Underwater Retrieval System

Team 24: Andy Le, Artiom Lisin, Jessie Preteroti, Ozzie Kirkby, Angelo Lu

The device will operate while fully submerged beneath the ice formation to avoid positional interferences. The device can be activated using a remotely transmitted acoustic signal and resurface by releasing a tethered, buoyant section of the device. This section will carry a microcontroller-based system with a GPS receiver and a point to point transponder to the surface. A handheld transponder can then be used to read where the device has surfaced, allowing for retrieval.

Operating Mode

The device and payload are deployed underwater and left for up to a year.

Recovery Mode

The remote actuator has received a signal, triggering the release mechanism. The spooled is then undone, and the passive buoyant top floats to the surface.

These sketches were then developed into the design below. The frame will be constructed of PVC due to its buoyancy capabilities when sealed with air and low-cost factor. Some of the design considerations and analyses are listed below.

Passive Buoyant Top

Release Mechanism

The release mechanism contains 100 m of braided rope spooled onto a rotating shaft. The shaft rotation is locked by a ratchet wheel and a pawl before deployment. Upon deployment, a waterproof servo motor will rotate the pawl to unlock the shaft; releasing the spooled rope.

Surface Communication

Two designs of the electronics were considered. The original design uses a HC-12 for surface communication and an ATmega328P microcontroller (MCU) for logic. The revised design uses a RFM96 for surface communication and an ATmega32u4 MCU. Conveniently, Adafruit sells a board that integrates both chips on a single prototyping board.

Both designs use the same GPS receiver (MTK3339 on a Adafruit Ultimate GPS breakout), have a similar on/off latching relay and linear solenoid activation mechanisms at the block diagram level. Due to logic level differences between the boards containing the ATmega328P and the ATmega32u4, supporting components, such as resistors, transistors, and diodes, are changed between the designs.

Both chips are similar in cost. The RFM96 is more flexible and has more features (i.e. automatic retry).

HC-12RFM96

433.4-473.0 MHz

433 MHz

~4Wh over 7 days, 5% active time

~3.5Wh over 7 days, 5% active time

~1 km range with line of sight

2 km range with line of sight, 20km theoretical with directional antenna and tuning

Proprietary communication protocol

Open source communication protocol (LoRa)

Serial bus to MCU

SPI bus to MCU

The ATmega328P and the ATmega32u4 both have similar software and electrical characteristics, and either are capable performing of what is required in this project. As we are utilizing a prototyping board the choice comes down to convenience, cost, and size. ATmega328P is the MCU in the popular Arduino Uno, the board originally chosen. This board is larger and more expensive than the ATmega32u4 powered Adafruit Feather 32u4, which contains the LoRa radio selected as well. For these reasons, the ATmega32u4 was chosen.

As a bonus, this development board is compatible with any of the “FeatherWing” line of accessories at Adafruit to add features easily if needed by the researcher (e.g. temperature sensing, data logging, etc).

Battery Monitoring

The voltage of a lithium polymer battery (LiPo) relates to the remaining discharge capacity of the battery through a known, temperature-dependent discharge profile. By monitoring the voltage of the battery, we can turn off the recovery mode electronics when the battery near empty to prevent damaging the LiPo through over-discharging.

A single LiPo cell is usually designed to operate safely between 3-4.2 V. A 3 cell (3s) LiPo battery, which the circuit was designed for, therefore operates between 9-12.6 V. Using a voltage divider made of two resistors, we can create an output that in the supported range of the MCU’s analog to digital converters (ADC), 3.3 V in this case. To be safe, we will use a maximum V_in of 13 V so we can detect unsafe over-volt conditions.

V_{out} = V_{in} \times \frac{R_2}{R_1 + R_2} \rightarrow R_1 = 2.93 R_2

(1)

Choosing resistors with higher resistances will decrease power consumption. However, we are limited by the ADC on the ATmega32u4, which is “optimized for analog signals with an output impedance of approximately 10kΩ or less.”

\left(\frac{1}{R_1} + \frac{1}{R_2}\right) = 10k\Omega

(2)

Solving equations (1) and (2), we are limited to a maximum of R_1≈39.29 kΩ and R_2≈13.41 kΩ. 13 kΩ and 39 kΩ resistors are widely available and can be used, corresponding to a constant draw of 0.23 mA at 12 V. Accordingly, the recovery mode electronics should be turned off when the ADC reads a voltage under 2.25 V.

Payload Frame

Frame Optimization

Since the payload sensor might not be buoyant, it will be held within a buoyant frame. This frame's size was then decided based on a buoyancy model derived from a force balance. Assuming that it will be a rectangular cuboid, the following model was able to be developed:

f_{L_1} (R, ρ_{pvc},ρ_{air}, ρ_{water},t,m_{payload}, L_2) = \frac{m_{payload}}{4\pi\left[ρ_{water} R^2-ρ_{pvc} (2Rt-t^2 )-ρ_{air} (R-t)^2\right]} - 2L_2

where R is the pipe radius, ρ_pvc is the density of the pipes, ρ_air is the density of the gas inside the pipes, ρ_water is the density of the water that surrounds the frame, t is the pipe thickness, m payload is the maximum weight the frame must support, and L2 is the lengths of the side. The following variables were considered to be fixed at the following values, which left only R and t as design variables.

Variable	Description	Value	Unit
$\rho_{pvc}$	Density of the PVC used	1380	$\frac{kg}{m^3}$
$\rho_{air}$	Density of the air sealed within the PVC	0.178	$\frac{kg}{m^3}$
$\rho_{water}$	Density of the water that surrounds the enclosure.	997	$\frac{kg}{m^3}$
$L2$	Side and height of the frame. Chosen since the maximum supported diameter is 10cm.	0.3	$m$

This leaves only R and t, pipe radius and pipe thickness, as design variables. Since McMaster Car only offers PVC is distinct combinations, there is a finite set of solutions. Using their Dark Gray Unfitted pipe , as the product, the following set of solutions emerges:

Thus, from the graph it is clear that the first four are infeasible in terms of practically since their lengths at those pipe specifications would all be larger than 3m which would be difficult for a single researcher to transport. This leaves the fifth and six pipes from the left, which have the following specifications.

Pipe radius [m]	Pipe Thickness [m]	L1 [m]	Cost per 5ft	Internal Pressure Rating
0.017	0.05	1.412	$8.64	630 psi
0.021	0.05	0.3484	$12.24	520 psi

Looking at the length, L1, it is clear that the sixth pipe combination is ideal, since it results in almost a 1m reduction of length (improving its ability to be transported). This increase in pipe radius comes at a larger cost and lesser internal pressure rating, however both do not pose any serious issues.

f_{L2}(0.021, 0.005) = 0.28 m

Based on this analysis the cuboid device holder must be at least 0.35m in length and 0.28m in height and depth. To provide a slight safety factor and nicer dimensions to work with, the numbers will be rounded to 40cm and 30cm respectively.