Debugging Refrigerator Problems

About a year ago, our freezer started generating excessive moisture inside, to the extent that, when the defrost heater was active, water would drip from some of the shelves and, worse yet, cause the ice drawer under the ice maker to freeze into a solid mass of ice.  We lived with it while trying a few “simple” fixes like checking the door seals and regularly emptying the ice drawer.

None of those simple fixes really addressed the problem.  That was, however, until we went away for a week at Christmas and happened to turn off the ice maker.  When we came back, the problem seemingly had gone away.  The ice left in the drawer was still in individual pieces.  Maybe the problem was with the ice maker?  Everything seemed perfect.  …Until a week later…  When the fridge was no longer cold.  Hmmm…  Back into debug mode.

A little bit of detective work I found that, although the fan in the freezer was running almost continuously, there was no air flowing into the fridge portion – the dampers were closed.  So, I bit the bullet and tore into it.  After a few Google searches, I decided to take the back panel of the freezer compartment off (very simple it turns out) to check the coil.  I found a solid block of ice around the coils.  Ah Ha!  The ice was blocking the flow of air, meaning no circulation, meaning no cold air flowing into the fridge side.  After some hot air from a hair dryer and some towels to mop up the moisture, the coils were unfrozen.  But why did they freeze?

Next I pulled the defrost heater.  The glass tube looked suspicously black.  Checking continuity, I found it was an open circuit.  Another Ah Ha!  No defrost heater = icing of the coils.  Also, no defrost heater = no melting of stuff in the freezer to cause the excess moisture.  At least one problem solved.  The defrost heater was replaced and all was well with the fridge.  Even the freezer seemed to be working better.  For a couple weeks.  Then the excess moisture came back again.

In the meantime, after the defrost heater issue, I decided to put a monitor node in the fridge/freezer so I could track temperatures via my home MRBus network (on a computer running MRBFS and a webserver).  The node uses was a wireless MRBW-RTS quad temperature monitor.  It uses cheap 2N3904 transistors to sense the temperature.  Each sensor can be placed many feet away at the end of a twisted pair of wires, allowing the transmitting node to reside in the fridge side (to make the batteries happier) while four remote sensors can be placed elsewhere in the fridge and freezer.  The on-board temperature sensor in the MRBW-RTS is used for one temperature point (bottom of fridge) while the other four sense locations use the 2N3904 remote sensors:

  • Bottom of freezer (same location as freezer thermistor)
  • Top of coils (adjacent to evap thermistor)
  • Top of freezer air duct
  • Top of fridge
  • MRBW-RTS and the sensor wires.  Note: A PRO XBee radio is used to get through the fridge walls plus several house walls between it and the receiving MRB-AP.  Even then, the typical receive strength (RSSI) reported by the MRB-AP is only -90dBm!  The entire node is enclosed in a ziplock bag to protect it from moisture.


The sensors are 2N3904 TO-92 transistors with their base-collector soldered to one wire of the twisted pair and the emitter soldered to the other wire.  The entire assembly was then coated with white liquid electrical tape to prevent moisture from affecting the measurement.  Fortunately, there is a hole between the freezer and fridge compartments with a manually adjusted valve that could be used to route the wires between the sides.

A picture of the entire freezer side:

After outfitting the fridge with the necessary telemetry, I started collecting data.  It quickly became obvious when the freezer went into its defrost cycle.  Each horizontal tick is 30 minutes.


The peak temperature at the top of the freezer and even the evap coils seems high, as does the duration (on for 30 minutes).  I have not yet found any definitive references on the peak temperature and duration for this specific model (GE GSH25JFXB), but the values I am measuring tend to be at the upper end of the ranges mentioned online.  This could explain the excess moisture.  Also, the rubber seals around the panel that covers the coils are deformed.  One of them was replaced when the defrost heater was replaced, but quickly deformed just like the one that was replaced.  This also leads me to believe the defrost heater is running too hot or too long.


Next step was to check the thermistors.  I checked both freezer thermistors in both an ice bath and in warm water, using one of the MRBW-RTS channels as a reference, especially for the warm water test.  In all cases, the thermistors, measured at the connector on the control board on the rear of the unit, measured correctly according to the table.


Sigh…  So what next?  So far, we’ve been back to just living with the problem.  All the online sources say to replace the thermistors when encountering a problem like this.  Although mine measure correctly, I may try replacing them anyway.

I would really like to get my hands on some technical data for the GE GSH25JFXB to determine if the defrost peak temperature and times are within spec.  If you know of any such data, please let me know.

Other options I’ve considered:

  • Replace the control board.  This is the other common “solution” suggested, but at $120 a board, I’d like to avoid that one if I can.
  • Fake the temperature rise using the data from the evap coil MRBW-RTS sensor to trigger a relay that shorts the real thermistor.  This should simulate a rise in temperature and cause the control firmware to terminate the defrost cycle.  However, it might also cause an error condition and doesn’t really solve the problem – just patches it.
  • Call a repairman.  Now come on, I’m an engineer. 🙂  Seriously, though, I’m not sure what they would/could do differently than what I’ve already done, aside from just replacing things to see if it fixes the problem.  But if you happen to be one or know one, I’m willing to be convinced otherwise…

If you have any other suggestions, feel free to leave them below in the comments.  I will continue to update as more progress is made.

