How much Bandwidth does Sense really consume?

I looked a bit in the forum and could not find the answer to such question. In essence, the Sense is an Analog-to-Digital Converter that collects data on either 2 or 4 channels (Main CTs + 2 Dedicated/Solar) and then uploads such data to the cloud via WIFI.

So, I did some simplistic calculations, with the following inputs/assumptions:

  1. Sampling Rate - 4 million samples/second (4 MHz)
  2. Data Resolution (# Bits of Analog-Digital Converter) 14 Bits
  3. Number of Channels - 2 vs. 4 (main CTs only, or with DCMs/Solar)

The results are in the table below. Does this mean that in the best case scenario (Using only the 2 main CTs) the Sense consumes 112 of my 200 MBPS upload bandwidth? That sounds extreme.

Sense Bandwidth

However, I might be reading the specs wrong. When Sense advertizes they collect 4 million samples per second, do they mean 4M samples or 4M BITS per second? If that’s the case, then the used bandwidth is fixed at 4 MBPS (Million Bits Per Second), and the data from the (2 or 4) CTs is simply multiplexed. This also means that (possibly):

  1. The data for the 2 main CTs is more precise (sampled higher) than when one uses 2 additional CTs for DCMs … or …
  2. All 4 CT channels are always sampled at the same/fixed rate (1 MHz), and, when no extra CTs are used, 50% of the data is either discarded before being uploaded, or still uploaded and the Cloud takes care of ignoring it (waste of bandwidth).

Thoughts ?

I got some basic information from the web site

Hi @drjb. We don’t send all that data to the cloud. Heavy processing and compression happens on the monitor itself. There is no precise figure for bandwidth required - it depends on how active your house is.

Long discussion. On this topic here:

And my speculation on the contents of the data stream back to Sense here:

One more thought on sampling as well - there are actually 4 or 6 values sampled via the Sense monitor 2 voltages and 2 or 4 currents via CTs. The unless you are having bad spike problems that would be detected by the Sense Labs detectors, the voltage changes “slowly and predictably”.

Thank you gentlemen, very useful/interesting discussion.

I had totally forgotten about the voltage (Need voltage to calculate Power). It is thus only fair to assume/infer that the Sense unit does in fact do some calculation (and compression) before uploading the data. This, in contrast with some earlier comments I’ve read on here that the box itself had no ‘smarts’. Though I do not recall the thread I got that from.

Now, if in fact the Sense also samples the voltage (not just the current from the CTs), I assume it does so from the two wires supplying it with voltage. This then begs the question: In most NA installations there are 2× 120V legs, and the Sense is fed from one leg only. Would there be situation where the voltages from the legs are DIFFERENT? In such case, wouldn’t the Power calculations be correct on one leg, and approximate on the other? … Bear with me, I’m getting somewhere with this.

During installation, I added two breakers on the least populated of the two legs in my electric panel. Such leg feeds primarily the bedrooms upstairs, which during the day, draw very little electricity. The other leg (more populated) is the one with the more interesting devices (HVAC, furnace, washer/dryer, oven, … ). I am not sure what the logic was when the house was built, but it seems the contractor decided that the two legs should each feed a different level of the house. I would have opted for a more balanced approach to ensure both legs would supply similar power.

All of this, to ask this question: If one is after better prediction/detection, wouldn’t it be better to feed the Sense from the leg that has the most ‘interesting’ devices i.e., kitchen, laundry, HVAC, … ???

Also, some additional thoughts:

  1. Is there a way to look at the live time traces from current/voltage that the Sense picks up? Not suggesting a ‘hack’ of the device, but should be useful for the curious among us.

  2. After energy monitoring and knowing which device consumes how much, I’m also interested in internet bandwidth monitoring. Most tools I found work with the specific computer they’re installed on. My son suggested a custom firmware for the router I’m using, but I’m hopeful there are tools already that do that. Any thoughts on Bandwidth monitoring … when there are about ~90 devices ?

@kevin1 Good point about the voltage. However, since the Sense is powered by its own breakers on only ONE leg … does it mean that it samples only ONE voltage … not both? This brings the total to 5 channels (not 6) ?

Sense monitor powers itself off of one leg, but samples both legs. That’s why you need to hook it to a (hopefully) dedicated 240V breaker. You can see different 1/2 second RMS voltages and power/current on each leg in the monitor Signals section. My voltages are balanced right now, but they often vary by a volt or two.

I’m also going to point out that you really don’t want a live trace of AC current and and voltages following the 60Hz cycle. You want an RMS reading like Sense presents in the Power Meter. If you do want to see readings for the leg splits, look at the Signals section and just divide the RMS power readings by the RMS voltage to get the current. The current hack is to use the informal API, like this gent did:

And if you want to do real bandwidth monitoring, the easiest way is probably to replace all your networking gear with Ubiquiti gear. That’s my plan once they update their UniFi In-Wall HD Access Points handle WiFi 6.

Thank you @kevin1, always very useful and detailed answers.

You’re absolutely correct about the two legs. I initially assumed that in a typical breaker box, the breakers on the left were from one leg, and those on the right are from the other leg (Wrong, I’m on a learning curve here). Then, when I replaced my electrical panel few weeks back, I recall that, in the back, the breakers were connecting to alternating rails. This makes perfect sense (now) and this means that, whenever two neighboring breakers are used, they take the power from one leg each. This did not make sense back then, but now I understand it better. My explanation above, that the legs are left/right is not correct. In fact, moving down a collection of breakers, those breakers alternate taking power from each of the legs. Essentially, if one needs 240V, all that is needed is to use two neighboring breakers. I recall when I installed my Sense few weeks back, there were 3 wires that need to be connected (2× 120V phases and one neutral). It all makes sense now.

This started as a curiosity from my end, but in the end, it’s becoming a worthwhile learning experience. Thank you @kevin1 and all for keeping this alive/interesting.

1 Like

Note: If you’ve just joined this conversation, there are many incorrect statements in my posts … I call those my ‘learning curve’, though it is bumpy and got some dead-ends at times. I could be thinking out loud at times … hope is ok with the community.


I recently went from an collection of old Apple Wifi devices to tp-link Deco mesh setup (x60).
One feature is that the dynamic up/down of each client can be viewed … not the greatest for statistics, but does give feel for the data rates from each.

The Sense, just watching it now:
up: lowest: 7 Kbps, highest: 29 Kbps, typical: 15-25 Kbps
down: lowest: 6 Kbps, highest: 73 Kbps, typical: 33-53 Kbps

I was a bit surprised that there was more coming down than going up … but then I thought about it, and the “down” (to the Sense) includes the traffic from the smart plugs, not solely from the Internet!

The house has a bunch of smart plugs in use, one dedicated circuit, as well as ~20 ‘detected’ devices.

That “highest load” seems to be not much different that my iPhone uses while watching the Deco statistics (the average iPhone usage is much lower, because I don’t have anything going on other than watching the screen ). It’s far less than typical laptop usage.

1 Like