"Hardware assisted" integration of Pentair Autotrol 760 water softener valve into Esphome and Home Assistant
Pentair Autotrol 760 is a controller used in water softeners/filters to govern softener regeneration based on amount of water used, hardness of the water, time of day and more. It features multiple cycles that run in sequence to do proper resin regneration (backwash, brine draw, rinse, downflow rinse, brine refill..)
- I wanted to see my water consumption in Home Assistant dashboard and generate statistics based on this data. The Autotrol controller monitors the water usage (hall-effect sensor + turbine with magnet) to count how many liters remainis before another regeneration is needed. Unfortunatelly there is no way how to extract the data. And I didn't want to use another water meter in-line.
- In my opinion the regeneration process is wasting a lot of water in two steps: At the beginning there is a 10minute backwash at the rate of 8liters/minute. The backwash is needed to "fluff up" (expand) the softener resin beads and to wash any dirt to the drain. In my opinion the 10minutes is too long and causes around 80liters of water wasted. Unfortunatelly the time is fixed and can't be changed in this model.
- Another wasted water is during salt draw and subsequent slow rinse of the resin bed. Once the brine tank is empty, the salt valve is closed and the resin is slowly rinsed to flush the salt into drain. Those two steps take around 70minutes in my configuration, while roughly first 15minutes is needed for the salt draw and the rest is for the rinse. As I observed the regeneration cycle, the rinse was sufficiently completed after 35 minutes (yes I tasted the expelled water to see if there is still salt being washed out). Even with all the salt flushed away the cycle took another 35minutes - wasting another 50+ liters of water. The timing set by manufacturer is in my opinion "too conservative" and causes a LOT of water wasted over the lifetime of the product. I wated to rectify this. I tried different "resin amount" and "salt amount" settings to lower this time but it didn't work as expected. By setting lower resin volume or lower salt amount, the cycle time got shorter but this also caused significant lowering of the brine refill time - resulting in insufficient amount of water(brine) for the next regeneration and improper resin regeneration.
By default in my configuration the softener used 250liters of water for regeneration. After implementing this modification the amount of wated water was lowered to only 130liters per regeneration and with more precise timing or TDS measurement it could be lowered even further!
By feeding the water usage into HA energy dashboard, you can see your monthly/daily/hourly water consumption
I decided to place ESP8266 microcontroller (Wemos D1 mini PRO) to emulate button presses of the Autotrol controller. If you press "square" + "button up" during regeneration, the controller advances to the next regeneration step. I exploited this and simulate button presses with the ESP8266 at the right times. I also read out the state of the motor that rotates the valve shaft to get the current cycle number. As for water consumption I also tapped off the signal from hall-effect water turbine sensor and count pulses in ESPhome
Warning
the disassembly is bit destructive as the casing is glued/ultrasonically welded. At the end you will need to glue it back together
Use plastic prying tools + picks to slide between two parts of the plastic. Continue around the edge - you will need to use a bit of force to break the plastic welds. Few drops of isopropyl alcohol seems to help with the separation.
After you are in, you see the bare board.
And the second side (after removing 4 screws):
Btw the square in the middle-top is supercap to keep RTC running and memory alive for few hours during power cut. After it gets discharged you must set clock again + you lose current "water remaining to next regeneration" as it is obviously stored in some volatile SRAM memory...
- Interfacing with the buttons is done by using small N-channel MOSFET transistor across the actual button. Buttons are NO with pullup to 3.6V. When pressed they short to GND. The MOSFET does the same when voltage from ESP8266 is applied to gate. I used breakout boards as the SO23 MOSFETs are too delicate for such purpose.
- Reading the motor status is tapped off at the TP 19 or R7. I placed small signal diode and 1.8k resistor in series. Additional pull down resistor (1.8M) is connected to mitigate the issue with essentialy floating input on ESP side. When the motor is turning, there is 3.3V going to the triac controlling the actual motor current.
- Water flow pulses are read from TP34. I again used small diode to mitigate any current flowing into the controller in case the ESP8266 misbehaving (during prototyping etc). Additional 1:1 resistor divider is implemented as the voltage signal is around 5.6V
Here is a pin mapping into the ESP8266
| Wemos D1 pin | Autotrol Controller |
|---|---|
| D1 | Square button |
| D5 | UP button |
| D2 | Regeneration button |
| D6 | Water flow pulse input |
| D7 | Motor input |
Using existing 3.3V power rail in the controller is not a great idea as the power path is not designed to handle any significant currents. The input 12V AC is half bridge rectified to around 17V DC and simple wimpy LM317 in SO8 package linear stabilizer is used (maximum current 100mA). The ESP draws around 80mA so this will not be feasible.
Using additional DC/DC or linear stabilizer is also not exactly ideal as I didn't want to load and sag those 17V with additional current as this voltage is used as power supply all through the unit as pull-up etc. And I am not sure how benevolent are those voltage margins...Also loading the transformer with only half-bridge rectified current is not so great either.
There is another full bridge rectified voltage rail used for chlorine generator (not supported in my unit but all components are populated). It has different ground and is isolated from the rest of the circuit using optocouplers etc. I used this to power my ESP through isolated DC/DC converter (rectified voltage is available at the second capacitor and test pads around, see images). I used URB2403MT-3WR3 but you can use any ISOLATED DC/DC with REGULATED 3.3V output with enough current capability and around 15-20V input. Mine has pretty wide 9-36V input
Warning
The DC/DC must be isolated as the GND from half-bridge rectified voltage and full-bridge rectified voltage must NOT be connected together!
I used silicone sealant (preferably neutral silicone) to fix all components in place. I don't like and don't advise using hot melt glue in long-term applications (although perfect for prototyping as temp. fix)! In my experince it does not work long-time and with thermal cycling gets deattached from surface
Warning
Paste tape to cover the gap between front case and LCD! Without doing so the silicone will squeeze between LCD and front panel and be ugly and visible!
Solder everything in place and then use silicone to set everything in place, let the silicone cure overnight...
Warning
Verify everything works fine before you glue the case back together! It is not easy to open and close after being glued shut...
I used few drops of acetone to all 4 corners of the case to glue/weld back together. I assume superglue will work fine aswell...
For the ESPhome code + explained principle of operation see /esphome