Update (July 20, 2013)

The fridge supposedly contains an adaptive defrost mode.  All the references to this mode on GE refrigerators mentions a pre-chill mode prior to going into defrost mode.


However, I don’t see any sign of this in my temperature graphs.  The purple line (bottom freezer sensor) hovers around 0F nicely, but never takes that 15F dive before defrost.  That might explain why things are melting.  (Note: in the plot below I shorted the evap thermistor at the control board as an experiment.  It certainly terminated the defrost earlier, but oddly, the 2nd defrost with no short was also similar…  Subsequent defrosts went back to like shown above – too hot.)


Update (July 20, 2013)

Tonight I caught the fridge right when it was beginning a defrost cycle.  So, I decided to try shorting the thermistor again and watching the whole sequence of events.  When the evap temperature reached about 45F, I shorted the evap thermistor at the control board.  The defrost cycle did not terminate immediately, so I left the short in place and watched.


The defrost cycle continued until about 15 minutes had elapsed.  Maybe there is a 15 minute minimum time?  Anyway, the heater turned off.  A 15 min dwell time elapsed, which is longer than the specified fixed 5 minute dwell time – maybe because defrost terminated so early?.  Then the compressor turned back on, followed by the fan ~15 min later.

Update (July 27, 2013)

This week has been one of waiting and watching.  After shorting the sensor last week, the defrost behavior continued to look good.  The peak temperatures were more what I would expect and the duration much shorter.  Also, the period of time between defrosts was significantly longer.  I also played with the freezer temperature control to see how low the freezer could go (Freezer = 9, reset to 5 on 7/23/2013 @ 20:00).


On 7/25 (18:38 to be exact) I turned off power to the fridge for ~30 seconds.  Suddenly, all the defrost cycles went back to their “bad” characteristics!


Hmmm…  Did shorting the thermistor cause the controller to go into some “limp home” mode?  (Thanks Nathan for that insight.)  The diagnostic mode can detect shorted thermistors, so maybe the normal control algorithm looks for it, too, and if found reverts to some default values for the defrost time ignoring the adaptive defrost feature.

So now I’m back to suspecting either the thermistor or the circuitry that reads the thermistor on the control board.  Since there is an order of magnitude difference in the cost of each, I think I will try replacing the thermistor first…

Update (August 3, 2013)

I did a bit more investigation to try and understand the control board.  First, each thermistor input has ~14.9k to GND.  This sets up a voltage divider between the thermistor (tied to 5V) and this 14.9k resistor to GND.  This voltage is then filtered and sent off to the microcontroller, presumably to an onboard ADC.  Watching the evap thermistor voltage during a defrost cycle and inferring the temperature, the thermistor temperature seemed to track well with the temperature reported by the RTS sensor next to it.  The point at which the defrost was terminated was in the neighborhood of 3.47V or ~70F.  This is equivalent to ~6k across the thermistor.

Next time, I installed a 12k resistor in parallel with the thermistor.  If my theory was correct, this should cause the defrost to terminate when the thermistor is 12k (12k || 12k = 6k) or about 40-50F.  Sure enough, the defrost stopped when the voltage reached ~3.48V and the temperature reported by the RTS was 52F, just as I expected.

Update (August 9, 2013)

Without much of anything solid to go on, I decided to try the good ol’ “replace things and see if it makes a difference” approach.  First, I replaced the evap thermistor.  If anything, this made the problem worse.  Now, the defrost stays on longer and ends up triggering a mode in the controller where the defrosts are no longer adaptive and just go every 8 hours or so.  Next, I replaced the control board ($$$) – no difference.  Doh!


First defrost is with the new evap thermistor.  Second defrost is one I forced (1 4 test mode) and has the new evap thermistor and new control board.  No notable difference.  Subsequent defrosts showed the same behavior.  Tonight I replaced the freezer thermistor (no good reason, but figured I should at least try).  We’ll see…

One thing I noticed is that the evap temperature reaches about 70F, then stops rising for a while.  After 20 minutes or so, it then spikes and triggers the control board to shut things down.  Odd.  Can’t explain that one yet.  But it does seem to explain why the defrost cycle is so long – the temperature is hanging out just below the threshold needed to stop the defrost cycle.

Update (September 1, 2013)

Back on August 11, I put a 24k resistor in parallel with the evap thermistor in an attempt to trick it into terminating early.  The ice melting problems have so far been solved.  No melted ice, no high temperatures or water in the freezer during defrost.


Below is a comparison of the defrost behavior before and after (respectively) the addition of the 24k resistor:



The defrost is clearly running much less often and the peak temperatures in the freezer (red trace) are lower.  Being paranoid, I checked whether there was a glacier forming around the evap coils.  Before a defrost, there was a fair amount of ice.  Uh oh…


So, I allowed the freezer to defrost and checked again:


All clear!  Phew!



  1. A very detailed presentation from where I had to learn a lot from your experience. I live in a city in Romania where there is no specialized maintenance service. I couldn’t find any documentation for my refrigerator. But I found the explanations very useful. Thanks, George

Leave a Reply

Your email address will not be published. Required fields are marked *