Since our high winds and storms the last couple weeks, I’ve been noticing my lights flickering a fair bit. No full-on outages, but it has been enough to briefly trigger some UPSs, although AFAICT not long enough to shut down my Enphase microinverters.
Is there any way to see the measured mains voltage at high resolution? Perhaps through the API? I’m hoping to see the dropouts (which are very short, milliseconds to tens of milliseconds).
Update: Oh! I just saw a power quality report in the Labs tab. That might be enough. Alas, it’s not showing any of the substantial number of dips I had a couple hours ago. Maybe it takes a while to be analyzed and posted? It definitely shows me having more than “most Sense users” (112x)!
Labs Power Quality runs on a delay. Typically refreshing every 72hrs. Sometimes it can be a week behind and other times about 24hrs.
You can see your real time voltage Settings → Sense Monitor, under signals.
All of us would like the power quality graph to be real time or near real time for situations like this. It’s been one of the most requested features over the past year or two. For some reason the decision makers at Sense have largely ignored our requests, which is strange because one would think this feature would be really useful to the utility segment.
Glad to know I’m not the only one. Unfortunately, based on what I read in the cited post, power quality monitoring really won’t give me what I’m looking for (although in my case it is showing something concerning: my L2 voltage has dropped into the 80s and 90s several times in the past 30 days; definitely something to tell the Utility when they come out to check on things).
I’d really like to be able to see the glitches that correspond to my lighting flicker. I suppose I can build a little device to do this with parts I have lying around. Super fun project, but a distraction from other, more important projects (lol).
I’m going to go back and refine Julia’s answer since you seem to be looking for very short term, nearly instantaneous variations from the norm. The 7 second rolling median measurement is for the “sustained” logging on top, on the right of the waveform (or so says the info bubble).
Please note that only 1 or 2 of these measurements are responsible for the logging event (excursion outside of +/- 10%). In my case, it is minimums below 108 volts on both legs that drove the logging. Notice also that two of the events happen within the same second, so I’m not sure why that occurs - two samples space narrowly in time, or two separate leg events.
If I follow the double event back in time, to 14:18:08 Pacific Time (vs Sense UTC time - Z), I can see the dip that occurs just a little while before we had a MAJOR power outage.
So going back to the original question. It looks like the top logs largest median sustained upward and downward excursions from 120V over 7 seconds. The bottom logs the number of “instantaneous” excursions beyond +/- 10% using a time period that is longer than a 60Hz cycle (because RMS voltage measurement requires 1 cycle or more). I’m guessing the time period for “instantaneous” measurements is the Sense base period of 1/2 second, but I don’t know for sure. @JuliaAtSense , can you get info on the sample period for detection in the bottom “instantaneous” dip / spike detection as well ?
I took a look at the CSV data yesterday, but Numbers.app on macOS made it too hard to convert the timestamp into a real date/time value, so I gave up. Since the data was only recorded at 1-second intervals, and it had that double reporting you mentioned, I was wary of its quality and didn’t want to put in a lot of effort right then.
The slow data presented on my page is interesting:
That DateTime format with the Z is a fairly standard output for time series data - It’s really just UTC time with decimal seconds, though from what I have seen of Sense Power Quality CSVs, none of the timestamps includes the decimal seconds. But that may mean also mean truncation that is independent formatting. Sense might have a more exact time for the event, but truncate / round. I wouldn’t necessary read it as now second samples.
In that same vein, the detection period has to be 17msec or greater because there is no consistent way to compute RMS (AC) voltage without seeing a full 60Hz cycle.
If you want to post or PM me the raw CSV, I can do a quick analysis run like the one below. I find this to be the most useful for seeing what’s going on with the power events - It shows what combos of L1 and L2 out-of-bounds are occurring. I haven’t seen a single case where either L0 or L1 have an event that have the leg both above and below the trigger range.
Yeah, I’m still discovering the challenges of Numbers - I want to like it, but keep on falling back on Excel and its infinite complexity (and capability) all too frequently.
Here’s a little fun - a gallery of User relative power quality issues. You are User 4 in lower right. The color represents the number of spikes/dips in each “quadrant”. Some users have a problem that has recognizable “fingerprint” (User 2). But, some have issues that move around (User 1 runs their own grid).
One more way to visualize the data - as a scatter plot vs. UTC Time of Day (not your local timezone) of the L0 and L1 “deviants”, the “instantaneous” voltages that have deviated the most from the 120V nominal, be the deviation high or low. You’re still the chart on the lower right. The square in the center is where all the points should reside.
Given the new 1 second RMS sampling period input from @JuliaAtSense (thanks !), the detection curve for triggering a Power Quality event would look something like this. Sense wouldn’t necessarily need to see a voltage below the 110V threshold for a full second to log it. A very low dip to near zero would trigger the detection if it lasted about 84msec. The second detection voltage curve for swells/spikes end also runs upward from 130V and is limited on the time side by whatever the max voltage Sense is able to measure.