Now that the boat is almost finished, I can start planning the launch. It’s reasonable to wait until summer when there’s more sunlight and better weather conditions. The shortest path across the Atlantic is from Newfoundland to Ireland, but it also has some unique challenges. It looks like I’m forced to choose between hurricanes and icebergs. There’s a tiny window between June and August when it’s safer to cross the ocean.

There are still interesting things to do within the next months and I will keep you updated. Here’s a short to-do list:

• Testing the boat. There will be extensive testing in a lake or river in different wind conditions and also some ocean tests before the launch.
• Collision avoidance. All ocean vessels transmit their position and course via AIS transponders. My boat will spot nearby vessels and change its course if it’s necessary to avoid them. Even if fishermen approach the boat, it won’t be an easy catch. The boat will do everything it can to run away!
• Sending pictures via satellite Internet – a very cool feature that will allow us see what’s happening with the boat. The boat will transmit real-time low-resolution pictures which will be published on the tracking map.
• Updating the website and tracking map, publishing design files and the code.

(Image source: noaa.gov, newfoundlandlabrador.com)

In the past few months, I have been writing thousands of lines of code to bring the boat to life. The boat can be controlled from a web browser over wifi and the program can be uploaded remotely. It’s useful for testing when the boat is close to the shore. I attach pictures of a simple user interface for route planning and reading sensor data such as:

• Compass heading and GPS position
• Wind direction
• Pitch and the angle of heel
• Voltage and power output of individual solar panels: three on the hull and two on the sailwing
• Voltage and current of all batteries: main, backup and one inside the sailwing
• Flood sensors in both hull compartments
• Temperature inside the electronics housing and inside the sailwing, battery and water temperature

Now that all sensors are tested, I’m finally going to join the hull parts together and apply an antifouling paint. It will be finished in less than a week!

In autonomous mode, the boat will follow waypoints that are being set by clicking on the map. The program is constantly reporting messages with sensor data for debugging.

In manual mode, the rudder and sailwing flap can be set at a specific angle. I want to experiment with different angles and get some feeling how the boat reacts.

When testing the boat in the water, I will control it from a powerboat using a waterproof tablet.

When I was simulating the boat movement in rough ocean waves, the compass output was too noisy and I wasn’t satisfied with the result. As the boat will be constantly climbing up a wave or surfing down a wave, the magnetometer is not reliable enough to estimate the correct heading. Fortunately, this kind of problem has already been solved in flying drones and I have implemented an open-source solution.

The new compass crunches the data from different sensors (magnetometer, accelerometer and gyro) and calculates not only the accurate heading, but also the pitch and the angle of heel. The boat knows how it is oriented with a 2-degree accuracy at any moment. Furthermore, the compass constantly auto-calibrates itself along the journey and adapts to the local Earth magnetic field. It’s a pretty advanced stuff. I have been testing the compass for a few days by constantly shaking it and logging the output. It seems the noise is completely eliminated.

There are four compasses and a dedicated microcontroller on the same board. If one compass fails for any reason, the microcontroller will automatically switch to the backup compass. The other two compasses are for experimental purposes. Each compass is actually a complex unit that consists of a magnetometer, accelerometer, gyro and a coprocessor that’s doing heavy calculations.

The board is potted in a thermally conductive epoxy to make it waterproof and mechanically robust.

With this compass, I solved another problem:

The compass must be located far away from high currents or anything that generates a magnetic field (solar panels, motors, magnets, etc.). Therefore it’s communicating with the main controller over a 1.5-meter long cable. The microcontroller that’s embedded in the new compass implements a communication protocol that’s more suitable for such a long distance (serial communication is for a long distance while the previous I2C protocol is reliable only at a very short distance).

The old compass is a tilt-compensated magnetometer and the noise occurs during fast up-and-down movements. The new compass is a more sophisticated solution that collects data from motion sensors and the noise is filtered out.

Over the past few years, there were several autonomous boats in which the autopilot stopped working just because some water entered the hull. It won’t happen with this boat. Because all the electronics is enclosed in a waterproof housing, it will remain fully functional even if the hull is flooded.

The electronics is enclosed in a waterproof polycarbonate case. The Linux computer on the left (Raspberry Pi) will record short videos or take pictures from USB camera and store them on an SD card. It may also send compressed pictures through a satellite modem on request. Because Raspberry Pi is power hungry, it will be switched on only when needed.

