Select Page
DUCK : A Device for Upper-limb Cycling Kit

DUCK : A Device for Upper-limb Cycling Kit

Team Gre-Nable is keen to present a project done by a group of students from INP Grenoble collaborating with members of our team. team Gre-nable has the ability to promote and use the outcomes of the study to answer all needs required by E-Nable members.


This post is an excerpt of the project’s final report presented by the students for their diploma and is available here in pdf format.



As part of the product development project our team [the students’ group] is working on the design of a device that helps a young child [Noé] riding his bike. He doesn’t have a fully functioning hand, which makes this a difficult task for him to perform. Our team needs to design a functional final prototype to help Noé ride a bike, which includes testing and prototyping in real conditions to ensure the safety and reliability of the final product.

Our team is tasked with developing this system for Noé because he does not have any fingers on his right hand, which makes him unable to ride a bike accordingly due to his inability to grip the handlebar. The project took place over the course of one academic year and we are currently at the final phase, which is the validation and verification of the final prototype. The group has been working with APF France handicap, an association dedicated to helping handicapped people and children like Noé, who also work in collaboration with Gre-nable, local E-nable members that specialize in creating functional prosthetics. The group has also been working in collaboration with two engineering school teachers Philippe and Marie-Laure [who are members of team Gre-nable].


General Overview

Our project started from a first attempt by the occupational therapist of designing an adaptation for Noé’s handlebar with thermoformed plastic sheets. This deviced proved to really help Noé in biking, but also led to the identifiation of some potential improvements.

Prior to the second meeting with the client, the occupational therapist told us that the main problem with the previous system’s function was Noé’s position and posture while on his bike. His elbow was too high compared to the other arm, creating an imbalance and bad posture resulting in poor alignment of his back. Since the system awkwardly positioned his arm, it was not serving him properly and didn’t have much utility as a consequence. 

Another important point we faced was to provide Noé with the ability to move his hand quite freely on the handlebar, even if his wrist is firmly maintained in the socket. This is the reason why we proposed a kinematic made of both a ball joint for free rotation, and a rotary sliding link which made it possible for the wrist to move up and down when cycling on uneven path. The analysis of this kinematics is shown on the picture below.


On the aesthetic side, it was also necessary to reduce the size of the system, principally the hand rest or plank, and the system housing as they were visually bulky and not very discreet for Noé’s preferences. In the end, though, we were able to correct the size and form of the final system to achieve a much better, working prototype, which is more adapted for Noé’s helbow and his back posture. Our system’s shape ended up resembling the shape of a duck, which is where inspiration for the name came from.

An overview of the new system should be briefly introduced prior to talking about the new functions and features. As we can see in figure below, the system has been modeled showing the new system in blue, mounted to a rough estimation of Noé’s handlebar. The device is simply slid over the righthandlebar and tightened using two screws to prevent it from slipping off or rotating.

For a closer look at each component we can reference in the image below.

The system features a plank with a ball in the middle, where Noé attaches his wrist when that part of the system is configured to his arm. This piece is able to move up and down via two guide arms (or rails) that slide within the larger housing component attached directly to the handlebars.

Our first prototype featured a stopper in the form of a pin that was installed through the bottom of the rails to prevent the rails and the plank from sliding up and out of the housing once installed, but new components were added to accomplish this in a more discrete way. Instead, internal stoppers were added along the tracks in the housing (shown in orange below). The L-shape of these stoppers catches on the new C-shape of the guides so that when the plank and guides are at their maximum height, they are not able to slide out of the housing. Not only did this save material, but it also removed some unnecessary clunkiness and potentially sharp or pointy aspects from the design that could have posed a safety issue.

The shape of both components were also strategically designed to hold springs on both sides at the same interface in this region, stacked on top of each other, where the design was then able to be more easily reduced in size.

The interaction of these components can be more clearly seen in next figure, where it will be discussed in detail.



Another set of new components includes covers on the sides of the housing, which is shown as transparent in both figures to more clearly see how the system has been redesigned. One purpose of the covers on both the left and right sides is to contain the guides, stoppers, and springs from moving during operation. Another aspect of designing these covers came out of necessity, since inserting the guides and stoppers would make it very difficult to install the springs in any other way. If one imagines installing the guides attached to the plank and then the stoppers, it would be challenging to compress the springs
enough to install them from the top of the housing.

The covers allow the user to first install each component from the sides of the system, where they can then all be nicely encased in a way that they will not escape during operation. At the same time, the springs have a much harder time accidentally coming out or becoming dislodged while riding. Figure below provides a side view of the system that makes it easier to visualize this concept (and it should be noted that the handlebar in both images is not shown to make the images less crowded). This image of DUCK also highlights how the movement of the rails work in this version.

