Furnace Fun

Just like every winter's day, I woke up and went to the thermostat to turn up the temperature. There in front of me was a boring, old device with the brains of a switch. I should upgrade that I thought and that day I ordered two Ecobee3's.

Installing these devices was straightforward. I didn't have the dreaded C wire where I installed them but they came with a nifty Power Extender Kit (PEK) to overcome that hurdle.

Life returned to a better version of normal. I could control my thermostat with my phone from anywhere I had an internet connection. And I could get automatic reports of how I was using my air conditioner and furnace. I could even say, "Siri, go to sleep" and Siri would set the thermostat to my nighttime settings.

Then something strange happened. I noticed that the thermostats kept turning themselves off an on. This was pretty annoying because the process took several minutes and I couldn't use the thermostat in that time.

"Darn, the PEK isn't working. Let me run some common wires up to the thermostats and see if that fixes things."

And I did. It sounds daunting to run wires through the walls of your house but as long as you already have one wire going where you want then it's not so bad.

It didn't fix the problem. A challenge! After much time on the phone with Ecobee and much time researching on the internet, I realize that the source of my problems are my two furnaces. For some reason, they are turning themselves off which in turn is turning off the thermostats. So why are they doing that?

After reading too much about my furnaces, I learn that they (and all furnaces it seems) have a protective feature called a limit circuit. This circuit has sensors on it that check that the furnace's conditions are correct for operation. For example, there is a sensor to make sure the flames are in the right place (important!) and that the input and output temperatures don't differ by too much.

As it turns out, my furnaces were both throwing limit circuit faults. But why? To figure this out, I turned to my Raspberry Pi. I decide to outfit it with five thermometers and a voltmeter built using an analog-to-digital convertor so that I could better monitor what was going on with my furnace. A wrote a web server using node that served up a d3-based website so that I could see the data. And low and behold, I found that when the difference between the output and input temperatures reached 70 degree F then the limit circuit fault would occur and the furnace would turn itself off.

Now that I had diagnosed the problem, fixing it should be easy I thought. Think again. I had the installers come out ($1400) to try to fix it. They said the problem was that the air conditioner coil was clogged and I needed to get it and the ductwork cleaned ($999). When that didn't fix it, I went online and got advice ($95) from Brian HVAC Guy but that petered out when things seemed to indicate that that furnaces were oversized for the house and the ductwork. Perhaps throwing good money ($3000) after bad, I added more and larger returns as well as an output exhaust to the furnaces.

And this morning I watched the thermostat turn itself off and on again. Sigh.

Frustration zeroed out by resignation, I let things be for awhile. But each time I walked by one of my $300 thermostats and saw it reset, that frustration account grew. With interest. Eventually, I returned to the problem.

I decided that I might be able to avoid the circuit limit fault by simply not calling for heat when the failure was about to occur. I decided to build a device that would measure the temperature difference between the air in the return duct and the air in the outgoing duct and that would use that to determine if a failure was about to occur. The device I built was Arduino-based ($12), had two thermistors ($0.20) to measure the temperatures and a relay ($0.60) to cut off the 24V call for heat signal sent from the thermostat. The device monitored the heat difference until it was above 70 degrees F, then it opened the relay stopping the call for heat. Then when the heat difference dropped to 40 degrees F, it closed the relay allowing the call for heat again. This worked well... for awhile.

In the meantime, I also considered the placement of the return vents. They had been attached at the smaller vents and so I questioned whether they could really be improving the amount of air being returned. It seemed like the smaller vent would still be restricting how much air could return. I thought that attaching them at the expanded duct at the interface to the furnace might be a better idea. So I did that. I thought it improved things slightly but it did not resolve the circuit limit fault.

One particularly cold morning, I woke up to find the upstairs furnace blowing cold air. This was new. With a limit circuit fault, the furnace simply turned itself off; it didn't blow cold air. I went downstairs to see what was going on and found to my dismay a new error code displayed. This one was a low pressure switch failure. This switch wants to detect a pressure difference between the combustion chamber and the exhaust. It indicates that exhaust fumes are being adequately drawn out of the combustion area. Some things that can cause it to fail are restrictions in the air supply, restrictions in the exhaust pipe, windiness or a bad sensor.

The easiest thing for me to try was to replace the sensor ($75). So I did this and it had no effect.

