constant communication error reports

[ELA]

I have been reading a lot of posts here since I joined recently about GFCI's being a potential problem in terms of signal sucking.

 

That really concerns me as I did not want to believe that was true. GFCI's are so prevalent that if they are a problem then I think I would have to give up totally on power line carrier transmissions.

I have several GFCI outlets in my house and so that prompted me to look a little closer at these.

 

Of particular interest was the following post by IndyMike.

Mike, can you provide any data to back this up?

 

Hello apostolakisl,

 

I see two problems in the following -

1) If this room was originally wired as a laundry, the circuit is likely protected by a GFCI (Ground Fault Circuit Interrupter). These devices sense current imbalances between the hot and neutral using a differential transformer. The inductance in this transformer can significantly lower X10 and Insteon signal levels.

 

 

I recognize that GFCI's have been around for a long time and the circuitry may have changed over the years. Within the past few years I have disassembled and tested Leviton GFCI's and so I can confirm that the Leviton 7599 receptacle is not a signal sucker.

 

The power leads pass through a core inside (only 1 turn) that does not present any appreciable inductive reactance to the 131Khz signal.

To be sure there would be some small amount of leakage though so I tested.

 

To be sure that my GFCI's were not signal suckers I tested one with a signal generator and oscilloscope.

I found no measureable inductive influence and only measured a 2ma current drain when excited by 5V p-p @ 131Khz.

[IndyMike]

ELA,

 

The GFCI itself is not a "signal sucker". It can (some models) increase the powerline impedance and allow downstream loads to absorb signal. With a load applied the GFCI essentially forms a voltage divider. Without a downstream load you would not have current flow and should not notice a signal drop across the GFCI. Caveat - Insteon devices themselves are "significant" signal suckers - on the same order as a .2uF cap.

 

I checked my basement GFCI using X10 and my ELK-M1. It's a 20A Pass and Seymour GFCI that powers my office and family room:

 

1) GFCI input - 2.5 V

2) GFCI output (loads removed): 2V

3) GFCI output (w/family room A/V Loads Installed): .2V

4) GFCI output (w/computer loads installed): 0V

 

Sorry, I should have tested this using some Insteon loads - maybe this weekend.

 

The construction of your Leviton GFCI may be very different than my unit (haven't taken one apart). I have GFCI's in my garage (don't know the brand) and have no problems with Insteon making it into this area.

[ELA]

IndyMike,

I looked at the GFCI both from possible capacitive effects (signal sucking) and inductive effects (increased line impedance) and resultant voltage division.

 

I loaded the output of the GFCI with 100 ohms and there was no measureable voltage drop across the GFCI at 131Khz.

 

As I mentioned the sensing core inside has only one turn of the AC line through it. (low inductance)

 

Your GFCI test data sounds rather strange>

"#2) output unloaded is less than input"???

 

Have your tried a simple resistive load?

[IndyMike]

ELA,

 

I cleaned up my original post a little, but yes - with my loads unplugged from the GFCI circuit, I saw a 0.5V drop from input side of the GFCI to output. Keep in mind that this is a live circuit with ~ 80 feet of romex and many J-boxes. Even with everything unplugged there is distributed capacitance and inductance loading the line.

 

As a separate test, I tried the following:

1) X10 Transmitter installed on the output side of the GFCI with all loads installed. ELK ESM1 registers 2.5 V - Transmitter is capable of driving 2.5 V into the loaded line.

2) Moved the transmitter to the input side of the GFCI - ELK ESM1 on output side. ELK registers < 0.1V.

 

From the above, I must assume that I'm taking the voltage drop across the GFCI. Based on your results, we appear to have a fundamental difference in the construction of our units. This is consistent with varying reports over the years related to GFCI problems. Some manufacturers products appear to perform well with X10/Insteon. Others, not so good. I appear to have the latter.

 

I did find that I had another 20A GFCI in the "parts box". I'll try running this one with the signal generator (likely this weekend).

 

Final note: Leviton was a rather high end provider of X10 modules. It's possible they designed their line of GFCI units with X10 in mind.

[Brian H]

I have seen those Arc Fault Breakers being a problem also.

Again it was some models where fine and other where problems.