The custom-built Navigation Controller on the right has been published on GitHub:
https://github.com/OpenTransat/OpenTransat-Navigator
https://github.com/OpenTransat/OpenTransat-Expansion

The Iridium satellite modem (RockBlock) is for transmitting short text messages like GPS position, battery voltage, power output of individual solar panels, data from different sensors, etc.

The polycarbonate case can be easily pulled out through the hatch in order to access the electronics.

The sailwing wireless controller will be located somewhere inside the hull as close to the sailwing as possible. It’s waterproofed with a heatshrink and silicone.

I find the color codes on the cables very useful to prevent wiring errors.

The second polycarbonate case protects a backup battery (LiFePo4 9.6V/2500mAh) that’s independently charged by a secondary solar charge controller. The backup battery will provide power for the secondary tracker that will keep transmitting the boat location even if the main power source fails.
The epoxy-sealed cable pass-through makes the housing perfectly waterproof for full submersion.

The waterproof connectors are used as switches to disconnect all power sources from the electronics.

The solar power sensor is a PCB potted in a thermally conductive epoxy. Its function is measuring power output of individual solar panels and reporting the data to the main controller. It must be kept outside the polycarbonate housing as the blocking diodes may get pretty hot. More details on GitHub:

https://github.com/OpenTransat/PowerSensor

This is the USB camera that’s connected to the Linux computer. It was built a while ago.

As the sailwing is free-rotating, I had to build a crazy concept where there is no electrical connection between the sailwing and the hull. A cable could easily overtwist and break. Waterproof rotary connectors have a high friction and unknown life span. All the power needed for the actuator comes from the small solar panels mounted on the sailwing. The actuator is controlled wirelessly by commands transmitted from inside the hull.

All the PCBs are waterproofed with a heatshrink and silicone, so accidental water leakage into the sailwing won’t cause any issues. Before the real mission, I will waterproof the batteries and fill the empty space with bubble wrap.

There are two 6.6V/1100mAh batteries connected in series that are charged by the solar panels. The total current consumption is only 9.5mA. If the actuator adjusts the flap 3-4 times per hour, it will add 1mA to the average consumption. The batteries can provide enough power for 4-5 days without any sunlight if the controller is constantly listening for commands. During a long series of cloudy days, the controller will enter power-saving mode which will increase the response time.

The controller processes the data from the radio receiver, checks for any errors, controls the actuator position and transmits a response which may include the battery state, solar power, etc. The radio link uses a low-level protocol to eliminate connectivity issues. The main controller inside the hull will repeatedly transmit commands until it receives the response from the sailwing.

In theory, carbon fiber blocks radio signals, but I have tested the controller inside the sailwing and it works.

About the hardware watchdog:
From my experience, a dedicated low-power microcontroller that resets the main controller can solve more potential issues than just a built-in watchdog timer.

Removable keel makes the boat easy to transport.

The keel is large enough to compensate high ocean winds.

There are aluminum profiles reinforced by fiberglass roving inside the carbon fiber shell. The outer shell and other parts were glued together with epoxy putty. The empty space inside the keel will be filled with water that can freely flow through small holes.

The aluminum/fiberglass profiles on top of the keel are inserted into the rectangular tubes which are part of the welded skeleton.
The mast that is sticking from the bottom of the hull will be hidden inside the keel.

The lead bulb for the keel was made by sand casting – a technique developed in the Bronze Age.
(Everything I do here is dangerous!)

The bulb model was 3D-printed in two symmetrical parts. The first part is placed on a flat surface and prepared to be buried in “green sand”.

The other half of the plastic model is perfectly aligned with the first half. The surface is lightly dusted with a powder-based release agent and again, the part will be covered with green sand.

Both plastic parts have been released from the sand molds. Now the molds can be joined together and the cavity between them will be filled with molten lead.

The lead is being melted in electric casting pot and poured into the sand mold.

Goodbye my dive weights!

Here’s the baby!

Another casting done in the same way.

The hull has been filled with a closed-cell PUR foam that will stay afloat forever.

As the foam is a good thermal insulator and the batteries generate heat when charging, there are several vents leading to the battery tubes to prevent overheating.

The two compartments that are not filled with foam will be completely separated from each other. Each compartment is equipped with a flood sensor that will transmit alert message in case some water gets inside.
The servo is wrapped with a rubber glove filled with a silicone grease as an extra layer of protection.

