensor data to the control unit, not really expected to do much more. Now the control unit could contain some power and possibly Ethernet.

It would be easy to use a 16x2 display but they are so common and so ugly. The wall box should be really basic IMO as the real power will be in the controller and your PC (web interface would be something)

Many PICs support LIN bus but it's single wire design is rated for only 30m

Another thought would be those low power Microchip ZigBee modules, they're $10US and use SPI and could be shut off except to transmit. The Thermostat head would only transmit on change and this could allow for a battery operated wireless unit.

ZigBee® Smart Energy Profile certified for full compatibility with Smart Energy meter or Energy Service Portal

this part is interesting ... I have no idea what it stands for .. is this the "Standard for the communication packets so they can understand each other" or this has to do with "energy consumption" and the problem you raised about power consumption of the sensor ...

Hardwire (24VAC) or “power-stealing” power option

what the .. "power stealing" ... ?!?!?!? I don't think google will show anything useful so if anyone can shed some light .. please do

anyhow .. if you use zigbee module only to "send data changes" .. it should last ..

the microchip zigbee modules 10$ .. donno .. never used one ..

as for the lcd .. you can always go wild and use GLCD + TOUCHSCREEN ... I got some from mikroe.com .. Easy Buy | Buy mikroElekronika Boards with a single click | mikroElektronika search for Graphic LCD 128x64 with TouchScreen .... the TouchScreen ( http://www.mikroe.com/pdf/touch_scre...cification.pdf ) works perfectly with mikroC ... have not yet had time to try to make it work with C18 (of course mikroC comes with closed sources support for it )

That would give you freedom to draw menus, buttons, be very user friendly ...

as for connecto to PC ... without reprogramming the device, the "input from PC" ain't gonna be any "more powerfull" nor easier to implement in PIC then a good menu .. IMHO

EDIT: with regards to network (point to multipoint or mesh) I opened Inexpensive Mesh Network ... might show up something useful if someone with experience show up