[ELA]

ELA,

 

Keep in mind that this is a live circuit with ~ 80 feet of romex and many J-boxes. Even with everything unplugged there is distributed capacitance and inductance loading the line.

 

Are you then saying that there is 80ft of romex on the load side of the GFCI?

Would my estimates of 10 ohms inductive and 1000 ohms capacitive be in the ball park (lumped values)? These are fairly small values and would mostly affect a reading taken at the very end of that line as opposed to direct at the GFCI output. (unless of course your GFCI is an attenuator as you have said.)

 

I did find that I had another 20A GFCI in the "parts box". I'll try running this one with the signal generator (likely this weekend).

 

Final note: Leviton was a rather high end provider of X10 modules. It's possible they designed their line of GFCI units with X10 in mind.

 

I would be very interested to hear your results of a bench test for comparison.

 

Then I think I would replace that Pass and Seymour

[ELA]

I have seen those Arc Fault Breakers being a problem also.

Again it was some models where fine and other where problems.

 

I would be very interested in hearing more details about Arc Fault Breakers being a problem. I am also quite familiar with one brand of AFCI (SquareD - Combination Arc Fault Current Interrupter).

I have dissected one and ran many experiments on it. I would expect it to be similar in terms of "Insteon signal attenuation"- to the GFCI ( = very minor).

 

If the reports are of Insteon signals causing a false trip of the AFCI, that would not surprise me as they are notorious for nuisance tripping (some brands bettrer than others).

 

I think the point I am trying to make is that the trail and error method of troubleshooting communication issues can lead to confusion as to the real culprit without a bench test for confirmation. I am very interested in IndyMikes GFCI behavior on the bench.

 

IM if you are reading .. I forgot to ask? Can I correctly assume you have tripped and reset the GFCI several times to attempt to "clean the contacts" ? These devices do have contacts inside that could conceivably add impedance. (A stretch I know )

[IndyMike]

IM if you are reading .. I forgot to ask? Can I correctly assume you have tripped and reset the GFCI several times to attempt to "clean the contacts" ? These devices do have contacts inside that could conceivably add impedance. (A stretch I know )

 

Not so much of a stretch - I have seen GFI's develop significant contact resistance due to nuisance trips.

 

I did check the voltage output prior to performing my tests: 121.9 Vrms fully loaded/122.3 Vrms unloaded. Cycled the CFI several times and re-tested - no change.

 

It does not appear that I was taking a drop at 60 Hz. I was taking a significant drop at 120 KHz. I'd estimate the total load on the circuit to be on the order of 1200 W.

 

I just realized that my signal generator will not drive any significant load at 130 Khz and my amplifier tops out around 45 Khz. What were you using to drive a 100 Ohm load at 130 K?

 

Short of borrowing an amp from work, I'll try a bit more sophisticated line powered test using the X10 transmitter and my scope. The Elk ESM1 is nice for relative testing, but the led bar graph defies any quantitative analysis.

 

Hopefully I'll be able to reproduce my previous results using the "box spare" GFI. If so, I'll disassemble it for some photo's.

[ELA]

IM,

Luckily my (work) signal generator only struggles a little at a 100 ohms load. You can tell it is being loaded but still maintains a 5V p-p output at 131Khz.

I ran a similar signal generator test on an Icon Switchlinc (as the load) & (100 ohms series Resistance to limit current) and got an expected ~44 ma demand at 5V p-p. My generator could not deliver the 200ma or so that might be required to drive the switchlinc alone.

 

Here is a picture of one of the GFCI units I worked on:

 

 

Notice the burn marks on the trip coil to the right That was a fun experiment.

[IndyMike]

ELA,

 

Ran some quick tests with my "box" Pass and Seymour GFI today that indicate it is not a problem. I injected X10 signals into the device and monitored the output with various loads. I did not see a significant drop across the unit (monitored my my ELK ESM1).

 

In reviewing my original configuration, I now realize that I was injecting signals through a "builder model" GFI. This is a service GFI located 6" from the panel. My destination GFI is 2' from the panel on the same leg (phase). It now appears that I'm taking the drop across the combination of the two GFI's. Since the "builder model" GFI was interfaced to the panel, I have no idea of the load - entire house reflected back to the GFI.