The rudder on such a small boat can get easily entangled in seaweed or fishing nets. Numerous autonomous boats were recovered with a missing rudder which suggests that the rudder is often underestimated. This very unique rudder combines two interesting features:

1. It avoids entanglement around the rudder shaft
2. It’s symmetrical, i.e. the force needed to steer the boat is relatively small

This rudder design is experimental and never tried before. Hopefully it will turn out well!

The rudder has a flange that fits in the cavity on the hull surface. The dust-proof seal prevents small particles from entering into the cavity.

The dust-proof seal is accomplished by using a thin flexible silicone sheet that closes all gaps around the flange.

There is a low-friction plastic bushing that holds the rudder shaft. The wateproof seal is located inside the aluminum tube.

The rudder shaft is made from a polished stainless steel tube that has a good adhesion to the rubber u-cups. U-cups are special o-rings designed for rotary shaft seal.

The u-cups are inserted into polycarbonate tubes which are glued inside the aluminum tube that is part of the hull skeleton. I didn’t insert the u-cups directly into the aluminum tube because aluminum doesn’t have as smooth finish as polycarbonate. It’s all designed in a way that I can easily replace the u-cups.

The stainless steel shaft is reinforced by a solid carbon fiber rod and attached to a carbon fiber sheet.

The rudder shaft is secured with a few layers of carbon fiber.

A rigid epoxy foam gives the rudder a nice hydrodynamic shape

As the foam is a bit fragile, it’s wrapped with a few layers of carbon fiber and fiberglass.

All communication modules have been waterproofed and placed under the deck while meeting these requirements:

1. The GPS and antennas must be far away from the sailwing as carbon fiber partially blocks the signal.
2. The communication modules must be at least 30 cm from each other in order to prevent interference.
3. The compass must be far away from any magnets, motors and high currents.

The solar panels are glued with silicone and two of them are secured with screws.

The original connector on the solar panel was not IP68 waterproof. I removed the connector, soldered a new cable to the pads, glued a laser-cut acrylic piece with silicone and potted the wire connections with epoxy.

The first three modules in the picture are waterproofed with a heatshrink and silicone. The compass is potted in a heat-conductive epoxy.

The modules are firmly attached to the deck with a fiberglass tape. I suppose they won’t need to be replaced for the lifetime of the boat. The cables are also secured.

All holes for screws and cables are potted with epoxy. This is the most reliable sealing method, much better than silicone.

The compass is far enough from any magnetic noise.

It can act as a backup wind sensor which is less accurate, but more robust. For example, I was told that the ultrasonic wind sensor may not work correctly in a heavy rain or when the boat is heeling too much. The algorithm will constantly compare the sailwing position with the wind direction. If there’s a big difference, it means that something is wrong with the wind sensor and the wind direction will be taken from the sailwing position.
Magnetic encoder is very sensitive to misalignment, so I had to come up with an alternative solution.

There will be 16 Hall sensors around the mast which will track a magnet mounted at the bottom of the sailwing. The accuracy is 360 / 32 = 11 degrees or even less with some calibration.

There is a small CPU that switches between the Hall sensors, crunches the data and sends the result through a serial cable to the main controller. The second cable is for programming.

The PCB is potted in a thermally conductive epoxy, making it perfectly waterproof and shockproof.

The sailwing is now equipped with a flap that is turned by the waterproof actuator. The beauty of this concept is that it requires a small force to adjust the sailwing even in high winds. Here’s a video of the actuator in action:

I decided to move the actuator higher than in the original design as it’s prone to entanglement. Also the flap shape was improved to minimize the contact with water when the boat heels. The boat will become uncontrollable at a 40-degree angle of heel as the flap will reach the water level. This will theoretically happen in very high winds (> 30 knots), but I must find it out empirically using a powerful fan.

The shaft is made from a stainless steel rod threaded only on the end. There’s a plastic bushing that minimizes friction.

Although the actuator has its own position feedback, there is a need for a more reliable position sensor. I teared down the same magnetic encoder I used for the rudder and here’s what’s inside…

I attached the magnet to the end of the flap shaft. Wow, the magnet fits precisely into the M6 nut after the nylon collar is removed! That was an easy job.

I made a 3D-printed housing for the encoder PCB that holds it right above the rotating magnet. The PCB will be potted with a heat-conductive epoxy.

The flap was made from a 1-mm carbon fiber sheet glued to a carbon fiber tube and wrapped with Kevlar. The result is a lightweight and semiflexible material. Semiflexibility helps saving the actuator from occasional water impacts.

