GBA Logo horizontal Facebook LinkedIn Email Pinterest Twitter Instagram YouTube Icon Navigation Search Icon Main Search Icon Video Play Icon Audio Play Icon Headphones Icon Plus Icon Minus Icon Check Icon Print Icon Picture icon Single Arrow Icon Double Arrow Icon Hamburger Icon TV Icon Close Icon Sorted Hamburger/Search Icon

Community and Q&A

Minisplit sizing

Tony Tibbar | Posted in General Questions on

When comparing specs from the same model line I noticed that the maximum BTU/h goes up as the tonnage goes up but the minimum BTU/h for heating, stays about the same.

Let’s compare 2 units for heating.
9000 BTU/h –> 5700-11910
18000 BTU/h –> 5620-18540

Assuming the same price, which unit would be the best choice with a 9000BTU/h house?
Just looking at the range I would say the larger unit because it the lowest setting is the same as smaller unit but it as more headroom for those rare very cold days.
But that’s assuming the units can regulate their output very precisely and not just 3 BTU settings.

Or is the ideal sized minisplit the one which has the modeled heat load about in the middle of it’s capacity range?

Short cycling is bad for efficiency and lifespan of the unit.
I wonder what happens during the days the heat requirement is only 3000 BTU/h.
Are units with a minimum BTU/h and equal HSPF the best for those days? Or is the difference so minimal it’s to be considered noise?

GBA Prime

Join the leading community of building science experts

Become a GBA Prime member and get instant access to the latest developments in green building, research, and reports from the field.


  1. GBA Editor
    Martin Holladay | | #1

    There are lots of issues here.

    1. If you have accurately calculated your home's design heat loss (using Manual J or a similar tool) and found that your design heat loss is 9,000 BTU/h, then you don't need more capacity than 9,000 BTU/h. Your design heat loss is the heat load on the coldest day of the year.

    2. You don't tell us your climate zone. If you live in a cold climate, you need to determine the heat output of your equipment at cold temperatures. It's possible that the equipment you are looking at is rated at temperatures that are warmer than your coldest expected outdoor temperature. For example, here is some information from Mitsubishi: Mitsubishi equipment has a heating capacity equal to 100% of its rated output at an outdoor temperature of 5°F. At -4°F, the heating capacity drops to 82% of the rated capacity. At -13°F, the heating capacity drops to 62% of the rated heating capacity.

    3. Other factors being equal, it's advantageous to choose equipment with a wide output range, since the equipment will usually be operating at partial load. As you correctly note, frequent cycling is less efficient than continuous operation.

    For a more thorough discussion of these issues, see How To Buy a Ductless Minisplit.

  2. Stephen Sheehy | | #2

    Tony: you should compare specs. The Mitsubishi and Fujitsu 9000 bthh minisplits can operate over a wider range than the Pioneer unit, with much lower minimum outputs and higher maximum outputs.
    In addition, they come with much longer warrantees. Finally the SEER and HSPF ratings are much higher.

  3. Tony Tibbar | | #3

    At Martin: Thanks for the link to the other thread. Very useful.

    At Martin and Stephen: I'm aware there are a great many factors that come into play when picking the ideal unit. My main question when writing my post was if there is a difference in efficiency and/or comfort when both units output, say, 7000 BTU/h.

  4. D Dorsett | | #4

    On the cheaper-mini-split end, the LG ArtCool Premier series modulates down to ~1000 BTU/hr out in both heating & cooling modes, and has respectable efficiency numbers too:

    It's important to have a reasonable handle on the actual loads, but the size & configuration of the space factors in too. The larger units may still modulate down to the same BTU/hr levels as the smaller ones, but it may not be as efficient at minimum modulation as the more right-sized unit due to the higher blower cfm and correspondingly higher power draw at minimum modulation. The higher cfm can also be a comfort & noise problem in small spaces, but can sometimes be useful for low load but larger or oddly configured spaces.

  5. Stephen Sheehy | | #5

    Tony- larger units from the same manufacturer have slightly lower SEER and HSPF numbers, so I assume that at a given output, the smaller unit is probably a bit more efficient.

    For what it's worth, my Fujitsu 9RLS3 operates from 3100 to 22,000 btuh in heating mode and from 3100 to 12000 in cooling mode, with SEER of 33.0 and HSPF of 14.2. If noise is important to you, it ranges from 23 db to 42 db.

  6. Tony Tibbar | | #6


    Yeah, I noticed differences in SEER and HSPF ratings.
    I wonder they tell the whole story because *maybe* the larger unit works more efficiently at really low temperatures. The smaller unit has to work very hard but the larger unit still manages the load with ease.

  7. Tony Tibbar | | #7


    I'm doing my best to get as acurate loads as possible. But besides of that I'm trying to learn how the sizing works for any given heatload. I'm making a summary of all info I found on this forum.
    Below I've quoted that summary and would appriate it very much if you/anyone could point out errors.
    I got the load at various temperatures by simply changing the design temperature.

