The Energy Detective Forums

General Category => Developer Support => Topic started by: GAR on August 22, 2011, 12:03:39 AM

Title: Theory of the In-Line Filter
Post by: GAR on August 22, 2011, 12:03:39 AM
110821-1502 EDT

Yesterday I received both an In-Line Filter and a Plug-In Filter from Energy, Inc. These are neither designed or built by Energy, Inc. They are labeled XPF for the in-line, and XPPF for the plug-in unit.

These filters are in X10 boxes with X10 instructions.

So far I have only looked inside the XPF and run some tests on it. For some information, capacitance value, I can read it off of the part. Other information is indirect. The circuit is a three stage filter consisting of three parallel tuned circuits in series, and two shunt capacitors to neutral. The shunt capacitors are 1.0 mfd, and the parallel tuning capacitors are 0.22 mfd.

Experimentally the paralel tuned circuits appear to be tuned to 150 kHz. A parallel tuned circuit at resonance is a high impedance. Conversely a series tuned circuit is a low impedance at resonance. Thus, form the 150 kHz resonance the inductance calculates to about 6.12 microhenrys. There is no direct capacitance from either input or output to neutral. The 1.0 mfd capacitors connect between the interior nodes to the neutral. Series resonance of 1 mfd and 6.12 microhenry about 70.4 kHz.

For some of my testing I used a 2.5 mH choke and a variable capacitor for a tuned circuit to determine TED and noise frequencies. TED does appear to be at about 132 kHz, and has a 10 millisecond duration. From what information I found X10 appears to use a carrier of 120 kHz.

Why use a high impedance input to output at 150 kHz instead of at 132 kHz?
At high frequencies, somewhere above 150 kHz, this looks like a capacitor shunting hot to neutral with some attenuation. At 150 kHz there is a high impedance from input to output with substantial attenuation from the shunt capacitors. At about 70 kHz the input is loaded by a low impedance.

Is this filter really an optimum design for the TED system, or just a readily available commercial product?

The data transmission time appears to be about 100 milliseconds for the 1000 and 200 or possibly more from the 5000. The signal level seems much lower from the 5000 than the 1000.

Title: Re: Theory of the In-Line Filter
Post by: GAR on August 22, 2011, 06:49:58 AM
110821-2014 EDT

I did not fully remove the filter from the box. So my assumption of the circuit is based on what I can see from the top side of the PC board and measurements.

Further experiments do seem to imply a minimum transfer function near 120 kHz. This is not the frequency of the maximum load (minimum impedance) on the source. It does seem to have a higher impedance at 150 kHz, but moderate transfer from input to output.

If your goal is to minimize X10 signals, then this may be a good filter for that purpose. For a 1 V input the unloaded output is about 0.05 at 120 kHz and 0.07 at 132 kHz.

Because I can not easily predict how this filter functions I opened up the entire board. The circuit is as I had assumed. Using a General Radio bridge at 1 kHz I read each of the inductors at about 4.5 microhenrys and this generally correlates with my previous calculations.

Thus, it does look like this may be a somewhat useful filter at the 120 or 132 kHz frequencies.

Neither the in-line or plug-in filter seems to have a great effect on the noise from one electronic ballast fluorescent. The frequencies with my crude tuned filter analyzer seem to be about 65, 100, 160 kHz. I would really expect 65, 130, and 195 kHz, or maybe it is 60, 120, and 180 kHz. I would need to do more accurate and controlled tests.

At this point I really don't know how to advise someone on an installation procedure that would provide a high probability of success. I expect that almost any installation will require filters.


Title: Re: Theory of the In-Line Filter
Post by: iteration69 on September 04, 2011, 08:36:36 AM
Before i started logging the data I noticed a lot of hiccups in the data stream. I threw the scope on the line and noticed a fair amount of noise, But this can be expected. Instead of using filters on the mains (which will be a terrible head ache) I simply moved the gateway down to the panel and ran a cat 5 cable directly to the gateway. I wanted to upgrade my switch to a 24 port anyway, and this was a good enough reason to run some more lines.

I also ran a small length of wire in order to power all the MTU's and gateway externally. This was more of a lazy factor than an operational factor.

Five months, so far, so good.