The sailwing is almost finished! The next step is building the flap that will control the sailwing orientation.

Both parts of the sailwing were joined with epoxy putty and a fiberglass tape. The solar panels will charge batteries located in the electronics compartment.

The sailwing has to be lightweight and strong enough to withstand high winds and storms, but robustness comes at the expense of weight. The inner skeleton makes the sailwing very hard to bend or break.
For maximum reliability, there won’t be any electrical connection between the hull and the sailwing. The actuator will be controlled wirelessly and powered by its own batteries located inside the sailwing. I will tell you more about it in the next posts.

The carbon/foam ribs were made from a foam core wrapped with carbon fiber and cut to shape with a table saw. All the parts are glued together with epoxy putty.

The sailwing is partially filled with a closed-cell two-component polyurethane foam which improves the structural strength. I decided not to fill the upper part in order to reduce weight. The lower part provides space for the electronics that drives the actuator, i.e. batteries, solar charge controller, motor driver and wireless receiver.

This is the inner side of the second part of the sailwing. The electronics will be accessed through a waterproof hatch that will be securely mounted with screws.

A lightweight and waterproof actuator is needed for adjusting the flap on the sailwing. The problem is, there is no such thing on the market – there are either small actuators, which are not waterproof, or heavy industrial products. So I tried improving a typical hobby actuator. The modified actuator has survived a few thousand cycles in a concentrated salt water which is enough for crossing the ocean. The flap will be adjusted no more than once per hour, so the actuator doesn’t have to be super strong and super durable, just waterproof. Here’s a video of the test:

Just a reminder where the actuator is located in the computer model. Even though it’s mounted above the water level, it has to be waterproof for full submersion as it will often get wet. Splash-proof is not enough!

The chosen actuator has a round rod which makes it suitable for waterproofing. U-cups are special “o-rings” designed for piston sealing.

The rod is sealed with four u-cups inserted into the polycarbonate tube. Normally, one u-cup is enough and two are for safety. I go crazy with four. The plastic bushings keep the rod concentric which is important for a good seal.

A fiberglass sleeve protects the body from both water and impacts. The electrical connector was replaced with a cable and the aluminum tube increases the strength of the rod end. Now the weakest part of the actuator is the rod, but this is the best I can do.

This is the worm drive with position feedback that was tested a while ago. It has survived over 3 million cycles under load, therefore it’s my winner for driving the rudder. I have published the hardware schematics & software on GitHub:
https://github.com/OpenTransat/ServoEncoder

Three shafts are mechanically coupled together: the rudder, motor and magnetic encoder for position feedback.

There will be a plastic cover on the back that will protect the electronics from accidental water contact. I will provide more information about sealing the rudder shaft later.

The boat will be powered by 8 LiFePo4 cells, providing a total output of 12.8V and capacity of 30Ah. That’s enough to actively sail for a week without any sunlight. The batteries are protected by impact resistant polycarbonate tubes which are made to be waterproof for full submersion. They will work even if the hull is completely flooded.

There are 4 batteries in each tube. The end caps are made from laser-cut acrylic disks glued together.

The tubes are pressure compensated with a Gore membrane vent.

Special care was taken for reliable electrical connections that must endure shocks.

The batteries are pretty heavy and are placed in the front of the hull as their weight will be compensated by the swept-back keel. There will be a hatch just above the end caps, so the batteries can be easily replaced. As batteries normally generate heat when charging and this part of the hull will be filled with a foam, the tubes are thermally coupled with the aluminum skeleton. I’m thinking about some better cooling options.

This is the last carbon fiber part made by resin infusion. The keel has to be exceptionally strong as it will carry the heaviest part of the boat – the lead bulb.

Just a reminder how the keel will look like. It’s swept back to prevent catching debris.

The foam plugs

The molds

The first layers are fiberglass and Kevlar.

The foam core is sandwiched between carbon fiber layers.

The infusion spiral is placed in a way to optimize the infusion process.

The resin starts to infuse the part.

The two sides match perfectly together! The keel will be reinforced by aluminum skeleton. Ideally, it should be as thin as possible, but robustness is a priority over speed. The surface is prepared for applying an antifouling paint.

The same process from building the molds to resin infusion was repeated to make a strong and lightweight sailwing.

