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

Community and Q&A

How does efficiency in a mini-split vary over various operating levels?

greenenvy | Posted in Mechanicals on

It is my understanding that a SEER rating is determined by simulating the operation of a particular make/model air conditioner over a hypothetical season, presumably assuming various temperature and humidity levels throughout the year.

I’m interested in learning about the efficiency of variable refrigerant flow (VRF) mini splits at various operating levels. This question has not been as relevant in the traditional fixed speed split systems that most of us have always used. Those binary systems were either on or off. With those systems, you were challenged with finding a system that met the load of the building with the least number of cycles during the hotter times of year.

Does anyone have any data on the BTUs per watt hour of these VRF units at the various operating levels? I am suspicious that they operate more efficiently on the lower end of the curve due to less friction at lower speeds, more coil area per unit of refrigerant etc. If that is the case, could there be an argument, at least for units with a very high turn-down ratio, to upsize the units over what would otherwise be prudent, to both more quickly meet the demand of the hotter days and have the unit normally operating at a lower end of its capabilities, but still without cycling?

It would be very interesting to see a graph of the BTU output on one axis and current draw on the other of some of the better mini splits out there to visually see this.

Since I live in Central Florida, I am far more interested in cooling performance of these units than in heating.

Thanks in advance for your thoughts.

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 resources on the web and on GBA that discuss this issue. You may want to start with this research paper: Long-Term Monitoring of Mini-Split Ductless Heat Pumps in the Northeast.

    In that paper, researcher Kohta Ueno notes, "Oversizing of MSHPs [minisplit heat pumps] can actually be beneficial: they modulate their capacity, and their highest efficiency is obtained when the unit is running at the lower end of its capacity range."

    That said, there is a caveat: oversizing can be beneficial, up to a point. Any minute now, Dana Dorsett will probably post a helpful and detailed comment providing guidance on how much oversizing makes sense, and how much is too much. The answer depends on the equipment under discussion: units with a good turn-down ratio behave better than units with a narrow capacity range.

  2. Expert Member
    Dana Dorsett | | #2

    OK, guess I've been summoned...

    Good insights into the part load efficiency of mini-split can be found in this 3rd party set of bench tests on a couple of venerable 1-ton mini-splits:

    New equipment has similar characteristics, but higher test numbers.

    Oversizing is beneficial only up to the point where the mini-split is cycling on/off a great deal of the time rather than modulating. When sizing a mini-split to a zone, you have to know four things:

    1: The 1% design condition cooling load, per a Manual-J calculation using agressive, not conservative assumptions. In other words, err to the low side rather than estimating high "just in case..."

    2: The maximum capcity as tested in the SEER test.

    3: The "rated" capacity at which the SEER efficiency numbers are more relevant. Mini-split efficiency is typically performed at a modulation level well below it's maximum output, but that difference varies.

    4: The minimum modulated output under SEER test conditions.

    Ideally you would have something like 25% or more margin to spare at the maximum output, which would mean it modulates below max even at the 1% design condition, but a minimum modulation well below half the design load, under 1/4 would be even better, which means it runs almost continuously, with little on/off cycling.

    As an example, say your aggressive load calculation came in at 10,215 BTU/hr @ 93F and you were trying to figure out which of the Mitsubishi FH series was the best choice.

    The 3/4 ton FH09NA tests at an impressive SEER of 30.5 assuming a "rated" peak load of 9000 BTU/hr @ 95F outdoors, 80F indoors. At maximum output it can deliver 12,000 BTU/hr of cooling, but it can throttle back all the way to 1700 BTU/hr under those conditions, per the submittal test sheet:

    That's about 17% of margin on your design load, but your load is above the efficiency test rated level, so it won't actually hit it's SEER number, since at least part of the time it will be running substantially over 9000BTU/hr.

    The next step up in the series is the 1-ton FE12NA, which has a maximum output of 13,600 BTU/hr which is about 33% of margin at design condition, and it can throttle back to 2500 BTU/hr which is still under 25% of the design load:

    It's SEER is "only" 26.1, but since it will be running below it's rated 12,000 BTU/hr it's as-used performance will likely be higher than tested, but whether it's efficiency would rise to that or exceed that of the FH09NA running a bit above it's rated output is an open question. Either one is going to be pretty reasonable choice.

