Calling for hot water, the recall

Project Logs

Collapse

Re-wiring
WJCarpenter • 04/04/2025 at 21:43 • 0 comments

Someone in the forum thread who owns a NaviLink unit was able to identify the connector needed for the RS485 interface. It's 5-pin JST-XHB or JST-XH. The "B" stands for buckle and is the little plastic piece that holds the connector firmly in place. For this particular equipment, the XH without a buckle holds pretty firmly. Someone also identified a source for the connectors with long pigtail wires for the 5 pins. They are readily available in lengths up to 30 cm.
That's the good news. I got some of those connectors with pigtails. I got the 30 cm length because I could always trim off some excess wire for tidiness. There are two bits of bad news. First, since I wanted to mount my device on the outside of the cabinet, I'd need to feed the cable through the knock-out grommet in the water heater case. That turns out to be about 40 cm from the socket on the control board where the cable plugs in. Second, the conductors inside the pigtail wires are pretty flimsy. It was really fiddly getting them stripped and then willing to stay in place in the screw-down terminals of the RS485 device.
I was able to get things assembled, which at least let me prove that the 12v power supply and the signals on the RS485 interface were as expected. I had to mount my device on the bottom of the cabinet, with wires hanging out and held in place by the cabinet cover weatherstripping. Yikes, like a barbarian!
Another common use for JST-XH connectors is in battery chargers for remote-controlled cars and other devices. There, the cables are called "balance cables" because of the way they work with mulit-cell lithium ion battery packs. As a trap for the unwary, the nomenclature of those cables uses a number that is one less than the number of conductors. So, for a 5-pin cable, it's called "4S". It's typically a single red wire and all the rest are black.
Balance cables are also readily available, but again in relatively short lengths. I didn't find any longer than the 30 cm cable I already had. But you can also buy cables that act as extensions, with a male connector on one end and a female connector on the other end. I bought a few of those in 20 cm length and strung them together to make an 80 cm cable. (When using these for charging batteries, there would very well be some rules about how long they can be, how many you can string together, etc. I don't think any of that applies for RS485 since it's intended to stretch over distances in noisy environments.) On the end toward the relay device, it was pretty easy to pull the connector off, leaving the metal pins that were already attached to the individual wires. Those made much better connections to the screw-down terminals of the relay, so I dropped my original pigtail cable out of the picture entirely.
Here is the final, mounted device. The Atom Lite on the RS485 base is on the right, and the relay is on the left. They are connected with a Grove connector cable (the shortest I had on hand). The device is mounted with some rare-earth magnets that I glued to the bottom.
For the relay contact wires, I just reached into my junk drawer an pulled out some wires that I thought would be easy to clamp into the screw-down terminal of the relay. They're pretty thick (blue in the photo) even though they need to carry next to no current. That also made them easy to bind to the screw terminals on the control board.
In my original implementation, I also had a simple momentary push-button wired to the control board in addition to the relay device. That was so I could troubleshoot things without going through any of the fancy-schmancy automation logic. After that first day, I never used that button again, so I removed it this time around.
Except for possibly pondering some of the additional information reported over the RS485 interface, this project is done. I don't intend to implement any additional controls from my electronics to the water heater.
Pump on, pump off
WJCarpenter • 04/04/2025 at 19:18 • 0 comments
My aim has always been to have the remote Atom Lite units blink yellow when the pump is running and stop blinking when the pump stops. Until now, I've been unable to do that, and I've just been using a simple timer that reflects the worst case value. It takes just slightly over 2 minutes for the NaviCirc cross-over valve to close when the water in my pipes starts out cool, so my timer was 2 minutes and 15 seconds. If the water is already hot, the cross-over valve stays closed and the pump hardly runs at all.

With these new RS485 reports, I still have not found a bit that definitively indicates that the pump is running. Some people have seen such a bit flip in one of the bytes, but they have different Navien units than I have (I think). That bit position does not change for me. There is a bit that I am using as a proxy for that. The bit indicates if hot water is flowing, which I call "Consumption active" in my sensors. There is also a report of the water flow rate. The bit for consumption active is set whenever hot water flows, which means it may or may not be due to the recirculation pump. If hot water is not actually flowing out of any faucet, shower, or appliance, then that bit is an accurate proxy for the pump.

