Hue indicator bulb

[jefflayman]

I have the Philips Hue system of light bulbs and hub. I had that system before I got Sense, so the fact that it integrates data into Sense was a bonus. A very small bonus, since my usage of all Hue bulbs comes to less than 1kWh per month. That is simply because the Hue bulbs I have are located where they don’t get much use, for example the front porch. I came up with a plan to get more mileage out of this integration and hopefully reduce my Other by identifying illumination demand using Hue bulbs.

To implement my plan, I ordered a couple additional Hue bulbs. I got plain white bulbs, which are 60 watt equivalent (800 lumen) LED bulbs which use 9 watts at full power. I plan to install these bulbs as replacements to 9 watt LED bulbs I already have. I will not use the remote control or dimming features of these smart bulbs, as I will continue to turn them off and on with wall switches. The purpose of these smart bulbs is merely to report to Sense when the light switch is turned on.

Hue bulbs are $15 each, so I recognize that this system is not a cost effective way to monitor usage. Here is the key idea, however, which makes my plan useful. I will place the new Hue bulbs in fixtures that have multiple bulbs. The smart bulb then becomes what I call an indicator bulb. If the indicator bulb is on and using 9 watts, then I will know that the other bulbs in that fixture are also on and are using their nameplate capacity.

I will install one of the bulbs in a ceiling fan which has three A19 bulbs. The other two bulbs in that fixture are also 9 watt LED bulbs. If this indicator bulb is ON, then I will know that instead of the 9 watts that Hue reports, I am actually accounting for 27 watts of usage. My indicator ratio is 3.00.

I will install the other indicator bulb in the basement. Illumination in the basement is two fluorescent-style LED fixtures rated at 48 watts each, plus a bare A19 bulb in an old-fashioned, ceiling-mounted, lamp socket. All three of these fixtures activate with one switch. When this indicator bulb is on, I will know that instead of the 9 watts that Hue reports, I am actually accounting for (9 + 48 + 48 = 105) watts of usage. My indicator ratio for this bulb is 11.67.

With each indicator bulb accounting for more than merely its own usage, I talked myself into spending the $15 per bulb. Now here comes the feature request. Could Sense add a user input field to the Manage tab of every Phillips Hue device which would accept numerical input of an indicator ratio? For the first bulb, I would input the value 3.00, while for the second, I would input 11.67. The other half of the feature request is for them to then multiply the value that Hue reports to Sense by the user input ratio before displaying the usage bubble on the Now tab. If the user does not input a value, the ratio would default to 1.00. The Hue bulb on my front porch would keep this default ratio, since it is not in a fixture with multiple bulbs.

[kevin1]

At the least, you can use data export to perform this multiplication of the indicator bulb usage.

[drjb]

Neat idea … however. From my experience, whenever I ‘mixed’ bulbs, I get slightly different colors, and to some, that is not very ‘appealing’. Unless of course you can adjust the Hue bulb to match the others’ color/temperature. Might as well get a colorimeter

[jefflayman]

@kevin1, I do plan to use data export and perform the multiplication myself. That will work for now, which is why I did not wait to purchase the new bulbs until after Sense takes me up on the wish list suggestion.

I had several people groups in mind when posting this idea. First is the Sense team, since they are always improving their product. If they like and implement the idea, then this concept would be convenient enough that people without the patience for data export could participate.

Second is forum users who have run up against the 20-device limit on smart plugs. I haven’t heard any limitation on the number of Hue light bulbs, so this idea could expand integrations for such people.

Third is anyone who wants to measure lighting. It has been noted that Sense will likely never detect an LED bulb because they don’t make a distinct electric signature. I have one native detection of a light, but that is the old-fashioned resistance filament bulb in my microwave (it was detected despite using only 26 watts!). I also monitor one LED lamp since there was an unused plug in my Kasa HS300. Other than this, all lighting was doomed to the oblivion of Other. With indicator bulbs, that need no longer be the case!

[jefflayman]

@drjb was right. My indicator bulbs arrived today and this picture shows what the 3-bulb fixture looks like with the indicator bulb installed.

Hue (the supplier of this smart bulb) calls light output from it Warm White. I’m not sure the name of the bulbs in my ceiling fan, but they look much cooler in comparison. I wasn’t happy with this appearance so I removed the indicator bulb from the ceiling fan and put back the original bulb.

If I were to purchase two regular LED bulbs in Warm White, then I could replace all the bulbs in my fixture. The result might be an acceptable color match, but drjb’s advice was fresh in my mind. Therefore, I located this new smart bulb in a single-bulb fixture that I use a lot. My indicator ratio will only be 1.00 for this bulb, which is a little disappointing. I can always come back to the idea of matching bulbs in the ceiling fan at some later date if I want more mileage from my smart bulb.

