Multi-circuit water floor heating controller

[elcaron]

Hi, I would like to get some input for a project I am currently planning.

Introduction
I just moved into a new apartment with central water floor heating, which is currently controlled by bi-metallic thermostats. Unfortunately, I am quite underwhelmed by the performance. Since the house is reasonably well insulated, and the stone floor has very high heat capacity, we have bang-bang controller cycles >1d, so some days, the floor is cold, some days it is warm, temperature under- and overshoots ... it is not pleasant. I could probably counteract this be limiting the flow, but then I would waste power in the high efficiency pump. (Essentially that should be equivalent to current limiting with a resistor.)
So something has to be done.

Unfortunately, I have only 3 power cable wires available to the thermostats, one color-coded PE. With this, I am not willing to install any digital controllers that rely on 3 wires for L, N, L-switched cables. It seems that I will have to cook up something myself.

About myself
I am a mainly theoretical physicist in research. My electronics skills are developed as hobby, but reasonably well backed by my profession. Firm in algorithms and coding.

DIY state of the art
I found e.g. this project. It is a nice project, but I would like to solve it little differently.

What is available
- European style (round 60mm) outlet sockets in each of 7 rooms that currently hold the bimetallic themrostats ...
- ... each directly connected with 3x1.5mm² power cable (blue,brown,yellow-green) to the room with the circuit distributor and the valves. Cable length <20m.
- 9 water circuits, each with a 230V 2W valve, opening and closing time about 3min. No half open state available and probably non desired for pump-efficiency reasons (see above). 3 of the circuits are in a single room and are currently controlled by one thermostate (valves wired in parallel).

What I have come up with by now
I would like to place a new unit in the thermostat sockets, each one equipped with an 0.96" oled screen, a BME680 and probably an ESP12. I would like to power the unit over the brown and blue wires with 5V (LDO to 3.3V at unit), and use the green-yellow one for single wire data transfer for long term compatibility (don't want to base it on wifi if I do not have to).
At the controller, I want to aggregate  the data and and implement a PID-style controller that controls all the circuits in ~30min PWM intervals to keep a relatively stable ground temperature.

What I not have made my mind up about fully
- I would prefer to keep the 230V valves. Due to the PWM, I would like to avoid mechanical relays. On the other hand, some SSR might have problems with the low current.
- I am not sure yet how to transfer the data bidirectionally over ~15m power cable on a single wire. Open-drain with MOSFETs and strong pullups? What protection should I place? Schmitt trigger? ESD protection?
Or would 1-wire work reliably with a driver like the DS2482?

I would be very happy if I would get away only one hardware version without revisions, so I would like to get this right on the first try. Programming will be interesting enough. Any ideas on the unclear parts? Issues? Improvements?

[Zero999]

Sorry, I can't be a huge help with this, since it's not something I've done before.

The one thing which stood out to me is using the green & yellow wire for data transfer, which would violate the wiring regulations anywhere in Europe. The green and yellow wire shall only be used as a protective earth connection, never for power or data. Rather than using the earth wire for data, it's possible to transmit it over the mains cable, which is perfectly safe, if done properly.

[coppercone2]

This is a problem i have too. I find myself opening the window with a fan right now.

My solution is gonna be to weigh it less for full room heating and compliment it with air vents lol. The floor is too much to monkey around with imo.

[alanambrose]

Hi, I'm in the middle of reverse engineering my own setup, which is an Uponor 'Smatrix Base' thing so I'm starting to understand these systems. They're actually medium complicated - in that they involve a mix of heating in the 5-20kW range; valves, actuators, flow limiters, pumps, filling, draining, by-pass, blending valves; fail-safes; overall thermostat limiters; flow balancing; pump exercise etc etc etc .... besides the electronics, and room thermostats see e.g.

http://www.pexa.com.tr/manifold/

I'm not sure how much control you have - is this a system you have complete control over, or are you piped hot water from some central source and you only have control over the valves and thermostats?

Bang-bang will work fine - my system takes 9 hours to raise the temperature 1 degree C. The stat resolution is 0.1 degree F (about 0.05 degree C) so bang bang should keep within say 0.2 degrees and switch maybe every 1/2 hour.