I expect that, most of the time when the remote buttons are pressed, other hot water is not being used. Sometimes that expectation will not be met. If the dishwasher is running or if someone is washing their hands while waiting for the hot water, that consumption bit will not reset when the pump stops. For now, I'll just live with that.

My Home Assistant automation for starting the hot button cycle is pretty much as it was before. I merely substituted my new relay device in place of the prior relay device.

Here's the YAML for that:
```
alias: Tankless hot button
description: Press the button, get hot water
triggers:
  - entity_id:
      - binary_sensor.hotbuttonkb_button
    from: "off"
    to: "on"
    trigger: state
  - entity_id:
      - binary_sensor.hotbuttonmb_button
    from: "off"
    to: "on"
    trigger: state
  - trigger: state
    entity_id:
      - binary_sensor.hotbuttonrelay_al_atom_button
    from: "off"
    to: "on"
conditions: []
actions:
  - data: {}
    action: script.tankless_button_cycle
mode: parallel
max: 10
```
The script "tankless_button_cycle", on the other hand, did undergo some changes.
- As I mentioned previously, I made the Atom Lite RGB LED on the relay device give the same feedback as the LEDs on the remote buttons. That's in place of the simple mono LED on the board of the prior relay device.
- Instead of blindly waiting a fixed amount of time, I now wait for the consumption bit to reset. There's still a 2 minute timeout on that waiting. That's to handle the case where hot water is actually being used at the same time. So, worst case, after 2 minutes the LEDs will stop blinking. It's also possible that the recirculation pump will still be pumping at the end of the timeout. I haven't seen that so far, and the 2 minutes is based on actual timing of the plumbing in my house.
- There is a 5 second delay before the wait to watch the consumption bit. That delay serves two purposes, both related to race conditions. The less important reason is that it give me a chance to see the flash of green on the relay device when any of the buttons is pushed. The green is for the 500 ms of the relay simulating the button push. More important is that I might start waiting before the consumption bit has initially turned on. The consumption bit might be set for a very short time (or maybe not at all) if the cross-over valve is already closed because the hot water has already arrived. In that case, with the delay, the LEDs will blink during that 5 second delay and then stop.
Here is the YAML for that script:
```
alias: Tankless button cycle
sequence:
  - type: turn_on
    device_id: 7c8657a0b747b7584ec3775223c14045
    entity_id: ea0d9e3eb5631aa2876a1112f6c8c8cc
    domain: switch
  - parallel:
      - data: {}
        action: esphome.hotbuttonkb_set_led_hot
 continue_on_error:...
```
Read more »
Sensor clean-up
WJCarpenter • 04/04/2025 at 17:45 • 0 comments
Things have kind of settled down for me on this effort, so I'm cleaning up and finalizing various sensor activity.
- Generally, I'm only reporting things that are actually interesting to observe.
- I'm not too interested in reports of water heater configuration items. For example, there is a bit that indicates whether the built-in display shows values in metric or imperial units. That doesn't affect the raw measurements, which are always in metric. If I wanted to change the built-in display panel to the other measurement system, I'd go to the water heater and do it there instead of via a Home Assistant action.
- I dropped the sensors for the controller and panel version numbers. I'm not sure they were being computed/represented correctly anyhow. In any case, there was nothing interesting to do with those constant values.
- There is a bit indicating if the water heater has power. I don't know how that could ever be reported false. If the water heater did lose power, my ESPHome device would also lose power (it's powered by the RS485 interface).
- There is a bit that some have interpreted as "recirculation is running". I have never observed that bit to be true for my unit. I'm leaving that sensor in place (in case I ever observe it turned on), but I'm changing the name to something more generic.
- Scheduled recirculation is mutually exclusive with Hot Button recirculation. A single bit indicates if a schedule is configured. I reduced the two previous binary sensors to just one. A separate bit indicates if recirculation is possible at all. I assume this means that the unit has the recirculation pump and circuitry. I'm leaving that sensor in place. (I'm not completely sure about this since the Navien has 4 different recirculation modes: always on, intelligent, weekly, and hot button. So, it's most likely that there are 2 bits dedicated to indicating which of those is in force. There is another thing called space heating. I don't know exactly what it is, but I think it's probably independent of those other states. The installation manual says you can't have recirculation and an air handler together.)
- For the sensors that report temperatures, I convert them to Fahrenheit using ESPHome sensor filters. (I also moved the divide-by-2 scaling to the same filters.) That's for my convenience and saves me needing to do it on the Home Assistant side. I also noticed that the inlet temperature value had a lot of jitter, so I added a throttle_average filter of 2 seconds to all of the temperature sensors. The set temperature for the water is a constant that can be changed via the display panel, but I'm leaving it in place as a convenient reminder. I changed the name to target temperature because "set temperature" can easily be confused with a command to set the temperature.
- Likewise, I used an ESPHome filter to convert the liters/minute value to gallons/minute.
- The relationships among kilocalories, BTUs, cubic meters, CCF, and therms for gas are a little more complicated than I knew before. Cubic meters and CCFs (hundreds of cubic feet) are mere physical measurements with only a difference in units. kcal, BTUs, and therms are measures of heat produced by burning the gas. The amount of heat you get out of a quantity of gas depends on the nature of what's actually in the gas. Like many gas companies, mine charges me based on therms even though the gas meter is measuring volume of gas. Each month's bill inlcudes a "BTU factor" for the conversion. That changes from month to month, which I assume depends on where they get the gas supply and their measures of the heat producing capacity of the gas that month(-ish). It's a bit of a bother to involve the monthly BTU factor into these sensor readings, so I'm going to continue tracking CCF for the time being. Here is part of a recent gas bill:
- I'm not sure about the conclusion of the Navien reporting current gas consumption in kcal as opposed to some scaled gas volume measurement, but I'm keeping it that way for now and...
Read more »
Software reconfiguration, part 2
WJCarpenter • 04/01/2025 at 22:11 • 0 comments
Unless a more comprehensive Navien component for ESPHome arises, I'll stick with the pure YAML approach, which is what I'm doing right now. I started with the approach seen in https://github.com/evanjarrett/ESPHome-Navien/blob/main/navien.yaml, but I made several functional and stylistic changes. That starting YAML is great for the reverse engineering task, but it's a bit cluttered for a more refined interface.

