Phantom usage being reported on unused channels

[wci68]

Working with one GEM:

Moving the offset to -3 made the issue much worse. I tried +3 and the issue remained, perhaps a little worse.

I tried every other value between -3 through +5 inclusive, and 0 seemed to be the best with all showing the same phantom consumption over time.


My COM firmware is 4.0 (both GEMs)
My ENG firmware is 1.49 (both GEMs)
The Current Constant on the Adv tab shows 221 (both GEMs)
The PT type shows as 238 (both GEMs)
I have all channels set to Micro-50, except for the two channels for my service which are S-200

Watching the Live Data tab shows that over time one or a few channels may randomly show 0.02A for one refresh cycle. One or a few others will do the same a few cycles later. This will continue, but over a period of time all channels slowly increase equally at least initially.

I left one of the two GEMs untouched and over the course of two days:
1) The unused channels did not continue to increase equally over time. Some were only a few Watt-Hours and others were nearly 20 Watt-Hours.
2) The sum of all the channels for the branch circuits, including the unused channels, was quite a bit higher than the channel for the corresponding service leg (over 300 watt-hour difference).

I did notice that GEM had its offset already at -3 so I set it back to 0 (I made a note of this if I need to return to this value). I did a 'Reset to Factory Settings' on both GEMs, reset my CT channel types, rebooted the ENG, rebooted the COM, and a 'Reset All Counters'. I did this on both GEMs and will let them be for a bit to see how they behave.

[ben]

Can you watch the real-time and see if something spikes on when the amps happen?

2) I'd expect this to be a little off, especially if you're using Micro-40s/80s. Instead of dealing with 1 CT error/channel error, you're dealing with all the error from a group of channels.

[wci68]

I've left both GEMs untouched for 12 hours and so far the results are puzzling.

The "phantom" usage does not happen in spurts - it is pretty continual appearing on different channels over time. I have been watching real time and do see one to three channels show 0.02A for one refresh cycle. A short while later another one to three channels may show this. And so on. The channels showing this appear completely random, or at least I have not been able to discern any pattern.

I still have the 1,4 jumpers installed in all of the unused CT channels.

Reminder that each GEM has the service leg and all branch circuits on that same leg. The power supply and PT for that GEM is connected to that same leg.

GEM 2 appears to have somewhat consistent increase in "usage" on all the unused channels. After 12 hours, the range is between 1.24 and 1.69 watt-hours, with one exception (noted below).

GEM 1 is truly puzzling. Of the unused channels, every other channel differs drastically: Half (every other channel) has very low phantom usage (0.30 watt-hour, pretty consistently) while the other half (every other channel) shows consistent usage much higher between roughly 1.52 and 1.97 watt-hours. Again with one exception.

Channel 32 on both GEMs seems to be an outlier - being an unused channel in both cases. Its usage is reported as being over 3 watt-hours in both cases.


For the difference in usage between the service legs and the cumulative branch circuits I would expect some difference yes, but this seems to be higher than expected. I'll be doing a little better analysis after the GEMs have been running a while longer. I would suspect "calibration" may help reduce this but for the moment I really would like to resolve the issue with the phantom usage on the unused channels.

I may have access to a scope this weekend and will see what I find. Are there any good test points on the GEM, such as for the PT output and power supply output? I'd like to check these while connected after checking their outputs isolated.

I did notice that my legs are not balanced - somehow I put two high-usage circuits on the same leg but that will be easily fixed hopefully this weekend. That may help with the voltage imbalance I am seeing between the legs (right now about 4 volt differential). I'll then compare the GEM values with those from my DMM and tune if needed.

[ben]

We've looked into it and we think there's a problem with the Live Data display. Ours shows a creep on the Watt-Hours, but when looking at the raw values they haven't changed.

We're working on a fix for that.

[wci68]

Hmm. That looks like it may well be the case.

Both GEMS have now been left alone for a full day.

On GEM 1 it does seem that the output data does agree with this thought. Using btmon I received this from the last seven (unused) channels:

Code: Select all

2016/02/05 19:33:39: Ch26:     -0.000003KWh (    0W)
2016/02/05 19:33:39: Ch27:      0.000000KWh (    0W)
2016/02/05 19:33:39: Ch28:     -0.000003KWh (    0W)
2016/02/05 19:33:39: Ch29:     -0.000000KWh (    0W)
2016/02/05 19:33:39: Ch30:     -0.000001KWh (    0W)
2016/02/05 19:33:39: Ch31:     -0.000000KWh (    0W)
2016/02/05 19:33:39: Ch32:     -0.000001KWh (    0W)

This definitely does not match the GEM Live Display (the first column after the channel number above is the cumulative value).