Building foam plugs using a CNC foam cutter

The finished plugs

The molds released from the plugs
Those black plastic sheets you see on the molds will form cavities on the final parts. The cavities will be fitted with small solar panels that will provide power for the actuator.

The molds covered with a white gelcoat

The first two layers are lightweight fiberglass and Kevlar.

A few layers of carbon fiber

The final layers will be removed after resin infusion. There is an infusion mesh and infusion spiral that will help propagate the resin through the fabric.

The resin propagates through the fabric under vacuum. The part on the top is already infused.

Both parts are released from the molds and are ready to be trimmed.

The aluminum skeleton inside the hull will hold all the critical parts together – the sail, keel, rudder and wind sensor. Precise cutting, milling and welding were beyond my skills, so different craftsmen and fablab did an amazing job helping me build a really robust skeleton.

The skeleton is secured with carbon/Kevlar and waterproofed with epoxy putty. The tubes that stick from the sides of the hull are for lifting the boat out of the water. They can also be used for attaching side floats and converting the boat into a trimaran (just in case).

CNC cutting holes in aluminum tubes

TIG welding aluminum. I’m just watching 🙂

The complete skeleton

The parts were released from the molds and trimmed. As you can see, the hull is coated with a transparent gelcoat in order to identify any dry spots just in case the infusion process fails. Everything seems perfect. The hull will be painted with an antifouling paint to prevent marine growth. The deck has a white gelcoat and the cavities for solar panels turned out really well! Both parts are surprisingly lightweight and rigid.

The yellow outer layer is Kevlar. It protects the carbon fiber hull from impacts and abrasion.

The rudder is designed with a flange that will fit this cavity in order to prevent entanglement around the rudder axis. This cavity was made in the same way as the cavities for solar panels on the deck.

The inside layer is carbon fiber

These cavities on the deck will fit solar panels with connectors and cables, making the boat hull streamlined.

Because carbon fiber partially blocks GPS signal and satellite Internet connection, it’s replaced with fiberglass and Kevlar in some areas of the deck. With this little trick, communication modules can be placed under the deck.

In this video, the molds were used to make strong parts from hi-tech materials: carbon fiber, Kevlar and a special foam core. The infused parts must be cured for 24-48 hours. Let’s hope they will turn out OK!

The molds are ready! Here’s a video showing how they were built. The most exciting moment was releasing the molds from the plugs.

This is exactly how the hull shape will look like. I will post a video showing all the steps how these plugs were made!

The top part will actually be more complex – there will be grooves for solar panels, but I will accomplish it by modifying the mold.

Just a quick update – manual hot wire cutting didn’t provide acceptable results, so I have put together this foam cutting CNC machine, making the cuts easily and precisely. The foam hull will be used for building the mold.

After countless hours of experimenting, calculating, fine-tuning, talking with experts from different fields and reworking, here is the final design. It looks realistic, but it’s just a computer model!

Key improvements:

• More streamlined and balanced hull
• Boat length is reduced to two meters
• Unique rudder design never seen before
• Improved keel
• Flap controlled by a waterproof actuator
• Reduced risk of entanglement

This rudder combines two interesting features – avoids entanglement and reduces the power needed to steer the boat. Don’t worry about the drag-inducing base – it’s easier fixed with a sandpaper than in CAD software 🙂

Here’s how it’s designed on the inside. The heavy duty aluminum skeleton holds all the critical parts together. At least the boat will stay in one piece.

Here’s a video of the small-scale model that has been built to test the concept. I’m really excited that this wingsail works surprisingly well. Now I’m confident in the design, so this experiment was the last one. It’s time to start building the ocean-going boat! I have a long todo list of design improvements based on the experience with this model.

This is a small-scale model of the boat. It has been built to test and tweak the concept before spending a lot more time building the bigger boat for extreme ocean conditions. It also makes software development pretty easy. I’ve got plenty of new design ideas while testing. The first tests went really good, but the boat still has some issues when sailing upwind or “tacking” − sometimes it can’t turn when pointing against a strong wind and current. Something needs to be improved here.

Different techniques were evaluated to build these two carbon fiber parts. I will give more details when building the ocean-going boat.

The hull has been filled with a closed-cell polyurethane foam. The side floats were made from XPS foam wrapped with fiberglass.

The two parts of the hull were joined using a Kevlar tape and painted with a special antifouling paint. This kind of paint doesn’t give as smooth finish as a gelcoat or normal paint, but it prevents subaquatic organisms from attaching to the hull.