Once everything is in place, the covers retain the components in conjunction with the stoppers, where the L and C-shaped features of the stoppers and guides create a retention point at the location in the dashed yellow box.

Again, this is how the plank and guides are free to move up and down with help from the spring, but do not become dislodged during use. To better imagine the movement, a black double-sided arrow is shown at the bottom of the guide rail. This is how the new design was able to reduce its size and become more sleek, which we will discuss in the next section.

Ball Joint Coupling – Tests and Experimentations

The ball joint coupling is the primary connection point between the bike and the rider.
These two parts, the ball and the socket, allows the rider to easily disengage with the system in case of a fall. It is therefore very important to size it properly in order to ensure the security of Noé. To reduce the lever arm, which provides better solidity and more stability for Noé, we changed our ball joint coupling system slightly. Instead of having more mobility in the socket, the ball will be the soft piece that can deform and enable more movement. The socket can now be integrated in a bigger part, saving space and materials. The socket is now completely solid, printed in ABS and does not have any mobility (completely rigid). Due to these changes it was necessary to size the system properly for Noé, so we then developed an experiment plan.


Several parameters have an influence on the force needed to separate both parts when
engaged, so we will call this force the “release force”. The radius of the ball and the socket
were made to the same dimensions to ensure good mobility.

Socket depth : Changing the depth of the socket will make the ball harder to remove, and because the system is round, making it deeper will allow more plastic to wrap around the ball, which will need to be squeezed more to be released. We need to have at least the depth of half a sphere (14.1mm), or its radius, to have a release force (and for retention, too). For better comprehension, we will say for example that a socket offset of 3mm from the surface has a depth of 14.1 + 3 = 17.1mm. Please see section view with 3 sample depths (3, 4, 5mm) for better understanding.

Softness of the ball : To change the softness, we changed the infill percentage when 3D printing this component. We could have changed the interior geometry of the balls but changing the infill % allows us to keep a resilient structure for the piece, and this makes it easier to classify the different balls after printing.

Materials used : Changing the materials for 3D printing will change the properties of the materials and how they influence the system. For the socket, we chose not to change it and keep it as ABS V2 (from Zortrax company) because we needed a solid piece. Having this piece made of a single material makes testing simpler, too. The ball is printed in SemiFlex, and there are different types of SemiFlex usable in GI-Nova such as: Ninja Cheetah and Zortrax SemiFlex.

There are more parameters that could influence the release force, such as the friction between both parts, the form of the system (not perfectly round) etc… But we chose to do our experiment plan on those parameters because we thought they were the main ones with the largest consequences on the release force.


Although the project in the end was successful in providing a useful product for our client, there are some points that could have been improved upon. For example, the design changes in version two of the system were well executed but could have been adapted for more adjustability. More specifically, the design aspect with regard to the adjustment of the height of the wrist could have been improved slightly. In the second version, the wrist is situated at nearly the system’s lowest position and is able to move up and down several millimeters. If the position of the L-shape of the stoppers was redesigned so that the point of contact between them and the guides was higher up, the height of the wrist and the amount of movement could be increased. With the covers, offering several different stoppers could make the system more adjustable since these pieces could be swapped and installed relatively easily. However, since the current stoppers worked well for Noé, it was decided to leave the design the way it was.

Considering the second design worked much better than the first and met nearly all of our client’s needs, we can definitely consider the system a success despite the small design improvement opportunities. Nearly all designs and products in general could be improved in some way, even if only to a minimal degree. It is good to recognize and reflect on these improvements for future projects that require similar methodologies, too, especially as our group continues studies in engineering and eventually more professional experiences.

In the end, the team was very satisfied to have met our goals when faced with the challenges presented by this project. Our group managed our time very well, and was able to equally distribute the work throughout the year in a way that made it possible to have a successful outcome. This has proven to be a very interesting subject to work on that taught us valuable lessons and skills that can be taken away and used in our futures, too.

Ski pole

Ski pole

Manon, hope for the 2030 Olympics ?

Our previous design of the ski pole aimed at Pierre-Luc, based on the principle of ball joint that enclosed his palm, does not seem appropriate to us in Manon’s case, because her clamp function can be performed by his thumb and his pinky. Of course the strength of this clamp is clearly not sufficient to hold a ski pole, so we keep the principle of using a socket build around her hand.

We therefore define new specifications intending the left hand palm to have the same feelings of touch as her right hand.

