BoatDetect

About three months ago, the Hackaday Prize 2020 Conservation X Labs dream team developed a clever 3d printed waterproof mechanical release for an underwater electronics enclosure. The mechanism releases the electronics enclosure from a heavier base once an event occurs. The enclosure then rises to the surface but remains tethered to the base by a rope. Once at the water's surface, a signal can be transmitted.

Their concept was exactly what the BoatDetect solution needed. A way remain underwater during the data collection period so the device can not be tampered with, but then rise to the surface at the end of the data collection period so it is easy to access the data card and replace the batteries. From then on it was a case of "Engineer see good idea; engineer copy good idea and try to make it better."

Largely, I believe I am on track to achieve a simpler, more reliable, potentially cheaper pressure-rated waterproof mechanical release for hobbyist underwater electronics enclosures. This design uses a standard PVC stop valve, a continuous rotation servo, four 3d printed parts, and eight small fasteners. When the servo rotates the PVC stop valve on one side, a linear plunger advances on the inside of the valve. To remain submerged, the stop valve plunger captures the end of a pole or rope that is anchored to the seabed. Once the servo is actuated, the stop valve plunger retracts, releases from the pole or rope, and the enclosure floats the to the surface.

It was not until I understood the problem in it's simplest form as "actuate mechanical motion on one side of a waterproof seal and cause mechanical motion on the other side of the seal," that I identified off-the-shelf PVC valves as a great already proven solution. Then it was a matter of finding the right combination of PVC valve and actuation method. PVC ball valves are cheap but cheap 5V servos and stepper motors do not have enough torque to turn them. Stop valves are the second cheapest option but require more rotational travel than a 180 degree servo and more torque than a cheap 5V stepper motor. So a continuous rotation servo is currently the cheapest actuation method I could come up with for the PVC stop valve.

Another challenge with this solution was encapsulating the outside of the valve in the waterproof enclosure while leaving the inside of the valve open to the water environment. This is opposite of how PVC connections are made so I had to make a simple modification to a PVC reducer coupling to make this work. I believe it will still seal properly with PVC glue. The result below shows the servo and outside of the valve on the inside of another larger pipe fitting that will eventually be sealed.

I am looking forward to getting the enclosure fully assembled and run some underwater tests to ensure the mechanism releases properly.

The new capabilities that BoatDetect offers are valuable for wide range of applications.

Boarder enforcement of vast ocean areas. BoatDetect devices deployed along the perimeter of vast ocean areas such as marine protection areas can be used to identify when and where boat encroachment occurs. Agencies can use this information to develop more efficient enforcement boarder enforcement strategies.

Measure boat activity levels in waterways. Similar to road tubes used to count traffic on a roadway, BoatDetect can be used to measure the level of boat activity in a waterway. Knowing which waterways are experiencing the highest level of boat traffic can help governments plan and implement infrastructure improvements more efficiently.

Measure boat activity levels near shorelines. When deployed near a shoreline, BoatDetect can help predict if and when boats are being used to access a stretch of shoreline. Public and private shoreline property owners can use shoreline boat activity information to develop infrastructure that either discourages or better manages boat access to the shorelines.

Correlate changes in an environment with boat activity levels. BoatDetect can provide quantitative measures of boat traffic for biologists and conservationists. With this information, they can better answer questions like "Does boat traffic near shoreline during turtle nesting seasons influence turtle population outcomes?" and "At what boat activity levels does significant shoreline erosion begin to occur?" 

I assembled the BoatDetect proof-of-concept prototype today and performed a few audio tests in my bathtub. 

The enclosure is a pool noodle wrapped around off-the-shelf 3in PVC fittings and pipe. A piezo diaphragm is super glued to the inside surface of an end cap fitting. As a leak check, I left the enclosure in the bathtub for about an hour. A the end of the hour I did not find moisture on the inside of the enclosure. So at this point I am confident that the PVC glue joints are watertight. 