Suggest checking out the commercial systems  (for residential) if only to see how they all hang together. Rolling your own won't be crazy difficult but there are some important things to get right to avoid overly heating your pipes etc. You'll see commercial systems with 24V or 230V actuator valves, rs485 or wireless data from sensors to controller etc. Maybe you can finesse a solution out faster and simpler - e.g. it sounds like your thermostats simply have too much 'dead time'.

Alan

[jbb]

This is interesting. I’d like to clarify a couple of things.

1) do you have all the valves in one location? You imply you want a central controller.
2) do you have access to both ends of the wiring without too much trouble?

I agree that yellow/green should remain protective earth for safety (and wiring compliance)


Getting power and data over 2 wires is a classic problem. What exactly do you want to send / receive?

Two schemes for two-wire power+comms which come to mind are a classic 4-20mA current loop (very popular in industrial plants and building management) and some 2-wire version of LIN (found this app note, don’t know how good it is).

For pretty much any such scheme, you have limited power available at the slave node, so I suggest a non-backlit LCD for the ‘thermostats.’ Reducing power dissipation here also reduces self heating and improves temperature measurement accuracy.

Example 1: the electronics in a 4-20mA device must consume <4mA. With a standard linear regulator, your budget is around 3.5mA @ 3.3V, which is enough to run a micro, small LCD (no backlight) temperature sensor and a couple of buttons.

Example 2: 4-20mA device with switching power supply. Assuming 12V drop across device, power budget is 3.5mA * 12V * PSU efficiency (say 75%). This nets you around 32mW, or approx 10mA at 3.3V. Still not enough to run an OLED

[jmelson]

Your thermostats need an "anticipator".  For heating, this is a resistor that is in series with the thermostat that causes it to warm up faster than the room.  It is wired in series with the thermostat so that it only heats it when the thermostat contact is closed to demand heating.  The idea is to shorten the very long heating cycles, and this will also shorten the off-periods between heating cycles.  The value of the resistor is low, so it only provides less than a Watt of heat.  You need to know the current draw of the relay or zone valve to compute the correct resistor value.
You can likely buy thermostats with this feature built in, possibly your thermostats have this feature and it has been mis-adjusted.  On 24 VAC thermostat systems, it may look like a shiny wire with a sliding contact on it.

If you go with an electronic thermostat, you need to implement the same function, possibly adding in a rate of temperature change to the calculation.

Jon

[elcaron]

Thanks for the answers so far.
Regarding the heating cycle: I have only installed a temperature sensor the day before yesterday, so I do not have full date. I have captured a full heating period, which started at 8am and 18.8°C (triggered while window was opened for ventilation) and ended at 4pm and 22°C. In the following 8h, the temperature dropped by only 0.5°C, so even with linear behavior, which there is clearly not, we are looking at 36h cooldown phase at current outside temperatures in Germany. I will only have more data in a few days, since I am having a party tonight, which will screw up the current cycle due to people and the fact that I opened the unheated wintergarden towards the livingroom that I am measuring.


@Hero999
I agree that PE has to be handled carefully. This is why I will not use it for commercial thermostats that require L,N and a 230V switching line over them. Did you see that this would be a complete low voltage line? It starts at the thermostat and goes uninterrupted to the central station, without touching any power wiring anywhere. I could instead use it for ground and connect it to PE (which I think is what some switching supplies do, so it should be fine). If anything, I thing having the 5V supply not floating would make it rather less safe.

@alanambrose, jbb
The warm water comes from a central heating unit with a dedicated ultra-efficient (dynamic) pump. All the valves are in one spot (see the link im my original posting, mine looks the same).
If you have ~30m bang-bang or PWM intervals is probably more a question of the implementation.

I want to send sensor values to the central station, and maybe info for the screen from the station to thermostats. If I go for wifi use with the ESP12s, I might also distribute the Wifi config from the central station.
The question with all data transfer solution is "Will it work over 20m unshielded, untwisted cable?" I don't think that I get two lines for differential signalling.

@jmelson
Mine don't have that and I am not going to solder it into mains myself. I don't really see the advantage in spending money on such a system. The thermostats will surely be 20-30€ each and I need 6. Cooking something up myself is well within reach and I could do so much more with that.

[elcaron]

I think I will invest the money and switch over to 24V actuators ... saver and easier to switch with mosfets. SSRs for <2W devices seem to be an issue.
I also think I will use ISO1540 IC for the lines, for protection and current sink capability.