GBA Logo horizontal Facebook LinkedIn Email Pinterest Twitter Instagram YouTube Icon Navigation Search Icon Main Search Icon Video Play Icon Plus Icon Minus Icon Picture icon Hamburger Icon Close Icon Sorted
Guest Blogs

Ductless Minisplit Performance During Cold Weather

Does it make sense to turn up the thermostat setting when the weather turns cold?

The indoor head of a ductless minisplit system is a small wall-mounted fan-coil unit that blows indoor air over a copper coil containing refrigerant.
Image Credit: Fujitsu

I tried an experiment this week during our cold snap. We’ve kept the door closed to the first floor ell (bedroom and bath) and let it run cold, because the Fujitsu wasn’t sized to heat that space too. I opened the door early in the cold snap, and let the heat pump go, leaving it set on 70°F. What I found was that overnight the main space went to 66°F, and the upstairs and back bedroom were 3° to 4°F lower.

My calculated heat loss in these conditions is about 24,000 BTU/hour, and the heat pump is rated at about 17,000 BTU/hour at about 10°F. You’d think it would not be able to keep up.

My heat loss number may well be too high, and the rated output of the unit may be quite conservative.

A variable-speed minisplit tapers its output as it nears the setpoint

One other thing I wonder about is that even though the room was not at the setpoint it seemed that the unit didn’t run on full output much.

My system has the temperature sensing built into the wall cassette, so it may be sensing a higher temperature than out in the room. It may make sense in severe weather to set the thermostat up to 72°F instead of our normal 70°F.

Unlike a boiler, these variable-speed units taper off the output as the space approaches the setpoint instead of always running at full bore, so that may be a disadvantage of a smart unit — it’s trying to stay at a more efficient operating point instead of making me as comfortable as possible.

Anyway, comfort trumped further experimentation and we closed the door again to the ell.

Modem and router energy usage

On a totally unrelated topic: lately I’ve tried to identify the electrical loads in our house that use power around the clock.

I am reluctant to have anything in the house that uses energy 24/7. With 8,760 hours per year, small usages add up. One such is the exhaust fan on the composting toilet — at about 20 watts it uses 175 kWh/year, or 4% to 5% of our energy usage.

Another is the cable modem for internet access we have from Comcast, and the Apple Airport Express wireless router. I’ve been measuring both. The router uses just over 3 1/2 watts, while the modem uses 6 1/2 watts. Together they use 88 kWh/year.

I imagine that if we were off-grid, we’d unplug them in the winter when solar power is short. Being grid-tied, we’re just lazy about it.

What’s using power 24/7 in your house?

Marc Rosenbaum is director of engineering at South Mountain Company on the island of Martha’s Vineyard in Massachusetts. He writes a blog called Thriving on Low Carbon. Marc teaches a 10-week online Zero Net Energy Home Design course as part of NESEA’s Building Energy Master Series. You can test drive his class for free.


  1. watercop | | #1

    It would be interesting to
    It would be interesting to repeat the experiment with a setpoint of 72, or perhaps even 75, during the next cold snap - let the Fujitsu really stretch its legs

    It would also be interesting to monitor its power consumption under various conditions of outdoor temperature and deviation from setpoint, if you haven't already done so.

    I share your pursuit of and disdain for 24/7/365 phantom loads. Without running around and unplugging all the various chargers and wall warts typical of a house of 5 highly digitally connected people, about the lowest we see is 120 Watts. Notable loads include garage door openers (14 Watts each), HVAC control power (12 Watts).

    At night, we run around 190 Watts owing to addition of some outdoor security and decorative lighting.

    For all of 2012 we averaged 25.75 kwh / day, so phantoms are a significant portion of the total.

  2. user-984364 | | #2

    24/7 power
    I've been running around measuring it all with a kill-a-watt, and came up with only 90W (that's including my web/email server and lots of other digital equipment). But my homemade whole-house monitor is saying more like 150W overnight; I don't know if there's something I'm missing or if my monitor is inaccurate at low loads. I'm wondering if there's an old HVAC transformer hidden in the wall somewhere. I also realized that even my UPS for the server has about 6 watts overhead, and in fact its overhead goes up the more I plug into it.

    Our largest always-on load is actually the radon fan at about 20W. Not much to be done about that I suppose.

    What do you use to accurately measure your whole-house, overnight load?

  3. davidmeiland | | #3

    Suggestions for circuit-level monitoring
    What are you guys using to monitor your panel (if that's what you're doing)? I need something that will handle several 30-50 amp 240v loads, and several 120s. I've been considering this one from Brultech:

  4. GBA Editor
    Martin Holladay | | #4

    Tracking down phantom loads
    I'm glad to see that there are others out there obsessed with phantom loads. Since I live in an off-grid house, I have to be concerned with these loads. They are battery-drainers.

