The best way to get Paulo Iscold to do something is to tell or bet him he can’t do it. In the winter of 2015, he was looking for a new project. His speed plane, Anequim, had set five world records the previous summer. A sketch of a twin-engine Reno racer showed up in my inbox, but that looked like some serious money, and he hadn’t found a supporter for that yet. Or someone willing to lend him a Merlin to go after Rare Bear’s absolute 3-kilometer speed record.
What he had found, however, was a potential client interested in gliders. Having designed and built with his students two record-setting speed planes, a Formula 1 wing for a Cassutt, and an Unlimited aerobatic plane, Iscold decided to travel to Germany between Red Bull Air Races (where he works as a race engineer for Kirby Chambliss) to talk to the undisputed masters of sailplane design about building a custom glider wing. Fortunately, they told him he couldn’t do it.
“Do you think you can do better than us?” asked a famous sailplane designer. Retelling the encounter, Amy Iscold, Paulo’s wife, says, “He called me up and told me the Germans said that what he was proposing was impossible, so he decided that would be his next project.” Iscold’s client in Brazil, Sergio Andrade, owned an ASH-30, a two-seat open-class glider with an 87-foot wingspan. After hitching a ride in the back seat, Iscold pitched the idea of making a custom set of wings for the ship.
In comparison to powered aircraft, with open-class sailplanes the flaps are used not only for takeoff and landing, but also to optimize the wing camber (including negative flap—or reflex—to reduce lift coefficient) for various lift conditions. Therefore, with every movement of the stick, turbulent gust, or rising thermal, the flaps need to be adjusted accordingly.
Iscold enjoyed his first hours flying with Sergio in his ASH, but the intense workload of cross-country flight, coupled with constantly adjusting for optimum flap angle, inspired him with the idea for a new project. Iscold’s plan was to develop a new set of wings for the glider, which would offer several advantages: increased aspect ratio and wing loading to optimize cross-country flight and, more significantly, a fly-by-wire control system, which would allow constant in-flight optimization of the wing camber along its total length without added workload for the pilot. Thus was born the Nixus Project (Nixus in Latin can be translated as “pushing ahead”).
Sailplane designers spend a lot of time focused on airfoils. Sailplane foils are tasked with being ultralow drag while operating at a variety of lift coefficients (high lift while thermaling, low lift while in cross-country flight). On top of that, due to the extreme bending moment induced by such long, skinny wings, the airfoils are generally quite thick in order to accommodate the necessarily beefy spar. Achieving all these objectives can be quite tricky, and adding variable camber via positive and negative flap only makes the job harder.
Wing molds were CNC’d from HDF. Molds weighed (literally) a ton and were discarded after the wing was closed out; the Nixus is a one-off project.
After reading several texts on sailplane design, Iscold decided to contact L.L.M. Boermans, a famed aerodynamicist and airfoil designer at the University of Delft in the Netherlands, to see if Boermans could advise Iscold on the airfoil design. “Actually, I designed my own airfoils [the wing uses five slightly different airfoils, becoming thinner and less cambered from root to tip], but Sergio asked for a consultant to check my work. That is when Boermans came in,” says Iscold. “I sent my work to him, and he sent a proposal back. I wasn’t happy and re-sent my proposal…After three or four times doing this, we decided to join forces. His [shape was used for the airfoil’s] upper surface, my [shape for the] lower surface. It is called the DUPI airfoil [the acronym is for Delft University/Paulo Iscold]. It is the first time that Boermans has shared a design with someone else,” says Iscold.
“He also changed the wing planform. I was going to implement a continuous change of chord [i.e., an elliptical leading edge], but I gave up halfway through and stuck to his planform. He added some really cool strategies to optimize the airfoil over the span. I wouldn’t have known how to do that, but now I know! He also designed the winglets.”
Custom autoclave built by Iscold with help from student interns. The autoclave reaches 250 F and 70 psi.
Now Just Build It
Once Iscold had the airfoil figured out, it was time to lay out a planform and structural design for the wing. In order to maintain good penetration for soaring, Iscold wanted to keep the wing loading higher than that of the stock ASH-30. This meant that the chord of the wing would have to be significantly shortened to achieve the required higher aspect ratio. That in turn would lower the maximum thickness of the wing. This meant that the wing spars would need to be exceptionally strong to deal with the bending moment at the root, and the skins would also need to be quite stiff to avoid aeroelastic effects (i.e., wing flutter).
Iscold solved these problems with his usual flair for accepting challenges that would defeat lesser mortals (and budgets, as this was a well-financed project). The wing skins would be fabricated from an ultrahigh-modulus carbon fiber. High-modulus carbon, while not as strong as standard-modulus carbon, is much stiffer, which is the crucial property for wings where the skins deal with aeroelastic loads, and bending loads are handled separately by a spar. Although the UHM fiber existed, it did not exist as a stock fabric, so Iscold convinced Hexcel to produce a special run of bespoke fabric at its UK factory.
The wing spars were an even bigger challenge. For these Iscold realized he would need massive spar caps. At the root, the upper cap was 7 inches wide and nearly 1.5 inches thick, made with 137 layers of unidirectional prepreg carbon. In order to achieve the material properties required, Iscold had to manufacture the caps in an autoclave. Autoclaves are essentially pressure ovens that allow parts to be manufactured at higher pressures and temperatures than can be achieved by ambient pressure/temperature vacuum bagging. Unfortunately, autoclaves are extremely expensive, and autoclaves big enough to cook the 25-foot-long main spars are probably only possessed by large aerospace firms.
