From time to time, we seem to experience a power "glitch", once a day, at roughly the same time -- usually ~3AM. (but, not every day... just "periods" when it manifests followed by periods where it is completely absent).
It's not a problem, for the most part, as everything is on UPSs, here (the microwave oven seems to complain the most as it isn't on a UPS and its damn clock often resets -- I long for the day when appliances have synchronized clocks or NO clocks!!!)
I assume this is some sort of switching transient that affects the entire city (?) -- or, at least large portions of it.
(our services are below grade so not likely caused by something physically interfering with the transmission lines)
I'm turning my attention to the design of the power systems for my current project and figure it would be prudent to put some line-monitoring capabilities into it (if only to let it anticipate such problems and plan ahead).
So, the questions are:
- how often to sample (to be able to catch transient events)
- maximum peak likely to be encountered
[Of course, I have to anticipate what the power conditions are likely to be in other parts of the market (US consumer and, separately, commercial/industrial) and not just rely on my own observations.]
I'm tempted to buy a line monitor just to see what they've done (in terms of hardware interface; the signal processing software won't be a problem). Recommendations? (again, two/three different markets, as above)
That sounds like somebody switching a generating set on or off. It is a process that can generate transients.
The famous Telsa grid scale battery in South Australia is reputed to be fast enough to cancel that sort of transient really fast (and half it's capacity is devoted to doing just that, which made it a really profitable investment.
The other source of that kind of transient is lightning strikes in the wrong place. They are less predictable.
We had repeated trips of the whole-house RCD every few days for months. Always at 7:58AM. ALWAYS. I turned off everything in the house that could know the time of day, and they still happened. Eventually I became certain that it was something to do with the power supply, and started harassing our provider.
They kept fobbing me off, saying I should replace the RCD - which at the time was brand-new, after a renovation where we got a new powerboard, and I told them that and insisted.
Eventually they admitted that 7:58AM is when they switch in some large PFC capacitors in preparation for inductive industrial loads coming alive. I told them that it was therefore their problem, and they arranged to send a couple of electricians around to fit a more tolerant RCD, at no cost to me.
When the sparkies came, they said they had been doing *hundreds* like this. And that's just for the households who figured it out, and complained long and hard enough to wear down the power company.
This was in 2000, before we started fitting RCD breakers on every circuit.
Just get a 'scope monitoring the line power supply. Point a video camera (preferably with continuous time display) at the screen and record a few minutes before and after 3am every day. You should be able to record the spike and the exact time it appears. Even a webcam on a laptop might do the job. The power supply company wouldn't be able to argue with that sort of evidence.
3am seems a strange time for transients. Some sort of fairly regular maintenance work at the generation station or distribution switching, perhaps?
On a sunny day (Tue, 29 Mar 2022 19:25:13 -0700) it happened Don Y snipped-for-privacy@foo.invalid wrote in <t20f2p$ob9$ snipped-for-privacy@dont-email.me:
The power network over here switches, sometimes 2 times a day, between networks it seems. This causes a short (usually less than a second) power dip. UPS takes care of that as far as computers go. Laptop has its on battery and will run much longer, For longer power failures (will keep watching satellite, radio, ham radio, etc) I have now a 250 AH 12 V lifepo4 battery and a 2 kW 12V to 230V pure sinewave converter. So if the UPS starts screaming when power goes for a longer time, then it takes a few minutes to switch the big stuff to that setup manually. Sure, the clock in the microwave will reset,,,, Happened a few weeks back, 2 hours no power, neighbor came to ask 'Have you no power too?' Well, yes and no, I was just switching to the big battery. Talk about clocks, last Sunday we went to summer time, had to adjust 13 clocks here Sure hope they get rid of that shit twice a year
The lifepo4 can easily power a microwave or cooking plate, or the TV for a whole evening..
Or better: use a low voltage transformer and a resistive attenuator, then capture the 1Vpp waveform on the LINE IN audio channel of a laptop. Capture and analyze using e.g. the free Audacity, it shows the waveform nicely. Much easier than trying to find a transient on a video (even when the scope did capture it - some scopes have a low acquisition frame rate and may miss a spike completely).
I'm not particularly concerned with *my* experience:
It's not a problem, for the most part, as everything is on UPSs, here (the microwave oven seems to complain the most as it isn't on a UPS and its damn clock often resets -- I long for the day when appliances have synchronized clocks or NO clocks!!!)
