House energy retrofit project 20
Roofing, day 4
[Click any image for a larger version.]
The next day the roofer arrived with parts to put together a specialized
"chicken ladder", which he hauled onto the shed-dormer and assembled up
top. Unclear why they're so named, as they don't really have
anything to do with chickens;
perhaps they're more footing assurance for someone who doesn't want to
just dangle from a fall-arrest rope fastened up top. They're available
and similar to the hook-ladder rig the roofer already had, but a little
more optimized for steep-pitch work. He opted to construct
his own out of wood, in part to make sure its rungs would clear the
standing seams, and wound up with something a little similar to
with the exception that it wouldn't simply hook over the top but
rather get temporarily screwed in.
They spread out some rubber roofing membrane underneath to protect the
panels before lowering this rig into place, fastened it in at the
top, and the roofer gingerly got onto it to test solidity. Once he was
sure about it, he issued the go-ahead to start siding up the cheekwall.
They didn't have a fancy cutting station for siding, but it turns out that
carbide cut-off wheels make perfectly acceptable slices through vinyl and
yield beautifully clean edges. The guys on the ground laid out all their
measurements and cut everything right there on that same side stoop that
so many other people [incl. yours truly] have used as a worktable. And
here they had to make some pretty funky angles to match the converging
The cheekwall siding didn't take the roofer very long at all; most
of his delays were in explaining to the other guys how to measure the
angles and transfer cut-lines between pieces and generate the sequence
of pieces he needed.
The last few pieces at the top were the most odd in terms of cutting;
no right angles anywhere.
This was one of the funnier shots from the job...
The workspace was getting tighter and tighter as well.
The last couple of bits went in, with one tiny triangle at the top capping
it off. I shot these from above, lying on the shed dormer and leaning
way out with the camera aimed back at the work area.
Here we also see a good cross-section of how the soffit hooks together,
very similar in model to siding. Right after this, the final piece of
rake trim metal had to go on over that and the wood before finishing up
The last bit of siding and the Z-bars to block the soffit void
basically got caulked in a fairly sloppy fashion, and I wasn't at
all convinced those spaces were truly sealed off. Both of these would
probably need to be screened over in some way before the next summer's
influx of insects.
With one cheekwall and its rooflet done, the chicken-ladder got pulled
and moved to the other side -- leaving several screw holes behind including
right through the panels. After all that talk about avoiding top-side
penetrations, this is how the roofer had chosen to attach his ladder.
Some of these were far enough down that they
would *not* necessarily get covered by the ridge cap. He swore
that he'd patch those -- for the holes through the Grace he simply
ran screws back into them all the way but that would have been wrong to
do on the aluminum. I really questioned this ladder attachment methodology
instead of using
a hook over the ridge -- this area of the ridge was a nice 90-degree angle
which would have given a pretty secure attachment, maybe one or two screws
to just tag it down. Instead he had put about ten of them into areas where
they maybe shouldn't have gone.
The west-end cheekwall got set up and sided in a similar fashion, so
I didn't bother with any shots of that as it was pretty much the same.
There was still a little bit of the day left afterward so the roofer
started some figuring for the shed-dormer, particularly how to locate
panels to land the stink-pipe in the middle of one. Flashing around
a seam is far less reliable, so any roof penetration really wants to
have the seams running on either side of it.
We agreed to just give up on the ATAS end-details entirely for the shed
dormer, and he'd just fab up a starter panel with an integrated drip-edge
to go over the side and get trim-nailed into the fascia. This seemed
so much simpler. A suitable piece was cut, and this time not just
buzzed apart with the angle-grinder. This needed a clean edge.
That was accomplished by marking up the piece and clamping it into the
dual-brake setup at the cut line, and then scoring right along the brake
nose a few times with a utility knife. A gentle bend was then applied and
un-done and re-done a few times until the piece broke precisely along
that score line. I had seen this technique used a couple of times on
things like door trim, which is thinner metal, and was intrigued to
find that it could also be used reliably on the stiffer .032 roof metal.
The additional bends were then the right angle to go down over the fascia,
and a little half-inch kicker flaring outward to form the drip-edge.
The panel needed a little trimming before being applied, and he wanted
to double-check that he'd calculated its width right to to accomodate
the stink-pipe placement. The upper end of this had to be tapered off
to match where the lower roof came up.
He wanted to get this first panel trimmed up and attached before calling
it quits for the day. This is how the split got put together.
