Do you have a question on Embedded electronics and controls? Ask our Expert, Paul Clarke

Mr. Clarke,

I have a relay and A/D board controlling my HVAC system at my house.  I have the software up an running, and it works fairly well, but I am having some trouble when the system needs to switch from heating to cooling or back.  There is a sensor in each one of 8 rooms, and if the temperature in any room is out of bounds, the system turns on the air handlers.  I have a database which sets the parameters for each room.  In addition to the sensor ID and a pair of time control parameters, there are four numbers associated with each sensor.  The first is a Boolean which determines whether the room is occupied or unoccupied.  The other three are temperature limits.  The first is the target temperature for the room when occupied.  When in heating mode, if the temperature in the room is lower than this, and the room is occupied, the room makes a call to the air handler.  In cooling mode the call is made if the temperature is higher than the target value.  The second number is the unoccupied minimum.  If the room is unoccupied, the system is in heating mode, and the temperature in the room is below this limit, the room makes a call.  The last number is the unoccupied maximum.  If the system is in cooling mode, the room is unoccupied, and the room temp is above the maximum, the call is made.  The problem manifests when the system has to decide to change from cooling mode to heating mode, or vice versa.  I've tried several approaches, and the best one so far is this:  There is a daytime and a night-time outside temperature target.  The daytime target is lower than the night target by three degrees, because the sun tends to heat the house, making it several more degrees warmer than ambient during the day.  Currently I have these set for 62F and 65F.  I then measure and record the outside temperature over the last 90 minutes.  I take the 15 minute average of temperatures 90 minutes ago, add the average tempertaure over the entire 90 minutes, and divide by two, allowing for a rapidly dropping temperature to have more weight than a slowly dropping one while still acounting for the fact there is a time lag in inside temperature trends due to the thermal mass of the interior of the house.  I then subtract the target temperature from the actual temprature in every occupied room and take the average, igoring any unoccupied rooms.  I take this number - which of course may be positive or negative, and add it to the weighted outside average.  If the resultant number is lower than the outside temperature target, then the system is switched to heating mode.

Do you have a suggestion for a better way to switch the system from heating to cooling or vice-versa?

Actually, no, there are two air handlers, each servicing four rooms.  (There are actually more than 8 rooms, but these are the ony ones that are monitored.)  Each room makes the call to a specific air handler, but there is only one variable which controlls heating and cooling.  It's not very wise to run the heat on one and the AC on the other, as the heat from one set of rooms tends to leak generously into the others.  I also have dampers and booster fans in the system, but their operation is not relevant to the question of when to cool and when to heat.  They simply turn on or off as appropriate during the conditioning cycle, and that code is working perfectly well, as is the code which makes the calls for each room.  It is the decision point for cooling vs. heatig that is not working well.  When  the sun has gone down and the house is cooling, depending on the exact numbers I deploy into the working constants, the system may either switch to heating too soon, leaving some rooms overly warm (the heater does not come on, of course, but then neither does the AC because the system has switched to heating mode), or else the cooling cycle goes too long and some of the rooms get too cool, albeit passively.  The reverse is true when the temperature transitions from cool to hot outside.

It may help you to take a look at the ststus screen from the thermostat programming utility.  You can view this at http://fletchergeek.homelinux.net.  You cannot see or manipulate the controls from your site (I haven't implemented logins, yet, and security is enabled), but it may give you an idea.  Ther is a little bug in the passive reporting that I need to stamp out. Your view does not show the correct occupied status.

Here is the live schedule matrix:

01 08 16 45 01 75 86 55
02 07 00 00 00 75 86 55
03 08 02 30 00 76 83 55
04 07 02 00 01 75 85 55
05 07 00 00 00 74 85 55
06 08 12 30 00 78 83 55
07 08 02 30 00 77 85 55
08 07 00 00 00 77 85 55

The fields are: Sensor number, Schedule area ( Sun, Mon, Tue,...Sat, Weekdays, Daily, Now), Hour, Minute, Occupied (1=yes), Target, Maximum, Minimum.  This data is fed from a master schedule which determines when the temperatures change and when the rooms become unoccupied.

I understand perfectly well about the holidays.  I am in the same boat myself, so don't be surprised if it takes several days for me to respond.

Hi Paul,

Was my question beneath you or something?

As you had a answer or sorts then I guessed you were moving forward.

To be totaly honest I would not have gone about this at all. However I'm a embeded electroincs engineer so have little transmsstion experiance - and these distances count as that.

I would have put it over RS485 which can handel the SNR at this distance, but maybe not the data rate. Optics is another option but costly, What about over a network, VPN or even RF? but how would you get the DC voltages there?

Your question is not beneath me, just not maybe my expert area I'm not one to just cast option on things I'm not 100% on.

If I may be so bold, it happens I am a qualified expert on transmission systems, so I hope I am not out of line if I interject something.  If this is a commercial or industrial application, fiber is really not that expensive.  If this is for a personal project, it's a little different matter, but even then it is not all that far beyond the pale.  Non-plenum duplex fiber cable can be purchsed retail in lengths of 5000ft or longer for about $0.30 a foot.  Of course that means one will have to splice up to 5 reals together.  At this rate, 25Kft will cost about $7500.  Of course this is more expensive than coax, but not hideously so.  I would not attempt to supply power to the end terminmation remotely.  Rather, I would obtain a local source of power, even if it has to be solar.  Ethernet media converters can be had foir under $200 each, and there are standard devices available to convert from analog to digital signals and back.  For more in-depth analysis, I would need more details of the exact nature of the signals.

If I may be so bold, it happens I am a qualified expert on transmission systems, so I hope I am not out of line if I interject something.  If this is a commercial or industrial application, fiber is really not that expensive.  If this is for a personal project, it's a little different matter, but even then it is not all that far beyond the pale.  Non-plenum duplex fiber cable can be purchsed retail in lengths of 5000ft or longer for about $0.30 a foot.  Of course that means one will have to splice up to 5 reals together.  At this rate, 25Kft will cost about $7500.  Of course this is more expensive than coax, but not hideously so.  I would not attempt to supply power to the end terminmation remotely.  Rather, I would obtain a local source of power, even if it has to be solar.  Ethernet media converters can be had foir under $200 each, and there are standard devices available to convert from analog to digital signals and back.  For more in-depth analysis, I would need more details of the exact nature of the signals.