So Iscold decided to make his own! Even if nothing else came of this project, the construction of the autoclave itself was a noteworthy achievement. After months of fabrication, the autoclave was ready, and Iscold was able to manufacture his eight spar caps. Having helped Iscold several times during the course of the project, I can only hope that he chooses a smaller project next time around. The inner spars were nearly 100 pounds apiece and were a bear to move.
The Heart of the Wing: Fly by Wire
The technological heart of the project was the fly-by-wire (FBW) system. The entire trailing edge of the wing, from root to tip, is composed of five separate control surfaces that can be individually moved up or down. The outermost surfaces are in fact the ailerons, and for safety’s sake, these remain under manual control. The other inboard surfaces are all servo operated.
“The fly-by-wire is just a different way to connect the stick to the control surface,” explains Iscold. “The same problems that we try to guarantee/avoid on a mechanical system we need to guarantee/avoid on an FBW system: mainly, the actuator should never break, and if it breaks, it should not break asymmetrically. It is that simple!” One should note though that whereas a traditional actuator is a control mechanism, the FBW is more appropriately described as a control system.
“Since the internal wing volume itself is small, the servos also have to be small,” continues Iscold. “I tried UAV servos, but the market is terrible. They want to know how much money you have before they quote you a price. Then I decided to go to robotics. A friend pointed me to the Dynamixel servos. Made in South Korea with Maxon Motors (Swiss), they have quite an impressive amount of intelligence already installed. So I decided to give these a try.”
Iscold bought several servos and wrote a program to run them continuously at 50% of maximum load for three months, 24 hours per day, seven days a week. During several visits to Catto Propellers to help Iscold on the project, I could see the little servo “pumping iron” whenever passing through the side office. “After 2000 hours of continuous operation, it showed a little free play, nothing more than that. Well, 2000 hours is not a bad TBO time!” says Iscold.
“I then decided to use two servos to reduce the current, and therefore the temperature, and probably increase the life even more. The problem with two servos is how to synchronize them. Luckily, the Dynamixel has a feature to synchronize servos that uses the torque (current) to avoid servo fighting. Again, I built a little device, tested it, and became super happy with the solution.”
The two-servo configuration also added redundancy to the position measurement, since Iscold now had two independent position sensors. This position measurement is crucial since it indicates flap angle. The Dynamixel uses an RS-485 bus and can provide a wide range of servo data to the FBW system computer including position, temperature, torque, voltage, and more. After sketching out a few system configurations, Iscold sent them to another project advisor, Dagfinn Gangsaas. Gangsaas has worked for Boeing, Lockheed, and now Embraer, and is in Iscold’s words “a master of FBW.” Based on their conversations, Iscold decided to go with a triple-bus system (in order to create redundancy with the physical connectors) as well as a triple-control computer, again for system redundancy.
Exterior of the access panel. This surface looks flat, but it actually matches the curvature of the wing. Beautiful machining by Nick Jenkins of Jenkins Aero.
In order to use this triple-bus system, Iscold had to develop one more item for the system, which he calls the “node.” Iscold explains: “The node is a small computer that sits on top of the servo and talks with the servo at a high baud rate (over 200 Hz). The node receives the triple information from the buses, takes a vote to make sure the information is reliable and consistent, and then talks with the servo.”
That is to say, if the node receives three separate readings for something like servo position, and these readings agree with each other (a small amount of discrepancy is allowed), then the system assumes everything is working fine and continues to relay pilot/flight computer inputs to the servo.
On the other hand, if the node receives two readings that agree, but a third that does not, it assumes the third reading is erroneous, and by “majority vote” continues to relay pilot inputs—albeit alerting the pilot that one bus appears to be faulty. Here you can see the brilliance/necessity of a triple-redundant system: In the case of dual redundancy, the system would not know which bus was faulty and which was still providing accurate information. By making the system triple redundant, the flawed bus is easily identified. One could argue that two buses could fail identically and thus gain faulty voting rights, but the chances of this happening are quite small.
The last permutation of voting possibilities is that each bus provides a different reading. In this case, the flight computer knows something is definitely wrong. Because a servo failure on one wing could result in asymmetric lift, in the case of an indeterminate vote, the system failsafe is to cut power to both pairs of servos (left and right wing), allowing corresponding left/right flap pairs to fly in trail. This alleviates any asymmetric roll condition.
“So the triple-computer, triple-bus, dual-servo system is the solution to avoid the interruption of the link between the stick and the control surface,” says Iscold. “To avoid the asymmetry, complex (but simple!) software was created to handle the failures coming from the servo, node, or computers. On top of that, each pair of servos (left or right) is connected to the same circuit breaker, and you can switch pairs on and off independently. This facet of the FBW is still in testing.”
Sergio Andrade at the controls of his ASH-30. Like Paulo, Sergio graduated in engineering and has always been fascinated by technology and aviation. In the ’80s, together with some friends, he founded a microlight aircraft company that produced over 500 units. He has been flying since 1968, but his passion for soaring is “only” 20 years old. “The Nixus has a lot of possibilities, both as a records-breaking glider (we hope) and as a flying laboratory to help research fly-by-wire and artificial intelligence applied to general aviation. This could bring benefits such as simpler construction, better efficiency, and improved safety,” comments Sergio.
Iscold notes that the node was developed in Brazil by two former students (Adriano Martins and Elias Jose) and manufactured in the U.S. by Ansync Labs of Sacramento. Paulo asked them for a quote, and they provided a sponsorship!
As of the time this article went to print, the Nixus has made four successful flights. The project is now a part of Iscold’s (and his students’) activities at Cal Poly where he serves as an associate professor of aerospace engineering. Congratulations to Paulo for another impressive engineering achievement, and we look forward to seeing how Nixus does when it’s finally cut loose to surf the mountain waves of the Sierra Nevada.