The rudder is made from carbon fiber, Kevlar and a steel rod.

The keel is made from two threaded rods and a foam wrapped with fiberglass. Finally, the foam was dissolved with acetone to decrease buoyancy.

The “bulb” is made from aluminum tube filled with lead and two 3D-printed plastic parts.

Waterproof servo and the electronics housing

Order in chaos

The completed boat

The navigation controller has been tested on a hobby RC boat which was navigating through multiple waypoints on a lake. Different approaches were evaluated to find the best algorithm working at high speeds. Stability at high speeds would be difficult to test with a sailboat as the winds in the middle of the ocean can be much stronger. The algorithm also calculates the heading from GPS just in case the compass fails.

I am taking a different approach before the next transatlantic attempt – every single component is thoroughly tested. I can’t tell if the boat will be completed this year. There’s no need to rush. The goal is to launch a capable boat that has a good chance of crossing the Atlantic.

Converting a toy boat into an autonomous boat

This is a 65-cm prototype of the wingsail. It’s made from a foam wrapped with carbon fiber. This kind of sail seems to be pretty heavy and mechanically complex when compared to a typical sail. However, the advantage is BIG – it automatically maintains an effective angle of attack using the small “tail” behind the large wing. It works much like an airplane wing. Now, the idea is to mount it on a small model boat and test how this little magic performs before building a large and robust wingsail.

The boat design consists of 12 carbon fiber parts that will be made by resin infusion – a modern technique providing the best strength properties. Building a mold for each part is a labor-intensive process that may take several days or weeks. Once the molds are ready, the carbon fiber parts can be made with ease. This will allow to build a whole fleet of boats in a relatively short time.
It’s a good idea to try it on a small scale first. The first experiments show satisfactory results even with some mistakes which will be avoided when making the final parts. Currently, I am building a 65-cm model of the boat as a proof of concept before improving the final design. You can see the test molds in the picture.

A month ago, I published a test of a worm drive that is going to do the most critical job – steering the rudder.

The motor made over 3 million cycles under decent load until it failed last week. You can see in the picture that the worm gear is completely worn off.

The motor is dead after 3 million cycles. R.I.P.

3 million cycles means about 7-9 months of continuous operation in the ocean. Crossing the Atlantic won’t take longer than 3 months and the $25 motor can be easily replaced after each attempt. Such a life span is comparable to industrial servos that cost 30-100 times more. The life span can be increased even further by adding extra lubricant into the gearbox or by improving the navigation algorithm to steer less frequently. For example, if the boat corrects its heading every 5-10 seconds on average, it can function for over an year or two.

Another great news is that my work has inspired a team from UK to try this steering system (worm drive + EMS22 encoder) on their ocean-crossing boat. Openness for the win!

There are three widely used materials for boat building, each one with its own advantages: fiberglass, carbon fiber and Kevlar. How to combine them properly to withstand the extreme ocean conditions? The way to find out is to test all possible combinations and examine their properties. Numerous tests have confirmed that the best composite is a combination of all three materials. Carbon fiber provides stiffness while the outer layers of fiberglass and Kevlar will protect the hull from impact and abrasion, e.g. rocks, icebergs, shark teeth…
The final result is amazing, it’s unbreakable!

Combining different layers of fiberglass, carbon fiber and Kevlar.

Examining the properties: weight, strength, stiffness…

A quality webcam has been teared down and the bare PCB has been sealed in a laser-cut acrylic case. The result is a robust lightweight camera that will be plugged into a Linux computer. The computer will be turned on on-demand, having a full control over the camera like taking pictures in different sizes or recording a video.

One of the critical parts of the boat is the hatch. It can be opened in order to access the electronics and it has to be perfectly watertight. The hatch in the pictures, which is normally used on yachts, has been mounted on acrylic “aquarium” for testing. It was left underwater for 7 days and there was not a single drop of water inside the aquarium.

Testing the hatch underwater

However, things get funny when the air trapped inside the hull expands during hot days and contracts when it cools down at night. Furthermore, if we want to use water pumps to get water out of the hull in case of emergency, the same volume of air must get in somehow. How to let the air in while preventing the water from entering the hull? There’s a really interesting device called a Gore membrane vent. It’s from the same material as outdoor Gore-Tex shoes. It’s surprising that you can blow air through the vent, but it doesn’t let any water in even under continuous pressure. The Gore vent fitted on the hatch has been tested underwater for a week, and again, no drops inside.