This is one reason I cannot understand today's evident desire for complex
residential rooflines -- it just makes it that much harder to match all
these pitch-split and transition details. Some of this might have been
simpler if I'd decided to bulk out the shed-dormer to the ends after all.
Too late now.
I got curious about the exact inter-panel distance, understanding that it
could vary a little from seam slop. Turned out to not land on nice
measurement boundaries -- it turned out to be just about 11 and one-sixth
inches. If it was exactly 11 per panel cycle I'd be at 132 inches over 12
panels here, not at 134. I eyeballed the roofer's first shed-dormer
panel and the distance to the stink-pipe and figured he had it pretty close.
We'd find out the next day.
Roofing, day 5
While up noodling panel measurements I also spotted this. What's wrong
with this picture? Where would water running under a panel go?
With one panel already down on top of this, it got fixed by lapping another
short piece of Triflex under the end of the first piece and above the
We had already gone through a conversation about underlayments and
drip-edges, where the sensible thing might seem to have the main Grace
layer lapping over the drip-edge piece, but almost everyone attaches
drip-edge *over* any underlayment present. The roofer defended the latter,
saying that it must be that way to prevent the ice dams that the ice &
water products are supposed to handle in the first place -- i.e. water
that backs up over the drip-edge piece won't be able to sneak underneath
to decking. In a superinsulated scenario the ice-dam problem should be
far less of a factor and I was thinking more of where inevitable under-panel
condensation was going to go, and realizing that proper installation
of the tri-flex was my last chance to have *that* handled appropriately.
floated this quandary
over at Green Building Advisor and got some helpful input, including a
link to a
detail from Grace
showing two layers of I & W lapped on either side of the drip-edge. They
don't specify the width of the lower layer but one assumes it would be one
3-foot-wide course rolled out -- enough to protect against ice dams on
most roofs, one would hope. So my substitute for that upper layer was now
the tri-flex, which isn't even specified as a water-resistive barrier but
at this point late in the game it's all I had. Therefore, its edge detail
needed to be perfect or I'd be getting more water into the plywood.
Especially since my lower Grace layer ended shy of the deck edge and
*didn't* drop over the fascia as shown in the diagram.
That same Grace detail document also contains another vindicating note:
Repair all holes left from removal of toe boards, roof jacks, etc.
Well, there you have it in black and white.
Much of the morning was spent prepping panels: cutting to length and doing
the lower-end bend, and lining up stacks of them ready to go up. Today
he only had two guys on site so once they got ready to install these
I handed most of them up from the ground.
Work proceeded fairly fast along the shed, seating and clipping panels
one after the other. The move to slide the panels up and seat them was
a bit risky -- crouching near the edge of the roof and *pulling* against
the end of the panel. Mostly he tapped them up into place with a
rubber mallet, extending the reach with that and keeping him a little
farther from tipping over the edge. The guy apparently kind of grooves
on the sketchy positions, though. It hasn't killed him yet.
They reached the stink pipe, which got a special panel made a little longer
and cut in two, with its ends carved out to lap around the pipe base.
The flashing fitting only needed about a half-inch trimmed off one side
to fit within the panel ribs.
Stink-pipe done, or as done as he was going to get it.
He had left a few gaps in the caulking around this, which I fixed later.
Here I was also checking that the rubber ring really had a solid seal to
the flashing piece, as water coming down here fast could conceivably
coast up the slope of the galvanized and try to dive under the lip.
At this point I had started walking around on the panels. This was one
of my concerns about the standing-seam -- would I be able to get onto my
own roof and move around without sliding off? I found that I was able
to just barely stick to them, at least well enough to walk around and
my extra margin of safety could be had by simply reaching down and
grabbing a seam. In this weather my feet were a little moister than
they'd be in cold, which I discovered much later made things a bit
For the record, shoes are *more* slippery -- at least the composition of
the rubber on the sneakers I had on hand. Shoe soles to stick to metal
roofing have to be just the right type of rubber formula, and that can
probably only be determined by experimentation. For example, the roofer's
shoes happened to be sticking pretty well, and his helper was sliding all
over the place. As a data point, Vibram is not bad for this purpose.
The last thing one wants to do, however, is sit down -- any kind of cloth
doesn't stick well at all, and immediately starts sliding down. Soft rubber
and/or flesh contact is about the only successful friction mechanism here.
Somewhere around the stink pipe they had run out of tri-flex, and instead
of waiting until they could get more of it the roofer just switched to
plain old roofing felt. Because it's what his daddy used for 40 years,
or something. I didn't bitch about it although maybe I should have; this
seemed like a casual compromise in materials that maybe wasn't right for
this job in the first place. At least *something* was still going over
the ice & water to keep the panels off it.
