Hacker News Reader

Comment by AlbertCory | Hacker News Reader

When I was in Google Patent Litigation, I volunteered to work on Nest because it sounded cool. So I spent a couple days over there (they weren't on the Mountain View campus) and sat in on some deposition preps. They were being sued by a Texas company called Allure, which was not a troll; they actually had a product that, at one time, you could have found on the Home Depot web site. I'm not going to reveal any legal secrets here, sorry.

I did a ton of research on this space, and dove into their source code. Literally all over the world, companies have been working on this problem for decades. How you know when someone is coming home so you can turn on the heat or A/C, or how the utility company can do it for you: all that is very, very well covered. I don't think there's anything fundamentally new about the Nest, and I don't have one in my house.

I actually stayed in someone's house that had two, so it's not as though it's completely foreign to me. I don't think that web-ifying your whole house is ever a good idea and I'm never going to do it. YMMV.

View on HN

I wish they could solve the basic use case of "Let me set a temperature and thermostat keeps the house at that temperature." I tried several versions of Nest, and they all inevitably decided on their own that they were smarter than me and generated some weird schedule despite my never commanding them to operate on a schedule. I just want my thermostat to hold whatever temperature I set it at, until I set it at another temperature. This shouldn't take AI wizardry to achieve.

It does seem like you could do that with a bimetallic coil and a little blob of mercury in a tube, like thermostats have done since the 1950s

> I just want my thermostat to hold whatever temperature I set it at, until I set it at another temperature.

If it only takes one of heating or cooling to do this, not both, a standard "dumb" thermostat can of course do this just fine. Just set it to "cool" or "heat" mode and set a temperature. But of course you will need to change modes as the season changes (unless you live somewhere where only one mode is ever needed).

If you want to not have to change modes when the season changes, it gets more complicated, because the simple control system in the standard thermostat is too dumb for that. For example, suppose you set your temperature to 75. The actual temperature rises to 76, above the setpoint, so your A/C comes on. It cools the house to 74 and then the thermostat shuts it off (you have to have some buffer so the A/C doesn't come on again almost immediately). Now the temperature is below the setpoint, so your heat comes on. It heats the house to 76 and then shuts off. Now the temperature is above the setpoint... You can see what's going on.

From a pure design perspective, the right way to solve this problem is an outside temperature sensor (and probably a sun sensor as well since sunlight is a significant heat input to your house), so the system can predict what will happen when both cooling and heating are off. Given reliable sensors, the control algorithm is fairly simple (automatic climate control systems in cars already do it). I have never seen this in a house, though. I'm not sure why.

I wonder if the quote is referring to balancing 2 systems. I don't have AC, but I use my heat to maintain a minimum acceptable temperature, not to maintain the temperature I find ideal. I don't want it to go a lot lower than the setting.

Surely a basic PID controller with reasonable tuning could mostly mitigate the problem you describe, which isn’t really a different problem than a thermostat that would attempt to turn the furnace on and off multiple times per second if the thermometer was fluctuating between 74 and 75 multiple times per second. Granted, tuning the PID controller implies loosening the “maintain the exact target temperature at all costs” spec.

> tuning the PID controller implies loosening the “maintain the exact target temperature at all costs” spec.

And if you loosen that spec, then, as others have already pointed out, just having two setpoints, a lower one for heating and a higher one for cooling, will be sufficient as long as the difference between them is large enough (probably 2 degrees or more would be enough for many use cases). That would be simpler than adding a PID controller.

This is solved in thermostats with hysteresis and even many DIY or close to it solutions like esphome[0] implement it. I know on my Nest when heat and cool mode is on the set heat and cool temperatures have to be two degrees away from one another, which is fine with me because for whatever reasons 67 in the summer feels great but I want close to 69/70/71 in the winter.

The linked esphome docs are actually great whether you care about esphome or not - they explain all of this very well and you can tell from the configuration values and samples what kind of logic goes into all of this.

[0] - https://esphome.io/components/climate/thermostat.html

Your summer and winter preferences may vary due to humidity levels. It makes a world of difference it how humans perceive temperature because of it.

