I have had sense over a year and almost all of my lights (that I have are Leviton Vizia Plus model switches running LED bulbs, most of those also have Zwave built in them (18) and are connected to a zwave hub (Smartthings) which makes them recognized in Alexa as devices thru Alexa’s integration with ST. When you turn these dimmers on, they come on slowly and power off the same way. I assumed that when I enabled alexa and network listening in sense that, I would pretty much use alexa to turn the items on and off a few times and sense would learn it. I also have ~ 8 outlets that are Zwave without the slow on/off but also have energy usage reporting built in it as well. But for some reason, my sense has seem to have pretty much found everything in the house except the Alexa devices and any light that has a slow on/ slow off dimmer. I think some people call them soft dimmers. Has sense found anyone’s soft dimmers or any device that is technically a virtual device in Alexa? Strange thing is that its found my fridge light and, dryer light and the light in my garage door opener.
That’s an interesting point about “manually” controlled devices being detected versus “virtual devices” (the same device, automated) not being detected.
Detection requires repeated use that generates a similar electrical signature. Flipping a panel breaker is not the same as flipping a wall switch and not the same as a smart switch and definitely not the same as rotating or sliding a manual dimmer at different speeds.
Detection is made more difficult when ons/offs are “ramped” (i.e. a slow dim on/off) and when those ramps may vary in length due to different modes. Sense is good at recognizing significant changes over the background “noise”. A ramping electrical signature is inherently more difficult to detect in a short amount of time because it’s stretching the amount of time over which the detection event is occurring. Technically, you move from being able to detect the event based on a model solely in Sense’s DSP to needing to looking at longer and more complex phenomena (patterns) that are only visible in stored data, and then need to be used to generate a more complex DSP model, if the DSP is even capable of seeing the longer event. This implies that longer (temporal) phenomena will take longer to develop for.
I’m pretty sure network detection (NDI) only works on smart TV protocols from a couple of makers. But even if Sense could detect the SmartThings network commands it wouldn’t be enough to “learn” your networked lighting devices easily for a couple of reasons:
- the reasons @ixu details above - mainly the slow ramp “hides” the on/off transitions
- zwave / Zigbee can have variable latency so network commands can be further separated in time from switching time.