Rather, I'm concerned with adding that monitoring capability to my product so *it* can learn/know what the "local" power conditions happen to be (it runs 24/7/365). This would help it decide how to adjust the load and schedule processes in anticipation of (or after detection of) outages.
For example, *real* outages, here, have been very infrequent. And, have have always been the result of equipment outages: cable segment failures, blown fuses in the neighborhood distribution station, etc. As these are relatively small events, from the perspective of the utility company, (unlike a storm knocking out power to a large portion of a town) they can be addressed with a small crew dispatched in short order. In each case, our outage has typically been less than ~6 hours (before they've rerouted power from "the other end" of the feed).
By contrast, when I lived in the midwest, we had outages pretty regularly (6 or more each year) and they persisted for longer periods.
So, siting my device *here* it could opt to maintain power for a larger portion of its componentry (it can selectively shed load) for a longer period of time and still hope to have spare backup capacity. In the midwest, it would have had to learn to impose deeper cuts as it expected longer outages.
It also lets the size of the (battery) backup be an independent parameter at the discretion of the owner instead of an implicit requirement of the design. A deployment with a bigger load looks like a nominal deployment with a smaller backup, etc.
[e.g., I have a ~2200W load that I can keep fully operational for ~2hrs. Someone else might not want to make that investment -- or might have a BIGGER load -- so the system needs to know how to make tradeoffs in scheduling its load]
Finally, being able to guesstimate the likelihood of an outage, based on past observations. For example, if power fails, returns and then fails, again, you'd be slow to reinstall load hoping instead, to let the power supply divert power to RE-charging instead of the load (which you expect to likely need to *shed*, again, RSN... when the followup outage strikes)
[This has been the pattern, here, with power coming back on-line. The crews will initially diagnose the problem, restore power -- and have the problem manifest, again, in short order ("Hey, Bill, the short is still there...")]
I'm hoping that by being able to *look* at the mains as they are powering the system, that it might be able to make some deductions about the possible cause (and duration) of the outage. Instead of being "surprised" by each new event!
Dunno. As I can't tell how widespread the problem is (it just this neighborhood? this side of town? etc.) or the system elements that lie between me and the "problem", I can't guess as to its cause. Is it something related to some other "nearby" power consumer and not the supplier? <shrug>
The UPSs all claim it to be "excessive rate of change" (or something to that effect... not "dropped cycle" or "low line" or...).
In parts of town with overhead distribution (which likely includes the high tension feeding us... somewhere), I can see large "switches" (but no idea as to how -- or why -- they are actuated/opened)
Again, the cause is likely out of my control (unless a local fault). I just want to be able to observe it in more detail than "gee, the lights flickered" or "ooops! power is out!"
Hence the question as to what the appropriate capabilities of that data acquisition (sub)system...
A transformer will have poor high frequency response and will filter out sharp spikes. Better to use a 100:1 capacitive/resistive divider across the AC line voltage. Verify the LINE IN audio channel is 600 ohms, then add a
600k across the top capacitor.
The neutral doesn't have to connect to the chassis of the computer. If you do, it will likely blow any GFI breaker that is on the ac line. Use GND for the return.
But are sharp spikes likely to be a problem? Any product which has been through regulatory testing will have demonstrated a good immunity to large short spikes, so it is much more likely that dropouts will cause problems. A small transformer will respond to frequencies of many kHz, so that should be enough. What will prevent it from measuring large spikes is core saturation but there are several ways of minimising the effects of this that have been discussed here in the past. Most products, especially those with switched mode power supplies, should survive dropouts of 10ms or less without any problems due to energy storage in their input capacitors.
You've missed the point of (my!) post. The goal isn't to "protect" the device -- OBSERVING the line won't do anything to make the device more resilient!
Rather, the goal is to try to understand the nature of the disturbance in light of other, "previously recorded (quantified)" disturbances and use that to predict what is likely to happen with the mains, "soon".
E.g., when our cable segments fail (each transformer is daisy-chained to the next transformer down the line -- unlike "taps" on an overhead distribution line), the center/current carrying conductor (it's a coax) shorts to the "shield" (neutral/ground). As such, you would expect to "see" this arcing event instead of a "clean" open.
You can imagine that power won't be coming back anytime soon in the event of such an observation.
A drunk slamming into a pole (overhead supply) would likely exhibit a different sort of "failure event".
A brownout, still different.