    The next step up in the series is the 1.25 ton FH15NA, which has a rated output of 15,000 BTU, which is nearly 50% above the design load, and a maximum output of 19,000 BTU/hr, which is 86% above the design load. But it's minimum modulated output is a whopping 6450 BTU/hr, well over half the design load. The 1% design condition or higher only occurs 88 hours in a typical year, and the average load is going to be less than half that, so the FH15 would be modulating less than half the time, and wouldn't hit hit it's still respectable 22 SEER.

    Worse yet, comfort will suffer, since there would be temperature under/overshoots when cycling, and lousier latent cooling (humidity) performance.

    Faced with this hypothetical my personal inclination would be to go with the FH09 despite the undersizing at the "rated" output level, since even aggressive Manual-Js sometimes overestimate the load. But the VERY low minimum modulation level will maximize the run times, maximizing overall comfort. And it still has substantial capacity margin for when it's hotter & more humid the than the design condition. But opting for the FH12 would not be a mistake, due to the still reasonable minimum-modulation, and with 33% of margin on the design load it'll beat it's test numbers.

    Other manufacturers have different min, max and rated numbers, and you need to know them all, along with your actual load to be able to estimate which will operate most efficiently and comfortably in your application.

  3. GBA Editor
    Martin Holladay | | #3

    I wasn't summoning you -- just honoring your dependability and helpfulness. Thanks.

    By the way, this would be a great topic for a future guest blog. Just saying.

  4. Expert Member
    Dana Dorsett | | #4

    I know... (great minds think alike?)

  5. Andrew_C | | #5

    Good grief. I don't know what Dana's paying job is (EE + slush skier, as I remember), and he doesn't advertise here, but if anyone is looking to hire someone to as an energy consultant, here's your start point.

    Dana: don't know if that's in your interest, just saying I appreciate your extensive contributions.

  6. greenenvy | | #6

    Thank you guys for your quick and thorough responses, with special thanks to Dana. I'm glad to see that my questions had been addressed through actual testing in the past. So, knowing what we know here, it is clear that high turndown ratios are highly desirous with mini-splits, especially in locals where there is a lot of seasonal temperature variation. Has anybody listed the turndown ratio (or min and max BTU output) of many of the more common makes and models in a web table or on a spreadsheet so that these values can be easily compared? It would be interesting to see how the turndown ratio correlates with the SEER ratings. I suspect that a high turndown ratio is at least contributory to higher SEER ratings.

  7. GBA Editor
    Martin Holladay | | #7

    There are two issues here:

    1. Does a ductless minisplit that happens to be oversized use less energy if it has a high turn-down ratio than if its performance range is narrow? (The answer is yes.)

    2. Do the SEER and HSPF standards accurately report the efficiency of ductless minisplits? (The answer is no.)

    We need new test methods that accurately report the energy use of ductless minisplits -- both for heating energy use and cooling energy use. These test methods have not yet been agreed on.

  8. Reid Baldwin | | #8
    has data on a variety of mini-splits. It focuses on heating performance rather than cooling, but it does show minimum capacity and maximum capacity.

  9. Dana1 | | #9

    All mini-splits sold in the US undergo HSPF and SEER testing for labeling purposes, and are required to maintain and publish the abbreviated results in a "submittal" sheet. Often the submittals are updated with firmware or hardware changes even when the model name doesn't change. This is a constantly moving target, and I don't know of any organization that maintains and constantly updates lists.

    Fujitsu has traditionally kept the minimum modulated performance if single zone mini-splits the same 3100 BTU/hr in both cooling and heating mode for most or all of their models, but the difference between the nominal-rated output and maximum output varies a lot. The minimum output of their 1.5 tonners is the same as thier 3/4 tonners.

    Most other vendors will have lower minimum output for the smaller units than the larger units, but it varies from model to model, series to series.

    Multi-split minimum compressor outputs are typically over 5000 BTU/hr from any vendor, which may be part of why their HSPF and SEER numbers are typically lower than the single-zone units.

    As far as I know, the Mitsubishi -FH06 & -FH09 have the lowest minimum-modulated outputs in the industry, both at 1700BTU/hr cooling, 1600 BTU/hr heating. The very high turn-down ratio and low minimum outdoor operating temp of the -FH09 makes it a pretty flexible option It can be 3x oversized for a 4000 BTU/hr cooling load from it's max 12,000 BTU/hr point of view, yet still has a very reasonable modulating range that 4000 BTU/hr zone. This unit is the exceptional- most don't have turn down ratios nearly that wide.

