Other Bubble Be Gone!

[MrMark]

I’m late to the discussion, but IMHO, the “Other” bubble is WAY more useful than the “Always On” bubble! Logically, “Always On” shouldn’t even be on the “Now” screen. By definition, “now” means: at the present time or moment. The “Always On” bubble is (supposed to represent an average over multiple days.

On the other hand, the “Other” bubble is (supposed to be) a real-time representation of the total present power consumption, minus the detected devices. Logically, the “Now” screen should include only the detected devices and “Other”, and all of those bubbles should add up to the wattage number in the upper-right corner.

Does this make sense to anyone other than me?

[kevin1]

I actually see the Always On bubble as useful for a good reason. If we are really applying “truth in advertising” to the labeling of things like the “now” in the Now visualization, we would need to call Sense native detections as “dynamic component of device”. That means that those detections will only ever show the dynamic (non-always-on) component of that device’s usage, because of the way Sense does native detections. Always On is important since the Always On component is left out of all of the native Sense detections, plus all the devices Sense never finds (many of which use mostly Always On power)

The Always On bubble and device is a useful diagnostic tool for reducing this collection of components across many, many devices in your house.

[MrMark]

Perhaps “Always On” is useful to someone, but it still has very little to do with “Now”. And when the bubbles don’t add up to the total current wattage, then that begs the question, where’s the missing wattage?

The Sense software is not at all consistent, but when it is showing an “Other” bubble (and no “Always On” bubble), the bubbles add up to the total. When the “Other” bubble is replaced by the “Always On” bubble, the bubbles never add up to the same number as the total wattage. In this example, the bubbles total 11,701 watts, but the actual total is 8,216:

As you can see, 3,485 watts are lost in the wash, and thus are not tracked anywhere within the app, and of course the “Now” screen is fundamentally inaccurate.

[kevin1]

And you don’t think that having that huge Always On isn’t something that you should take aim at ? It looks like you have two devices on DCM that have huge Always On components - that’s kind of an unusual situation, but shows an important difference between DCM and smartplug devices vs native detctions.

• Native detections are guaranteed NOT to include Always On components.
• DCM and smart plugs do, by their very nature, do include Always On component

For you, the smart thing for Sense might be to subtract off the known device Always On components from the whole house Always On so they are not double counted. I would like to see the same enhancement.

[MrMark]

I can’t reduce those “huge” SCV loads to any great extent, so there is nothing to “take aim at”…they are what they are. I have them on Flex sensors because that portion of the power cost is a tax write-off for me. In reality, they are not actually “Always On”, but Sense misidentifies them as such, so there’s nothing that I can do about that either.

But that doesn’t mean that I don’t want to be able to monitor and track the 3,485 watts that Sense is not showing or recording. Even if those watts were “Other” instead of “Always On”, at least that broad category would have a history, whereas “Always On” does not.

I don’t mind that Sense wants to guess at always on loads, but I hate that it displaces the “Other” bubble and corrupts the “Now” screen. And FWIW, even before I had the SCV loads and the Flex sensors, “Always On” was always incorrect. It was a much smaller number, and a much smaller percentage of the total, so it didn’t seem so obtrusive and out of place on the “Now” screen, but it was still wrong. In those days, I always had an “Other” bubble, so at least the bubbles added up to the total wattage, whereas now they do not.

If one has a few hundred watts of “Always On”, it can almost be fun, kind of like a scavenger hunt, trying to track down those elusive loads. Unplug a DVR and then wait a few days in breathless anticipation in hopes that Sense will show a 20-watt lower “Always On” number (yay!). But when an incorrectly-estimated non-real-time bubble dominates the “Now” screen, causing it to be plain old wrong, then its neither fun nor useful.

[kevin1]

Thanks for the additional info. It would be really interesting to see a previous 48 hour view of your whole house Power Meter, plus a previous 24 hour Power Meter view of both of the devices on the Flex Sensors, plus the Now at that moment in time. That would give a better picture of why Sense is seeing the Always On values for all three.

Always On for the whole house and for each flex sensor are based on formulas that use the time history of measurements for those devices so I’m not sure why you would say they are wrong.

How are Always On and Standby calculated?

[MrMark]

Well, here are the Now and Meter screens, but sadly there is no easy way to see power meters for individual device from a real computer.

The bubbles are wrong wrong because no combination of bubbles adds up to the displayed real-time wattage. There’s just that big “Always On” bubble, which has nothing to do with “Now”, and also fails as a substitute for “Other”.