The other downside of this
was that where he had been trying to minimize stapling into the tri-flex,
with the paper he went absolutely batshit crazy with the tacker. He said
he wanted to be sure that the stuff wouldn't tear off under them walking
In retrospect, this was just fucking lame. More holes in the Grace
underneath, and more *steel* to scrape against the underside of the panels.
He should have brought more Tri-flex that morning or run out for additional
supply that afternoon. Well, you read it here first. We'll see how this
all works out twenty years down the road.
The last of the panels were handed up, by the helper kid instead of me this
time. The length was really convenient -- try doing this with shingles!
By day's end, the guys had nicely roofed themselves into a corner.
And at that point, they *ran out of panels*. With three or four left to
go. Apparently the roofer hadn't quite planned right for accomodating the
wing-walls and using partial panels in various places, and hadn't ordered
enough from ATAS. The remaining ones would have to be backordered, meaning
a substantial delay until the roof could actually be *finished*.
However, they still had some other parts to work on so they'd be back
the next day to finish up what they could.
Meanwhile, our discussion about the shed-dormer end pieces and the fact
that the cut-offs from its panels were substantially long inspired me to
get going on the air-conditioner hutch project. Here was my chance to
make its little roof match the big roof, which would just be hilarious.
I made sure he left me a few of the cutoffs, where he would have otherwise
added them to the aluminum he recycles. The shed-dormer overhang didn't
quite cover the condenser and plenty of drip had been landing on the outer
edge of the box every time it rained. The enclosure is designed to be
fairly weatherproof but having that much water hitting it that hard
couldn't be good for it long-term, so I wanted to work toward getting
it under better cover soon.
I already knew the dimensions I wanted for the cover -- 2 x 4 feet, and just
a simple single surface would do. While the guys were paneling the
shed-dormer, I brought various materials out into the backyard
and started noodling how to put this thing together. It would be built
almost entirely from the project's scrap wood I saved.
As the top would have a tilt, I wanted to pre-plan for that in the
supporting framework. One of the 2x4s was a little warped and I figured
that one end would already match the desired angle and the other would get
forcibly untwisted a bit with a framing clamp as the top got screwed together.
It worked out okay, and with that pre-stress on it the wood would likely
straighten right back out over time.
This was the vague concept, although supported on posts a little
higher over the condenser box. Enough to keep rain off, but not mess
with airflow. And obviously it would need a waterproof upper skin.
That evening I started putting the rest of the hutch top together,
first making the roof panels for it. I used the score-and-break cutting
method, bending the panel over a sharp edge after slitting the rib
parts, and it worked quite well. The only other tool needed was a
short steel straightedge to sit down inside the panel to run the
Development of the rooflet went pretty quickly. I bent the drip-edges
at either end on the brake in the backyard, and before it all went
down onto the hutch for real, I ran a layer of his vaunted tar-paper
underneath it as underlayment.
The panels really do click together in a really satisfying way.
This short video demonstrates that.
In running out of panels, the roofer had also just about exhausted the
supply of hold-down clips -- there were about eight left from that big
box, and I didn't want to swipe them for this in case he needed them
up top. But it was easy enough to bend up a few functional clips
out of more of the roofing metal and whitney-punch some holes for screws.
Here's an end section of how it would go together, using my own cutoffs from
the drip-edge pieces as a demo of the seam interlock. Again, constructed
entirely from scrap, even down to the screws themselves which were probably
some dropped ones that I had picked up off the ground with the magnet.
Roofing, day 6
I showed the roofer my little creation the next morning, and he definitely
approved. I felt sort of like the small child with the plastic toy workbench
and the tinkertoy wrench and screwdriver, going "look, I can do like daddy
does!" But just watching the real roof go together had already taught
me a lot about working with these materials, and I think I was doing
a pretty precise and well-built job on it even if the methodical attention
to detail would take me far longer.
The roofer had bigger things to attend to, one of which was fabbing up
our custom gable-end caps for the lower rakes. Here's the promised
section picture of one -- pretty simple, really. The stock end-cap piece
was ripped a little narrower and the upper part given a 90 downward to
come inboard a measured amount and clamp down over the endmost roof rib.
They ran these in where needed, and the fiddly bit was shaping a closure
at the lower end. Here it's in process, with a bit of excess hanging off
in various directions that got trimmed off and bent more neatly.
