A common occurrence in our house is bath time.
Another common occurrence is not having enough water for afore mentioned bath.
Our hot water is provided by a system boiler/unvented cylinder arrangement and I use a Nest to control the hot water (more on that later). The current schedule has it running the boiler for thirty minutes each day.
What is very annoying is that on some bath days we have ample hot water and on others, we don’t. We end up boiling the kettle to make up the difference. I don’t like the idea of just turning on the hot water as we might just be heating water that is already hot enough.
How much hot water do we have?
The first thing to figure out is how much how water we actually have. As I’ve mentioned, one some days we have ample and on other days we run out. Why does this happen?
Here are a few of the main factors that spring to mind:
- The duration of the showers in the morning.
- The ambient temperature in the loft.
- How much washing up we’ve done.

The image above shows the basic operation of an unvented cylinder. Cold water enters the tank at the bottom and hot water exits at the top. The coil inside the tank is connected to the boiler and is responsible for actually heating the water in the tank. The whole thing works off the principal that the hot water will sit on top of the cold water and be pushed out the top.
My plan, therefore, was to attach temperature probes to the cold water entry and hot water exit and see what I could measure.
I was also interested in trying to measure the effect of the loft temperature on heat loss.
The Sensor
With my requirement of three sensors, I set about building a circuit that included an ESP-8266 and three DS18B20 temperature probes.
The DS18B20 uses something called the OneWire protocol, which means you can connect multiple sensors in parallel and use a single GPIO to read all the values. I found an excellent tutorial on the subject here – https://lastminuteengineers.com/multiple-ds18b20-arduino-tutorial/


For power, I wanted to use an off the shelf power supply. As these are mostly 5v, I added an LV1117V33 LDO. This little device accepts up to 7v input and outputs 3.3v, which was perfect to power the D1 and the sensors. To this, I connected a barrel input.

Software
I had originally planning on writing my own software to read the values from the sensors and send via MQTT to my HA instance. I actually did write my own software and even got the auto-discovery working, but as I wanted a web UI running on the device, so I could change easily change WiFi and MQTT settings, going down the custom road made little sense.
I was vagly aware that the excellent Tasmota Firmware (which I use in lots of smart sockets) did have some form of sensor support, so I did some reading. It turns out that not only did it support the DS18B20 sensor, it supported multiple ones 🙂
I flashed it onto the D1 mini and, after setup, navigated to the Configuration page.
You can select Module type as Generic (18) to gain access to the various GPIO pins on the device. I had connected the data wire of my sensors to D4, which is GPIO2. I choose teh DS18x20 option.

Once I hit save, the device rebooted and, lo and behold, the temperature started flowing into the UI!
Housing
Knowing how dusty my attic can be, I wanted to house the electronics in something, just to keep it clean and tidy. I had some Hammond enclosures (https://www.hammfg.com/electronics/small-case) and one of them seemed perfect.
I drilled four holes. One for each of the temperature probes and one for the power connector. The probes are held by three IP68 glands. Not necessary at all, but I just had them and thought they’d suit.



Once I had it all assembled, I plugged it in. It was at this point I wished I had some sort of indicator light to tell me it was powered or connected to the internet. I had to wait a minute, but eventually the Tasmota appeared on my list of connected clients.
Installation
I have a pretty standard unvented cylinder. There are two pipes entering at the bottom, the cold water inlet and the hot water return to the boiler. The hot water outlet is at the top.




The third probe I just left hanging in the air, supported by a piece of furniture. My loft really is full of junk!
The data

From the three probes, I could have a guess as to which probe was were!
The 15.8 was the ambient temperature in the loft, with the 19 and 48 degree measures pointing to the inline and outlets.

My hot water turns on for 30 minutes each morning (30 mins because the Nest Thermostat won’t let me do anything shorter). I looked at the temperatures when I got up around 7am and I could see the top temp was at almost 60 degrees.
Next Steps
I’m going to let this setup run for a few weeks and note the days we run low/out of hot water.
Combined with better control of my hot water (I’m thinking of replacing my Nest’s control), I think there is an opportunity to reduce the time my hot water needs to actually be turned on.
Other things might be possible, such as extra hot water when we have guests over (using their connection to the house WiFi as an indicator of presence). Or ensuring these is hot water for washing the dishes at 6pm each night.
I’ll post an update, once I’ve got some findings!
Update 24th Feb
With the sensor installed for over a week, I pulled a graph from Home Assistant.

The green line, which reaches the highest values is the outlet temperature. You can see the rather pronounced peaks and troughs. The orange line represents the inlet temp and that does spike a little, each morning, when the heating is turned on. The red line is the ambient temp of the loft. It does vary by a few degrees.
From the graph, it’s very obvious when we’re very low on hot water!
I also find the temperature loss over the course of the day to be interesting. This is obviously a function of the ambient temperature in the loft, but also, I suppose, of the temperature of the incoming water.
I’m in the process of upgrading my Home Assistant instance and I hope to be able to retain more data than I can do right now. I’d like to be able to compare summer months to winter months, so I need a few months of stored data to perform that comparison.
More to come!