One counter-argument might be – well, you never know when anyone will turn on their AC, either, at least not minute-by-minute.  The grid is a balancing act; unpredictable, random loads have the same effect as unpredictable, random generators.

To which one might then counter yes, but there are so many AC units out there, they average out, more or less, turning on and off at random times and smoothing things out in aggregate.

To which the solar advocate might reply OK, then with enough solar the peaks and valleys of generation should cancel out too, as clouds move out of one area into another.  Does this seem likely out in practice?

To find out, I grabbed 5 minute data from about 40 Enphase systems in the twin cities on a highly variable, sporadically cloudy day.  Because we don’t yet have a whole lot of solar here, and I didn’t want the one or two large commercial systems in the group to swamp the smaller residential systems, first I normalized them all to a % of their max output.  (This might be cheating a little, but with a lot more systems randomly distributed in size and geography, the swamping-out effect should be minimiized.)  Here’s what just 4 of those systems looks like; each is indeed pretty messy and unpredictable at the 5-minute range:

Then I averaged all of the systems.  Here’s what the average looks like, compared to one of the individual systems:

It appears that things certainly do smooth out when we look at geographically distributed systems.  If I were a grid operator, I might feel a lot better about that.

The caveats might be that this is a very wide geographic range – I grabbed systems from all of the twin cities and suburbs.  And that’s probably larger than the various sub-grids within the cities; what the variability is within those subgrids is, or how this solar variability affects them, I’m not sure.  And of course my initial normalization of all systems to the same size could be argued with.

There have been much more rigorous papers and presentations written on this as well, see for example “Quantifying PV Power Output Variability” by Thomas E. Hoff and Richard Perez in 1999, and “Implications of Wide-Area Geographic Diversity for Short- Term Variability of Solar Power” by Andrew Mills and Ryan Wiser at LBNL in 2010.  But with the advent of 5-minute monitoring from systems like Enphase, I wonder if even better results could be found from this wealth of data.

[1]  I’ll submit that a sporadically cloudy day is more trouble to a grid operator than a generally cloudy day.   We often know if a day will be cloudy well ahead of time, and that doesn’t yield the minute-to-minute variations of a sporadically cloudy day.  The grid is better, I think, at responding to these longer-term variations.

The above changes are before Xcel proposed a slightly lower increase; it’s unclear how that will affect residential customers at this point, I guess.

Today marks the 1000th day of operation for the 2.53kw solar array on our roof.  Since we’ve installed it,

• 8.9 MWh of energy has been produced (almost 9kWh/day on average)
• 6 tons of CO2 have been avoided
• $1,100 in electricity costs have been avoided ($850 if I don’t count the optional fee for wind power I’ve subscribed to for years)
• 70% of our electrical consumption has been covered
• No maintenance has been required (other than optional snow removal)
• My inverter manufacturer (Enphase) has gone public
• My panel manufacturer (Siliken) has gone bankrupt.  (But warranties remain intact, in theory)
• Utility rebates from Xcel have dropped from $2.25/watt to $1.50 since we installed.

We’ve recently embarked on a plan to renovate the house, and may end up with a new flat roof on the back.  If so, I hope we can have space and budget to add another 1kW or so.

Minnesota has also continued to move forward on solar energy, with Ikea briefly holding the state record with a 1.1MW array on its roof, later eclipsed by a 2MW utility-scale system in Slayton, MN.  My utility, Xcel Energy wanted to eliminate the solar rebate altogether, but it was extended at a lower rate of $1.50/watt for now – appropriate, I feel, so that the solar industry has some planning runway, but rebates drop along with system costs.  There has also been some discussion of a 10% solar standard for Minnesota.  Solar Works in Minnesota!

(Image link goes to Amazon)

Read any collection of energy tips, and you’ll get pretty much the same thing.  Turn off lights behind you, unplug your cell phone charger, put your AV center on a power strip, etc.  The problem with these ideas, IMHO, is that you have to remember to do them, everyone in your family has to do them, and you have to keep doing them indefinitely to keep getting any benefit.  And face it, it’s hard to change behavior.

There are a few things I’ve found which make it simple and automatic to save energy, without adding too much hassle to daily routines, and without any significant expense or installation trickiness.