I really wish I could have my thermostat's set point use "heat index" (a straightforward function of temperature and humidity) instead of just temperature.

I am perfectly comfortable (and therefore want my system to turn off) after my AC has been running long enough to drop the humidity a bit, even if the temp has only dropped a fraction of a degree. As humid air leaks in, I want the AC back on even if the temp has barely increased a degree. Basically the heat index in my house swings enough to change my comfort and therefore could be a useful trigger, even if the temperature doesn't swing nearly enough to be a useful trigger.

The Nest even has a humidity sensor, but it only gets used in very specific (hardly useful) situations.

I considered writing a Works With Nest app for this. Then that platform disappeared in favor of Works With Google, but I never found the time for either...

> I really wish I could have my thermostat's set point use "heat index" (a straightforward function of temperature and humidity) instead of just temperature.

You could try wrapping your thermostat in a wet sock.

I'm not sure my wife would love that, and I'm not sure how to keep it wet at all times (apply water of a known temperature??) but I am intrigued: how does it work? The temperature of the sock will somehow correlate with the heat index of the ambient air?

It's mostly just a joke about wet-bulb thermometers: https://en.wikipedia.org/wiki/Wet-bulb_temperature

Actual web-bulb thermometers would either require manually adding water when you want to make a measurement, or have a large tank of water that gets continuously wicked onto a cloth which is wrapped around the thermometer. The general idea is that, if the surrounding air is dry, the water in the cloth will evaporate and cool the thermometer below the temperature of the air itself (the "dry-bulb temperate"). This is just the same evaporative cooling effect that makes sweating and swamp coolers work. If the surrounding air is very wet (humid), very little evaporate cooling will occur, and the wet-bulb temperature will be very close to the dry-bulb temperature.

I'm not sure how well a thermostat that only used the wet-bulb temperature would work. To actually calculate relative humidity, you would need at least a wet-bulb and dry-bulb measurement. But given that relative humidity in most places probably doesn't swing that wildly, I imagine that just finding a comfortable wet-bulb temperate for your thermostat might work okay in practice.

This is brilliant, thanks. I do see how additional calculations between wet and dry (or just a hygrometer, which the Nest does have already) would be ideal, but as you say, using just a wet-bulb does seem a bit better than using just a dry-bulb in terms of approaching my goal, even if it's not quite all the way there.

Rigging a wet sock (with constant wicking from a water source) onto my actual wall-mounted Nest Thermostat would never fly, as I said, but I have no qualms about doing this experiment with my Nest Temperature Sensor, being a small and relatively cheap thing that I can easily steal for a while...

According to a psychrometric chart [0] it looks like if the dry-bulb temperature stays locked at 70F, while the RH drops from 60% to 40%, the wet-bulb temperature will drop from 62F to 56F. Or if the dry-bulb temperature stays locked at 80F while the RH drops as before, the wet-bulb temperature will drop from 70F to 64F. That's linear enough for my purposes.

Therefore if my conventional (dry-bulb) set point is 75F in the summer, and I'm most comfortable around 50% RH combined with that 75F, then I'll try a set point of 63F with my Nest using the remote Sensor wrapped in wet cloth!

Apparently I also need continuous airflow over this wet-bulb thermometer. I wonder how much.

[0] https://hvacologist.com/2022/02/21/wet-bulb-temperature-vs-d...

If you’re willing to spend the money, I’d recommend getting a ZigBee or z wave thermostat(or an EcoBee, as it has local control through HomeKit) and then you can combine the temperature readings with your own humidity/temperature sensors - as mentioned above, ESPHome/Home Assistant may be overkill, but would work great for this kind of setup

Humidity definitely plays a part in my preferences but where I live there are 100 degree temperature swings and interacting with the elements outside leads to (for me) more extremes inside.

When I get back from a run in August at near heatstroke levels I will almost never cool down if it's not set on the "chilly" side. Conversely, in winter when heatstroke is replaced with near frostbite I am chilled to the core and need much higher heat to ever warm up.

