How conservative is Manual J?
![](https://secure.gravatar.com/avatar/5e1179c5ab1c60e9709077fd1045e284?s=52&d=https%3A%2F%2Fimages.greenbuildingadvisor.com%2Fapp%2Favatars%2F200x200%2Fk.png&r=g)
HVAC installers oversize equipment. For example, the 550 square foot in-law apartment that I rent in western New York has a 70 kBTU/h furnace. The place has bad windows, not much insulation, and sits on an unconditioned garage, but still. Way oversized.
Best practice to avoid oversizing is to run an aggressive Manual J. I did my level best at this using the freeware at CoolCalc.com. With aggressive input data, it gave me a design heating load of 17.6 kBTU/h at 0 F.
My question: How conservative is Manual J itself?
I did an experiment to try to answer that. On a cold night (not design-conditions-cold, but 23 F), I recorded my furnace’s ON and OFF times for two hours. I calculated the duty cycle,
duty cycle = (sum of ON time) / (total experiment time),
then multiplied it by the furnace capacity. That gave me an estimate of the heating load at 23 F outdoor: 6.7 kBTU/h.
When I change the outdoor design temperature in CoolCalc to 23 F, it gives me a load of 11.2 kBTU/h. That’s 67% higher than what I measured.
Is Manual J usually that conservative?
GBA Detail Library
A collection of one thousand construction details organized by climate and house part
![](https://www.greenbuildingadvisor.com/app/themes/greenbuildingadvisor/dist/img/modules/detail-library/thumb1.png)
![](https://www.greenbuildingadvisor.com/app/themes/greenbuildingadvisor/dist/img/modules/detail-library/thumb2.png)
Replies
I don't know about Manual J but timing furnace run times is a good way to get a ballpark heat loss per degree F. I did it in 2010 on our house for 8 hours on a still and foggy steady state temperature day. The furnace efficiency was known at 95%. I came up with a heat loss of 360 Btu's per degree F. This matched very closely my manual calculation and does jive with the yearly metered gas readings. The day was very foggy so solar gain was not a factor so it should be quite accurate as to straight heat loss.
If you do a nightly setback you want to make sure the house has had adequate time (6 hours) to get back up to temperature before timing furnace run times.
Thanks, Doug. I kept the indoor setpoint constant during my two-hour furnace timing experiment (no night setback). There were no solar gains (nighttime) and minimal internal gains (mainly from the fridge; lights and computers were off; no showers or cooking). I may repeat the experiment on a colder night to get a data point closer to design conditions.
From your experience of furnace timing agreeing closely with your calculations, it sounds like maybe I'm getting a "garbage in, garbage out" error with the Manual J software. I had to guess at some of the construction details, so I might have entered some bad input data.
I might also try another Manual J app if there are other free options.
Kevin; When you say apartment are there some common walls and not exterior walls to be considered?
It's a stand-alone unit above an unconditioned, detached garage/storage space. No shared walls.
I built an art studio for a very good friend in 2013 in Minneapolis. The new studio is above a 22' x 22' unconditioned garage. The only connection to the main house is a stairwell from the upper level so the thermal envelope is on 6 sides. Quite a few windows for natural light, R-30 walls, R-65 ceiling, R-50 floor and very airtight. I ran an 8'"x 10" heat supply between the stair stringers from the furnace that split into (4) 6" supplies fed through the floor web trusses. Far and away the most comfortable room in the 1978 built house. I laid a digital thermometer on the floor while visiting (12-2018), 70F, same as the thermostat setting.
With regards to your inputs did you account for the fact that the floor was not insulated?
@Doug, sounds like a great setup for a studio! Cozy, lots of natural light.
@John, I did account for the uninsulated (or anyway, poorly insulated) floor.
I reran Manual J on another free calculator, LoadCalc.net. That one gives more control over input data than CoolCalc.com. By making increasingly aggressive assumptions about the wall, ceiling and duct R-values, I managed to drive the 23 F outdoor / 66 F indoor load estimate down to 9 kBTU/h. That's in the right ballpark, but still 34% higher than the 6.7 kBTU/h that I measured in the same conditions.
Do these Manual J calculators include the 40% oversize factor that ASHRAE recommends? (Dana Dorsett wrote "For comfort and efficiency, ASHRAE recommends that heating equipment be sized at 1.4 times the design heat load" in this article: https://www.greenbuildingadvisor.com/article/out-with-the-old-in-with-the-new.)
>"Do these Manual J calculators include the 40% oversize factor that ASHRAE recommends?"
Manual-J does not insert an oversize factor.
The purpose of a Manual-J is to estimate the peak loads for a variety of reasons that go beyond merely picking the heat source, and it doesn't try to insert an oversize factor. ACCA (the source & publishers of Manual-J) is a completely different organization than ASHRAE, the organization that recommends the 1.4x oversize factor. ASHRAE has it's own methods for calculating loads ( https://www.ashrae.org/File%20Library/Technical%20Resources/Bookstore/preview-load-calculations.pdf ), but they don't auto-insert an oversize factor either- that part is left up to the discretion of the owner/user/engineer.
I've seen some crazy load numbers for less conventional homes from coolcalc, enough times for me to inherently distrust it. In one case more than 300% of a fuel-use calculated load, but haven't been privy to the inputs. I've never used it myself. Like Kevin, even with fairly aggressive input numbers loadcalc load numbers I've seen usually overshoot measured reality by more than 25%.
I've seen a handful of Wrightsoft Manual-Js done correctly by professionals that came in pretty much on the money, not more than 10-15% over what was observed, which is within measurement error when dealing with occupied homes. I've also seen Wrightsoft Manual-Js that were clear cases of "garbage in = garbage out", with the contractor's thumbs all over the scale.
Great info! Thanks, Dana.
Remember than Manual J is about design day load - which might have much higher wind/infiltration than when you did your test.
Temperature adjusted measured data and Manual J results aren't supposed to match - unless you measure with design day winds.
The wind question is interesting. I haven't see an input for peak wind speed.
Certainly peak wind speeds vary geographically? Even vary significantly in a microclimate. I am thinking of cities with tall buildings.
Presumably there are areas where peak wind speeds don't happen with temp extremes. Hurricanes (and tropical storms) don't occur when peak temps occur. Certainly solar gain tends to be down during storms.
Here in NC, we rarely have high wind speeds in the winter. Chicago - totally different story.
>"I haven't see an input for peak wind speed."
Just like 99% & 1% outside design temperatures, most Manual-J tools infer the peak wind from the location from a built-in database.
Good point about wind and infiltration, Jon. My furnace timing experiment was on a still night. When I axe the infiltration rate in LoadCalc, its prediction is within 10% of my measurement. I'd say that's pretty much spot on, given the measurement noise. (The measured duty cycle is very accurate, but the manufacturer's reported furnace capacity has some uncertainty, which propagates through to the load estimate.)
If the Manual J design conditions include both extreme temperature and extreme wind speed, I guess that could lead to pretty conservative load estimates. A 1% temperature event coinciding with a 1% wind event seems pretty unlikely.
I agree. Using a peak (or 1%) wind speed wouldn't be appropriate. But neither would calm.
IMO, one should collect 10 or more measurements and then plot them. Fit a line through the upper points (supposedly the ones with higher wind).