An example – My wife bought me a very nice espresso machine for my birthday about a year and a half ago; when it’s on, it’s using a lot of juice.  The group head has a resistive heater, and the boiler is keeping the water very hot.  I’d get up, make coffee, and then I’d be good for a while.  Almost invariably, I forgot to turn the machine off when I was done, and there it sat, drawing hundreds of watts, keeping itself warm for the next coffee I wasn’t going to make any time soon.  And worse, in the summer, it was adding heat to the kitchen.

Enter the device pictured above.  You press a button on the side, and it turns on and stays on for 30 mins, 3 hours, or 6 hours.  I set mine to the 30min setting, so when I make my coffee, I push the button, make the coffee, and walk away.  It turns itself off in 30 mins.  Done and done.  The beauty is, I have to take the action when I want something good to happen (like make coffee) – it’s a positive action, not “remember to do this when you are done.”

The other appliance I could use this device on is my stereo amplifier.  It pulls tens of watts even when not playing music.  Having the timed outlet set to 3 hours or even 6 would stop me from accidentally leaving it on over night.  A geekier and more satisfying solution, since my music source is a Logitech Squeezebox digital music player, would be to hook it to an X10-controlled switch, and tell the music server to turn on the amp whenever the device is active, and turn it off when it’s not.  I haven’t done either of those things yet, and sometimes the amp is on all night.

Of course, there are other “set it and forget it” ideas like setting computers to sleep when idle, installing motion-detector switches for lights (I prefer ones like this [amzn] which are manual-on, auto-off), programmable thermostats (if you actually program them), etc.  But I thought the device above was pretty slick for controlling random devices I’d rather not leave running when I’ve forgotten to turn them off.

Have you found any other unusual but simple, set-it-and-forget-it devices or ideas to help control unnecessary energy use?

The top line numbers:

• 2011: 839 therms, 4538 kWh consumed = 99.4 million BTUs
• 2012: 687 therms, 4577 kWh consumed = 84.2 million BTUs

On the face of it, that looks like quite an improvement, but the vast majority of the decrease was in natural gas, which is influenced by the weather.  We can use Heating Degree Days and Cooling Degree Days to quantify air conditioning requirements, and from degreedays.net, I find:

• 2011: 7357 heating and 1128 cooling degree days (@65F base)
• 2012: 6390 heating and 1238 cooling degree days (@65F base)

So we had about 13% fewer heating degree days, and consumed 18% fewer therms of natural gas, our primary heating fuel – fairly strong correlation, but maybe some improvement.  I don’t measure cooling energy independently (yet) so I can’t correlate that at all.

On the solar production side, we had this from our 11 panel, 2.53kW array:

• 2011: 3306 kWh, or 72.8% of our consumption
• 2011: 3352 kWh, or 73.2% of our consumption

Not a big change there.  So other than the weather, we didn’t have significant changes in our home other than slightly older kids, and the addition of a mini-split heat pump for AC & some heating.

I’ve mostly run out of ideas to make this old house more efficient.  Every light is CFL or LED, we’ve insulated the walls & attic, replaced the windows, etc.  We are considering some renovation, and that will probably change our energy use if we go through it.  I’ll keep you posted!

I iterated over the entire linux-2.6 git tree as of today (~3.7-rc8), and counted all commits (excluding merges), then counted up every Reviewed-by tag in all commits.  (I realize that this isn’t the only measure of review quality.  More caveats below).  It’s  possible that some commits might have more than one Reviewed-by tag, but I assume that’s rare overall.  I found:

4.7%: 313,848 commits, 14,631 reviews

Breaking it down further, for some of the top-level directories here’s what I found:

 0% sound: 12151 commits, 68 reviews
 1% virt: 451 commits, 8 reviews
 2% crypto: 888 commits, 23 reviews
 3% block: 1643 commits, 61 reviews
 4% security: 1675 commits, 68 reviews
 5% drivers: 149445 commits, 8107 reviews
 7% fs: 26747 commits, 2138 reviews
19% mm: 5858 commits, 1134 reviews

Changes Over Time

However, things seem to be improving.  Here is the total ratio of “Reviewed-by” tags to commits for each release since 2.6.13:

Perhaps Reviewed-by is coming into vogue; I find that interesting, since there has been no formal push for using Reviewed-by more often, as far as I know.  What if we look at those top-level subsystems only since v3.0?