You'd just set a low point and a high point far enough away from each other that the thermostat wouldn't cycle back and forth between heat and AC. You could also add a timeout so that the heat cannot turn on for at least two hours after the AC cycle just finished. During November, we always seem to get crazy temperature swings where we run the heat one day but need to run AC the next. I'd love a dumb thermostat that didn't require me to manually switch over to heat or AC mode.

Is it not possible to forecast temperature changes rather than installing a bunch of extra sensors? If the device knows its location and has an internet connection, it can examine the day's weather forecast, know where the sun will shine, and know how cloudy it will be.

You'd still need correction for how sunny/shady house walls are

Good point. I wonder if the temperature sensor in the thermostat would be able to help correct that error.

E.g., if the thermostat expected a temperature of 75 F at 12 PM but it was actually 78 F, then you know the house received a lot of sunlight and can correct accordingly.

The rate at which your house warms while the heat is on is an indirect measure of the effect of both the temperature of the outside of your walls and the amount of solar radiation coming through your windows.

As others have indirectly pointed out, for the range of circumstances a given house is going to encounter, the linearizing assumptions in a standard PID controller are probably sufficient. If tuned properly, the "differential" portion of the PID controller will shut down the heater earlier when it's warmer out and/or there's more solar energy coming through your windows.

> If it only takes one of heating or cooling to do this, not both, a standard "dumb" thermostat can of course do this just fine. Just set it to "cool" or "heat" mode and set a temperature. But of course you will need to change modes as the season changes (unless you live somewhere where only one mode is ever needed).

Why ? What would be so hard in detecting "every time I don't do anything temperature goes up therefore summer and I need to use AC" ?

> If you want to not have to change modes when the season changes, it gets more complicated, because the simple control system in the standard thermostat is too dumb for that. For example, suppose you set your temperature to 75. The actual temperature rises to 76, above the setpoint, so your A/C comes on. It cools the house to 74 and then the thermostat shuts it off (you have to have some buffer so the A/C doesn't come on again almost immediately). Now the temperature is below the setpoint, so your heat comes on. It heats the house to 76 and then shuts off. Now the temperature is above the setpoint... You can see what's going on.

All would it take is to put a gate "if I used cooling to correct temperature last time, don't use heating for few hours or until it drops to higher difference and vice versa". You're trying to portray it in way more complicated way that it is.

> From a pure design perspective, the right way to solve this problem is an outside temperature sensor (and probably a sun sensor as well since sunlight is a significant heat input to your house), so the system can predict what will happen when both cooling and heating are off. Given reliable sensors, the control algorithm is fairly simple (automatic climate control systems in cars already do it). I have never seen this in a house, though. I'm not sure why.

Too simplistic. Which side of house you monitor, is it in shade or does it show temperature when sun heats up the wall ? Which room is closest to "hot" wall/window?. It would be a lot of work for very little improvement

This is solved with two set points, 68° for heat and 78° for cold, and appropriate clothing.

> If you want to not have to change modes when the season changes, it gets more complicated

There's an additional complication for multi-story houses that I didn't mention before. In multi-story houses, when you're cooling, you want most of the ducted airflow to be directed to the upstairs outlets, to take advantage of natural convection (cooler air sinks); similarly, when you're heating, you want most of the ducted airflow to be directed to the downstairs outlets (warmer air rises). That requires adjusting dampers in the ducts, usually just downstream of the output of your HVAC unit. Unless that adjustment can be automated (which would require automatically controlled motors on the dampers, something I haven't seen marketed, though again I'm not sure why), your airflow will be suboptimal for either heating or cooling if the system automatically switches between modes.

Switch controlled open/close dampers are common enough.

A typical suggested application is controlling the airflow to a room using a thermostat to open or close the damper, but they can be hooked up to anything with a switched output.

Can these dampers stop somewhere short of full open/close? Typically one wants full open for one season and mostly closed but not fully closed for the other.

Apparently? I'm not an HVAC guy, just poking away at understanding my current system and thinking about what I might change.

As an example, this damper has an adjustable bleed rate, with a 2 wire input for switching the state:

https://www.resideo.com/us/en/pro/products/air/forced-air-zo...