    A minisplit works efficiently when it's modulating. If not efficiency, lifespan and comfort drops.
    Modulating means it's (almost) always running.
    A minisplit archives that goal by increasing/decreasing it's output.
    The most ideal, but not existing, situation is when the heat load is always within the modulation range. At the bottom of this page the specs of two 0.75 ton units

    The room has a 0F design temperature and 12,000 heat load.
    That simply means the unit must be able to provide 12,000 BTU/h at 0F.
    The capacity of the unit is of no importance as long as it still modulates at that temperature.
    Both units can handle this load.

    At 20F the load is 8572 BTU/h and well in (modulation?) range of both units.
    At 30F the load is 6857 BTU/h and well in (modulation?) range of both units.
    At 40F the load is 5143 BTU/h and well in (modulation?) range of both units.
    At 50F the load is 3429 BTU/h and well in (modulation?) range of both units.

    From that point on things are changing....

    At 52F the load is 3,086 BTU/h
    - The Fujitsu unit starts cycling because the required output dropped below 3,100 BTU/h.
    - Still well within range of the Mitsubishi unit.

    At 61F the load is 1,543 BTU/h
    - Both units are cycling now.

    Fujitsu 9RLS3
    0.75 ton 9000 BTU
    - 3100-12000 BTU cooling - Nominal 9000
    - 3100-22000 BTU heating - Nominal 12000 at +47F
    - 15,400 BTU/hr max at +5F,
    - 14,000 BTU/hr max at -5F,
    - 11,100 BTU/hr max at -15F.
    Fujitsu also has a low temp version but I can't find the specs for that one.

    Mitsubishi MSZ-FH09NA
    0.75 ton 9000 BTU
    2800-9000 BTU - Nominal 9000
    - 1600-1800 BTU max at 47F- Nominal 10900 at +47F
    - 12200 max BTU max at 17F- Nominal 6700
    - 10900 max BTU max at 5F

  8. Stephen Sheehy | | #8

    Another factor to consider when concerned about when a unit starts cycling is whether the house is a high performance, low load building. If it is, you'll never need to turn on the heat at 60 F and rarely at 50 F because the house will retain heat from the sun, bodies and appliances.
    I'm in Maine, 99% design temp is 0 F, 12000 btuh load. Except for a several hour stretch in early May when we hadn't seen the sun for a week, our mini-splits have been off since April 12. During that period, we had nighttime lows in the 30s but the house never cooled below around 70F. The only sources of heat during that period were the sun, two people, a dog and a cat, plus the usual appliance loads.

    As for the difference between units from a given manufacturer, I really don't know if efficiency changes much at very low temps. The units are very similar and the specs are very close, so I am inclined to not worry about any differences.

    As far as choosing between Fujitsu and Mitsubishi, for us It came down to the local installer we were dealing with. I think you can't go wrong with either.

  9. D Dorsett | | #9

    Tony, what you're referring to as the "nominal" output is actually the "rated" output under which it's HSPF efficiency is tested. To use that value as the modulation level for test, the unit must have at least that much capacity at +17F under conditions proscribed by HSPF protocols as well.

    If your heat load at a design temp of 0F is 12,000 BTU/hr the unit has to be able to have a bit more than that much capacity to accommodate defrost cycles. If your climate is super-dry in winter with ultra low dew points at 0F dry bulb temps you can design right to the wire with the max capacity number, but in most places you would need a bit of cushion.

    The max capacity @ +5F on Mitsubishi submittal sheets is under fully saturated rime-icing conditions, where most of the heat is from the phase changes from water vapor to ice, but that maximum capacity is much much higher than the average capacity under those conditions. But with drier air the average capacity will still be a very large fraction of that number.

    If your load is really 12K @ 0F, the FH09NA clearly isn't going to cut it for you, but the FH12NA probably will, as will the 9RLS3H.

    A smaller unit doesn't "work very hard" to deliver the heat. The blowers, coils and compressors are smaller. The larger coils mean the larger unit can deliver the heat at a lower modulation level, but that doesn't necessarily mean it's efficiency is always higher, since it could also be using more power even at that lower percentage of it's much higher max. At ridiculous oversizing the higher draws bigger blowers and compressors eventually trump coil oversizing. Oversizing by more than 50% (the max-capacity at design temp to design-load ratio, not the "rated" capacity to design load) is about the limit of where efficiency from oversizing starts to go down. but up to that point the part load efficiency benefit of the larger coils gives it a little something as long as the minimum modulation isn't too high. There is no simple universal model that works across vendors and model series- it's a bit of a judgment call.

    Mitsubishi submittal sheets used to have a COP @ +17F at some modulation level below the "rated" value, which was sometimes a useful reference point. But current submittal sheets don't seem to have that. On their newer submittal sheets they do offer up the AHRI Efficiency COPs at both +47F and +17F, which would be at the rated output under the test conditions (not saturated rime-icing conditions, but a more reasonable dry-bulb +17F, wet bulb +15F, which is about 70% relative humidity outdoors.)

  10. Tony Tibbar | | #10

    The min-max output is tested at +47F
    The rated output at +17F
    The HSPF rating is tested at +17F

    The 9k Mitsubishi unit clearly is to small. But the larger units have bad minum output.
    12k starts cycling at 49F, 15k starts cycling at 32F.
    When I first posted my findings I doubted between Mitsubishi and Fujitsu because they both have their strong points. But after you pointed out my big error I recalculated and now Mitsubishi seems out of the question. Now I'm doubting between 9RLS3 and 9RLS3H.