It’s also wrong because if one backs out the SCV loads, the actual “Always On” is closer to 700 watts. That’s what Sense showed for the previous year. The “Now” screen above hints that it would be either 306 watts or 2580, both of which are wrong. By a lot. Nothing adds up to reality.

[kevin1]

Thanks for the screenshots - enough to confirm what I thought earlier.

• First off, you’re in a relatively rare situation where you have high use devices, both on Flex Sensors / DCM that both have very substantial Always On components, and almost no dynamic power usage Now (per the numbers on the two SCV bubbles).
• I’m guessing from the numbers that you have at least one more device that has a very large Always On component since your two SCVs only account for about 4900W of Always On. I’m thinking there is probably another 2000-2400W of Always On stemming from another device.
• The reason your bubbles add up to more than current usage is pretty simple - the Always On usage for the SCV devices is being double counted. That is a current shortcoming of the current Sense accounting, but not such a big deal for most devices on smart plugs and DCM.
• So again, the Always On numbers aren’t wrong, but they are being double counted. I’m sure if you include the app Power Meter view of the SCV devices, it will be clear that they are almost entirely Always On, at least at the moment in time you captured.

[MrMark]

@kevin1

First off, you’re in a relatively rare situation

So Sense people have told me me. And yet I know a lot of people who have similar setups, so I have to admit that I am surprised that it would be so rare among Sense users.

@kevin1

I’m guessing from the numbers

Yes, one can only guess.

@kevin1

the Always On usage for the SCV devices is being double counted

Surely I am not the only one who sees that as flawed.

@kevin1

the Always On numbers aren’t wrong, but they are being double counted

And for some reason, you do not see that as wrong?

@kevin1

it will be clear that they are almost entirely Always On

I know for a fact that they are not always on. Fair to say that they have various components, some portion of which is almost always on, but some things often on is not the same as “always on”.

And none of this addresses the fact that the bubbles don’t add up, nor does it address the large number of watts that are missing from the Now screen. It also blows off the fact that a two-day average has very little to do with “Now”.

If you’re not a Sense guy, you certainly do sound like one!

[kevin1]

No @MrMark, I’m just a user who has invested some time and energy in understanding how Sense works so I can get the most out of it. I try to share my knowledge with other users in the community. Always On components can be very confusing concept, especially since a lot of folks have preconceived ideas about what that means, but in Sense-land, it is defined by the house behavior over the last 48 hours and device behavior over the last 24 hours.

I did suggest in a previous post that Sense should fix the Always On double counting between smartplugs / DCM and the house Always On, so I’m the same page with you there.

It would be useful to see the 24hr device Power Meter charts for your SCV devices as well as a little bit more about them, if you are interested in understanding more.

[MrMark]

Sitting at a real computer with a four monitor setup, I find it incredibly sad that the web app is so inferior to the mobile app. Dinking around on my phone just seems like a massive waste of screen real estate and time. But griping aside, but here you go:

Not much to see really…at this zoom level, it’s pretty much impossible to see 5-minute interruptions, even though there are several on each circuit.

[kevin1]

@MrMark,

Thanks for the device Power Meter graphs.

This chart below explains why Sense sees your bottom graph as having an Always On near 3240, even though you might have one or even two 5 minute interruptions of 0W during the 24 hour period.

It’s a chart of the 1% bin calculation that Sense uses for the Always On values vs. the number of minutes (per 24 hours) of interruptions where the power usage is 0. I have run the calculation for 3 different distributions for the non-zero or non-interrupted minutes.

• blue - mean of 3241 with standard deviation of 10W (closest to your second plot)
• yellow - mean of 3241 with standard deviation of 100W
• red - mean of 3241 with standard deviation of 1000W

Bottom line is that with your current distribution of usage, you would need 15 minutes of interruptions before Sense would see your device as something other than Always On. And if there were 15 minutes or more of interruptions (presumably at, or near zero), your Always On component for that device would drop to 0W.

Again, you can debate whether the 1% bin is the right function, but the results are indeed explainable.

[MrMark]

Well, that’s interesting at a purely academic level, but no matter what the math is behind it, the “Now” screen is still off by thousands of watts. And not to be argumentative, but I had the whole thing shut down for half an hour one day, and that did nothing to the Always On number, so maybe that’s how it’s supposed to work, but that’s not how it’s actually working.