On GEM 2 I see similar where the output data does not match the GEM Live Display, however there is still some slight phantom accumulation:

Code: Select all

2016/02/05 19:36:47: Ch25:     -0.000321KWh (    0W)
2016/02/05 19:36:47: Ch26:     -0.000247KWh (    0W)
2016/02/05 19:36:47: Ch27:     -0.000327KWh (    0W)
2016/02/05 19:36:47: Ch28:     -0.000264KWh (    0W)
2016/02/05 19:36:47: Ch29:     -0.000349KWh (    0W)
2016/02/05 19:36:47: Ch30:     -0.000232KWh (    0W)
2016/02/05 19:36:47: Ch31:     -0.000354KWh (    0W)
2016/02/05 19:36:47: Ch32:     -0.000224KWh (    0W)

This is less than a half watt-hour per day but is still 2 orders of magnitude higher than GEM 1. For this I won't be concerned but I'll still see if I can get it closer to the other GEM.

Thanks Ben for continuing to look into this!

[ben]

You might try the amp offset adjustment on GEM2 again to see if there's anything in there.

[wci68]

I first "calibrated' both PTs. The two were about 0.2 volt different in reading the same leg (for testing), and both read slightly low. Bumping the PT Type by 1 from 238 to 239 brought both closer to the value measured by a scope, a WattsUp!, and a Fluke 189.

Adjusting the amp offset did indeed make an improvement. Moving from 0 to +3 helped quite a bit on GEM2. Interestingly on GEM1 moving from 0 down to -2 seems to have helped even make improvement there. This GEM originally had an offset of -3, but I don't know if this could have been intentional.

[ben]

I first "calibrated' both PTs. The two were about 0.2 volt different in reading the same leg (for testing), and both read slightly low. Bumping the PT Type by 1 from 238 to 239 brought both closer to the value measured by a scope, a WattsUp!, and a Fluke 189.

Adjusting the amp offset did indeed make an improvement. Moving from 0 to +3 helped quite a bit on GEM2. Interestingly on GEM1 moving from 0 down to -2 seems to have helped even make improvement there. This GEM originally had an offset of -3, but I don't know if this could have been intentional.

We set the offset to whatever seems best here. There isn't really a method to determining the proper offset besides trial/error.

[wci68]

Makes sense - thanks Ben.

I have both GEMs pretty close with minimal phantom usage now. I do see a pattern between both units:

GEM1:

Code: Select all

2016/02/10 13:58:37: Ch26:      0.000016KWh (    0W)
2016/02/10 13:58:37: Ch27:      0.000008KWh (    0W)
2016/02/10 13:58:37: Ch28:      0.000029KWh (    0W)
2016/02/10 13:58:37: Ch29:      0.000005KWh (    0W)
2016/02/10 13:58:37: Ch30:      0.000014KWh (    0W)
2016/02/10 13:58:37: Ch31:      0.000007KWh (    0W)
2016/02/10 13:58:37: Ch32:      0.000063KWh (    0W)

GEM2:

Code: Select all

2016/02/10 13:59:41: Ch26:      0.000016KWh (    0W)
2016/02/10 13:59:41: Ch27:      0.000006KWh (    0W)
2016/02/10 13:59:41: Ch28:      0.000019KWh (    0W)
2016/02/10 13:59:41: Ch29:      0.000004KWh (    0W)
2016/02/10 13:59:41: Ch30:      0.000012KWh (    0W)
2016/02/10 13:59:41: Ch31:      0.000006KWh (    0W)
2016/02/10 13:59:41: Ch32:      0.000024KWh (    0W)

The 'even' numbered channels seem to increase at a higher rate than the 'odd' numbered channels. In particular the last channel (32) seems to increase the fastest. I had thought I noticed similar previously. I do still have jumpers (1-4) in all of these unused channels.

The values are all very small so I won't worry about them. As I add circuits I'll remember to clear the counters for the new channel first and I'll ignore the rest of the unused.