The Gore membrane vent mounted on the hatch for pressure equalization

Rudder failure is a common issue in autonomous transatlantic attempts, therefore the rudder control requires a special attention. If we expect a rudder movement every 3 seconds for a duration of 6 months, that makes over 2.5 million cycles! That’s more than any servo manufacturer can guarantee (besides some very special and expensive products for medical applications). To make things even worse, turbulent water flow may be kicking the rudder from side to side which shortens the servo life span. Common linear actuators have a life span of only 50 thousand cycles and we need 50 times more.

It seems hopeless, but…

I have been testing a simple worm drive connected to a magnetic encoder as seen in the video. It’s loaded with weight to simulate water resistance. Two million cycles so far and no signs of wear (!) completely exceeded all expectations. Of course, real ocean conditions may be different. Let’s give it a try in the upcoming transatlantic attempt.

As usual, it’s open source on GitHub:
https://github.com/OpenTransat/ServoEncoder

As unbelievable as it may seem, all PCBs you have seen in the previous posts were manually assembled. The following pictures show step by step how prototyping is being done without expensive equipment.

Step 1: Preparation. Collecting all the parts and laying them out on the table for convenient access.

This is the bare PCB as it arrived from the manufacturer.

Step 2: Applying solder paste.

A simple “stencil printer” is used to hold the stencil securely in place while applying the solder

All exposed pads are evenly coated with the solder paste.

Step 3: Placing all components with a vacuum pick-up tool and a firm hand.

All parts are laid out on the PCB. Note that they don’t need to be perfectly straight as the reflow process will take care of it.

Step 4: Let’s cook the goodies.

The reflow oven is made from a toaster oven and a controller that follows a specific temperature profile.

This is the finished PCB. It looks perfect and it works the first time. With a little experience, assembling a PCB like this one takes about 2 hours – one hour of preparation and one hour of placing components.

The boat controller is finished and it’s going to be tested on a hobby model boat. The Navigation Board revealed in the previous post is designed to be universal for any autonomous boat. In the picture below, it’s mounted on top of Expansion Board that will provide additional functions more specific to the boat. This modular design allows to make quick modifications on the Expansion Board while keeping the more complex Navigation Board unchanged.

What has been added:

• Watchdog timer based on a low-power microcontroller that monitors the “heartbeat” of the main CPU. If the Navigator doesn’t respond for a minute or it reports any kind of failure, the watchdog will cut power for a few seconds. This is how we usually fix problems. 🙂
• BFC (Big Freaking Capacitor) to suppress noise.
• Easy connections for all additional hardware components. All wires will be directly soldered to the pins.
• A couple of MOSFET switches to turn on particular hardware only when needed: Linux computer with a camera, satellite modem, AIS receiver (for collision avoidance), RC receiver and water pumps. Wondering what the water pumps are for? They will pump water out of the boat if flood sensors detect big trouble…
• The Power Sensor will measure voltage and current of each solar panel separately. It will be interesting to see which panel is shaded by the sail at any particular time.

Shared with love:

https://github.com/OpenTransat/OpenTransat-Expansion

https://github.com/OpenTransat/PowerSensor

This custom-designed PCB is the brain of the boat. It will be responsible for the most important tasks like navigation, collision avoidance and sending messages. Here are some interesting features:

• The microcontroller is Atmega2560 running at 16 MHz. It has been chosen because it’s widely used in development boards like Arduino Mega and there’s a vast amount of available libraries for various sensors.
• It can be programmed over Wifi. This will make life easier as there is no need to open the hatch and connect the cable for programming. The boat can be tested in the ocean while staying connected to a laptop located on the shore or motorboat. With a long range Wifi antenna, the program can be uploaded even hundreds of meters over the air.
• Tilt compensated compass. There will be also an auxiliary compass that will be mounted far enough from magnetic noise.
• Accelerometer & gyro. The motion sensor will give us a better idea about what the boat is actually doing, especially during storms and high winds. It will report pitch & roll, average wave height, etc.
• Battery power sensor.
• 2 MB flash memory for data logging.
• Experimental 256 KB FRAM memory which is faster than flash.
• Temperature, pressure and humidity sensor. It may discover potential problems inside the waterproof housing such as leaking, overheating or need for pressure compensation.

Yes, it’s open source as promised! Schematics on Github:

https://github.com/OpenTransat/OpenTransat-Navigator