The basement lights are already a mixture of florescent-style bulbs plus a single bulb, so color difference there is irrelevant. At this location the bulb will function as an indicator to monitor usage, just as expected.

I turned on the wall switch and watched in Sense for my new smart bulb. It takes about 5 seconds for the bubble to appear. Turning the switch off, the bubble takes a full minute to disappear, although the integration reports status as N/A within 20 seconds. This feels clunky compared to Kasa smart plug integrations or native Sense devices. For perspective, however, many room lights are used for either minutes or hours at a time, so measurement delays in seconds are not numerically significant.

This post from last year confirms that Hue is not part of the 20-device limit. It says the Hue bridge can address at most 50 bulbs.

Question about Phillips Hue integration

A Hue bridge is limited to 50 bulbs and we can (currently) only integrate with a single Hue bridge. Expanding beyond this has not been prioritized as I can only think of one user who has expressed a need for additional bridge integration. I use Hue in my home, and did so before our integration. They’ll show in the Sense app exactly as you have them set up in the Hue app. So, they can be grouped as one room or broken out as individual bulbs. For instance, I have a ceiling fan with four bulbs i…

[jefflayman]

@drjb suggested a colorimeter to help match bulbs. He was kidding, but it can be fun to keep a joke going. The cheapest meter costs $200, which is too much to spend on a lark. There are apps for everything these days, so I kept looking. One web site said its author got results nearly equal to his colorimeter using a photography app pointed at something called a gray card. The app and the card together cost less than $20, so I was in.

Here is the photography app looking at the gray card under illumination from my Hue indicator bulb in its single-bulb fixture. Color temperature is 3000K, as highlighted with green. I tried holding the card at different angles and distances from the bulb. In almost all cases, the reading remained 3000K, though occasionally it switched to 3200K. Without the gray card, readings varied more, ranging from 2800K through 3400K. Apparently the app measures increments of 200K.

I tried the same thing at my ceiling fan. The three original bulbs register 3200K in almost every position. This shows that bulb color difference is barely noticeable using my crude technology, but that a detectable difference does exist. It also shows that the human eye is pretty discriminating. drjb was right again!

[drjb]

@jefflayman This is hilarious (in a good way of course)… If your profile didn’t say it, I would have guessed you’re an engineer as well (I am mechanical/electrical). I first started thinking about a colorimeter few months ago after Covid hit. The company I work for allowed/encouraged us to borrow desk monitors to work from home. So, I have 2 HP monitors from work (identical) and one Asus (my own) in the middle. The colors of course are NOT the same (just looking at the windows background). I started then exploring such colorimeters but I could not, in all honesty, justify the expense for something I’d rarely use. Plus, I do not think these monitors are fancy enough to adjust color temperature. I’d need to check further. Correction: The monitors do have color correction, but it’s not perfect.

I also explored doing a DIY Colorimeter with Arduino, but that required more ‘effort/time’ than I had available. Now however, with what you’ve done, you’ve just revived that interest. Going to check the Arduino forum for anything recent.

All the Best

[jefflayman]

My indicator bulbs are sometimes using 8w and sometimes 9w, which is unexpected. I run them at 100% brightness all the time, just as explained in the opening post. Does anyone have an explanation? Here is some data:

The behavior occurs both at bulbs. Above is the one in a single-bulb fixture, which runs continuously for most of the day. Most of yesterday registered 8w, but at the end of the day Hue began reporting that the bulb used 9w.

The basement bulb has similar changes in wattage. The plot above is for the last hour of usage yesterday. Incidentally, I’ve discovered that we are in the basement only for snatches of time. Sense reports that over the last 10 days this smart bulb is on for an average of 3 minutes at a time and is used about 15 times per day. Apparently we run down to check on laundry or store something, but don’t stay long. Come to think of it, that is accurate.

My first thought was that since both bulbs switch from 8w to 9w at approximately the same time of day, perhaps something was unusual about my electric quality at that time. I brought up the above plot from Labs. It does not appear unusual to my eye. I had seen readings of both 8w and 9w prior to yesterday, but did not pay much attention since I thought it would settle on either one or the other. It has not.

Currently the single-bulb is on and just started reporting 9w. When I first turned it on, it was using 8w and stayed that way for an hour. In a typical household, a watt more or less would not matter. Since these are indicator bulbs and I am multiplying by a ratio of up to 11.67, however, it makes a difference. Can anyone help me identify this wattage variability?

