Project Logs

Collapse

• Passive Gas Meter Reading

  ydrip • 09/09/2023 at 21:08 • 0 comments

  One of the larger goals for this project is to build an ultra low power platform to measure a variety of sensors. The next logical step after water meters is measuring gas meters for a few reasons. There are no passive devices on the market that do this for 'dumb' mechanical meters as far as I know. Second, natural gas companies usually charge months after a home owners usage (similar to water), which can lead to surprise bills. However, gas meters work a bit differently than water meters, which presents new challenges.
  

  Source: Wikipedia
  

  Basic mechanical gas meters use a diaphragm that expands and contracts with the flow of gas. This in turn moves levers and gears to eventually turn a mechanical dial indicating the usage. There are no magnets in this design so I decided to measure the change in the earth's magnetic field caused by the moving metallic parts instead.
  

  In the previous two post I discussed how the earth's magnetic field was adding a variable DC offset depending on the orientation of the sensor. In this case, I'm using it to my advantage. I'll do a more detailed write up on how I modified the design to do this in the next post. For now, check out the video demo above.
  

• Ultra Low Power Design with Mixed Signal Programmable Logic

  ydrip • 08/31/2023 at 01:19 • 0 comments

  The most interesting aspect of YDrip is likely its programmable logic. This is where a lot of the heavy lifting happens, including tasks such as leak detection and revolution counting. While it's possible to use a low-power microcontroller as an alternative, the drawback is that it would significantly increase power consumption. This isn't ideal for a product designed to remain operational over many months. Some microcontrollers, like Atmel and TI's MSP, come with customizable hardware blocks capable of performing basic functions like hardware-based counting. However, their capabilities are somewhat limited. This is why I ultimately decided to opt for a complete hardware solution.

  My initial approach was to use discrete logic chips until I found the GreenPAK product line from Renesas. Specifically, the SLG46826 model. It's reconfigurable, energy-efficient, and cost-effective ($1.60 for individual units). The package includes features such as analog comparators, counters, logic gates, oscillators, and more.

  Analog to Digital

  
  

  Starting from the left, the sine wave from the TMR sensor is fed into one input of the analog comparator. The other input is a reference voltage of 1600mV, which is the midpoint of the signal. Any voltage from 32mV to 2V can be selected in increments of 32mV. In earlier designs I used two comparators. One for each differential output of the TMR sensor and XOR'd them together. This provided no benefit so I removed the second one. Using a differential signal would help if the difference between them was amplified, but I was not doing that at this point. Here is an example of a differential signal:

  The red signal is the difference between the two signals above it and is twice as large. This improves the signal to noise ration and shows why differential outputs are used.

  Pulse Counting and Leak Detection

  The high/low output of the comparator splits into two paths. The top path measures the total water usage. It consists of two programmable 8 bit counters. When they overflow that triggers a third counter configured as a one shot counter to send a short wake-up pulse to the ESP32S3.
  
  The bottom path is the small leak detection logic. It low pass filters and counts the pulses. This filter is configurable via I2C. That value gets stored in two 8 bit counters and another one shot counter is used to send a leak interrupt signal.
  

  Wake-Sleep Controller

  A critical part of achieving a long battery life is duty cycling the magnetic sensor. It draws about 300uA. The wake-sleep controller can duty cycle power to the sensor and have the analog comparator only sample during 'on' times. This brings the system power down to only 55uA! See the previous post about power consumption. The main clock for the system is brought in from an external 32kHz RTC because the internal low power 2kHz oscillator is a little too slow. 
  

  Clock Divider

  Finally, a series of flip flops are used to divide down the 32khz clock to 64Hz. This slow clock is used to detect long running leaks by the low pass pulse filter.

  Conclusions

  By utilizing a mixed signal programmable logic IC this design was able to achieve a standby current of 55uA while also maintaining a variable wake-up time and leak threshold. Almost all of the logic blocks have been utilized in the SLG46826 making maximum use of it's capabilities.
  