The ESPHome UART bus component is intended to be used as plumbing by other components. In the pure YAML case, there is no other component. The UART bus debug configuration (unrelated to DEBUG logging level in ESPHome) gives direct access to the underlying I/O buffer. That is just right for extracting information and updating template sensors based on that data. Luckily, the RS485 packets for the Navien are newline-delimited, which lines up with how the UART bus debugging wants to see things.

This is the configuration part of the uart block (doesn't include the decoding logic in the lambda, which I'll describe later):
```
uart:
  id: uart_bus
  tx_pin: 19
  rx_pin: 22
  baud_rate: 19200
  data_bits: 8
  stop_bits: 1
  parity: NONE
  debug:
    direction: BOTH
    dummy_receiver: True
    after:
      delimiter: "\n"
    sequence:
      - lambda: |-
          UARTDebug::log_hex(direction, bytes, ' ');
          UARTDebug::log_int(direction, bytes, ' ');
```
When DEBUG logging is enabled in the overall configuration, all RS485 packet bytes are logged as hex and decimal values. I changed the delimiter to a space because I find that easier to read.
```
[10:56:54][D][uart_debug:114]: <<< F7 05 50 0F 90 2A 45 00 0B 01 0C 02 17 00 62 5D 28 00 00 00 00 05 00 00 CB 14 00 00 2C 02 E7 08 B1 50 14 00 E0 01 00 00 00 00 AA 48 00 00 00 00 58 F7 05 50 50 90 22 42 00 00 05 49 62 5D 2C 00 00 00 00 00 00 A0 BE 00 20 00 00 00 00 E0 01 01 00 00 02 B9 00 00 00 00 00 5A
[10:56:54][D][uart_debug:176]: <<< 247 5 80 15 144 42 69 0 11 1 12 2 23 0 98 93 40 0 0 0 0 5 0 0 203 20 0 0 44 2 231 8 177 80 20 0 224 1 0 0 0 0 170 72 0 0 0 0 88 247 5 80 80 144 34 66 0 0 5 73 98 93 44 0 0 0 0 0 0 160 190 0 32 0 0 0 0 224 1 1 0 0 2 185 0 0 0 0 0 9
```
I mainly adopted the decoding techniques and names from https://github.com/dacarson/NavienManager/blob/main/Navien.cpp, though I temporarily gave the sensors IDs and names prefixed with "dc_". For tidiness and performance reasons, I removed things that did not yet have an interpretation. Where things appeared to be booleans, I made them template binary sensors instead of template numeric sensors. For the controller and panel version information, I made them template text sensors (though they don't completely match what I see on the water heater's display). For the sensors I'm currently observing, I took them at face value, though I am pretty sure some of them are either wrongly decoded values or are misinterpreted as the wrong kind of information; that will be sorted out in a later step. For now, I am avoiding doing conversions of things like centigrade to Fahrenheit, though I do the trivial sorts of conversions needed for the Navien-specific conventions.