 0% sound: 3341 commits, 32 reviews
 1% crypto: 185 commits, 2 reviews
 3% block: 303 commits, 10 reviews
 3% security: 406 commits, 15 reviews
 5% virt: 111 commits, 6 reviews
12% drivers: 40747 commits, 5054 reviews
15% fs: 5362 commits, 814 reviews
27% mm: 1193 commits, 325 reviews

mm/ continues to win, and improve the reviewed % as well.  fs, drivers, and virt have improved their percentage lately, too.

Filesystems

fs/ is second, and of particular interest to me.  How’s it doing since the beginning of the git tree?

1% fs/affs: 117 commits, 2 reviews
1% fs/bfs: 72 commits, 1 reviews
1% fs/coda: 143 commits, 2 reviews
1% fs/debugfs: 85 commits, 1 reviews
1% fs/fuse: 399 commits, 5 reviews
1% fs/gfs2: 1233 commits, 13 reviews
1% fs/logfs: 124 commits, 2 reviews
1% fs/minix: 91 commits, 1 reviews
1% fs/ncpfs: 121 commits, 2 reviews
1% fs/nfs: 2731 commits, 36 reviews
1% fs/ocfs2: 1578 commits, 29 reviews
1% fs/omfs: 57 commits, 1 reviews
1% fs/pstore: 64 commits, 1 reviews
1% fs/reiserfs: 512 commits, 7 reviews
1% fs/sysv: 95 commits, 1 reviews
2% fs/9p: 369 commits, 8 reviews
2% fs/autofs4: 216 commits, 6 reviews
2% fs/btrfs: 2607 commits, 55 reviews
2% fs/hostfs: 96 commits, 2 reviews
2% fs/hppfs: 46 commits, 1 reviews
2% fs/jbd: 172 commits, 4 reviews
2% fs/nfsd: 1358 commits, 33 reviews
3% fs/devpts: 57 commits, 2 reviews
3% fs/ecryptfs: 403 commits, 16 reviews
3% fs/ext2: 314 commits, 10 reviews
3% fs/ext3: 516 commits, 19 reviews
3% fs/jbd2: 222 commits, 8 reviews
3% fs/lockd: 376 commits, 14 reviews
3% fs/ubifs: 471 commits, 16 reviews
3% fs/udf: 306 commits, 11 reviews
4% fs/nls: 23 commits, 1 reviews
4% fs/quota: 116 commits, 5 reviews
4% fs/ramfs: 70 commits, 3 reviews
5% fs/ext4: 1629 commits, 93 reviews
6% fs/hugetlbfs: 131 commits, 9 reviews
6% fs/proc: 934 commits, 62 reviews
7% fs/ceph: 673 commits, 51 reviews
16% fs/cifs: 1741 commits, 286 reviews
50% fs/xfs: 2376 commits, 1204 reviews

Any fs not listed above had < 1% Reviewed-by tags.

I’m interested primarily in xfs, ext4, and btrfs for my day job.  How have they done over time?

(XFS exceeds 100% at times when many commits have more than one Reviewed-by tag.)

The Caveats

Now, granted: all this measures is the number of aforementioned Reviewed-by tags per commit in the tree; the absence of a tag does not necessarily mean there was no review. And, in many cases, it may be that the maintainer has done a review, and it’s unusual for the maintainer to add Signed-off-by as well as Reviewed-by.  And finally, presence of the tag does not say anything about the quality of review. So take this with as big a grain of salt as necessary. If you’re the maintainer of one of these subsystems, and Reviewed-by is not what you use, I’d like to know (and to know why that is). Meanwhile, this seems like the best measure we’ve got.

Finally, who are the top ten individuals with the most Reviewed-by: tags since the git tree’s inception?

