Do the Sense CTs know the actual voltage on a per-device basis to calculate wattage based on the actual amps * actual volts per-device?
I am doing some generator planning and finally have an extremely accurate Fluke meter. I was taking current readings and comparing to Sense. Excluding inrush (which Sense does not sample often enough), I see differences higher than 10% in some cases.
While measuring my HVAC system, the sense is showing about 1,000W less than when I measure per-circuit with my clamp meter. I compared the main Sense power meter UI before / after the HVAC system started.
It occurred to me that Sense might extrapolating or averaging based on ALL power draw seen on the CT (i.e. the actual amperage readings per pole/leg).
Can someone clarify? It would be much appreciated.
@bill1,
The CTs on each leg supply current information, the taps on the 220V breaker supply the voltage. The Sense samples all four voltages and currents somewhere around 1M times per second, per the specs on Sense device and the underlying chip hardware (2 TI ADS7946s each sampling 14bits at 2MSPS). That’s likely much faster than the time resolution of a Fluke True-RMS meter. From what I can tell, most Flukes only update the display 4 times/sec at most.
The real question is what does Sense do with those 1MHz snapshots of power and energy.
I suspect some pre-processed version of the raw data is sent back to the mothership for machine learning
The mothership likely time averages energy data into half-second values that are presented in the app…
The app might also present near-instantaneous power data in the tags it shows in the power meter. The values in those tags sometimes exceed the the measurement of the Sense power meter by a fair amount and may represent the max inrush current for power on of motor and other inductive loads.
I don’t see how the Sense is sampling that often when it never displays even a fraction of the actual inrush in the UI. The Fluke 376 is showing 110 A for the inrush which is close to the nameplate LRA rating of the compressor. Sense never sees more than 20 A after many on/off cycles looking at the device level stats and the UI showing the total data for the CTs. This pattern repeats itself for other motors as well. It’s not unique to my heat pump.
I’m actually not that worried about the inrush measurements. I don’t expect the Sense to measure what takes a $400 meter to display. It’s a consumer grade product at a much lower price point. (The Sense’s measurements are similar to what I get with cheaper clamp meters and that’s fine IMHO.)
The reason for my post is trying to understand why the Sense is under representing the wattage values. This is why I was asking how it calculates the wattage. I can’t tell if the data it collects is inaccurate or if the calculations on the back end are inaccurate.
@bill1,
A few things you should think about when you use a current probe vs. Sense
The Fluke doesn’t measure current independent of voltage. You really only get a current measurement with no sense of phase with respect to voltage. So any power calculations you (or the meter) do are based on a assumptions about the exact voltage level with a phase angle of 0. Very likely with inrush current, which is highly inductive, the real power is much less than the apparent power since the power factor is far less than 1 for the first 100-500msec. Remember that 1 AC cycle is about 17ms.
Know your meter - Take a closer look at the inrush current measurement specs for the Fluke 376. For the 381, the spec is:
“The Inrush function takes approximately 400 samples over a 100 ms period (slightly more than 5 AC cycles) and calculates the starting current envelope.” That’s only a sample every 250 microseconds, about 200x slower than the Sense. The big issue is that the sense is displaying average power on the Power Meter for a 1/2 sec (500ms) period.
The voltage drop upon inrush is not so significant that the true current is only 15 A. It’s a 3.5 ton heat pump compressor. Sense outputting a value of 15 A isn’t logical. It’s obviously wrong and you don’t have to do any math to come to that conclusion.
It really doesn’t benefit anyone if the Sense takes samples millions of times per second when nothing useful happens with the data. I can understand the need to restrict the UI to display data at a half second interval. But, Sense’s software isn’t displaying the maximum value read within 500ms or even the average value within 500ms, from what I can tell. The software behaves like a cheap $25 Chinese meter you can buy on Amazon. I think it’s displaying an absolute value at the 500ms mark, which is worthless for understanding inrush.
Again, I don’t care about the inrush nor do I expect Sense to be accurate in this case. My problem and reason for the post is to come to some understanding why ALL wattage representations seem to be off a bit. I am looking for someone who can explain how the wattage values are calculated with respect to different voltages (which is where I assume something is going wrong).
Sorry I got caught up in the discussion of inrush, when your real concern seems to be steady state measurements before and after the AC unit turns on. In steady state, I would still question your Fluke methodology for computing real (Vs. apparent) power with a highly inductive load. If you are simply multiplying RMS voltage (based on 120VAC) by RMS current obtained from the meter, you would be calculating apparent power rather than real power. Sense delivers real power measurements by virtue of its high sample speed. And real power is what you end up paying your power company for.
One more thought on inrush, Bill. I’m fairly certain that Sense is providing some insight into inrush once a device is identified. You can see in the photo below that Sense spots a 1051W spike from my Wine Fringe even though the 1/2 second average trace only shows a 460W spike based on the 1480W scale. My real questions are:
What is the Sense sample period for that 1051W value ? 1 60Hz cycle ? 100ms like the Fluke ?
What inrush sampling period is best for sizing generators ?