    I won't be commenting on the GBA site for the next week. I'll be on vacation until March 2. I'm leaving GBA in the capable hands of Rob Wotzak.

  5. dankolbert | | #5

    Dave - here's a Maine company a friend of mine owns that has helped us out on a couple of jobs.

  6. JonathanTE | | #6

    Smart power strip or timer to control phantom loads?
    Marc, two thoughts come to mind for lazy control of those modem and router loads. First is to plug them -- along with the computer -- into a "smart" power strip, which will automatically shut the modem and router fully down whenever you turn off the computer. Boot the computer back up, and the modem and router will automatically turn on at the same time.

    The dumber version of lazy would be to connect them to the wall via an outlet timer. If you are off the interwebs consistently from, say, mindnight to 7am, you set the timer to be off at those times. Of course the timer uses up 0.5'ish watts (what I measured with a Kill-a-Watt years back, if memory serves), so to be worthwhile it has to be shutting down the modem and router long enough each day to compensate for its added load. If both objects are connected to the same timer, that won't be hard to accomplish with even just a few hours of down time.

    A possible advantage of the outlet timer option is that--in my experience--modems take a little while to get up to speed in being fully connected to the internet, longer than it takes my computer to be up and running. Having the modem (and router) turn on in the morning before you tend to be on the computer gives it the necessary time to coordinate with the ISP and be more seamlessly at your beck and call. The disadvantage is that it won't save as much energy because you'll set the timer "conservatively" outside your computer use time range, and it will turn on the modem and router even during vacations when you are away from the house and have no need at all for them.

  7. watercop | | #7

    I've installed several TED
    I've installed several TED systems. The one PITA with TED is that its signals across home power wiring are sensitive to noise superimposed by power supplies. Other than that, TED works well, and seems to be highly sensitive and accurate, to well within 1% of utility meter.

    Our range, dryer and big screen all use just a watt or two on standby. The whole house genny trickle charger uses 10 Watts or so, somewhat reduced by a solar panel.

    Other loads folks may not consider are door bell transformers, hard wired smoke alarms, clock radios, and the myriad of device chargers incidental to the digital age.

  8. user-984364 | | #8

    Down to the Doorbells actually asks how many doorbells you have when modeling your home. :)

    Here's a post I did pretty long ago about everything I could measure. Most of it is from a Kill-A-Watt, but some is from eyeing the whole house monitor and turning the appliance on or off (including the doorbell).

    Some of it has changed since then - I need to do an update, but it is surprising how so many little "insignificant" loads can really add up.

    I'm probably going to go with the Brultech monitor soon - seems like a relatively inexpensive way to get several circuits monitored, and I hope it's more accurate for the whole-house load. Since we have solar, we have to look at net grid input/output as well as current solar production to get the house load.

  9. Tonyparker | | #9

    Building energy efficient home
    Building energy efficient home is one of the ways to reduce the effects of global warming. Home sweet home, therefore, set your house as good as and as comfortable as possible in order to live comfortably.

  10. Marc Rosenbaum | | #10

    @Curt - I agree, if we get another snap with temps under 15F... may not happen this season.
    @Eric - that's an impressive list of standby power draws. I'm confused by the refrigerator numbers - seem quite high for standby, I don't recall exactly what mine used when I had it on a power monitor but I recall feeling it was not worth worrying about. Your standby numbers, times 8,760 hrs/yr, are almost as much as my fridge uses.
    Also, I've seen garage door openers over 10W, yours is pretty low at 3W.
    @Various - I went to the Brultech site and it wasn't clear to me if there is a cost to look at the data online. The eMonitor has a monthly cost :-( but on the flip side has a very usable online interface. Has anyone used the Brultech?
    @Curt - I used an early TED and it was really a PITA yet what annoyed me more was that that the company was completely unembarrassed that they had major bugs and were charging me to be a beta tester. Have they improved? By way of comparison, I've found Energy Circle (vendor) and Powerhouse Dynamics (manufacturer) to be very responsive companies.
    @Jonathan - the timer is a interesting solution. Does the order that the router and modem come on matter?
    Thanks all!

  11. user-984364 | | #11

    My fridge numbers
    Marc, I reported the fridge numbers confusingly, sorry. For the refrigerators I listed the average draw over time while operating, not pure standby. (I mentioned it in the post, but maybe not prominently enough). True standby was indeed unmeasurable. I averaged it so I could basically find out "If I left for a week and all HVAC was turned off, what does the house load look like?" Sorry for the confusion. I'll re-measure the garage door opener to be sure; it is only a few years old, so maybe better than average.

  12. user-984364 | | #12

    Marc, I don't know for sure if there is a Brultech reporting solution which requires a subscription, but if I get it I plan on extracting data over the serial port (which seems fully supported and documented) and doing my own graphing/reporting via other means. There are open source utilities out there to send the data to various endpoints.