When the furnace failed in this way, it went into a mode where it wouldn't produce heat for three hours but would continue to blow air. Power cycling the furnace caused it to produce heat until the failure occurred again. As a result, on very cold mornings when this failure happened (and the failure did seem to be correlated to low outside temperatures) I went downstairs and power cycled the furnace. After I got tired of the trudge up and down the stairs, I used an ESP8266 computer to create a remote switch. This nifty little computer has WiFi built in and can act like a web server. So I wrote a program that had it serve up a web page with a button on it that when pressed would open a relay ($9) connected to the furnace's power for 10 seconds and then close the relay thus restarting the furnace. So when I felt the furnace blowing cold air, I would go to the web page and press the button. The furnace would restart, heat the house and often not exhibit the failure again that day.

Despite this workaround, I checked the intake and exhaust pipes as best as I could and save for a few leaves that I removed, they seemed fine.

One morning, the low pressure switch failure kept occurring even after restarting the furnace. Again frustrated to the point of desperation, I called a furnace repairman to take a look. (Yes, calling for help had become a desperate move.) A fellow named Kirby showed up to help me. The failure was occurring when he came and so that was good. But with the top door of the furnace open it did not occur. When he closed that door, it happened almost immediately. Aha! After that, he disconnected the air supply and just let the furnace breathe the air from the basement (which further research seemed to indicate was fine if not very efficient). This seemed to fix things! Yay! He left. Later, the error recurred. OMG. Boo. But at least the failure did not seem to be occurring as often.

While Kirby was here, I also talked to him about the limit circuit fault on the other furnace. He suggested that I could change the sensor to allow for a higher temperature before the fault occurred. The sensor is simply a switch that triggers when the temperature reaches a certain level - 160 degrees F in my case. That is a very low temperature and certainly doesn't represent a fire hazard. So I went ahead and ordered a 250 degree F switch ($7.66) but I did not install it.

I decided to take a better look at the exhaust pipe. I bought an endoscope ($36) to help with this but the pipe looked fine. I even cut the exhaust pipe open to see it at every point and it was completely clear.

I wanted to better diagnose the pressure switch issue and thought I could do this with a manometer ($82). I haven't quite figured out how to use this to do that but I was able to measure the duct pressures with it which I believed were the cause of the circuit limit faults. For posterity, this is what I measured for the upstairs furnace:

Location	Pressure
Before Filter	-.2 wc in
After Filter	-.36 wc in
Before Coil	.56 wc in
After Coil	.24 wc in

I don't remember all the rules of thumb I learned online, but I do recall that nothing about these numbers seemed alarming. So no help there in solving the problem.

One morning I woke up and the temperature downstairs was 15 degrees below what the thermostat was set at. For some reason the device I had built to stop the call for heat when a failure seemed imminent was not re-enabling the call for heat. I disconnected it rather than debug it and decided to install the higher temperature limit circuit switch I had bought on Kirby's advice. I did then then carefully monitored the furnace in operation. The output air temperature stayed at a steady 145 degrees F while the input temperature hovered around 73 degrees F. This is near the maximal difference recommended for the furnace so that seemed fine. And importantly, the furnace ran non-stop for several hours until the temperature rose 15 degrees F. Hallelujah! This is how a furnace is supposed to work! And with no crazy devices attached!

That same morning, the upstairs furnace was in its failure mode as well. (It seemed its failure was not only correlated with the outside temperature but with the indoor temperature as well. That is, when it was warm downstairs then it was more likely not to fail upstairs.) In any case, I had woken three times in the middle of the night with cold air blowing on me and had remotely restarted the furnace. Inspired by my success with the other furnace I decided enough was enough. I decided to change the remote switch to work automatically. I wanted the switch to be able to detect that the furnace was in a failure mode and to automatically trigger the furnace restart.

The furnace's failure mode was indicated by an LED. When the furnace was operational, the light blinked between a medium level and a high level. When the furnace was in an error state, it turned off and then blinked at a high level a certain number of times to indicate the error code. It then repeated. Unfortunately, the leads of the LED were not accessible so I couldn't measure its voltage directly. So I decided instead to use a photoresistor to let my smart switch "look" at the LED. The algorithm was to read the voltage across the photoresistor 10 times per second and count the number of times it dropped below some threshold that indicated the light was off. (The light was only off during an error state.) Then when this count exceeded 100, the furnace's power was turned off for 10 seconds. This device worked perfectly and kept the furnace from blowing cold air on us at night. It also kept the temperature at the thermostat's setting.

Of course this was not a fix and I continued to consider the problem. Another hint that I got while installing this smart switch was the the error did not occur while the bottom door was open. It then happened immediately when the door was closed. So along with what I saw with Kirby, it seems that leaving the top door open or the bottom door open is enough to keep the error from occurring. Something to consider.