Sense is a cool idea, but in my experience over the past few years, the reality isn’t nearly as cool as the idea. There are too many system outages. The Sense device has awful Wi-Fi sensitivity on 2.4 GHz, and it can’t do 5 GHz at all. I had to install a dedicated access point just to keep the Sense device from locking up every time it lost the WiFi connection. In a place where any other device can manage 50 Mbps on 2.4 Sense also can’t do WPA3. There are gaps in the data, potentially related to the system outages. In my particular case, there are dozens of devices, amounting to thousands of watts, that are not being tracked anywhere in the app. The interface on a real computer is inferior to the mobile app, and the mobile app isn’t that great. It just seems like a crowd-funded project that never was quite finished.

Putting aside most of the problems above, if anyone were to ask me, I would tell them that the Now screen should be dedicated to things happening in the present tense. I would tell them to put “Always On” somewhere else (anywhere else!), and then make the rest of the bubbles add up to the current power consumption. Pretty simple concept, really: Now is now. I don’t object to the idea of an Always On calculation, as long as it’s not on the Now screen.

When I first got the Sense device, I had the feeling that Sense was a work in progress. I was really looking forward to all the great things to come. Two years later, I’m pretty well convinced that the work is no longer in progress, and my confidence in those great things to come is very low. I fear that what you see is all you’ll ever get. If Sense had developers actively working on the software, glitches would be fixed in hours or days. I mean, bad math is a simple fix. I can’t see that anything has ever been fixed. I used to do software development for CAD, and I quite often added multiple new features in a single day. Bugs were fixed in minutes, not days, weeks, months or years. That’s what active software development looks like.

Does this seen negative? I’m sorry, but I feel negative about it. We can go back and forth in this forum, and at some level I would enjoy the conversation with you, but I honestly don’t believe that anything that I say here is going to result in changes for the better. I’m just venting on you (which I don’t want to do) because there is no way to communicate with the development team, if indeed there actually is such a team. I noticed that the firmware version changed, but I can’t see that it made any difference. And there is no published changelog or release notes, so who knows?

[kevin1]

Thought I had offerer something more useful than an academic exercise - a way to get closer to your desired outcome. I’m fairly certain that if you drop your usage on one of those two DCM devices for a little more than 15 min you will see their Always On drop to 0. But that doesn’t fix your issue the second you turn your device back on. And the house Always On needs a drop to zero of a little over a half hour to hit zero for the next 47 1/2 hour.

I have added your request to the Wishlist. Feel free to like it if you really want a fix.

Avoid Double Counting of Always On between DCM/smartplugs and House Always On

Smartplugs and DCM measure both the dynamic and Always On components of the connected devices. The Always On component from these devices gets double counted since it is also part of the house level Always On. This can result in off-putting numbers in the Now bubble display. https://community.sense.com/t/other-bubble-be-gone/14987/23 Please subtract device Always Ons from the house Always On for a more accurate remaining Always On.

[MrMark]

I didn’t mean to be unkind, or whatever negative thing(s) I was. I do appreciate your efforts and the conversation.

@kevin1

a way to get closer to your desired outcome

I’m sorry - I must have missed that part…(?) Obviously I don’t want to have to power down my world for over half an hour every day just to fake Sense into showing the actual wattage. Was there something else?

@kevin1

Feel free to like it if you really want a fix

I do, so I did. Do you really think it will make a difference?

And in any case, thanks for talking to me.

[kevin1]

@MrMark, no unkindness taken. I get frustrated by things that don’t work the way they are supposed to, too. My feedback doesn’t fix the problem, but if you trust the calculations, you can redo the arithmetic for the Now display.

• Your Always On after compensating for the double counting is approximately
  7465 - (1636 + 3224) = 2605W (still big)
• Your Other with double counting for those devices removed is 7771 - 2605 = 5166W.

As for the Wishlist - not sure of the odds now that it is posted, but the odds are far lower if I don’t post.

[MrMark]

I understand what you’re saying, but there is also the issues that “Always On” is a relative concept. In my mind, nothing is always on, and precious few things, like router, Ether switches, phone system, and security system hardware, are even almost always on. I only wish I had an Other bubble. That seemed a lot more relevant and correct to me.

The “Other” bubble should be the sum of every device that Sense can’t detect. And sadly, it can’t detect a 24,000 BTU mini-split A/C, nor the variable speed fan on my central HVAC system, which are my biggest loads other than the SCV circuits. It also can’t figure out the (5) LED light fixtures in my garage, which amount to a total of around 400 watts. It’s surprising to me, because all of those fixtures are all identical, they have a rather unique startup signature and a very predictable and consistent run wattage. Seems right up Sense’s alley, but no.