Scan of the left hand, equipped with a muffle

Pour ce nouveau projet, nous essayons de nous passer de l’étape moulage, en réalisant un scan de la main in-situ, en position de maintien du bâton de ski. La main étant équipée d’une moufle qui sera ensuite bien adaptée par la couturière de la famille.

For this new project, we are trying to work without the molding stage, by performing an in-situ live hand scan, right in the position of holding the ski pole. The hand being equipped with a muffle which will then be well adapted by her family’s seamstress (grand’ma).

The scanning operation is not as easy as expected, but after 3 attempts we got a good quality mesh file.


Another evolution of our process, we will not transform the resulting mesh (STL format) into a B-rep file for importing into our usual CAD software (Onshape).

Once the STL is imported into an ONshape’s “part studio”, an enveloping surface is made around the mesh, made of multiple ‘Spline’ curves (generalization of Bezier curves) drawn over successive cut planes. These curves, building a group of sections, each of them wrapping the glove, are then connected (joined) to each other by ‘lofts’. A loft is a surface obtained by interpolation between the different curves, in the form of a NURBS surface (cf wikipedia).


A pile of parallel planes will slice the fist holding the pole. All plans are referenced on remarkable points of the muffle.
On each plane, a closed curve is drawn to surround all boundaries of the displayed STL, thus creating a section. By interconnecting the successive parallel sections by lofts (3D loft function), we obtain the blue envelope (as a surface) which will then be transformed into volume using the “thicken” function.
After having created thickness to the developed surface, having cut the end of the socket to allow the thumb to come out of it, the socket is ready for integration into the new holding system.

Highlights of the device including a (simple) system.

The new device is therefore no longer based on the ball joint principle but on a two-element system :

  • an element on the stick (host)
  • a detachable socket for Manon’s hand.

The cohesion of the two elements is achieved by neodymium magnets, powerful enough for the stick to follow each hand’s movements, but detachable enough to allow the release of the socket in case of fall.

The host frame is attached to the stick. The vertical side of the host is on the outside to allow ejection of the interlocking (inwards) in case of any fall. On the inner side of the host, we can see the location where the small circular neodymium magnet aimed at vertical support will be screwed, and the rectangular magnet, more powerful, dedicated to lateral support.
L’emboitement est “collé” sur le bâti grâce aux forces d’attraction des deux aimants.
The hand-socket is “glued” to the host-frame thanks to the forces of attraction of the two magnets.
  1. The correct positioning of the interlocking is ensured by a centering dome, and a calibrated location
  2. The holding of the interlocking on the frame is the attractive forces of the main rectangular magnet (40x40x4)
  3. The second magnet (circular) facilitates vertical support and centering of the hand-socket.
-The first prototype validated the functionality of holding the ski pole and its test on a ski slope confirmed our technical choices.

Some small improvements were made to give more room for the thumb and the final version of the interlocking was printed using semi-flexible material (BASF Fusion, with shore mark 65D).

The success of this new concept quickly attracted other parents. So we reviewed (cleaning) the design scripts, so any new requests would be made quickly. The frame (HOST on the drawing) is almost generic, its adaptation to the hand-socket is minimal. On the other hand, the hand-socket being 100% adapted to the size/aspect of the hand and the type of agenesis of the child, its design will be a little more touchy.

Ces adaptations de système à la main d’un autre enfant nécessitent de maitriser l’outil de conception CAO, mais n’est pas aussi compliqué qu’il y paraît. Les fichiers STL du système développé pour Manon ne seraient d’aucune utilité pour un autre enfant. Par contre, nos développements sont open-sources et disponibles sur la plate-forme Onshape, et nous sommes toujours prêts à donner un coup de main 🙂 [© E-nable France]

These system’s adaptations for another child’s hand will require some skill using a CAD design tool, but are not as complicated as it seems. The STL files of the system developed for Manon would be of no use for another child. On the other hand, our designs are open-source and available on the Onshape platform, and we are always ready to give a helpful hand 🙂

Let’s keep in touch.



Multi Tool Holder with Ball Joint

Multi Tool Holder with Ball Joint

Pen/Pencil MTH Holder Evolution


During the delivery of our first MultiToolHlder [MTH] dedicated to writing, it came to us that our method for positioning the pen was not necessarily that desired by the wearer of the prosthesis.

Every person who writes has his own habits, so we have to design an MTH that will suit needs well without the need to design a new support.

We got inspired by GPS mounting system that we used to suck on our car windshields.

No sooner said than done, the specifications were simple to write:

  • to redesign the existing pen sleeve
  • to graft a ball joint on the sleeve
  • to redesign the socket whilst grafting a self-tightening screw to lock the ball joint.