(Click the "Install" tab to access installation guides)

As a kid we had a very simple (mercury) thermostat which had two levers -- one for heating set-point & one for cooling set-point. I recall it working quite well.

Always thought the old mechanical bimetallic strip solution was particularly elegant.

In effect, two strips of metal were bonded together forming a coiled "spring." Because each metal had different coefficient of thermal expansion, the spring would expand or contract to different lengths depending on the temperature. Place the electrical contact at the temp where the heater or HVAC should turn on & bam.

No electronics at all, just a temperature-dependent mechanical switch.

Exactly.

I had a problem with the temperature in my house often being wrong, so I bought a smart thermostat. Now I have ten problems.

they still exist, that's what cheapo thermostats still use. i've wired up a couple thermostat controlled extension cords over the past couple years for a dog house and chicken coop using them

Also, they usually had a tiny magnet at the tip of the bimetallic strip to make their decisions path-dependent (more "sticky"/less flip-flopping).

> I just want my thermostat to hold whatever temperature I set it at, until I set it at another temperature.

You can set your own schedule with Nest, you don't have to let the AI take over. Mine is fully on my own manual schedule.

I've set my own schedule with nest, and it still reset it to its own weird (and far too warm) schedule

You need to turn off learning mode, otherwise it eventually overrides your schedule.

Personally I think it's significantly better than what was common ( in the UK ) before, but the tech is dime a dozen, and new houses should probably come with underfloor, and control per room, though most still don't.

I had Hive before and prefer that; being able to boost the heating to a certain temperature for 30/60/120 mins is a common use case for me that Nest doesn't do.

You can disable the "smart" behavior. It's in the menus. I did. It's fine.

You could have turned that off before using it.

they all inevitably decided on their own that they were smarter than me and generated some weird schedule despite my never commanding them to operate on a schedule.

Seems to be a Google habit. I regularly get "commute traffic reports" several hours before or after I leave home or the office. Of course there's absolutely no way to set the alert schedule yourself.

Does a simple analog Honeywell thermostat not do that?

It turns out the cheapest thermostat at the store fulfills your exact requirement :)

You think there’s nothing fundamentally new about the Nest, but it arguably changed the home smart thermostat industry and people love it. People loved it even before Google acquired them so you can’t even argue it’s because of cost or marketing.

They define do have something fundamentally new. I think it was exactly what was described in the video - easy installation, intuitive design, and right timing.

I think you are looking at only what's available in the Home Depot. Connecting to your HVAC system is "fundamentally" simple and lots of companies have done it.

"Home automation" is also not a new concept. X10 came out in 1975:

https://en.wikipedia.org/wiki/Home_automation

What Nest had is consumer-friendliness, which I give them full credit for. Not surprising for Tony Fadell, who came out of Apple.

Oh look, now he's into crypto:

https://www.msn.com/en-us/news/technology/man-behind-the-ipo...

Maybe he'll bring his magic to that, too.

In the 1980s I spent several years defending Honeywell against some programmable thermostat patents filed in the late 1970s by one Burness C. Hall, who had aspirations to start a company. There was a ton of prior art out there. (The parties eventually settled.)

https://patents.google.com/patent/US4200910A/

I think you’re overestimating the popularity. Only 11 million were sold by 2011.

https://www.cnet.com/home/smart-home/nest-says-it-has-sold-o...

How often do you change your thermostat? On its own that statistic means about 3% of Americans have bought one, which, when you consider that it's more expensive than the traditional kind, you only really need one thermostat per household (maybe two or three if your house is extra fancy and has multiple zones), only a homeowner would consider the purchase, and thermostats are exceptionally long-lived devices that rarely actually need to be replaced, that's pretty good.

Property managers likely buy many more than one per property - it’s not likely that 3% of Americans decided to buy a Nest if you consider that apartment buildings would buy these in bulk

I would consider property managers pretty unlikely to buy anything but the least expensive thermostat available.

just look at the alternatives, they're horrible. The Google Nest thermostat has revolutionized the thermostat ecosystem in the same way that the iPhone revolutionized the cellphone industry.