Etc. Unless you can claim that those "details" carry no information, then one would want to preserve as much data as possible.
[I can detect an outage just by counting 60Hz cycles and "timing out" in ~20ms; doesn't tell me anything other than "power is out!"]
I'm not limited to just observations of the mains but can also factor in
*other* observations (e.g., it is summer and ACbrrr loading, on the network, is high; or, it is winter and an ice-coated branch may have taken out an overhead line)
OK. I can see why measuring disturbances that won't affect your devices could be useful. The question then becomes one of how much measurement bandwidth and resolution is necessary to get this unknown information and how best to acquire it and process it. As you don't know what you are looking for about the only bandwidth constraint is going to be the bandwidth of the power transmission system subject to avoiding frequencies high enough to have long-wave broadcast transmissions. That suggests an upper frequency limit of maybe tens of kHz. Of course, arcing in nearby wiring could have much higher frequency components. Maybe its worth looking at how arc fault interrupters detect arcing? John
There are a fair number of publications (research) that describe faults in the transmission network (because the utilities have an interest in minimizing these). How those are reflected *into* a "point of observation" that *I* would easily monitor is an issue. (telling me what's happening on the high-tension side of my local feed isn't the same thing as telling me what I will see "locally")
But, that ignores:
...as well as other failures, noise sources, etc. E.g., a fault in a large load that eventually takes out the local branch circuit can be just as "bad" for your source of power as a system-wide power failure.
Tens of Hz is enough to (quickly) detect *outages*. Will tens of KHz be sufficient to differentiate between different line faults and transients? How much wire do you assume separates the observation from the fault?? <shrug>
[And, we've not addressed the magnitude of the transients... 10X? log scale?]
That's a good idea. But, again, they are only looking for arcing and not the general case of "disturbances"/line quality. E.g., in an industrial setting, I'd expect the mains to be a lot dirtier than in a domestic setting. So, you'd have to "learn" when that "dirt" is significant and when it just has to be tolerated.
And, they are obviously targeting "cheap" detection methods; I'm not so constrained.
I'd like to use a laptop for a control application in an unattended location, instead of a big PC and a UPS. But a PC bios can usually be set up to restart the PC after a hard power failure, and it seems like laptops can't do that for some reason. I'd love to find one that does.
It's relatively important to have an average-AC-voltage measure, because some wiring issues can be diagnosed that way; you'd want to sample rectified/filtered V at tenth second or so, but only log 'events', not keep all the data. There's a few percent variance allowable, of course.
Load changes and loose connections or lightning strikes, or diurnal overvoltage because of xformer tap decisions could show up. Sites vary, but expect 2kV for miliseconds, in case of lightning.
More important surges or loads will cause zero-cross timing shifts which tie into odd clunks from motors, so I'd think to log also any deviation from 'normal' 120 (or 100) zero crossings per second. Maybe store a few dozen cycles before and after an 'event', gives you an idea of the transient. Motor starts without zero-voltage-switching have dimmed my lights and several times rebooted a Linux box exactly when I hear the buzzsaw.
I need a Windows machine. If power fails for a full day maybe, a laptop will run out of battery power and shut down, but can't be set up to restart when power comes back. At least none that I can find.
Strange.
Maybe some laptops can be set up to deep-sleep when batteries get low, but wake up when power comes back. If one could ride out a few days that way on remaining battery power, it would be OK.
But those would be "perform once" types of calculations; the wiring isn't likely to change "dynamically".
OK. I assume we're still speaking of consumer deployments...
Hmmm... I hadn't thought of keeping much "context".
The FNET folks sample at ~1.5KHz (I don't recall the precision) and try to detect anomalies in the grid by coordinating observations from geographically dispersed measurement stations. I assume some of their algorithms could be applied to local "single observation" points.
As long as the "disturbance/transient" doesn't affect (or predict!) power availability (as seen at the output of MY power supply), I'm likely not too concerned with it. I can carry most of my (distributed) loads over short durations, regardless of the investment a particular site may have made in "backup" capability.
I'd have to be more concerned about sites that power individual nodes "locally" from PSUs that aren't as robust. I will know that this is happening (*has* happened) and can take some measures to prevent loss of functionality. E.g., don't dispatch processes to nodes that may not have reliable power at a particular site. Or, force the OS to checkpoint every process running on those nodes (which adds overhead).
Ideally, I'd not want to take any prophylactic measures unless needed as they translate into "overhead"/waste.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.