The structure of the code generated by ESPHome is mostly a large and sophisticated polling loop. Best practice is to avoid any long delays in a synchronously executing component. In fact, if anything takes longer than 30 ms, a warning is written to the console. I'm not sure what happens in practice, but in theory some timing-based operations could be delayed or missed. There are two sources of delay in the UART bus debugging section. First, the very act of using the UART logging methods (with nothing else) is enough to bring the timing to 50-60 ms. That can be easily suppressed by setting the overall logging level to something less verbose than DEBUG. I typically use INFO for otherwise-chatty stable code. Second, the logic for decoding the bytes and publishing updates to the template sensors can add another 100 ms or so if DEBUG logging is enabled,...
Read more »
ESPHome implementation alternatives
WJCarpenter • 03/30/2025 at 20:42 • 0 comments
Besides the collaborative reverse-engineering effort documented in the forum thread, there are at least 2 concrete ESPHome interface configurations. They differ a bit in approach. The github repositories for both include documentation of the packet formats. They expect bytes to be read/written over the UART interface, implying a separate component that translates the microcontroller's logic levels to RS485 signals.

The repository https://github.com/evanjarrett/ESPHome-Navien takes the approach of defining everything inside a YAML file. Debug logging within that configuration facilitates the iterative discovery process for reverse engineering. Most of the packet data bytes are made available as simple values, but some well-understood items are given semantically meaningful names. Some ideas for this repository seem to have been taken from or inspired by https://github.com/esphome-econet/esphome-econet.

The repository https://github.com/htumanyan/navien takes the approach isolating the low-level manipulation into an ESPHome custom component. That simplifies the YAML configuration and, in the long term, is more consistent with how other components are usually referenced in ESPHome. However, it comes at the expense of being a bit more cumbersome for iterative discovery. Only a handful of data items are exposed as sensors and switches for use in YAML. Adding more requires manipulating the code for the custom component and making related changes to the YAML.

There is also a non-ESPHome implementation in repository https://github.com/dacarson/NavienManager. It uses Arduino environment but is still useful as an information source.

My short-term aims are these:
- Only read from the water heater; never write. That's just a safety measure. If something goes wrong, it can't be the case that I sent a command that gave the Navien brain damage. The only thing I'm really interested in commanding is the press of the Hot Button, which I am doing with the tried and true relay technique.
- I want to be able to see when the recirculation pump turns off. That will let me give better user feedback for my remote button presses.
Taking those into account and the current state of the two GitHub repositories, I plan to start with the pure YAML implementation. I'll observe for a while and then decide what to do.
Wires
WJCarpenter • 03/30/2025 at 19:17 • 0 comments
In the forum thread I mentioned, other people have already figured out the connectors and wiring arrangement for the Navien board's RS485 interface. For example, in this post, user aruffell provides this clear picture of a custom cable they created:
The connector in the picture is a 5-pin JST-XH, though the slightly preferred connector is JST-XHB (the "B" is for "buckle", the integrated clip that keeps the connector from coming out). Both have been reported to work without issue.
In this post, user tsquared shows the wiring with the connector inserted into the socket on the Navien board:
RS485 uses differential signalling, and there is an A and a B signal line. Only 4 pins of the JST-XHB connector are used. Combining the info from the above pictures, the pin assignments are:
1. 12vdc
2. B
3. A
4. (not connected)
5. Ground
In addition to the RS485 wiring, there is also the relay wiring for the Hot Button feature. That's somewhat simpler since it's just 2 wires and orientation doesn't matter. They're connected to the relay pins Common and NO (normally open).