• V2.5 Prototype - Adding Mag Sensor Amplification

  ydrip • 08/24/2023 at 19:35 • 0 comments

  Sponsor

  
  

  This version of YDrip was kindly sponsored by PCBWay. They have been my PCB vendor for version 1 and 2 so I'm thrilled they decided to sponsor the manufacturing for v2.5. I've used them in the past because their website provides real-time updates with the status of your product and their support engineers provide great design feedback to make sure their are no manufacturing delays. The ordering process is also extremely easy with their KiCad plugin, which automatically packages your board design, opens a browser, and adds it to your shopping cart.

  This was my first time ordering 3D printed STL parts from PCBWay and I'm very satisfied with the result. The 0.8mm lip that connects the two halves of the enclosure printed perfectly along with all of the ribs for rigidity. Their part finishing is also well done with no signs of support marks or warping. The PCB quality is what you've come to expect with no flaws and good enough dimensional accuracy to friction fit into the case. Thanks PCBWay.
  

  Electronic Design Changes

  The ALT021-10E is an ultra sensitive magnetic sensor (0-0.25 mT). It's so sensitive that I've been able to use it without an amplification stage which has reduced BOM costs and complexity. Unfortunately, I've run into a few problems.

  1. The earth's magnetic field can interfere with measurements depending on the device position.
  2. Some water meters have very weak magnets that are not measurable with the current design.

  To fix these issues, I decided to add an op-amp to amplify and offset the output signal. I chose the SLG88103104 from Renesas which has a stunning 375 nA quiescent current per channel. It also has a shutdown pin which I can duty cycle with the sensor, but it almost wouldn't make a difference in the total power budget.
  

  The opamp is setup as a differential amplifier with both sides of the wheatstone bridge sensor going to the +- inputs of the amplifier. (Ignore the resistor values. They are just place holders)
  

  A better design would use an instrumentation amplifier to avoid loading the wheatstone bridge which could reduce accuracy if not take into account. One way to mitigate this is by using much larger input resistors than the bridge resistance (10k in this case). Here is an example from the datasheet. This app note from TI goes into more details about the tradeoffs for each approach.
  

  This is great, because instrumentation amps cost more and drain more power. The addition of the amp solved my initial problems, but I ran into new ones.

  New Problems...

  Up until now I've been reusing some parts like the mag sensor since they were out of stock at the time. This meant I didn't have to deal with device to device variances which would impact my amplifier design. Silly mistake.   
  

  The ALT021 wheatstone bridge DC offset variance is +- 20mV. We can see that in the datasheet and following plots.

  Green & Yellow = Sensor 1

  Pink & Blue = Sensor 2
  

  Sensor 1 has a difference of 27mV between differential pairs and sensor two has a difference of 9mv. Here are the same devices rotated by 90 degrees.

  There is a 2mv - 4mv change due to the earth's magnetic field.
  

  This last plot is with a magnet spinning near the sensors.
  

  In a perfect world both signals would be centred around Vcc/2 or 3.3V/2 = 1.65V with no magnetic field. The problem starts when you try to amplify the difference between the two differential pairs. The difference in DC offset between sensor 1 and sensor 2 will also get amplified resulting in an output signal with an even larger offset. Here is a simulation with sensor 1.

  The bottom two lines are 50mV sine waves with 180 degree offset at 1.41V and 1.68V (sensor 1). We can see the output signal (green) is offset by about 2.6V. Let's take a look at sensor two which...

  Read more »

• Low power small leak detection

  ydrip • 07/19/2023 at 15:25 • 0 comments

  I have made modifications to the programmable logic to identify minor leaks and trigger the ESP32-S3 to alert the user, which will have a significant positive impact on battery life. Software-based detection can identify large leaks when the ESP32-S3 wakes up to report regular usage. For instance, if the usage remains consistently high for an extended duration, however this is not implemented yet. All variables can be customized through the YAML file.

