"Dirty/Noisy" Power Signature Devices


It’s been mentioned in other threads that certain devices have a dirty/noisy power signature and prevent sense from accurately identifying other devices. Plasma TVs have been mentioned as well as GeoThermal pumps. Are there other types of devices that create these noisy signatures?


@andy mentioned his well pump was identified as a culprit by tech support.


Yes indeed, and Sense engineering says that they can’t do anything meaningful with my home because the Franklin Electric constant pressure pump (its signature varies dynamically depending on how much water is being used) is so “noisy”. The pump is a synchronous motor and the controlled generates variable frequency power for the pump, in addition to variable power draw. So it’s a jittery line with lots of wobble riding on a variable magnitude base.

Visually it’s absolutely clear and identifiable, their chief architect picked it out immediately, but the algorithms aren’t designed for such behavior.

And, since it’s both my domestic water source and my 500’ deep geothermal well water source (which provides for heating and cooling, as well as domestic hot water), the well runs frequently enough to pretty much obscure my entire home.


Good info, thank you Andy. You’re the person I was referring to with the geothermal pump. I’d like to identify all categories of noisy power signatures if possible so people speaking out about their experiences is helpful. I’m curious if all variable pumps are considered noisy since they have dynamic signatures. Hopefully others speak up about their “noisy” devices so we can compile this information.


I want to be sure that the “geothermal” isn’t causing any confusion. This is a domestic well pump. Actually, this pump only incidentally provides water to my geothermal. Its primary function is domestic water.

That’s because we implemented a “standing column” (one deep well, instead of 2/3) geothermal, where the very same well provides domestic water and the geothermal heat pump water. Anyone with this kind of pump, regardless of geothermal, would have the same issues with Sense.

The manufacturer for our pump is Franklin Electric and the model is SubDrive 745. We love it, because it maintains the water pressure within about 2PSI and we don’t get any fluctuation at faucets, showerheads, etc, like most other well water systems. I don’t know if other manufacturers of constant pressure pumps would have the same kind of controller signature.

With assistance of the Sense data scientists, we determined that whenever our constant pressure pump is running, there is a 2.5-3 second ramp up, to 250 watt consumption at low flow rates and up to 750 watt at 15 GPM (which is what our geothermal needs), and sub second saw tooth oscillation in power consumption. That’s seriously confusing their analytics.

Visually, both on the Sense power monitoring and on my oscilloscope when connected to my WelServer, it’s pretty neat looking, particularly when you watch while adjusting water flow. Very distinctive to a human, disaster for Sense algorithms.


For purposes of this discussion, we might also want to adopt slightly different nomenclature:

  • signature = 500msec or so on and off signatures recognized by current Sense algorithms
  • SIGNATURE = full power envelope of a device from on-event through to off-event.

I would imagine that noisy devices have SIGNATUREs with the following characteristics.

  • On for a significant % of the time - obscuring other devices
  • Substantial power peaks - 500W or more, that would obscure other devices
  • Continued variability in both time and amplitude domain - no real steady state constant power consumption.

Though I would think that devices with some what predictable amplitude / time cycles would eventually be recognized and cancelled in the future, as Sense works on longer window (SIGNATURE) identification.


I have very variable power draw from two devices. The first is my desktop computer that has a range of power draws that can change by 80 watts (140 to 220) in a few seconds that depend on the tasks the cpu and graphics have to accomplish. There is essentially no pattern to the variable power level I can detect. The second device is the washing machine with it’s direct drive motor. Every movement of the direct drive creates a different power demand and while the pattern is repeatable, the power draws vary depending on the load and where the wash cycle is. I doubt either of these devices will be detected anytime soon. Fortunately, they are not always on and Sense has been successful at identifying many devices.


I tracked down noise on my line yesterday after turning off breakers and watching the power meter. It turns out to be my work laptop (newer dell) which is on a docking station and power supply (older unit). It seems like very small changes in load become spikes on the power meter. I’m going to see if I can get a new power supply that is less noisy.