  13. watercop | | #13

    Marc - I haven't had to use
    Marc - I haven't had to use TED tech support in awhile...when I did it was passable. Manuals aren't bad and user forums bridge the gap as well.

    I've been leaning toward E-Guage in order to monitor many more circuits as well as avoid monthly internet fee, but I haven't tried one yet.

  14. Marc Rosenbaum | | #14

    proliferation of products is great
    I hadn't even heard of Brultech and eGauge. It seems as though the biggest advantage of the eMonitor is the fact that there is no monthly subscription cost. Other than that, I've been very pleased with the eMonitor product, so I look forward to others' posting their experience with these other products.

  15. user-1135248 | | #15

    To try and pull back to the
    To try and pull back to the original topic...

    Your results aren't surprising, as it sounds like the only mechanism to
    move heat around your place is convection and some amount of internal-structure
    conduction in the test scenario. Assuming you're in a well-insulated
    envelope that should yield more even temps across different parts of the
    space but as I'm finding in my own retrofit it certainly isn't perfect. And I've
    got a *ducted* system, albeit with somewhat wimpy ducts going to the upstairs
    and no return other than down the staircase so it runs about the same 3-ish
    degree delta cooler too.

    What are you counting as "full output" on the minisplit? Fan speed may not
    be the only indication, you'd have to also look at compressor current.
    My Daikin definitely ramps down during cooling, but in heat mode it comes
    right up to its full 7 or 8 amps and stays pegged there until the
    setpoint is satisfied -- however, the fan speed *does* vary between high and
    low during that time. Probably has to do with the capacity the system can
    actually provide vs. the design not wanting to blow cooler air into the space,
    so if the indoor coil isn't quite keeping up a lower fan speed will allow that
    air to at least emerge at a higher temp.

    There's really no reason for a heat pump to regulate capacity on heating,
    as it's not trying to balance latent vs. sensible loads. The Daikin folks
    seem to have considered it that way, anyhow.

    The other experiment would be to heat with pure resistance for a test period
    instead, for an [expensive!] COP of 1.00, and measure that. I've been trying
    to do that with my system but there are some airflow and basement-slab
    nonlinearities that are making the data a little ... odd.


  16. user-1005777 | | #16

    Heat pump low temperature performance
    I have a Mitsubishi 12000 btu unit (MSZ-FE12NA) installed in October of 2011. The minihome is 16 feet by 68 feet with 2 bedrooms and 1/2 bath at one end and the master bedroom and main bathroom at the other. The unit is installed in the living room that has a cathedral ceiling.The coldest temperature it has endured is -16.5 F (that is a minus) The living room was down to 65F in the morning, but I read the temperature at the input to the unit with a remote laser temperature sensor and it was at 71F. I then put the unit into "powerful" mode, that ignores the sensor temperature. The room returned to 70F within minutes. The end room thermostats did cycle that night at 61F to bring on the baseboard heaters. This was the first time they have come on since the heat pump was installed. I believe that the heat pump could have kept up if the controls were better. Several options come to mind.....the cheapest being a remote temperature sensor at floor level that is averaged with the sensor in the high mounted wall unit. Software could then choose the best option for remediation.

    The best option would be an outdoor temp sensor and manual input of heat loss per 100F delta T. The software could then heat to that need. If the house gets too warm, reduce the heat load requirement in the equation.

  17. user-1118336 | | #17

    heat pump fan/compressor speed in heating mode
    Hobbit/Roger/others: Defrost cycle (or avoiding frost-up) contributes to the apparent odd fan speed vs. outdoor unit energy input you're seeing. Mini-split manufacturers have different methods to prevent frost problems, so the answer for a fujitsu may not be the same as for Mitsubishi, LG, Daikin or other. These outdoor units are designed to run well below freezing and use mfgr-specific/creative cycling to prevent frost build-up; not your father's heat pump that shuts down around 40F. Personal experience is similar to Roger's: my Mitsu MSZ-09 indoor fan seldom runs anywhere near full speed, but never short of providing great heat (although room temp usually runs 3F greater than the remote's tstat setting).

  18. user-939142 | | #18

    Measuring the juice
    One small trick with the computer and its components is to run the peripherals off the computers internal power supply rather than using all the bricks. Most the bricks for printer or modems want around 5V or 12V and the computer power supply outputs just that. And when the computer is shut off, those devices lose power too.

    For measuring phantom loads, you really have to get the voltmeter out. I have a whole house measuring device and it is just not accurate enough when you start nit picking one watt at a time.
    I like my Efergy whole house device as it is highly mobile so you can move it around to separate feeds, share it with your parents, etc. Once you measure and get an idea, you are unlikely to need to keep the thing monitoring forever. You can take the several hundred dollars in savings and buy insulation or LED bulbs.

Log in or create an account to post a comment.



Recent Questions and Replies

  • |
  • |
  • |
  • |