  10. charlie_sullivan | | #10

    Reid, that's a great database. It shows that some LG models have an even bigger turndown ratio than the Mitsubishis--those with outdoor units LxUxxxHxVx, e.g LAU090HYV1 and LSU120HSV4.

  11. greenenvy | | #11

    The minimal operating levels are clearly very important in model selection.

    In the US Department of Energy study provided early in this thread by Dana and found at, it says:

    "The manufacturer claims the heating capacity can vary from 3,100–24,000 Btu/h. However, despite loading the compressor to achieve minimum operation based on the manufacturer recommendations, this minimum capacity was not observed in the laboratory. Thus based on the behavior observed in the laboratory, the unit would need to cycle on and off to meet a load lower than approximately 5,000 Btu/h."

    It makes you wonder how much these numbers can be relied on. Does anyone have an educated opinion on the accuracy of the figures provided by the manufacturers beyond the testing done here?

  12. Expert Member
    Dana Dorsett | | #12

    The difficulty of measuring the BTU output of a variable speed unit with any accuracy is pretty tough, and more difficult still at very low speeds. It's hard to say just how accurate the manufacturers' numbers are on the mininum output end, and it's not clear where error bars are on the Ecotope engineers running that bench test.

    They were unable to reproduce Mitsubishi's HSPF numbers with their test setup, which could indicate anything from a defective unit to manufacturer BS to measurement error on their part. Without more time & budget for figuring it all out, they simply commented when their numbers divereged from the published numbers, which was the right thing to do.

    This wasn't a directly a DOE sponsored project. The NEEA (a consortium of electric utilities in the Pacific Northwest) hired them to run this bench test along with in-situ monitoring of hundreds (thousands?) of mini-splits as part of a larger program. The budget for bench testing has to be orders of magnitude smaller than the setups used by the manufacturers, and I'd be inclined to believe the manufacturers' numbers on the min-modulated output are more accurate than Ecotope/Herrick Laboratory numbers. It's not that the manufacturers don't have incentive to fudge the numbers here & there, but the minimum modulated output isn't one of those places I would expect them to intentionally skew things by very much.

    Some detail on the test setup used can be found here:

    Starting on p29 they discuss a variety of factors that may account for discrepancies between their tests and the manufacturers' published performance data.

  13. greenenvy | | #13

    From reviewing the spec sheet on the Gree Crown 9000 BTU unit, it looks like it beats the Mitsubishi on the cooling side for its turndown ratio and matches it on SEER, making it a [potentially] excellent choice for those of us in the south and in areas where they spend more money on keeping cool than keeping warm.

    For cooling, the Mitsubishi min/max BTU output is 1,700/12,000 (7.05 ratio) and the Gree min/max is 1,535/12,966 (8.44 ratio) for cooling. In heating mode, the Mitsubishi min/max is 1,600/18,000 (11.25 ratio) while the Gree min/max is 2,388/13,648 (5.71 ratio).

    You can buy the Gree for about $1,200 online vs $1,600 for the Mitsubishi. Of course, $400 is nothing if the cheaper unit is going to have less reliability. I doubt that we are going to get many or any Gree testimonials here but there are probably many readers with an opinion about their Mitsubishi units. Have you found them to be highly reliable if installed properly?

    I believe that some of the A/C supply houses down here in Florida carry Gree so I'm sure that some contractors will quote them. It is my understanding that Gree is the actual manufacturer for some or many of the other brands. They claim that that are "the largest specialized air conditioner enterprise in the world", what ever that means.

    Finally, I took note of the Gree because of its very high SEER and high turndown ratio on cooling. It is possible that other models have very good turndown ratios on the cooling side like the Gree but don't do so well on the heating side. Not too many units out there can claim SEER ratings north of 30, however, like both of these units but it sounds like the SEER rating is only part of the story.

    Gree specs:

    Mitsubishi specs:

  14. Dana1 | | #14

    Daikin, Mitsubishi and Fujitsu are all first-tier Japanese manufacturers with first-world quality control & manufacturing. Distributor support in N. America varies, and can be important. If you have to wait 30 days for a back-ordered repair part to get shipped the factor it's a problem.

    While Gree is the largest manufacturer of air conditioning equipment in China, sub components are often manufactured in lower cost SE Asian countries with less stringent tolerances. This can be an issue with some of the second tier Japanese & some Korean vendors as well.

Log in or create an account to post an answer.


Recent Questions and Replies

  • |
  • |
  • |
  • |