213     Reviewed-by: Jesse Barnes <jbarnes@virtuousg---.org>
227     Reviewed-by: Chris Wilson <chris@chris-wil---.co.uk>
238     Reviewed-by: Josh Triplett <josh@joshtripl---.org>
245     Reviewed-by: Mike Christie <michaelc@cs.w---.edu>
255     Reviewed-by: Dave Chinner <dchinner@red---.com>
284     Reviewed-by: Michael Chan <mchan@broad---.com>
362     Reviewed-by: Pieter-Paul Giesberts <pieterpg@broad---.com>
435     Reviewed-by: Roland Vossen <rvossen@broad---.com>
478     Reviewed-by: Arend van Spriel <arend@broad---.com>
486     Reviewed-by: Christoph Hellwig <hch@---.de>

hch, somehow I am not surprised. :)

Edit: Everything else is bad!  Don’t get this device.  Really.  It’s the buggiest piece of junk I’ve had the misfortune of trying to run on my network.  I’ll leave the rest of this post here for posterity.

Unfortunately it doesn’t seem to do SNMP and there’s no way to monitor traffic.  On a whim I figured I’d look into the firmware and see if it offers any hints.  Just a note to the interwebs about how I got it unpacked:

First I ran binwalk to see what’s in it; that’s a cool tool I didn’t know about!

# binwalk-0.4.5/src/binwalk -m binwalk-0.4.5/src/magic.binwalk FW_X2000_AnnexA_2.0.04.007_20121009.bin

DECIMAL       HEX           DESCRIPTION
-------------------------------------------------------------------------------------------------------
256           0x100         Squashfs filesystem, little endian, non-standard signature,  version 4.0, size: 1746252031 bytes, 1243 inodes, blocksize: 0 bytes, created: Wed Jul  9 20:36:00 2025
352           0x160         LZMA compressed data, properties: 0x5D, dictionary size: 65536 bytes, uncompressed size: 65536 bytes
15110         0x3B06        LZMA compressed data, properties: 0x5D, dictionary size: 65536 bytes, uncompressed size: 13944 bytes
...

Ok, so maybe squashfs starting 256 bytes in.  So dd that out:

# dd if=FW_X2000_AnnexA_2.0.04.007_20121009.bin bs=256 skip=1 of=FW.squash

But squashfs tools won’t touch it.  What’s in it?

# hexdump -C FW.squash  | head -n 8
00000000  73 68 73 71 db 04 00 00  53 c6 73 50 00 00 01 00  |shsq....S.sP....|
00000010  78 00 00 00 01 00 10 00  c0 00 01 00 04 00 00 00  |x...............|
00000020  3f 1c f1 1d 00 00 00 00  19 6f 68 00 00 00 00 00  |?........oh.....|
00000030  11 6f 68 00 00 00 00 00  ff ff ff ff ff ff ff ff  |.oh.............|
00000040  b0 15 68 00 00 00 00 00  11 3a 68 00 00 00 00 00  |..h......:h.....|
00000050  2c 65 68 00 00 00 00 00  fb 6e 68 00 00 00 00 00  |,eh......nh.....|
00000060  5d 00 00 01 00 00 00 01  00 00 00 00 00 00 3f 91  |].............?.|
00000070  45 84 68 34 8a 09 0a 40  62 ae 9d db 78 c5 ce 30  |E.h4...@b...x..0|
...

Ok, that’s the “shsq” magic…

I tried the various unsquashers from the firmware mod kit but those failed too.  In the end, I grabbed the squashfs-4.2 tools from sourceforge and applied this patch:

diff -Nurp squashfs4.2/squashfs-tools/Makefile squashfs4.2-hack/squashfs-tools/Makefile
--- squashfs4.2/squashfs-tools/Makefile	2012-11-30 12:26:45.958576463 -0600
+++ squashfs4.2-hack/squashfs-tools/Makefile	2012-11-30 12:24:21.969566639 -0600
@@ -26,7 +26,7 @@ GZIP_SUPPORT = 1
 # To build using XZ Utils liblzma - install the library and uncomment
 # the XZ_SUPPORT line below.
 #
-#XZ_SUPPORT = 1
+XZ_SUPPORT = 1

 ############ Building LZO support ##############
diff -Nurp squashfs4.2/squashfs-tools/squashfs_fs.h squashfs4.2-hack/squashfs-tools/squashfs_fs.h
--- squashfs4.2/squashfs-tools/squashfs_fs.h	2011-02-11 09:49:24.000000000 -0600
+++ squashfs4.2-hack/squashfs-tools/squashfs_fs.h	2012-11-30 12:24:49.310566558 -0600
@@ -270,10 +270,10 @@ struct meta_index {
 typedef long long		squashfs_block;
 typedef long long		squashfs_inode;

