File	Date	Author	Commit
avr	2012-10-16	Paul Fox	[4caf60] switch from sputstring/sputchar/sgetchar to put...
blog	2012-06-06	Paul Fox	[70210f] revised for republishing. grrrrr.
doc	2011-05-15	Paul Fox	[8419e4] don't bother doing averaging to create the watt...
host	2012-09-25	Paul Fox	[4e60a7] fix formatting
plot	2012-09-25	Paul Fox	[c753af] switch calculating plot-based minima to daily m...
AAA_rehosted_at_github	2013-04-27	Paul Fox	[8ed1b8] irmetermon now hosted at https://github.com/fox...
Makefile	2011-05-11	Paul Fox	[87f05e] add diff_install target
README	2013-04-27	Paul Fox	[1a07eb] irmetermon now hosted at https://github.com/fox...

Read Me

irmetermon now hosted at https://github.com/foxharp/irmetermon

irmetermon -- Paul Fox, April 2011

Many modern electric meters deployed in the US now have wireless
connectivity to enable the electric utility to read them from a
vehicle in the street, rather than requiring someone to approach
each meter physically.  While in principle one might be able to
get access to the meter's data via the electric utility's radio
interface, a) the equipment would be expensive, and b) the data
stream is likely (hopefully!) encrypted.

Whether for testing convenience, or required for calibration
purposes, these meters also all seem to be equipped with an
infrared LED that emits one pulse for every watt-hour consumed. 
(Think of this pulse as the modern equivalent of the spinning
disk that existed in older meters.)  Rather than trying to
use the radio interface, it's far, far, easier to simply watch
for pulses of the little IR led.  This project does just that.

I've played with two circuits for detecting the IR pulses using
an IR phototransistor.  The simple schematics appear in the doc
subdirectory (schematic.breakdetect and schematic.adc).  There's
also a screen grab of a scope trace of the IR pulse in
doc/irpulse.png.

One circuit, which interfaces directly to a serial port, is quite
simple, and the software to gather the input pulses equally
trivial -- the software looks for ANSI "break" conditions on the
serial port, which will occur once per pulse.  (Using a unix tty
driver, "break" conditions can be made to appear as if a NULL
character was received on the port.)  Unfortunately, I wasn't able
to make that trivial circuit work across varying ambient light
levels.  I'd be pleased if someone could make this scheme work --
the absence of a microcontroller makes this method much simpler,
and therefore more accessible to more people.

The other (more successful) circuit uses an Atmel AVR micro to
monitor the phototransistor's output, and does edge detection in
software to detect the pulses.  This makes it easy to ignore
changes in ambient light.  Again, a serial interface is used,
though in this case actual ASCII messages carry information about
the the pulses.

The host side software in either case logs the number of pulses
(each representing a watt-hour consumed) per minute.  Charting
software (using gnuplot) makes those results visible on a
webpage.  The host s/w is split in two:  read_ir.c (or
read_ir_breakdetect.c for the simpler circuit) reads data from
the serial port, timestamps it, and forwards it across a pipe to
ir_meterlog.c, which does the logging.  Daily logs are kept on
both a 1 minute and 10 minute basis.  (Disk is cheap, and while 1
minute logs are very interesting on a daily basis, it probably
makes more sense to keep 10 minute logs long-term.)  In addition,
a simple file is continuously maintained which records the
current power consumption, based on the timing of the most recent two
pulses.

There are more detailed construction and operation notes in the
doc directory in "Installation".

paul

----------

Thanks where thanks is due -- I was inspired to tackle this by
some earlier investigation into the IR LED by "Dave K", on his
"Tech Blog":
  http://davestech.blogspot.com/2008/02/itron-remote-read-electric-meter.html
as well as by "blalor" who posted some scope trace pictures:
  http://www.flickr.com/photos/blalor/3876765214/in/photostream/