In over two years, it has detected one set of incandescent lights, and one set of CFL’s, although the CFLs sneak by quite often, probably because their cold startup signature looks different than their warm one. It usually puts up a bubble for my 750 watt halogen floodlight, but it always mistakes it for something else like a dishwasher or dryer.

It can’t figure out my refrigerator, probably because of the inverter technology. It can’t figure out my washing machine, probably because of the variable speed motor. It hasn’t identified anything with a variable speed motor. It hasn’t detected any computer, printer, monitor or TV, even though some of those things get turned on every morning and off again every evening around the same times. And yet it gets tiny things, like my Makita battery charger.

I tried WEmo and Kasa smart plugs, but that worked out badly. It seems that neither can accurately measure high-efficiency computer power supplies. And both are unable to return to their previous state following a power failure, so after repeatedly wandering around my house to reset all of them on a day when the power was glitching all day long, I returned the lot of them for a refund. If Sense could have learned the devices via the smart plugs, it would have been worth putting up with them for a while, but that’s not how it works. One of those great things yet to come.

But I digress…yet again. Sorry.

[kevin1]

Can understand your frustration. I’ve been trying to post a better set of expectations around on this forum, based on experience and knowledge about how Sense works.

Forced Learning?

Tons of similar threads on this forum on that topic - the short answer is no, you can’t force learning and “teaching” Sense what a device looks like in isolation is counterproductive if you want it to recognize the device under all kinds of conditions in the house. If you want to understand more, read this: Realize also that with the current “native detection” there are certain classes of devices that are likely to be discovered and other that are far less likely. Sense is working on new ver…

We might see a broader range of supported devices as Sense rolls out progressive detection, but their initial scheme that relied on immediate detection of on/off clearly wasn’t going to fly for most electronics, variable speed motors (electronics) and inverter appliances (electronics).

I would differ with you on two points. All of my Kasa smart plugs maintain their on/off status when they are powered up again, so you might have had a bum unit. And they do measure power accurately with 1W resolution, so I’m not sure what level of accuracy you were looking for on your high-efficiency power supply. I monitor AppleTVs that have 1-3W of standby power usage very consistently. I can’t see why you would want to monitor power usage less than that in the household scheme of things unless you had hundreds of such devices, in which case I would gang them all onto a single Kasa device.

[MrMark]

@kevin1

We might see a broader range of supported devices as Sense rolls out progressive detection

Is that a thing that is likely to happen any time soon?

@kevin1

you might have had a bum unit

If so, then I had a bunch of bum units. And their support guys told me that there was no way to make it automatically return to the previous state. If it works for you, that seems great, but it didn’t work for me, and the manufacturer was unable/unwilling to help me make it work.

But even if I could have solved that problem, there is still the issue of them not being able to accurately read the consumption of the best computer power supplies. It worked okay with cheap ones, but not with “Platinum” or “Titanium” PSUs. This is important to me, because building efficient computers is part of my livelihood. And I also tend to have more computers running at any given time than the average guy, so it would be great to be able to see how much power those computers are using.

Both smart plugs brands had problems with any kind of reactive load. They metered restive loads, like incandescent light bulbs, almost exactly the same way as both Kill-A-Watt and Fluke meters, but anything with a motor or a switching power supply was dramatically different. I regret that I didn’t document it…in retrospect, it would have made a fun video for my YouTube channel.

I’m not all that concerned with tiny loads, although things like Apple TV boxes are surprisingly numerous, and they all contribute to that super-irritating “Always On” bubble, so it wouldn’t be a bad thing to be able to monitor them.

[kevin1]

Not sure when progressive detection is coming. I know Sense started ramping up on it in Feb 2021:

2021 Data Science Update

Once again I’m going to challenge your assertion about smartplug accuracy. The Kasa plugs are absolutely accurate at measuring Real Power, which is typically what residential customers are billed for. You seem to want to measure Apparent or Reactive Power, which are both legitimate measurements for other contexts. But just to be clear, Sense measures and itemizes Real Power, so it wouldn’t make sense to mix in Apparent Power for some items. Appreciate why you can’t use Kasa for some of your power supply needs.

In this explanation, the author uses the term “True Power” instead of Real Power, but this is one of the better explanations of each.