  Here is what the Home Assistant Card looks like

  
  Source: https://github.com/YDrip/ydrip-esphome/commit/77c73d2ae2ac85a2539d3d88513e001614a7b9a4
  

• Green Hacks Contest Submission

  ydrip • 07/03/2023 at 23:05 • 0 comments

  
  

  I wanted to share that this project will be submitted into Hackaday's Green Hacks contest. This opportunity came at the perfect time because the costs for the development of version 3 of YDrip will be significant. This version will be weatherproof and potentially capable of reading diaphragm-style gas meters (some testing is still required). To achieve this, I am planning on injection molding the enclosure, which has a significant upfront cost.
  

  The theme of the contest is to:

  Build a hardware solution that recycles electronics, makes energy harvesting or storage more efficient, or contributes to the sustainability movement.

  YDrip contributes to the sustainability movement by giving home owners and business the tools to reduce water consumption. It does this in two ways. The first is by detecting leaks before they turn into costly problems. The second is by giving users real-time water usage data to empower individuals to make informed decisions to conserve water. By visualizing consumption patterns and setting goals, individuals can modify their behavior and adopt water-saving practices. This can include water decisions about whether it is time replace that appliance or how long to water the lawn.

  A Unique Solution
  

  The YDrip project addresses a gap in the open source market by offering a versatile solution. Its low-cost and low-power design, combined with programmable digital logic, enables it to be repurposed for various applications. For instance, the hardware's fundamental elements can be utilized to monitor gas meters, and by adding an IR LED, it becomes capable of measuring smart electricity meters. This flexibility and adaptability make YDrip a valuable option for a range of monitoring needs.
  
  

• V2 Prototype - Source Code Release

  ydrip • 07/02/2023 at 01:31 • 0 comments

  You can find the initial release of the ESPHome plugin for YDrip on Github today. The software is basic, but functional. It puts the device to sleep when connected to ESPHome. The wake up period is based on the number of water meter magnet revolutions i.e consumption. Leak detection is still a work in progress.

  Version 2 of the hardware and enclosure can also be found on Github.
  

  
  
  

• V2 Prototype - Power Measurements

  ydrip • 06/30/2023 at 22:43 • 0 comments

  It can be difficult to estimate battery life for IoT devices, but it's still a good idea to take measurements during development to see if you're even in the ballpark. Commercial water meters that are used by cities last 10-20 years. Some can even use the rotation of the magnet to harvest energy. YDrip won't have that advantage due to the farther mount distance. WiFi will also be a challenge because it wasn't designed for low power, however WiFi 6 and the new ESP-C6 may help with that in the future. For now, I'm targeting months of usage on a single battery. Let's see if that's possible.

  The first measurement shows the current draw (at 4.5V) while the ESP32 wakes up from sleep and transmits the water usage. Average current is 110mA over 6 seconds with fast connect enabled. This is fairly high and will limit how many times the device can update the server, but this is expected. There is lots of room to optimize here.

  Read more »

• ESPHome and Home Assistant Integrations

  ydrip • 06/24/2023 at 19:56 • 0 comments

  
  

  Demonstration of ESPHome and Home Assistant integration.
  

• Water Meter Test Setup

  ydrip • 06/23/2023 at 22:54 • 0 comments

  This is the test jig I use for testing YDrip when I need to play with different variables like rotation speed, mounting angle and distance. It consists of a dissembled water meter magnet attached to a DC motor. The brass housing sits on top of both to simulate a real instillation. The The ATtiny412 counts the revolutions using a hall effect sensor and acts as a source of truth.
  

• V2 Prototype - Battery Powered

  ydrip • 06/19/2023 at 19:12 • 0 comments

  The main goal for V2 was making it battery powered which turned out to be extremely difficult. First, the PCB and case was expanded to make room for three AA batteries.
  
  Then I started replacing/adding a few components:
  

  • RT9080-33GJ5 regulator due to its low quiescent current.
  • PCF8563 RTC as a low power clock source for the circuit and time keeping while the ESP is sleeping.
  • ESP32-S3 replaced the discontinued S2 module
  • RGB LED + lens for user feedback
  • MB85RC64TA FRAM for storage
  • ALT021-10E magnetometer

  
  The major hurdle was finding a replacement for the LIS3MDL magnetic sensor. Having already exhausted my off the shelf sensor solutions I had to make something from scratch.

  Read more »

View all 12 project logs