[drjb]

Not knowing exactly how this works, I’m tempted to speculate. There are at least two reasons I can think of. And, the deviations you’re seeing could be either real or the result of some on-board processing.

1. Possibly some change in the temperature of the unit causes it to draw more/less current
2. Rounding of displayed values.

#1 is rather obvious. #2 is like this: suppose the wattage varies between 8.8 and 9.2 Watts throughout the day (that’s about 5%, which is not unreasonable). Now, depending on how that consumption information is transmitted from the Hue lamp to Sense, and how Sense rounds the number (or truncates it), one could easily see the deviations you observed.

I’ve noticed that in all my devices/bubbles, the Watts numbers are always integers (no decimal point). Could it be that Sense only ignores the decimal part and only reports the integer? It’s certainly way cheaper to do integer math in such devices.

Thoughts ?

[kevin1]

My 2c. With the Hue integration, the power consumption comes from a formula for each bulb type. Philips Hue once posted an API app note that gave some insight in to the formulas for different bulbs, some non-linear function of brightness setting and supply voltage. So you need to ask whether there could be minor variation in either of these, because as @drjb suggests, there is nothing to the right of the decimal point here, only rounding, so the sensitivity really depends on the number of decimal places in the voltage and brightness data. If power is measured to a few decimal places, a 1% change in supply voltage could cause the 12% change in power usage you are seeing. Even though you’re power looks pretty stable, you could easily have a 1% difference in there.

[jefflayman]

I looked for the post Kevin references, which sounded helpful. Although this may not be exactly it, I found a web app from the manufacturer at Energy consumption - calculate yours | Philips Hue US. I tried three values (5%, 50% and 95%) and output was (1.0w, 3.2w and 8.5w) respectively. Reported values have one more decimal place than Sense does, so I agree that rounding is a prime suspect. Thanks for your input, guys!

While searching online, I also ran across a couple other interesting sites. https://atomstar.tweakblogs.net/blog/17888/measuring-calibrated-hue-energy-usage-via-bridge, has a single row of data for the E27 White bulb. The interesting thing here was that the smart bulb uses 0.3w when idle. That comes to 3 cents per month if electricity costs $0.14 per kWh. My indicator bulbs are not powered when the light switch is off, so will not have any idle usage.

Your smart bulbs aren't dimming the way you think they are. Here's why

The above article explains why there is a curve instead of a line in what Kevin remembers. The focus of that article was comparing lumen output to the brightness setting. This is not the same as energy usage versus brightness setting, which is what I was looking for. However, it was interesting anyway. Hue series bulbs have this curve, taken from the article referenced above:

[jefflayman]

Here is an idea to resolve the variability in wattage reported by the indicator bulb (8w versus 9w). A side benefit of this idea is simpler math. This idea replaces the part in my opening post about Sense asking users for an indicator ratio.

Instead, the app should allow users to input information about any associated bulbs. It could say, “Total wattage of any bulbs that light together with this smart bulb = ___w (zero if none or not applicable)” Using the same three examples given in my opening post, input would be 18w at the ceiling fan, 96w at the basement, and 0w (default) at the front porch.

With this solution, Sense would add the user constant to whatever is reported by Hue before displaying the bubble, but only if the Hue value is greater than or equal to 1w. This is simpler to code because it requires only integer addition, rather than multiplication by a floating point number followed by rounding. As noted by @drjb, integer math is easier to implement in digital devices.

[drjb]

If I understand this correctly, you’re asking Sense to ‘upgrade/include’ in their App the ability to do such calculation (scaling). I recall reading somewhere that Sense does (or might?) provide such calc already for their dedicated DCMs. This is done for devices that use 240V, such as ovens, ACs, and clothes dryer. I can’t find the link, but I think depending on how one wires up the CT, there is a way to include a scaling factor (2×) in the calculation for overall drawn power, when only one of the legs/phases is being monitored. If that’s the case, then such scaling option is technically already available in the App, and simply needs to be extended to devices other than Sense’s own CTs.

Can one of the mods (@kevin1) kindly comment on this and provide a link to the calculation I mentioned. I do not recall whether the scaling is done within the App or on exported data, for those brave/patient enough to do their own tallying-up work in Excel.

[kevin1]

Yes, the DCM measurement does include a 2x multiplier under the hood when DCM CTs are configured to measure 240V power in either 2 x 240V or 1 x 120V / 1 x 240V modes. But I suspect that this multiplier is far from portable to other data entry points. I suspect that this multiplier is a dedicated factor inside the Sense Monitor DCM firmware, not something that lives in the app or Sense backend.

