I just completed a gut renovation of an 1880's era carriage house with 13" thick masonry walls. Roof has 5" of poly iso foam and existing skylights. Walls were sprayed with 2-1/2" closed cell foam plus R-13 fiberglass. Floor has 2" to 3" of XPS foam (the original slab was not level so we built it up with foam). Then on top of xps we put in a radiant system embedded into a new gypcrete pour.

As it just got cold the customer is noticing drafts. The thing is I know that house is tight. I installed all the windows. The foam was sprayed up on to the roof sheathing behind the studs and down to the XPS foam. Pretty sure there is a 2" foam air barrier all around. No drafts felt before it got cold.

The house has cathedral ceilings 16' high at the peak. My theory is that it's convection currents. Am I fooling myself?

I have the same thing in my turret with high pitched ceiling but mined is on the second floor with an open stairwell.

You might take a infrared thermometer or camera and find exactly where the temperature differences (required for convection) are. Or a blower test to be sure it is airtight.

Cheaper yet would be a smoke stick to trace the "currunts" he is experiencing.

Of course, one could always calculate the assembly r-value of the roof structure with 5" of poly iso foam and see how much cooling was going on up there at the peak. Does that meet code for cathedral ceilings?
Is that sheets on TOP of the roof sheathing?
Is there something in the rafter cavity?

What is the r-value of "existing" skylights these days? More than one at what size? 2' X 4'? That's 16 sf or 24 sf or ...more at what r-value?

Typically the windows will cool the air mass next to them. It then drops, forcing the warmer air up the middle, which is replace with even warmer air, forcing this mass back to the windows.

Try the smoke pencil with and without the widow coverings closed.

Say you have a high wall with lots of windows. What's the solution to a sheet of cold air circulating downward? Perhaps a continuous register along the floor (creating a counteracting upward flow).

Jon - that is the solution for forced air.
The normal solution for radiant is a large blade ceiling van turning slowly to stop the rise in the middle and gently send it back to the windows.

Some designers will really stack the tubes under the windows to try to negate the drop.

Air will always move when heated. A cathedral (sloped) ceiling exaggerates this because it funnels air, the higher the point the more aggressive the air flow.

Thanks for your replies. The were some real budgetary concerns with this project so we could not do a lot with roof. I made about $20/hr. GC-ing and providing construction financing, thereby going into debt. The things you do for friends....

Please.

It is infinitely more likely that the radiant floor isn't running much allowing cold air to move en-mass from the windows in particular. This will be more noticeable in the shoulder months when loads are low and the heating system is not running much.
This is a common problem with system that do not employ weather sensitive control (outdoor reset). ODR will keep a radiant panel warmer longer and lower the heat load while lowering the operating cost. If you don't have it, get it.

No substitute for proper design.

Moving air in winter is rarely the answer. The lack of moving air is one of the benefits of radiant heating. If you are hot, moving air is a benefit. You can't have it both ways.

Warm "air" rises, "heat" travels at the speed of light in a straight line.

"Skylights", in the plural, as in more than one? A lot of skylight concentrated on one side or the room is enough to create convection currents strong enough to be felt, since it creates a sinking column of air on one side of the room, and a rising column on the other. Skylights tend to be lossier than windows too, but as the installer you should be able to dig up the U-factors of the actual units.

With a lot of window you on one side of the room you can have similar issues, especially if they are lower-performance windows. This phenomenon happens starting at about 30F (and lower) in my radiant-heated family room with the 15' cathedralized ceiling, where one side of the room has about 100 square feet of U0.5 window. (Im strongly considering low-E storms for that side, for both energy use and comfort reasons.) In my case it's easy to trace the convective flow with a smoke-pencil. If you sit near the windows you can feel the draft, but it's a cool cascade of air falling toward the floor, not leakage at the windows. (The solution is to stay at least a foot away from the window. :-) )

The skylights were there from a previous renovation. Anyway, it seems like it's simple physics; warm air rises, cold air falls. He has 6 skylights about 20 sf. so it's a fairly large cold surface way up high. The greater the temp differential and the greater the distance between differentials the faster the air movement.

It looks like radiant panels under windows reduce drafts and panels anywhere on the same wall improve the distribution of MRT. I didn't find anything about how effective extra tubes in the floor are in stopping the convection draft (but it seems like it would be at least partially effective).

cool cascade of air falling toward the floor

This will happen regardless of window quality. As you suggest it is a matter of degree. Your solutions is brilliant as usual.

The definition of comfort is the lack of discomfort.

Storms perhaps, but matching output as closely to the load as possible produces the best results for us.