Despite the roof not being complete, they wanted to get as much of the
ridge cap closed off as they could. To prepare, the roofer had to cut
about seventy small sections of z-bar to sit down into each pan on
both sides along the ridge.
These were further prepped with pieces of the double-sticky butyl tape,
to help them seal down against the panels.
Another look at the felt paper here with lots of staples in it, not all of
which were even fully seated.
The two of them started working along the ridge installing the z-bars.
We were all wondering why ATAS forces their materials customers to cut
all the z-bars for this application on-site, rather than supplying pre-cut
ones already accurately matched to the panel widths. I could see site
cutting introducing the possibility for a lot more error if not done
From the street, something about the scene made me immediately focus on
the fact that there were guys working on top of the roof. Maybe it's all
the parallel lines converging in perspective, that immediately draw the
eye upward. In fact I noticed this effect several other times where a
glance at the front roof seemed to always bring features of the ridge
line into prominence.
The roofer cut a separate little section of ridge-cap to go over
the wing-wall part, figuring to integrate that into the main center one.
Meanwhile, the helper installed the foam backer pads behind the z-bars.
Then they started install the main ridge cap. It's a simple bent piece
with two little open hems that are supposed to hook over the protruding
ends of the z-closures and shelter a fairly generous volume of everything
under the cap. It gets final-fastened in the middle of each z-bar with
a pop-rivet for extra holding security.
Except that the cap piece wasn't fitting right -- it appeared to be too
wide to actually clip in. Either it was intended for a 90-degree ridge
which this at the shed dormer isn't, or the cap had gotten slightly
squashed in the shipping box, or the z-bars had been mounted too high.
In theory, there was some spec measurement for distance from peak
for the z-bars that they had followed. Maybe it was a combination
With one side already riveted down, the roofer grabbed the lip and
hung his ass way overboard to try and pull the cap up against
the z-bars while the kid got a rivet or two in. With this initial
realignment done, they could then continue down the line.
This felt a little ghetto to me; the ridge cap should have fit better
from the beginning without being forced.
Even more ghetto was the junction between the small end ridge-cap
section and the main one. This is where those slightly elevated ends
of the ridge became a problem. The roofer tried to pretty up the metal
with some slitting and bending, but if you look at this carefully it's
a total reverse flash with the end piece sluicing water right in under
the main part of the ridge cap. The end piece isn't even over its own
z-bar. The whole area looked like total ass from the front, with the
main ridge cap sort of bent oddly all the way along.
The roofer gave me the impression that a lot of this was semi-temporary
and would get fixed up better later after the final panel delivery happened,
but other than some finalized fastening at the ends I didn't see what else
could really be done.
I was having decidedly mixed feelings about the roof in general. On one
hand standing-seam metal would be a great long-term solution and totally
beat the pants off any asphalt shingles, and the new assembly contained
several waterproofing layers backstopping each other so that I *should*
never see water in the attic again. And overall the roofer had been
doing a pretty
nice and robust installation job, even accomodating my what-if questions.
On the other hand, across the various trade segments involved there was a
whole lot about the roof that could have been done better -- without any
particular extra work on anybody's part, just by paying more attention
to stuff they already knew how to do.
Fortunately a lot of what's been bitched about has to do with areas out
at the edge of overhangs, where problems are less likely to affect the
house itself. But I'll need to keep an eye on the soffits and siding
for any evidence that water might be going astray.
About three-quarters of the ridge cap got put on before the roofer ran
out of things really needing to get done before the extra panels came in.
He stuck some tarpaper over the remaining open panel ends at the top
and that and the Grace could keep the unfinished parts sufficiently
water-resistant for the intervening time.
We actually got really lucky on weather over the week most of the roofing
happened -- basically no rain the whole time. It rained a couple of
days later, and you can bet I was right out there with an umbrella and
the camera checking drip patterns. *Finally* the water path behind the
trim metal had been mostly closed off, and the drip was all coming off
the end of the overhang like it should. Yay.
Check the big pic for odd inconsistencies in how the gable closures were
terminated at the lower corners. The guy that showed up for one day did
one of them, and the main roofer did the other.
The drip pattern off standing-seam was interesting, especially on the
lesser-pitched part. Water would sort of pile up along the length of the
panel, and then as some part of it started rolling down it would trigger
a little cascade that gathered speed on the way down, the result being
each separate panel end peeing substantially outward in amusingly
random little bursts.
Eventually there would be a gutter under this whole thing and then it
wouldn't matter how it drained. But for now it was still all landing
on the heat-pump. I really needed to get that hutch finished.
Next section (21)