Software reconfiguration, part 1

WJCarpenter • 03/29/2025 at 04:39 • 0 comments

As I mentioned, my first step is to ignore the RS485 stuff and get the hot button mechanism working with this new hardware setup. Because I used ESPHome, things are mostly abstracted. Here are the changes I made based on testing the new assembly:

Changed the node name to "hotbuttonrelay-al" ("al" for Atom Lite) to prevent a collision with the existing hardware.
Modify the board definition and the referenced GPIO pins for the button and the relay trigger.
I don't have an overly convenient way to wire in the BME280 climate sensor, so I just deleted that definition along with the I2C bus. (I have plenty of temperature sensors around the house, so no biggie).
The Atom Lite doesn't have the usual on-board LED of typical ESP8266 or ESP32 development boards. Instead, it has a more sophisticated RGB LED. I redefined that as an ESPHome RGB light object using the fastled platform. (The compilation of the fastled libraries always gives a pile of warnings, but they don't seem harmful so far.)
The mini relay board does not have an LED for visual feedback, so I configured the Atom Light LED to turn on and off colored green with the relay activation. Since the Atom Light is the same component used for the remote hot buttons, I cloned that bit of YAML to have it use the same hot and cold lighting effects. The relay activation uses 500ms. All the other handshaking and activity happens faster than that, so I put in a 600ms delay before the hot lighting effect starts. That lets me see the brief green signal for the relay activation.

Here's the YAML so far:

# https://hackaday.io/project/202744-calling-for-hot-water-the-recall

substitutions:
  node_name: hotbuttonrelay-al
  RELAY:  'GPIO26'
  LED:    'GPIO27'
  BUTTON: 'GPIO39'
  log_level: 'DEBUG'

esphome:
  name: ${node_name}
  on_boot:
    then:
      # delay to allow other boot time stuff to settle down
      - delay: 5s
      - script.execute: set_led_cold

esp32:
  board: pico32

wifi:
  ssid:      !secret wifi_ssid
  id:        !secret wifi_ssid
  password:  !secret wifi_password
  power_save_mode: high
  fast_connect: on
  manual_ip:
    static_ip: !secret hotbuttonrelay-al_ip
    gateway:   !secret wifi_gateway
    subnet:    !secret wifi_subnet
    dns1:      !secret wifi_dns1
    dns2:      !secret wifi_dns2

logger:
   level: ${log_level}
api:
  encryption:
    key: !secret hotbuttonrelay-al_apikey
  reboot_timeout: 60min
  services:
    - service: set_led_hot
      then:
        - logger.log:
            tag: 'hotbutton'
            level: INFO
            format: "service: set_led_hot"
        - script.execute: set_led_hot
    - service: set_led_cold
      then:
        - logger.log:
            tag: 'hotbutton'
            level: INFO
            format: "service: set_led_cold"
        - script.execute: set_led_cold
    - service: set_led_off
      then:
        - logger.log:
            tag: 'hotbutton'
            level: INFO
            format: "service: set_led_off"
        - script.execute: set_led_off

ota:
  platform: esphome
  password: !secret ota_password

switch:
  - platform: restart
    name: "${node_name} Reboot"

  - platform: gpio
    id: i_relay
    name: '${node_name} relay'
    pin: '${RELAY}'
    restore_mode: ALWAYS_OFF
    icon: mdi:hot-tub
    on_turn_on:
      - light.turn_on:
          id: bright_light
          red: 0%
          green: 100%
          blue: 0%
      - delay: 500ms
      - switch.turn_off: i_relay
      - light.turn_off:
          id: bright_light