[jefflayman]

I had to learn the hard way about color-matching bulbs. My post above describes measuring the light color using a do-it-yourself colorimeter. Based on that measurement and on the words “warm white” from the Hue product description, I bought some 9w LED bulbs with the label warm white and 3000K color temperature. I put them in the ceiling fan fixture beside the indicator bulb. Drum roll, please. I turned on the light switch.

Disappointment! It still didn’t match. In fact, it looked just like it had with the original ceiling fan bulbs. Could it be that those original bulbs were 3000K bulbs? With all this swapping of bulbs, I decided to look closely at one. Below is a picture of the original bulb from my ceiling fan. It has a label just above the threaded base. That label includes the color temperature, circled in green.

I also looked closely at my Hue indicator bulb. It has a similar label, which again includes the color temperature. The photo below shows that the Hue bulb color temperature is 2700K. No wonder the color didn’t match with my ceiling fan bulbs!

I looked around my house and found that all the LED bulbs have labels just above the threaded base. I found two that are labeled 2700K and installed them beside the indicator bulb. While not a perfect match if you scrutinize the light output, it is good enough for casual inspection. Success! Now I can finally use my Hue bulb as an indicator per the original plan.

The moral of this story is to read instead of trying to figure it out. If anyone else installs indicator bulbs, they can go straight to the correct bulb type by looking at the labels. The DIY colorimeter results were not sufficiently precise for purchasing light bulbs, not to mention that setting up such measurements is extra work. Playing with the technology was sort of fun for me, but the rest of you can avoid my learning curve.

[drjb]

Don’t feel too bad about the ‘had to learn the hard way’ … a large percentage of men (more than women) have some mild color blindness (be it the pink/grays or else), and that does not help. That is why most of us learn the hard way … and then invest in fancy electronic toys to give us numbers we can use

A bit of background: I got into these color temperatures ‘details’ with an old hobby of mine … Salt-water fish tanks. No matter which T5 bulbs I picked, my aquarium never looked as ‘bluish’ and crisp as those in the stores … Then came the PAR values, the color temperature … and then it became too complicated to keep track as not only you have to buy the correct color, but also of a specific make, otherwise there are bound to be variations. Oh, and the colors change a bit when the bulbs age …

Good study though. Thank you for sharing.

[jefflayman]

Since I am the person who started this thread, I wanted to report back to the community on my experiment. I consider it a success! In the six-week period just prior to this experiment, Other was 20% of my usage, and over the most recent six-weeks it is now 5%. There were half a dozen Sense native devices detected during that same period, so I can’t take credit for all of that change. Yet I will say that expanding the usage of my Hue integration definitely contributed! How? This graph tells much of the story.

Here is how I interpret this data. Weeks 47 through 52 represent a baseline of lighting accounted for prior to the experiment. As I described above, there were a couple Hue bulbs plus a lamp via HS300, and one bulb natively detected. These together were accounting for a daily average usage of 0.1 kWh.

I added Hue bulbs incrementally during weeks 1 through 5, so they make their appearance in the chart one or two at a time. The bulb named Basement has the highest indicator ratio by a large margin, while Porch, Corridor and Office are all single-bulb fixtures with an indicator ratio of just one. Taken together for the period following week 6, my overall indicator ratio is 2.0. This ratio is visualized here:

If you consider it a poor investment to purchase 12 smart bulbs to chase down 0.6 kWh per day of electrical consumption, I am with you. Tracking energy has become a hobby, and hobbies get expensive. There are three other locations in my house where I could install Hue bulbs to track even more usage, but I know they will not measure much energy consumption as I have already picked the low-hanging fruit.

Aside from merely measuring energy consumption, the experiment provided a couple side benefits. I asked Sense to tell me if the Basement light has been on for more than 20 minutes. This usually means someone has forgotten to turn it off when they left, so I can go turn it off and actively avoid wasting energy. Before bed I can check in Sense that all the house is dark. Plus, it’s cool to see all those bubbles on my Sense screen in the evenings!

A final observation from the above data is a general decrease in lighting as time goes by, especially if you discard outliers in Basement and Kitchen usage. I attribute this to increased natural lighting available as winter becomes spring. Daylight savings time began at week 12.

[jefflayman]

Two years ago I started this thread to request a feature related to the Hue integration. Discussions here helped clarify the feature request and also branched into other lighting topics. I’m writing again to update the community on my experience with Hue indicator bulbs.