-#define ZLIB_COMPRESSION	1
-#define LZMA_COMPRESSION	2
-#define LZO_COMPRESSION		3
-#define XZ_COMPRESSION		4
+#define LZMA_COMPRESSION	1
+#define XZ_COMPRESSION		2
+#define ZLIB_COMPRESSION	3
+#define LZO_COMPRESSION		4

 struct squashfs_super_block {
 	unsigned int		s_magic;
diff -Nurp squashfs4.2/squashfs-tools/unsquashfs.c squashfs4.2-hack/squashfs-tools/unsquashfs.c
--- squashfs4.2/squashfs-tools/unsquashfs.c	2011-02-28 16:27:06.000000000 -0600
+++ squashfs4.2-hack/squashfs-tools/unsquashfs.c	2012-11-30 12:28:31.793690803 -0600
@@ -1460,10 +1460,10 @@ int read_super(char *source)
 	 */
 	read_fs_bytes(fd, SQUASHFS_START, sizeof(struct squashfs_super_block),
 		&sBlk_4);
-	swap = sBlk_4.s_magic != SQUASHFS_MAGIC;
+	swap = 0;
 	SQUASHFS_INSWAP_SUPER_BLOCK(&sBlk_4);

-	if(sBlk_4.s_magic == SQUASHFS_MAGIC && sBlk_4.s_major == 4 &&
+	if(!swap && sBlk_4.s_major == 4 &&
 			sBlk_4.s_minor == 0) {
 		s_ops.squashfs_opendir = squashfs_opendir_4;
 		s_ops.read_fragment = read_fragment_4;

And after that hackityhack, it does unpack. And what do you know, it does have busybox:

# ls -l bin | grep busybox
-rwxr-xr-x. 1 root root 323640 Oct  9 01:38 busybox
lrwxrwxrwx. 1 root root      7 Nov 30 12:25 cat -> busybox
lrwxrwxrwx. 1 root root      7 Nov 30 12:25 chmod -> busybox
lrwxrwxrwx. 1 root root      7 Nov 30 12:25 cp -> busybox
lrwxrwxrwx. 1 root root      7 Nov 30 12:25 date -> busybox
lrwxrwxrwx. 1 root root      7 Nov 30 12:25 dd -> busybox
...

And apparently runs the Linux kernel:

# ls lib/modules/
2.6.30

So, ok, Cisco/Linksys, where’s the GPL notifications etc?  I haven’t seen anything obvious, but their GPL download center doesn’t seem to have anything for this box.  Is it a GPL violation?  I sent a request through their normal process, we’ll see what happens.  :(

Sadly, no hints about SNMP in there, either.

Edit:  Although the “GPL code center” has no download for this box, I did obtain source via a request, and apparently there is a source offer buried deep on the setup CD.  I’ll see if the source builds, next, I guess.

And I had a heck of a time making that work.  hdparm -y would quickly spin them down, but using hdparm -S to set an idle timeout seemed to have no effect; I had been trying to use hdparm -S 241 to set a 30 minute spindown time, and I had no luck whatsoever.  With the drives spinning, the server used more like 30W.

Mostly through trial and error, I found out that if you set a lower spindown timeout, i.e. hdparm -S 3, the drive will spin down in 10 minutes.  So unlike most drives which follow the timeouts as specified in the hdparm manpage, this one seems to spin down in 10 minutes for low -S values, and never for higher values.  I don’t know what the threshold is, but for now, hdparm -S 3 is doing the trick for me.  After doing this, the drives finally managed to be idle most of the time, when not serving up media:

I also used the idle3-tools to turn off the idle3 behavior and save the drives from the massive load/unload cycles; I don’t know if that’s warranted or not, but I did it anyway.

On a related note: The Western Digital Red [amzn] “NAS-optimized” drives might actually use a bit less power than even the “Green” drives.  They use 4.4W (vs 5.3W on the Green) for read/write, 4.1W (vs. 5.5W) for idle, and 0.6W (vs. 0.7W) for standby/sleep.  (Comparing to the WD20EFRX and WD20EZRX models here).  If one of the Green drives in the mirror goes, I might consider replacing it with a Red.

