[sz4kerto]

We've moved to an new apartment (house) and we had to do a full renovation. It doesn't have modern insulation and I calculated that for the time being the ROI on insulation isn't worth it. It's a multi-floor semi-detached house and I wanted the best comfort and the most economical heating possible.

In particular: stable and individually adjustable temperatures for bedrooms and living rooms; underfloor heating in some rooms (bedrooms), radiator-based heating in some others (living room), and combined UFH+radiators in some others (where UFH might not be enough during extreme colds).

I thought I can just pay someone some money and they'll set up the controls for me. It must be a simple exercise, right?

I could not have been more wrong. After spending a few hours of understanding the setups that "experts" have recommended, I figured out edge cases where they would be either wasteful or uncomfortable (meaning: unnecessary and inavoidable temperature overshoots or undershoots, etc.). I had many-many rounds with Honeywell, Tado, Siemens, etc. and every single one of them had _major_ issues.

The renovation got a bit stuck because of this, but the plumbing was ready so I wanted to see whether the pluming and pumps are working, at least. So I connected the pumps and valves to "smart plugs", i.e. Zigbee-controlled plugs, so that I can see that they turn on. They did, which got me thinking...

Right now I have $20 Zigbee temp sensors sprinkled across the house, $30 smart plugs and relays driving valves, pumps and the boiler, and Home Assistant is controlling the whole thing. Everything works perfectly and I could implement some features that simply no system would have done out of the box, for example in rooms where there's combined UFH and radiators I can drive both heating systems when the target temperature is far from the desired (so that the room heats up quickly) but as the room temp is getting closer to the target, the radiators are turned off so that UFH dominates heating (more comfortable and more energy efficient than radiators). In rooms with radiators, temp is +- 0.4 C within target, in rooms with UFH, it's +-0.1C within target.

[darkwater]

Yeah, the automated/remote controlled heating system world, and also the ringbell world is basically a giant scam, since they are updates on world that also scammed you in the past. I cried when I shelled out so much money for my Tado device, but even a dumb bTicino device costed in the hundred of euros realm, and it's just a sensor + a small LCD display and a designed-in-hell menu system to program it. And the same happens in Ip-based ringbells. A Doorbird will cost you hundreds of euros for what is basically a webcam plus some nice metal casing and a shitty software, but it competes with analog systems with optic cables etc that cost basically the same or more.

[cameronh90]

Off the shelf systems aren't only optimised solely for efficiency. They're made to be simple enough that an installer with half a day's training can do a few multiplications and additions to set the parameters that'll give you a tolerable percentage of optimal in the situations that equipment is specified for -- while still being understandable by the next guy. This nearly always means things are a bit oversized and inefficient to account for the things that the simple models are missing.

Almost everything in engineering is like this, not just heating. It's pretty rare that something is fully optimised.

[KennyBlanken]

> We've moved to an new apartment (house) and we had to do a full renovation. It doesn't have modern insulation and I calculated that for the time being the ROI on insulation isn't worth it.

You calculated wrong, guaranteed. Most likely, you wildly underestimated fuel/electricity costs.

> After spending a few hours of understanding the setups that "experts" have recommended, I figured out edge cases where they would be either wasteful or uncomfortable (meaning: unnecessary and inavoidable temperature overshoots or undershoots, etc.).

Instead of thinking "the entire HVAC/heat industry are idiots who can't do any of this right", maybe you should take a look in the mirror and consider that your assumptions and/or criteria are wrong.

For example: under/over shoots in a modern HVAC or heating system will not cause any "waste" or discomfort. 1-2 degree F in overshoot does not mean the space will lose appreciably more heat than if it had perfectly regulated at the setpoint. You also don't want a system that responds instantly. Let's say you open the door to receive a package, and you're signing paperwork, etc. You close the door. The air in the room is substantially cooler.

Should the heat turn on?

I bet it does in your home...but the correct answer is no, because the air will warm up rapidly from all the objects that were at the temperature of the room. Thousand-plus square feet worth of surface area...