 

More investigation to be performed. I plan on pulling the builder model gfi and running the same tests. if a see a drop across the unit, I'll bring the scope in to quantify.

 

Notes:

1) why can't things ever be simple.

2) At the moment, it appears that I owe an apology to Pass and Seymor

[ELA]

IM,

I appreciate your follow up.

Troubleshooting these systems is complex and it is easy to get confounded. I know cause I do it all the time ( the get confounded part)

 

Insteon is in dire need of better diagnostic tools.

I have such great sympathies for people who experience communications issues and only have the trial and error method to rely on. I feel grateful to have some additional tools to be able to take a closer look at problem suspects.

[fitzpatri8]

I'm all for increasing the ease of installing and troubleshooting device control, but I still can't see how your project will be any improvement over the signal diagnostics already available in the ISY or HouseLinc 2. Of particular concern is that, if I understand the project correctly, you aren't trying to log data over an extended period of time.

 

Perhaps it would help if you stepped through how you would run a test and fix a communications problem your way as opposed to using the tools already built into HL2 or the ISY.

[ELA]

I'm all for increasing the ease of installing and troubleshooting device control, but I still can't see how your project will be any improvement over the signal diagnostics already available in the ISY or HouseLinc 2. Of particular concern is that, if I understand the project correctly, you aren't trying to log data over an extended period of time.

 

Perhaps it would help if you stepped through how you would run a test and fix a communications problem your way as opposed to using the tools already built into HL2 or the ISY.

 

fitzpari8,

I will respond in the appropriate thread, and where you first expressed your skepticism for "my project"

[IndyMike]

ELA,

 

It's been an interesting day of testing. I've had to abandon my Elk ESM1 in favor of the scope as the elk did not did not really provide quantitative information. From what I've seen so far, I have to change my original hypothesis - "modern GFI's" do not appear to present a series impedance proportional to the the current load. This may have been true at one time, but I have tested three different devices and they all appear to add insignificant impedance when presented with a simple load.

 

To be short, I had added simple loads up to 1000W to my GFI's and have not measured significant 120 KHz attenuation (my testing is with X10 transmitters). This agrees with your signal generator measurements.

 

Nonetheless, I do measure up to 15 db signal loss on a test circuit when using complex loads. While trying to rationalize this I realized that these complex loads can present a significant line to chassis capacitance that is insufficient to trip the GFI but does add load to the 120 KHz signal. These loads appear to imbalance the current transformer in the GFI and produce both phase and signal attenuation within the device.

 

As an extreme test, I added a .1 uf load to chassis on the output of my GFI. At 60Hz the GFI was happy (nominal 4.5 ma draw under the trip level). The circuit would trip during an X10 communication @ 3 Vrms (245 ma) with and attenuation around 42 db (measured prior to the trip).

 

The above is rather instructive in that I have been a part of forum posts involving X10/Insteon communication intermittently tripping GFI's/AFI's. Hind sight being 20/20, this appears to make sense now.

 

In short (again) this is not at all a simple problem. It involves complex impedance's and power to chassis capacitance's that are not readily characterized. I do not have the equipment to perform a line to chassis impedance test at 130 Khz. All I can do at this point is fabricate rough test cases to bracket the problem. My test cases so far indicate that the GFI can introduce both phase shift and signal attenuation when presented with a complex load.

 

If you have any further ideas on how to characterize these load conditions, I'm all ears.

 

IM

 

Edit: IEEE article dealing with common mode power nuisance trips of GFCI's (and capacitance to chassis). What I am missing is the GFCI circuit and the possible 130KHz loading due to differential current flow.

 

http://www.jonathankimball.com/pdf/gfci_ecce.pdf

[ELA]

IM,

I appreciate your follow ups and dedication to testing for results.

 

From extensive testing I have done with AFCI's it does not surprise me that current sensing cores designed for 60hz are much more sensitive at 131Khz.

 

There is a specification for GFCI's that dictates the required maximum trip response times. I believe most trip faster than the required time. Whether or not the GFCI would trip might then be related to the duration and/or repetition rate of the message.