I had a chance to do another Sense experiment today. Yesterday, I replaced our Sharp under-counter microwave oven. The existing one was only about 10 years old, but seemed to take longer to heat things up and had a portion of the interior from panel that appeared to be melting. Because it had to fit in a fixed sized island opening without a faceplate, we had to replace with the same exact same model (according to my wife, Sharp has some patents on under counter models and manufactures all of them in the world, even for pricey brands like Viking). My first question after the replacement install was whether Sense would immediately detect the new microwave or whether another round of training would be needed. Any guesses ?
Here’s a hint - even though it’s a new microwave oven, it doesn’t have the new inverter-type power supply. The microwave working bits, from my tear-down, are old school, straight out out of the 1970s: a huge 15 pound high voltage dual coil transformer, an enormous capacitor, and the microwave magnetron. In fact, the only concession to modernity vs a 1940s microwave generator is the solid state diode on one leg of the capacitor.
ps: After a few normal heating cycles, I’m going to look at Power Meter waveforms before and after to see if there was any power usage differences before and after.
My guess would have been “Yes”, but for way less technical reasons. Sharp is a reputable manufacturer, affording proven and stable internal industry-standard electrical components.
Awaiting your outcome, regarding your comparative energy efficiency vs its 10-year old predecessor. Enjoy your new microwave!
I’m not sure I’ll be able to measure the relative efficiencies of the old vs. new, because I didn’t do experiments to see how much power the old microwave was delivering to water inside the microwave. People actually do experiments to see how many degrees their microwave warms 1 liter of water in x minutes at full output. Didn’t do that, but anecdotally, my wife thinks the old microwave was at only 3/4 of it’s original capability.
But I am interested in whether the power usage has changed. It’s going to be harder than than one might expect, for two reasons. First, Sense has it’s “trained” view on power level in the device view which doesn’t exactly correspond with real usage, but somewhat close. Second, the background noise in my house makes it hard to measure, especially for the old device, since it’s in pieces already.
Here’s the Sense Device Power Meter for a run of the old microwave.
Both of them show power usage of ~1830W. But they seem almost too close…
If I look at the new microwave’s power usage in the house Power Meter, I see something a little different. If I do the arithmetic between the base and the plateaus for these runs, I get a number around 1670W depending on exact measurement points. But if I sum the numbers coming out of the Sense added tags, I get an average of 1821W for on, and only 1609W for the off transition. I’m not going to question Sense’s math right now because I think those tags reflect a more detailed, multidimensional calculation (a complicated power ramp-up, and the difference between real and apparent power).
I have a second Sense coming but that is already spoken for, for another purpose. Getting my Traveller HS110 back in action (Kasa and Sense see it, but Sense hasn’t added to devices yet) and hoping that the real power draw is closer to 1.6KW. Looking at the spec sheet for this microwave drawer on the web, I see that it has slightly different specs:
And investigating the IEC test procedure, it appears that it is done with water, not popcorn (bummer)
The 14.5A would suggest a higher power input level of 1740W, closer to the number observed in the Sense tags and identification. But given the size of the input transformer and capacitor, I’m betting we are now into the slightly esoteric world of real vs apparent power. If I look at the schematic for the “drawer” and edit to see what the plug sees when the microwave oven relay kicks in, it looks like this:
Update: I still don’t have a Sense waveform for the new microwave, but the Kasa app with my Traveller HS110 tells me the power is about 1640W. It bopped around, a little below and a little above in multiple readings. So the Sense main Power Meter was closest to matching the HS110. BTW, the oven uses about 3W when off (standby)
OK, it took me a while to realize that my newly reconfigured Traveller HS110 had actually been added to the Sense devices. I had renamed the Traveller to “Traveller HS110” in Kasa, but Sense kept the old name, so I didn’t recognize it. I had to go into the the old device and erase the old name and then the new Kasa name percolated through.
Here are the results of Sense vs. HS110 for a basic full power microwave heating cycle.
Close, but 10% different. Gotta do some thinking and research on this one. I know that the sampling, and averaging/reporting are different between the two devices.
Sense samples in the 1 megahertz range, HS110 in the 2.5kHz range.
Sense Power Meter averages/reports on 1/2 second interval, smart-plugs on a 2 second interval (though my reporting is longer than default)
Not so clear how Sense derives “predicted waveforms” for detected devices. If I believe the data in the power meter tags, the Sense meter sees a larger power usage value coming out of the “on” transition. The electrical engineer in me tells me that maybe the Sense transition detection isn’t correcting for a 0.90 reactive power factor when the microwave first fires up.
One more comparison before I move the Traveller to my other “non-detected” microwave. Here’s the same Sharp microwave oven in Defrost mode, both via the Sense detection and HS110.
Sense detection - Same differences. Due to the differences in measurement approaches, I trust the HS110 more for actual power measurement, but trust Sense more for exact time resolution for ons and offs.
The HS110 is sampling data every second, but only reporting results of every 6th sample (most normal Sense installs, every 2) back to the Sense. When the defrost microwave cycle is only 18-20 sec (by listening to it), that makes the power cycles look uneven.
The HS110 is actually doing a “ground truth” measurement of the real power (vs. apparent power) consumed by the microwave. Sense is actually doing a machine learning prediction of power usage.
The HS110 also pick ups the small amount of power used by the oven fan and microwave circulator, as well as the 3W going to the controller, while the Sense does not. That is to be expected from such a small power user.
Now that I have my Traveller H110 in action, I’m going to look at my other microwave oven, an older KitchenAid, that was originally installed in the house as a built-in 20 years ago. We replaced it with the Sharp drawer in a remodel 10 years ago, but it still worked so it’s our second microwave - very useful.
Similar specs to the newer one that replaced it: 1000W output vs 950/1000W for the newer one. 13.0A max vs 14.5A max on the newer one.
But Sense has yet to identify this one. It’s used less frequently (about 1/3 the frequency), but is also the old school transformer type (no inverter here). If I look at my Traveller HS110, it looks about the same as the newer one in power level as well.
One more cycle - opening and closing then door with a 18W bulb. At first I though this microwave was using 21W in standby, then realized I had left the oven door open downstairs.
Why would the 10 year old unit be reliably detected for nearly forever, and the replacement unit drop in and be immediately detected, while an old but similarly-specced unit NEVER be identified/detected ? If I look at the main Power Meter for the on and off signature off the oldest unit, I can see some clues. Even though both microwave models hit about the same power usage on the Power Meter and from HS110 measurements (1630W), the oldest one (KitchenAid) has a several second power ramp at the start. The start-up might even look like several different different transitions to Sense, instead of one continuous one.
Just to put it in perspective, here’s another Power Meter screenshot at the same scale for the microwave (and its doppelgänger) that was identified correctly by Sense. At the half second resolution Sense displays, and the same scale, the ramp differences aren’t as apparent as I originally thought, and there really are two transitions tagged in the Power Meter for both “on” ramps. It would be very interesting to see both these microwaves in the higher resolution and many dimensional “feature space” that Sense uses to identify devices, to discern what leads one to be identified, but not the other.