[quickthrowman]

> I could not have been more wrong. After spending a few hours of understanding the setups that "experts" have recommended, I figured out edge cases where they would be either wasteful or uncomfortable (meaning: unnecessary and inavoidable temperature overshoots or undershoots, etc.). I had many-many rounds with Honeywell, Tado, Siemens, etc. and every single one of them had _major_ issues.

Temperature hysteresis is unavoidable with a conventional thermostat, but you can reduce it with PID controllers. Most commercial building automation systems use PID controllers extensively.

My guess is that the residential options from Honeywell, JCI, Siemens, Trane, Carrier, etc are focused more on one-size-fits-all applications, whereas commercial BAS systems are more or less bespoke designs for a specific building (using commodity sensors and controllers). I work with all five of the aforementioned companies on building automation projects, FWIW.

[rightbyte]

Sounds like a nightmare for a future buyer to operate.

Some people are unlucky enough to buy homes where a machine engineer designed the boiler setup and the boiler room have enough valves and manometers to like operate the engine of Titanic.

I guess programmers are the new sinners in this area nowadays.

[sz4kerto]

If we ever sell this (which we don't plan to), I know what to install (it'll be quite good, just not this perfect). I have it in a cupboard (a Siemens Connected Home thermostat system), the downside of that is that the combined UFH+radiator rooms will be less comfortable.

(But still more comfortable than 99% of the houses I've been in.)

I haven't mentioned in the parent comment but as a test I've dismantled the HA system and installed the Siemens system and it works well, just not 'perfectly'.

[rightbyte]

Sounds well thought out. Many seem to forget designing for replacing.

In generall I think all these IoT systems will be a major headache as they age.

My thermostats on the radiators are 45 year old by now. That is kinda the expected service life we are used to.

[evrimoztamur]

The existing professional setup was also a nightmare, what gives?

[rightbyte]

Well first of I am envious and I would want to do something similar.

If I inherit a heating system I want it to be all mechanical except maybe the control system for any heating pump.

[ajolly]

I'd like more details on your home assistant setup as I'm trying to optimize mine.

Btw, you can use $5 LYWSD03MMC thermometers with ble or zigbee.

[sokoloff]

For every John Siegenthaler and Dan Holohan, there are thousands of mechanical engineers and tens of thousands of plumbers who are happier to slap in a $20K 4-hour boiler retrofit. There’s not enough extra money in catering to the 0.1% of homeowners who care about the details.

[dzhiurgis]

A lot of this is so easy with AI now. Just need some confidence and patience to work with AI lol.

[sokoloff]

It’s a service or human engineering problem rather than a mechanical engineering or knowledge problem.

Contractors today put in over-sized equipment, set flow temps higher than needed, undersize emitters for aesthetics and cost, and run pumps at full speed to avoid callbacks for “it’s too cold”. You can’t afford the windshield time to drive over to tweak the system to extract maximum performance, because homeowners don’t want to pay for it and will go with the lower bidder enough times that your premium AI-powered service will struggle.

I’ve tweaked my reset curve 12 times the first winter and 2 more times since then (counted from my spreadsheet).

Realistically, if I gave up on the last 3% tweaking, we could have lived with it after 2 post-install tweaks, but at least one of those had to wait for seriously cold weather snap to fine-tune the low end.

My spouse would happily agree that the house is finally very comfortable and noticeably more than before. She’d also tell friends who asked that there were a few days the first winter where the house wasn’t warm enough and needed an adjustment. People who heard that story might conclude that their neighbor’s guy who never has a callback for “too cold” is a safer bet.

Over time, I think even the best mechanical contractors will start to lean towards avoiding callbacks and do that by running the system 5-10°C hotter than “correct” engineering requires. That’s still better than today, where flow is set to 80°C, pumps to max, and the thermostat cycles 4 times an hour but the house is never too cold.

[amne]

a bit off-topic: Are you running a single boiler and if so, how are you mixing UFH with radiators given there's a ~20C difference between the recommended temps for the two?

My knowledge is that for UFH you run at temps between 40-50C and radiators run at 60-70*C.

[sz4kerto]

UFH has mixing valves, so it runs on 38 C and radiators run on 55C. Single boiler.