    The only reason I'm doubting between Mitsubishi and Fujitsu is that the Mitsubishi starts cycling at higher temps. But that's with the 9k units. Compared to the bigger units Fujitsu also wins.
    The Fujitsu also seems a much better guarantee at low temps. It likely will even handle the1% that's below 0F.

    I understand what design temp is but I find it a bit problematic to find the exact design temp. On I can find weatherstations quite close to my spot on earth.
    But that are the very extremes. Do I have to check all the graphs in the winter months and guess which extremes I have to ignore? Or is there a neat site for that kinda info?

    BTW thanks for the LG suggestion. I'm not ignoring that one. I just focus on one unit right now to avoid this thread getting more complicated as needed.

  11. D Dorsett | | #11

    Not quite it. The HSPF protocol tests performance at both +17F and +47F, and the seasonal average HSPF is estimated, based on the COP at those temperatures using a standardized mathematical model for a heat pump. Efficiency @ +47F is generally higher at lower load as long as it's modulating but they are not allowed to test at a modulation level lower than what it can actually deliver a +17F. The fact that the actual load is lower than the tested modulation level at +47F means that the as-used efficiency of a mini-split at +47F is often higher than the COP at the tested modulation level.

    The units of HSPF are BTUs per watt hour, so with 3.412 BTU/watt-hour. Thus an HSPF of 12 BTU/watt- hour is the same as an average COP of 12BTU/watt-hr /3.412 BTU watt-hr= COP 3.52

    If you take the submittal sheet for the FH12NA, which per the model tests at an HSPF of 12.5:

    The COP @ +17F is 3.34. If that were the seasonal average COP the HSPF would be 3.34 x 3.412=11.4, which is well below it's modeled HSPF.

    The COP @ +47F is 4.2. If that were the seasonal average COP the HSPF would be 4.2 x 3.412=14.3, well above it's modeled HSPF.

    The seasonal average using the model is somewhere in between. The labeled HSPF12.5 is a seasonal average COP of 12.5/ 3.412= 3.6, higher than the tested COP @+17F, but much lower than the tested COP @ +47F.

    The model assumes roughly a US climate zone 4 average outdoor temperature/humidity. Your as-used HSPF depends on your actual climate and how reasonably sized (or how optimally oversized) the mini-split is for the load. When running flat-out most mini-splits, even those with very high modeled HSPF won't do better than a COP of 2.5 @ 47F, which would be an HSPF of 8.5, and at 0F no better than 2.0 (HSPF 6.82). But if you size it where it's modulating most or all of the time it can pretty much hit or sometimes beat it's numbers even in climates colder than zone 4.

    In very low load houses the cycling on/off chews into the as-used HSPF to well below the tested/modeled HSPF, but at very low loads the efficiency matters less.

    Even though they are both "rated" at 9000 BTU/hr cooling, the Fujitsu 9RLS3 is a "bigger" unit than the FH09NA, probably due to larger coils on both the outdoor & indoor units, and aren't directly comparable. Both have higher cooling capacity than 9000 BTU/hr too, but if they rated it higher their SEER efficiency numbers would suffer due to the lower efficiency at the higher rate. Whether sizing for cooling or for heating, it's more important to look at the maximum capacity numbers relative to your loads than the "rated" numbers. Where they decide to rate capacity is a marketing and design choice by the manufacturers. They are allowed test and rated something with a max capacity at +17F of 20,000 BTU/hr at some lower level, and will do that if the goal is to be able to deliver some capacity at -15F, which seems to be what Fujitsu is doing. The -15F max capacity of the RLS3s are approximately the "rated" cooling capacity of those units. This is a marketing consideration more than it is a technical one. The 1.25 ton 15RLS3H is pretty comparable to the 1.5 ton Mitsubishi FH18NA, more so than the 1.25 ton FH15NA. But that's primarily due to where the two manufacturers decided to test & rate the equipment.

    If you don't have 25 years of hourly binned temperature data there's no way to just look at a few years of Wunderground data to divine the 99% bin. In most places it's possible to look at ASHRAE or ACCA estimates for a few nearby listed locations and triangulate a bit to come up with an estimate for your location.

    You'll note that the 99% bins in the ACCA list don't always agree with those of the ASHRAE list, a function of how many years of data and exact climate stations they might be using.

  12. Tony Tibbar | | #12

    Thanks again Dana!
    I'm gonna study your reply tomorrow because I spend way to much hours in front of my PC already.
    But a quick question before I go offline.

    I triangulated the design temperature data. Nothing really nearby.
    I used 3 airports that are ~55 miles from my location.
    My house is in a semi-rural area. A neighbor at 40ft, another at 300+ ft and wide open fields on the other sides.
    The airports are open too, but mostly near quite large cities. Should I apply some correction factor?

Log in or create an account to post an answer.


Recent Questions and Replies

  • |
  • |
  • |
  • |