I was getting quotes for some heating/cooling projects at home and, while discussing the control panels, I expected that Nest would be an option. I was surprised and found it comical when the salesperson claimed that the Nest control panel was not supported. This is an instant non-starter for me.

I've recently been wanting to change the three Nest (v3) thermostats in my house to something else. I'm looking for something that has a 1 degree (F not C) hysteresis, and can be controlled via WiFi, but not the cloud. (I don't like the idea of the government changing my temperature settings.) So far I haven't found anything, but I did find a few that allow local control when the cloud cannot be reached. It should be easy to put them in a separate guest network with no Internet gateway.

> I'm looking for something that has a 1 degree (F not C) hysteresis...

Maybe not what you're looking for, but if you're just looking for a 1F degree lag in temperature change, you can also try just increasing the heat stored in your house's thermal mass. For example you could add something with high thermal mass (i.e. concrete) to the place you want higher thermal lag, and position it so it soaks up a lot of solar radiation. That'll buffer out your temperatures.

Thanks for that. Probably not practical for my needs. The Nest "Smart Learning Thermostat" tries to save energy a little too aggressively for my taste. I think another brand would do better. In addition to the "Lag" being too wide, it also seems to have a low sample rate for the temperature. My master bedroom will often go to five degrees above the setpoint because the Nest will keep the heat on thinking it's still cold. It's really complete garbage. (I've measured the temperature separately to confirm it's a Nest issue.)

Often turning on and off heaters (and more especially heat pumps) is harder on them than keeping them running a bit longer? Were you sure it was actually wrong about the temperature and not just intent on not having to cycle on and off a ton?

Well I don't know for sure, but the displayed temperature on the Nest would only update every few minutes. My independent measurement showed that it had reached the setpoint about two minutes before the Nest updated its reading (to show that it had overshot by several degrees). It's possible that the Nest is trying to reduce the cycling, but if that were the case, it shouldn't affect the displayed temperature.

I installed an American Radio Thermostat unit when we moved into this house. It has both a cloud and local connectivity.

Yes! American Radio is what I'm leaning toward. I wanted a CT-80, but they have discontinued it (confirmed with them). I guess I'll go with the CT-50. The CT-80 had a nice feature in that it could support both Z-wave and WiFi, but apparently they had too many bugs and pulled it.

I think you'd need more than a thermostat to keep the temperature stable. This is the main selling point of variable speed systems and I'd imagine cheaper brands would have been selling "Stability of a variable speed for the price of a single stage!" packages with such a thermostat if it existed.

the HVAC people call this Swing but Ecobee calls it 'threshold' and you can have as little as 0.5F of it. It's gonna cause more wear on the equipment though, and you gotta keep in mind that furnaces will ignore the thermo until a cooloff period has elapsed anyways.

For local WiFi, anything that's HomeKit compatible should work, but you'll likely lose out on something by restricting its access to the internet. For example, my Ecobee can be controlled locally via Home Assistant (using Home Assistant's HomeKit Controller), but it gets outside weather (used to figure out how early to turn the heat up to meet the schedule requirements) via some cloud service. Plus the HomeKit interface, at least when used in Home Assistant, is not as full-featured as the native (cloud-based) one.

If you desire 100% local control, it's worth looking at Zigbee and Z-Wave thermostats. Using Home Assistant can give you full local control without requiring a vendor's hub and app.

You might try something from Honeywell, which is the brand that most HVAC contractors use. The stock Trane and American Standard thermostats are also white labled Honeywell I believe. Nonetheless, they use a concept of "CPH" or cycles per hour tied in with some PID circuitry.

I go for longer run times and use a CPH of 2, so the thermostat tries to run for 15 minutes followed by 15 minutes off. In mild weather, it is more like 8 minutes on and 20 minutes off. I was skeptical about replacing my Nest with it, but I think it holds the temp more comfortably. Plus you can add wired and/or wireless sensors to various rooms in you house so everything stays more balanced.

I used a Honeywell smart thermostat and a Nest and the UX for the Nest is definitely better. The functionality is similar though.

If the only feature "smart" thermostats had was that I could adjust the temperature with my phone from bed I would consider it money well spent.

Reply on news.ycombinator.com