What these units do is move heat.  Amazingly, they can move it in either direction!  In the summer, they move heat out of the house, obviously enough, just like any A/C.  But in the colder months, they can move heat into the house as well.  Even when it’s cold outisde, you ask?  Yep!  How is that possible?

There is heat outside even when it’s “cold.”

Even when it’s freezing outside, there is still heat. Really!  Until we get to absolute zero, there is still heat present which can be moved around.  Don’t believe me?  Consider your freezer.  Let’s say you keep it at 0F.  How does it stay at 0F?  Well, if it ever gets to 1F or 2F, it extracts heat from the inside of the freezer, and moves it to the outside (i.e. your kitchen).  So yes – it’s moving heat from a very very “cold” place and warming up a warmer place as a result.  The mini split heat pump works this same way.

My Fujitsu ASU12RLS2 / AOU12RLS2 indoor/outdoor units can actually extract heat from the outdoors even when it’s -5F, although the efficiency diminishes as the outdoor temperature drops.  Air source heat pump efficiency is expressed by the HSPF (Heating Seasonal Performance Factor), which, via Wikipedia, “is a ratio of BTU heat output over the heating season to watt-hours of electricity used.”  HSPF is expressed in BTU/Watt-hour, and my unit has a rating of “12″ meaning that for every Watt-hour of electrical energy it uses, it moves 12 BTUs into the house, on average, over the heating season.

Over 100% efficiency

BTUs?  Watt-hours?  Ok that might not mean much.  But get this: Watt-hours (Wh) and BTUs are both expressions of energy, just expressed in different units.  There are 3.413 BTUs in a watt-hour of electricity.  So for every 3.413 BTUs of electrical energy we input, we get 12 BTUs of heat into the house, for a multiplier of about 3.5x.  How’s that for efficiency!  By comparison, a simple resistance space heater is 1:1, providing 1 unit of heat for every 1 unit of electrical energy input.  The difference is that a heat pump moves heat rather than creating it directly, and is therefore able to do so with more than 100% efficiency.

So it’s 350% efficient, that’s great and all, but natural gas is cheap, and electricity usually comes from coal – does it make sense from either an environmental or a cost point of view?  I created a spreadsheet on Google Docs to take a look.  An abridged version is here:

Looks like a win

With the following assumpti0ns (take a look at the spreadheet to alter them): 82 AFUE boiler, $0.71/therm natural gas, $0.10/kWh electricity, a 12 HSPF heat pump, and 1161 lbs CO2 per MWh, it looks about break-even on operating costs, but about a 30% reduction in CO2.  Of course things like energy costs and carbon intensity vary by region; for carbon intensity numbers for your grid region you can look here (I actually took numbers from my utility’s annual report).

So that looks pretty good, and in fact if we may have even a bit better outcome, because:

• We have solar on the roof and buy wind energy, so our CO2 intensity should be lower
• We’d probably use the mini-split in the shoulder seasons, when it is more efficient and the boiler is less efficient

The one downside, right now, is that we have just one of these things, at the top of the stairs.  It’s a point source of conditioning so distribution is something of an issue.  But it’s really impressive how much heat it puts out on a chilly night, with not much of a spike in the daily electrical energy graph.

Edit Nov 15 2012: There’s a very cool calculator at buildinggreen.com which lets you easily compare any two heating fuels.  Give it a shot!

2012 was a hot one. We’ve had at least 31 days over 90F, something which hasn’t happened since 1988, and July 2012 was the second warmest July on record in the Twin Cities. This is a departure from the norm; for example the average number of 90F days is 13 – we almost tripled it this year.  The 1930′s homes in our neighborhood weren’t built with AC, and retrofitting ductwork is tough.  And although I obsess over our energy use like some people obsess over Kim Kardashian, we do have to be able to function on those sweltering days.   Last year we struggled through the hot week with a single 8,000 BTU window unit upstairs, which was cumbersome, noisy, ineffective, and inefficient.

Enter the Mini-Split

This year we decided to give in and install something more efficient, permanent, attractive, and convenient – a Fujitsu ductless mini-split AC (our model is here).

