ESPCLOCK V2.0 - Design


I previously started a project to hack an analog clock to sync with NTP. I call the project ESPCLOCK, because it uses the ESP8266 chip (or more specifically, the ESP-12E development board for the ESP8266) to integrate with the analog clock.

From a usability viewpoint, I like the fact that there is no need to mess with timezones during setup. Basically when setting up the WiFi, the setup page grabs the timezone from your browser automatically. So setup is as simple as entering the WiFi credentials, and the clock will start keeping accurate time and handle daylight saving automagically.

However, in its current state, there are two issues with the clock:

1. The battery life is awful. Connected to a 2400mAh USB battery pack, it can only last for about 27 hours. A typical analog wall clock will run for at least a year on a single AA battery.

2. To deal with possible power loss, it writes the current clock time to EEPROM every second. Since the EEPROM has a limited number of write cycles, this is clearly not practical. We can of course simply not write to the EEPROM, but that means having to setup the current clock time after every power loss.

I have been thinking about tackling both problems in version 2 of the ESPCLOCK. While waiting for the parts to arrive, I am going to document the initial design I have in mind.

Instead of using only the ESP8266, which is rather power hungry, I am thinking of delegating the clock module to the ATtiny85. This is a diminutive chip that has a normal power consumption of 2.5mA and a sleep mode consumption of only 0.5µA. It is also extremely low cost, priced at slightly over $1 if purchased individually.

The idea is to have the ATtiny85 drive the hands of the clock, as well as communicate with the ESP-12 via I2C. This can be achieved via the TinyWire library. It will only need to respond to 3 commands from the ESP-12:
  • Get current clock time
  • Set current clock time
  • Set current NTP time
The ATtiny85 will effectively act as the "brain" of the clock and automatically catches up with NTP time if necessary. Given its power consumption parameters, it will theoretically last for about 200 days on a 2400mAh battery (assuming 200ms of uptime, and 800ms of sleep time per 1s cycle).

To deal with the EEPROM issue, a 0.47F capacitor will be connected in parallel to the power supply. The ATtiny85 will constantly monitor the voltage of its power supply. When power loss occurs and voltage drops below a certain level (say 3V), the capacitor will provide the ATtiny85 with enough residual power to write the current clock time to the EEPROM. That way, it can keep track of the current clock time accurately in the event of a power loss, without maxing out the write cycle budget of its EEPROM.

The ESP8266 will sleep most of the time, wake up every 30 minutes (probably make this configurable), get the current NTP time (and convert to local time), send it to the ATtiny85 via I2C, then sleep again.

I have ordered another ESP8266 development board, the Wemos D1 Mini, because it seems to have better unmodified deep sleep characteristics (low current draw of 170µA during deep sleep) compared to the ESP-12E. Assuming an average current draw of 70mA over a 10s active uptime (this is just a wild estimate; I could be totally off the ballpark here), it should theoretically be able to last 180 days on a 2400mAh battery.

Putting the two results together, the entire concoction should last about 95 days on a 2400mAh battery. Put it another way, a 10,000mAh battery should be able to drive the clock for a year on a full charge. Given the low prices of USB battery packs, or even a pack of 4 x 2500mAh NiMH rechargeable batteries, an Internet-connected analog clock that runs for a year on batteries alone seems less a pipe dream!
Wemos-D1-Mini
Wemos-D1-Mini
Wemos D1 mini

Comments

  1. Interesting. I've been working on a similar project, similar to this but with the addition of a RTC (DS3231). At the moment I've gotten total power usage down to an average of 0.75ma and with some design changes I think I can get less than 0.5ma in total. See https://github.com/liebman/AnalogClock. I found that the motor controller that I am using to drive the clock is a power hog and am working on some alternatives, BJT h-bridge or direct drive from the ATtiny possibly with PWM if that helps reduce power. Have you played at all with PWM driving the clock?

    ReplyDelete
    Replies
    1. No, I haven't. The designs I have been playing with always use software to drive the 2 wires originally connected to the clock's quartz crystal. I figure if this mechanism is able to run on an AA battery for a year, it's a good starting point. So my current thinking is to get the uC power consumption down to a minimum, and also to be able to save the current clock state when the battery starts running low. That way, once a new battery is connected, it has enough information to sync the clock back with NTP time.

      Delete
    2. I found that pulsing the clock with 3.3v uses a lot of power. My first option, using the motor controller make the pulse timing, and the faster adjust timing much easier by using 1.5v or 1.8v to drive the pulses. I also found that driving direct with 3.3v uses that about same amount of power and presents issues with the tick and adjust pulse widths not operating properly. My current experimental setup plan is using the PWM capability of Timer1 on the ATtiny85 and this weekend I hope to be able to run some tests. (I use a web server on the ESP and I2C to the ATtiny85 to be able to adjust the pulse timing on the fly). I like the idea of detecting powerful and saving the current clock position but wonder how much extra power that requires... I'll have to look at the ATtiny85 data sheet again.

      Delete
  2. Having some thoughts on NTP sync clock movement... Here is an idea which may possibly solve battery issues:
    Instead of quartz clock movement, what if you use a radio sync clock movement like DCF77 or WWVB signal receiver. That type of clock movements comes with a price range within 10 to 15 USD, have mostly the same pysical structure and have exactly the same structure on syncing the clock automatically but use DCF77 instead of NTP. So, on clock PCB, cut the wires of the ferrite antenna to prevent receive of DCF or WWVB signals, and connect the ESP8266 or ATtiny85 to the board which receives the NTP time from WiFi, converts the message to DCF signal type (coding only) and injects to the clock PCB. So the clock module still thinks this signal is a DCF signal. ESP doesnot need to work all the time, only once every let's say 24 hrs; get the NTP signal, convert and send to clock module. Only 1 AA battery lasts from 1 to 1,5 years on the DCF clock module itself. Any thoughts?

    ReplyDelete
    Replies
    1. Btw, the clock module itself is keeping track of the hour-minute-second hand all the time, so there's really no need to concentrate on clock side of the business.If only i had the electronics knowledge on hacking the DCF77 clock PCB :)

      Delete
    2. It's an interesting idea. Thanks for sharing!

      I don't think it's possible to "inject" the time into the clock PCB as you put it, since what's on the PCB is probably firmware that reads signal off the DCF77 receiver and decode the information.

      I did think it is probably more feasible to build a mini DCF77 transmitter and convert the NTP time to DCF77. Since our own transmitter is very near to the receiver, it's probably not a problem overwhelming the real signal (assuming it is present in the first place).

      However, after doing some reading on DCF77 transmission, I find it beyond my field of understanding and seem to require advance knowledge in RF transmission, antenna design etc.

      Delete
  3. Sorry for the consecutive posts; but here is a link to a sample module:

    https://www.aliexpress.com/item/Radio-Controlled-Wall-Clock-Quartz-Automatic-Movement-DCF-Germany-RC-With-Metal-Hands/1902348532.html

    ReplyDelete

Post a Comment

Popular posts from this blog

Adding "Stereo Mixer" to Windows 7 with Conexant sound card

Roomba navigation algorithm

Hacking an analog clock to sync with NTP - Part 5