The above graph is from Sense daily data export as manipulated in Excel. The orange color represents natively detected device usage. Native devices only account for 6.4% of my lighting energy usage. That is good because it means I have switched almost entirely to energy-saving LED technology. The down side is that Sense can’t detect LED so most lighting was lumped into Other. Based on two years’ data, Hue and the indicator bulbs allowed me to reduce Other by 24%.

Data recorded by Sense directly from the Hue integration is gray in the graph, while values calculated using an indicator ratio are light blue. The last color, green, applies to Christmas lights only, which are on a smart plug with a defined schedule, so are easy to estimate by multiplying the programmed hours of usage by the wattage, measured manually just after installing the lights.

@drjb suggested above not to mix bulbs. In April I replaced the two “dumb” bulbs in my 3-bulb ceiling fan fixture with Hue light bulbs. This was done to access the dimming capabilities of smart bulbs rather than because the indicator system wasn’t working. Whatever the motive, his suggestion not to mix bulbs ended up being correct. However, I still believe in indicator bulbs for other cases, such as my basement where the indicator ratio is 11.67.

I have new electric usage data on that basement light! As described above, it consists of two 48W LED fixtures and my Hue bulb, all activated with one switch. Sense found this combination as a native detection in August of 2021 with energy usage of 87 watts after the startup spike. As noted above, nameplate capacity of the fixtures is 48*2+9=105 watts. This discrepancy could mean Sense is wrong, but it could also be that fixtures only draw their nameplate at startup. I tend to believe Sense on this because it is based on actual measurement. Another reason to believe Sense is the graph below which compares the native value with my estimate from the indicator bulb.

The detection efficiency in this graph stabilizes at about 36% after Sense finishes refining its model. The wattage difference between nameplate and measured is 83%, so further explanation is needed. I have a hypothesis based on the observation in post 5 above. There I observed that the bubble takes a full minute to disappear after the switch turns off. Since my basement usage is for lots of short periods (post 8 above), this might be enough to explain the rest.

[jefflayman]

My last post graphed a consistently low correlation of Hue versus Native for daily energy consumption by my basement lights. One explanation was wattage, but that accounted for just a small part of the difference. I hypothesized that the time it takes for the bubble to disappear after the switch turns off could explain the rest.

A hypothesis is supposed to be a testable explanation. Testing my idea was possible using data available in the Sense app. The data needed is not part of export so required manual input. That data is the Times On count by day, found in the Usage section of the device page as shown below. Such data fits perfectly with daily usage data exported from the web app.

I created two columns of manually entered data: one based on the Hue indicator bulb associated with the basement lights, and one from the Sense detection of the same light circuit. A third column in the spreadsheet computed the difference between the Times On count for each day. The pivot table below summaries these differences for 300+ days in 2022, beginning at week 10 after Sense finished refining its model.

The difference was zero for 125 days. Said another way, 40% of days had the same count of Times On from both sources. Remaining events show good clustering about zero. About 20% of days had positive values, meaning that Sense did not detect one or more events. Sense sometimes has trouble disaggregating usage but does remarkably well with my basement light.

Negative values in the table above are real events that Sense counted but Hue did not. Hue failed to count them because those events occurred in close succession. Sometimes I leave the basement but return immediately because I forgot something. If I turn the light switch on within about 1½ minutes of turning it off, Sense never hears about that cycle. In this respect, the native detection is better than the integration. I would give this particular native detection 5 out of 5 stars!

Having the Times On data allowed comparison in another way. The chart below plots daily average energy usage per Time On. Note how the indicator light never had any data less than 3 Watt-hours per time on, since 2 minutes is the shortest time it can register. The native device has more realistic data as small as 0.7 Watt-hour, which feels right for short trips to the basement to grab something. I like how both distributions have similar shapes even though the scales are quite different.

Based on the factors discussed above, I applied two adjustments the data. The first changed the indicator ratio based on wattage from Sense rather than nameplate. The second adjustment subtracts (Times On) multiplied by (1 minute and 20 seconds of apparent usage after the switch is off) for each day. The result is shown below.

Improvement is significant, but I’m disappointed that the trend line is still only 65%. If anyone has ideas, I would appreciate suggestions.

[kevin1]

Very cool. Love the meticulous analysis. I’m assuming you have two different detections, one for Native Basement and one for the Indicator Bulb ? You might want to try comparing on times for those two as well on a daily basis, using the same kind of pivot table. That takes the power out of the equation and might also give you a better idea about how well your Hue short cycle correction is working.

ps: Didn’t realize the Hue Integration has that short gap/cycle miss.