These units come in various configurations, but simple ones like ours have an outdoor compressor/condenser similar to central AC and a single indoor air handler that hangs on the wall.  Refrigerant, power, and control lines run back to the outdoor unit.  Because it’s a matched set, and because there are no ducts to blow through a hot attic and leak air, they can be very efficient – ours is 25 SEER, 13.8 EER for cooling.  For comparison, Energy Star standards for central AC require >=14 SEER/ >=11 EER.

We weren’t going for perfect comfort, just something to keep it tolerable, and keep the upstairs cooler at night for sleeping.  One consideration with these units is that they are a single point of conditioning, and distribution can be an issue*.  We installed only one unit at the top of the stairs, figuring some cold air could fall and hot air could rise.  If we’d been going for more consistent whole-house comfort, at least one more head would have made sense, and we may still do that in the future.

In the end, we selected a single 12,000BTU (1 ton) unit after doing a whole-house heat gain analysis to determine the necessary cooling capacity.  I contacted the fine folks at The Neighborhood Energy Connection in Saint Paul to do the analysis.

*other configurations have a small unit above/below the living space with short duct runs to different rooms to solve this issue, at the cost of some efficiency.

Effectiveness & Energy Use

Overall, it’s been pretty good.  We put it in just before a perfect storm of hot:  103F days, both sides of the family visiting, and a birthday requiring use of the oven to bake a cake.  That was a struggle, but it kept things reasonable – around 80F downstairs – and much drier, which makes a huge difference.  Distribution was a bit of an issue; it’s hard to cool a hot kitchen with a unit on the 2nd floor.

Here might be a more representative day: Sept 11 2012 had a high of 93F.  We closed the house and set the unit to 74F at about 11:00AM, and turned it back off around 7pm.  The bigger spikes are likely the espresso machine.  ;)  Stuff later in the evening is cooking & running the dishwasher.

The graph shows indoor & outdoor temperatures, solar PV power production, and household power use.  Outdoor temp peaked at 93F at 4pm, and indoor temps on the first floor peaked at about 76.5F around 1pm.  On the energy side, we used 13kWh and produced about 13.5kWh.  If we’d anticipated a string of hot days, we would have just left it on 24/7 to keep things cool and dried out.

In July 2011, with the window unit struggling to keep up, we used 33kWh-38kWh per day for the hot days.  In July 2012, with the mini-split running 24/7 on the hottest days, we used between 19kWh and 28kWh per day – quite an improvement.

It Heats, Too

The unit can also work in reverse, as an air-source heat pump, with an HSPF of 12.0, meaning for every BTU watt-hour it consumes, it moves 12 BTUs of heat into the house.  Converted to Coefficient of Performance, this indicates that it transfers about 3.5 units of energy into the house for every unit of energy it consumes – or 3.5x more efficient than a resistance space heater.  It maintains rated heating capacity down to 20F outside, and continues to function even in -5F weather.

I haven’t yet decided if we’ll use it much for heating, but it might make sense in the “shoulder seasons” when our 83 AFUE boiler would be operating infrequently and at lower efficiency.  Because these electric units can heat and cool so efficiently, they are often used in the design of well-insulated net-zero or near-net-zero homes with PV installed.

The Verdict

Overall I’m pleased with this thing.  It’s quiet, efficient, attractive, and made the hottest weeks quite tolerable.  It helps that we’ve done air-sealing and insulation of our home, so the single unit is better able to meet our cooling load.  I may experiment with sun control screens on the south windows next year to cut down on solar heat gain.

It’s handier, so we did run AC more days than we did last year, which is a risk if we’re trying to conserve energy.  And although some studies have shown that window ACs actually save energy for that reason, in our case we used 588kWh in July 2011, and 488 kWh in July 2012, and July 2012r was hotter:

                    Cooling Degree Days
                    (Base65)   (Base50)
                  Month Dept. Month Dept.
                  Total from  Total from
                        Norm.       Norm.
July 2011          433   174   898   179
July 2012          475   199   940   202

Less energy + more comfort = win.  We might be tempted to put one more unit on the first floor next year.  I’m not sure how that will affect overall energy use, but because the home’s cooling load will be the same (and we won’t need a bunch of fans to blow the cool air downstairs) I wouldn’t expect it to rise significantly – we’ll see.