EFFECT OF PHASE ON THE RESPONSE OF X10 DIMMER MODULES AND ON THE DIM
STEPS REPORTED BY A CM11A INTERFACE:  A study which resolves several
issues with dims and microdims, their effects on various X10 dimmer
modules, and discrepancies reported by a CM11A which receives these
functions over the power line.

           by Charles W. Sullivan

(Originally posted to Usenet newsgroup comp.home.automation on
29 Dec 2005.  Revised 21 July 2007.)

INITIAL OBSERVATIONS
--------------------
The number of dim steps reported by a CM11A when a dim (or bright)
function is received over the power line has what at first appears to
be a random component.  The steps for dims sent by most X10
transmitters - CM11A, TM751, RR501, Mini-or Maxi-Controller (although
not CM15A) - and reported by a CM11A can be fitted to the formula:

  Dim steps = (11 * N) + X

where N >= 0 and X is either 2 or 3 (rarely 4) with approximately equal
frequency.  The CM11A is limited to reporting a maximum of 210 steps.

With a CM11A as the transmitter, N is the programmed dim_level (1-22)
minus 1.  With the other transmitters, N depends on how long the "dim"
button is pressed but the same formula holds true.

Two CM11A units plugged into the same power strip always report the same
value of X.  But two CM11A units plugged into receptacles on different
phases (+/-120 to neutral) of the AC power line always report opposite
values for X; if one report 2, the other reports 3, and vice-versa.

X10 control signals are transmitted over AC power lines by sending 
an encode sequence of short bursts of 120 KHz at the zero crossings of
successive half-cycles of the AC waveform.  The above observations lead
to the hypothesis that the reported value of X is determined by whether
the dim power line code is received beginning with the rising or falling
zero crossing.

TEST SETUP
----------
To test this hypothesis, a CM11A test setup was rigged to always
transmit the dim signal beginning on the zero crossing at the same
half cycle.

The secondary of a 9 Volt AC transformer (wall wart) was connected from
Signal Ground to the Carrier Detect (CD) pin of the serial port to which
a CM11A is connected. (The CD line is not connected in a standard CM11A
serial cable.)

Software (Heyu) was modified:  After sending the dim function code,
e.g., "0xE0 0x64", and receiving the checksum from the CM11A, the final
0x00 "send-it" byte is deferred until a few milliseconds after the CD
line becomes inactive.  With the 9V transformer secondary in phase with
the AC power line, the dim signal transmission should then start on the
rising zero crossing.

This test setup successfully removes the apparent randomness in X.  A
receiving CM11A on the same AC phase as the triggered sender now always
reports X as "3" in the above mentioned formula.  If the phase is
reversed by plugging the wall wart transformer upside-down in the plug
strip, X is always reported as "2".

Note: The enhancement to Heyu to allow triggering on the zero crossing
for a specific half-cycle has been implemented in the standard Heyu
distribution with the -tr and -tf command line switches.  These are
described in the EXPERIMENTAL STUFF section of man page heyu(1).

EFFECT ON STANDARD X10 DIMMER MODULES
-------------------------------------
(Note: A major redesign by X-10 of the LM465 hardware and firmware
occurred around the beginning of 2007, however the part number remains
the same. The redesigned units now share many of the features of the
LM14A 2-way Lamp Dimmer although they remain only 1-way, i.e.,
receive only.  What follows applies only to the older LM465 models.)

It has previously been observed by myself and others (Dave Houston for
one) that when transmitting microdims (i.e., programmed dim_level 1) to
a standard X10 Lamp Module, e.g. LM465, that about 150 microdims are
required to span the range from full brighness to fully dimmed.  The
question arises how this number relates to the supposed 210 brightness
steps in the module.

Microdims were sent in a loop using the above test setup with a 3 second
delay between transmissions.  The voltage across a 100 Watt lamp
connected to a lamp module was measured with a Fluke 77 multimeter,
which is specified as measuring "true RMS".

With the in-phase trigger (CM11A reports 3), the number of microdims
required for full dimming of a LM465 was 223 or 224.  With the
out-of-phase trigger (CM11A reports 2), the number was 112 or 113.
Similarly to the LM465, the same results were obtained with a 1980's
vintage X10 dimmer (with thumbwheel housecode/unit switches on the
sides) and with a Radio Shack branded module from about 1997.

These results suggest that the CM11A correctly reports the out-of-phase
microdims as two steps, but mis-reports the in-phase microdims as three
steps instead of only one step.  They also show that the number of
brightness steps in a standard X10 lamp module is actually about 223 or
224 rather than the supposed 210.  (The only official X10 document I
know of which mentions the number of module brightness levels is the
CM11A protocol.  I suspect that number came only from the 210 step
reporting limit of the CM11A.)

With 150 untriggered microdims, one would expect on average about 75
in-phase single steps and 75 out-of-phase double steps for a total of
225 steps.  So these results are consistant with previous observations.

EFFECT ON X10 LM14A MODULES
---------------------------
The effect of phase on a LM14A 2-way module was measured by sending an
extended status request after each dim transmission and recording the
resulting extended preset level (1-62) reported by the module.

Sending _contiguous_ triggered dims via programmed dim_level 1-22 after
an extended preset dim 63 command yields two very different curves of
extended level versus reported steps.  Both are approximately linear.
The out-of-phase curve reaches extended level 1 at a programmed
dim_level 15 (CM11A reports 156 steps). But the in-phase curve is still
at extended level 7 at that dim_level and doesn't reach extended level
1 until dim_level 17 (CM11A reports 179 steps).

Curiously, unlike the LM465, the in-phase and out-of-phase LM14A curves
for sequences of microdims are point-for-point identical, reaching
extended level 1 after 154 transmissions.  The number of microdim
transmissions required for a change of 1 extended level alternates
between 2 and 3 along the curve.

Charles W. Sullivan
Greensboro, NC