I then ran two qualitative audio tests with and without a parallel resistor and diode limiter on the piezo diaphragm output signal. During the tests I used two approximate high and low intensity sounds in the water to confirm that the microphone output signal was sufficient to be picked up by the recorder. I created the high intensity sound by hitting two rocks together. And the low intensity by slashing the water with my hand.

It was easier to hear the quieter splashing sounds on the recording playback without the parallel resistor and diode limiter installed. 

Next step is to make a few small enhancements to the prototype to make it easier to use out on the lake. I am very much looking forward to capturing some real boat motor audio with this device and analyzing the data to understanding what boat motor audio signature could be reliable for boat detection.

In parallel with getting this device out on the lake, I am looking into a method to generate sound at known frequencies and power levels for an evaluation of the piezo microphone sensitivity. Then back to the lab ("cough cough bathtub") to perform this testing.

The current detection strategy for the BoatDetect device is make-or-break on the assumption that boats are the loudest sound source at unique frequencies in an aquatic environment. The proof-of-concept prototype that I am working towards is meant to collect the audio data needed to either validate or invalidate this assumption. If invalidated, it is possible that an alternative strategy for boat detection could be developed from the test data. 

The proof-of-concept prototype is simply a piezoelectric diaphragm connected to a battery powered audio recorder. All in a buoyant waterproof enclosure. The current plan is to start recording, place the device in the water, and drive an recreational motor boat by the device at different distances. Then collect the device and analyze the sound data. Getting data at different distances will help determine at what distances the motor boat remains louder than the ambient noise and can be detected by the device.

Current plan for the buoyant waterproof enclosure is to simply wrap pool noodles around a 3in PVC pipe. 

Two of the most challenging features of any in situ environmental measurement data logger, like the BoatDetect device, is how to retrieve data from the device and how to replenish it's power source. The technologies to achieve these tasks can become particularly complex and expensive when deploying the device underwater. 

To meet the current cost target for the BoatDetect device, the strategy is to have the user locate the device by GPS coordinates (initially recorded by the user at deployment), physically retrieve the device, download the data from a memory card, and replace or charge the batteries at every battery recharge interval. While I generally understand this to be the correct strategy for the BoatDetect device, I believe a significant change to the device is needed to make this process more convenient for the user. 

Up until now, I have been proposing that the BoatDetect device be tethered to the bottom and float underwater at a distance from the surface. The main inconvenience of this approach is that the user has to visibly locate the underwater device and "fish" the it out of the water to retrieve it. If for whatever reason the device came untethered from the bottom or shifted location from initial deployment, the user might waste a significant amount of time looking for a BoatDetect device that simply is not at that location anymore. And if the water is too murky, the user has no chance of retrieving the device once deployed.

The operational inconveniences of the current approach have convinced me that a surface level indicator of the device's location is required. Even more so, placing all components at the water surface in a "buoy" would make the batteries and memory card immediately accessible and open up many opportunities for future developments such as visible and audible warnings to boaters, solar panels, transmitted GPS location, and wireless data transfer. The main challenges that I anticipate with a buoy solution is that the device now needs to be visible to boaters at night and is more likely to be tampered with. This change in approach will increase the material cost of the device beyond the initial $40 USD target, but the reduction in operational costs is likely to offset the material cost increase. 

A common method for vessel detection (primarily for collision avoidance) in the marine industry continues to be AIS (automatic identification system) in combination with radar. As part of AIS, each vessel transmits position, course, speed, and a unique signature that is received by other nearby vessels. As you can imagine, these systems are very accurate but also very expensive.

Hydrophone arrays and advanced camera systems are some of the lower resolution solutions available for detecting and tracking boat traffic from a single vantage point in areas like bays and waterways. Both are still relatively expensive, require a significant amount of data processing and consume enough power to require an expensive solar energy/battery system.

The BoatDetect device delivers a low power boat detection method at a fraction of the cost of existing systems. The caveat being that the BoatDetect device only has the resolution of detecting whether or not a vessel is in the vicinity. Even so, this level of resolution can still be useful to conservation groups and individuals who operate under a limited budget. 