# Local test button. This goes through the relay logic, 
# so it's not an analog press of the water heater button.
binary_sensor:
  - platform: gpio
    name: "${node_name} Button"
    id: i_button
    pin:
      number: '${BUTTON}'
      inverted: true
      mode:
        input: true
    filters:
      - delayed_on: 10ms
      - delayed_off: 10ms

light:
  - platform: fastled_clockless
    id: bright_light
    name: ${node_name} LED
    chipset: SK6812
    pin: '${LED}'
    num_leds: 1
    rgb_order: GRB # required
    default_transition_length: 0s
    icon: mdi:led-on
    effects:
      - pulse:
          name: fast_pulse
          transition_length: 0.25s
          update_interval: 0.25s
          min_brightness: 20%
          max_brightness: 99%

script:
  - id: set_led_hot
    then:
      - delay: 600ms  # give the relay blink a chance to show
      - light.turn_on:
          id: bright_light
          red: 100%
          green: 60%
          blue: 0%
          effect: fast_pulse

  - id: set_led_cold
    then:
...

Physical assembly
WJCarpenter • 03/29/2025 at 02:50 • 0 comments
The RS485 unit and the relay unit both have LEGO-compatible holes in their bodies. I decided to make a 3D-printed slab with a few LEGO studs in the right places to take advantage of that. I started with this nice custom LEGO component generating script: https://github.com/cfinke/LEGO.scad. My requirement was a little unusual, so I made a couple of local modifications to that OpenSCAD script.
- The standard height for a LEGO stud in that script is 1.8mm. I wanted them taller to give a more snug fit, so I changed that to 8.0mm. That's no good for normal LEGO stacking, but it works great with the M5stack stuff.
- I change the code a bit to suppress the generation of the studs except in the four locations where I wanted them. I spent a little time trying to see if I could parameterize that, but OpenSCAD makes that at best kind of klunky and hacky. If I later figure out a reasonably elegant way to do that, I'll send a PR to the author.
The script does include a parametric scaling factor for the diameter size of the studs. Through trial and error, I settled on 1.01 (ie, 101%) to allow for filament shrinkage and still have a snug fit.
Here are the bottom sides of the components along with the 6x6 base I created.
And here is what it looks like when it's press-fit together:
I think it makes a pretty tidy bundle. I haven't yet decided how I will mount this bundle to the water heater, but I'll probably use the refrigerator magnet technology I tried before. This bundle is a bit lighter than the bundle from the earlier project.
Hardware decisions
WJCarpenter • 03/29/2025 at 02:34 • 0 comments

At the water heater end, requirements are minimal. Just about any ESP32 variant would fit the bill. I decided to use something from their Atom series since they are cheap and I already have some on hand. Specifically, I'll use an Atom Lite, which has two physical buttons and a single RGB LED.
They have a couple different RS485 interface components that will work with the Atom series. The cheapest option is their RS485 to TTL Converter Unit.
For a tiny amount more, their ATOM Tail485 - RS485 Converter for ATOM makes for a tidier package since it plugs directly into the Atom Lite.
In the end, I didn't choose either of those. I spent a few bucks more and chose the ATOMIC RS485 Base.
The reason for that choice is because it leaves the 4-pin port on the Atom Lite exposed. (It also leaves the USB-C port exposed for power, though once it's wired up to an RS485 circuit, that circuit can power everything.) I'm not sure if the reverse-engineered RS485 conversation will let me remotely "push the button" to call for hot water. For insurance, my first step will be to implement the hot button feature using a relay, as in the previous project. This Mini 3A Relay Unit interfaces to the Atom Lite via that 4-pin port.
The total cost for these three components is under US$20, which is not too bad for a one-off project.
Project motivation
WJCarpenter • 03/29/2025 at 02:12 • 0 comments

My original project has been operating for a while in a satisfactory way. I do still have one unsolved problem (knowing when the pump stops pumping), though for a while I've had some ideas about how to tackle that. In the meanwhile, some clever and industrious people over in the Home Assistant forums have been making progress on reverse engineering the wire protocol between Navien's NaviLink unit and the tankless water heater. I think that will give me the answer about the pump (not sure). To find out, I'll need an RS485 interface to the water heater control board.
If you look around for an RS485 interface board (as some of the folks in that forum thread have done), you come across some components from M5stack. Their stuff is particularly appealing for one-off projects since it mostly comes in reasonably attractive packages. I already have a little box full of assorted M5stack stuff, so I decided to make use of that and a few things I had to order from them to re-implement the hardware.