I have transitioned to a PVC pipe fitting enclosure for the BoatDetect device. The previous enclosure, a 1000ml wide mouth water bottle, did not have a flat surface to mount the piezo diaphragm to. I considered pouring a small amount of epoxy resin in the bottom of the 1000ml water bottle to create a flat surface, but decided against it because of the added cost and hassle of working with epoxy. At $9.60, the PVC enclosure is only $3 more expensive than the 1000ml water bottle. It makes up for the added cost with a wider access port and is potentially easier to source than the bottle. The wider access port allows a little more space for the electronics assembly and certainly will make it easier to access the batteries and SD card. 

I superglued the piezo diaphram to the inside surface of the bottom plug. It did take me a while to find a PVC fitting that had a flat surface to mount the piezo to. When I get around to developing a build-it-yourself guide, I will recommend that builders start by finding a PVC fitting with the flat surface and build up the rest of the enclosure around that fitting as I have done here. 

I have some reservations whether the DWV (drain, waste, vent) threaded connection will be be able to hold pressure at depth. As the name "DWV" implies, they are not used for high pressure PVC installations. On top of that, the threaded connection would also likely require that a thread sealant be applied and inspected every time the device is opened up. The alternative is a wingnut compression pipe plug which does have some design history with other DIY underwater enclosures. I just checked price and looks like the compression pipe plug is actually a little cheaper than the combined price of both the DWV threaded plug and threaded spigot fittings. It would require a small length of PVC pipe to fit in but that ought to be cheap. Well then, the compression pipe plug is roughly equivalent on cost and a clear winner on both pressure rating and ease of installation/removal. I will buy the small section of pipe needed this week and get the enclosure glued together for testing.

I completed a BoatDetect breadboard data logger prototype with all the major functional features. It includes a real-time clock module, SD card module, and an Arduino Nano. 

At a high level, the system repeats this sequence of events during operation:

1. the Nano enters sleep mode
2. the real-time clock module wakes up the Nano after a configurable amount of time
3. the Nano reads an analog input pin
4. the Nano writes the timestamp and analog input value to a file on the SD card
5. repeat steps 1 to 4

The sleep/wake up strategy reduces the power consumed by the Nano. 

I plan to implement a few more power reduction strategies in the coming weeks:

• add a transistor on the SD card power, only power the SD card when needed
• accumulate logged timestamps and analog input data in SRAM, write this data to the SD card all at once (reduces the amount of time that the SD card is powered)
• remove the power LED from the Arduino
• remove the voltage regulator on the Arduino and replace it with a regulator that is more efficient at lower currents
• remove the power LED from the real-time clock module

Overall, I am very pleased with how this prototype operates and excited by it's simplicity. I was able to get everything wired and working in a little over three hours. I leaned heavily on the Arduino data logger designs that others have posted online.

The first iteration of the piezo diaphragm microphone circuit includes:

• piezo diaphragm used as a microphone (represented as the AC source and 15nF capacitor)
• high pass filter with 250Hz cutoff
• 40dB amplifier
• low pass filter with 1kHz cutoff
• half-wave rectifier with a smoothing capacitor

The microcontroller will wake up at 2-4 second intervals, measure voltage at the rectifier, and write the analog voltage reading and RTC timestamp to the SD card. 

Parts are on order and should arrive in 30 to 50 days.

After several hours of internet research I have been able to reduce the total material cost of the BoatDetect system from $37 to $24 ! The resources available online for developing a low-cost, low power Arduino data logger are just fantastic. This device may be my own design but it is certainly built on the shoulders of hackers across the internet that have shared their knowledge freely.

The most significant changes are:

• switch from a Arduino Nano to an Arduino Pro Mini 3.3V for reduced power consumption
• reduce micro SD card storage size from 4GB to 128MB for reduced cost and potential for reduced power consumption
• Tiny85 microcontroller no longer needed because of low operating current of Arduino Pro Mini
• reduce the number of 3400mAh 18650s needed from four to two and still be able to operate for four weeks
• removed 3D printed components to reduce cost and avoid need for access to a 3D printer