According to our development process, our design is naturally open source, available through Onshape the online application at the following URL:

The folder can also be found with the Search (target) function targeting the Public domain, with the string: ” team : MultiToolHolder“.

Visitor can discover the assembly as we simulated it, to correctly place the different elements.


and the designs of the parts constituting the MTH are grouped in the PartStudio directory “PenHolder_rotule “.

As usual to modify the file, it will first be necessary to make a copy in your personal space, folder that you can then change to wish, including to adapt the MTH to your target nesting.


The Socket

This element is the heritage of a previous development that we described in our post:

At the end of this socket must be created and “solder” a thread that will tighten the ball.


Design of the thread

The aim is to realize a cylinder, whose periphery will be dug (extrusion with removal of matter – extrude remove)a triangular profile corresponding to the thread, sweept along a helical path, image of the screw.


Le profil en triangle (jaune) dont le plan de construction est normal au chemin hélicoïdal, va définir un volume qui sera retiré de l’enveloppe cylindrique de la vis.

Step #2

Extrusion du cylindre complet de la vis, puis extrusion (remove) du profil du pas de vis suivant le chemin hélicoïdal.

Etape #3

On creuse l’intérieur de la vis pour y insérer la rotule, puis on “conifie” le début de la vis pour permettre un effet de serrage (l’écrou aura une conification inverse).

Etape #4

On fragilise la vis avec 8 fentes pour créer des lamelles un peu souples qui emprisonneront la rotule lors du serrage.

Etape #5

Enfin, on assouplit la base de chaque lamelle pour faciliter la flexion et le serrage autour de la bille.

3D Prints

Socket printing

The socket is a complex shape, to ensure the printings with high quality, printing will be done vertically, the socket laying on the flat section of the screw using supports to maintain a smooth appearance of the socket.

To ensure that the thread is properly stick on the printing mirror, a trick is used which consists in the existence of a small extruded flat part (thickness = 2 layers) in the same plan as the end plane of the thread. This piece (trick) is exported to the slicer at the same time as the socket, so it is she who imposes the collage of the whole on the plate.

Printing of the socket

Toute la complexité de l’impression réside dans l’impression de la rotule qui est orthogonale avec l’axe du manchon.

Dans une première version, le manchon avait été imprimé en appuyant la section coupée de la rotule sur le plateau. Lors d’un essai avec un bénéficiaire, la section du raccord entre la rotule et le manchon s’est cassé suite à un essai de rotation du manchon sans dé-serrer l’écrou. Essai concluant de résistance des matériaux et de la capacité de serrage de l’écrou!.

Nous avons donc orienté l’ensemble pour que la rotule, le raccord et le manchon soient imprimés dans le même plan.

Simplify3D, génère des supports de qualité qui se décollent sans laisser de traces. L’expérience montre qu’une épaisseur de couche de 20/100ème génère une rotule suffisamment précise pour l’utilisation.

Printing of the screw

Quelques essais d’écrous en PLA ont mis en évidence la présence de frottements (PLA sur PLA) importants lors du dé-serrage ce qui rend l’utilisation moins aisée pour une personne n’ayant qu’une main valide.

Un essai avec du filament Iglidur, (fabriqué par Igus) réputé pour ses qualités de frottements réduits, confirme le choix. L’écrou sera donc en Iglidur (, c’est cher mais on peut en demander quelques mètres en échantillon.

Par contre, l’impression n’est pas triviale, une température élevée pour la buse (260 à 265°C) avec un plateau à 70°C, et une vitesse d’impression faible (20 mm/s) comme pour du flex.

Printing of the inserts

Les stylos, crayons, pinceaux … ayant tous des diamètres différents, il faut donc imprimer un jeu de bouchons de diamètres différents, avec du filament flexible. Nous utilisons deux filaments : ninjaFlex et SmarFlex.

A partir d’un design paramétrable (paramètre  Pen_diameter dans le Part Studio ‘manchon_a_rotule‘), on exporte autant de bouchons que l’on veut pour constituer le jeu. Pour cette livraison, les diamètres choisis sont 9,5 mm, 8,5 mm, 8mm et 7 mm (crayon de papier courant).

et pour terminer, une coupe générale du MTH à rotule assemblé;

… and finally the Multi Tool Holder on stage.

Nathalie ré-apprend à se servir de sa main droite pour écrire, pour dessiner. Les réflexes vont revenir rapidement.


Afin que Nathalie puisse tenir des outils de plus petits diamètres, Patrick imprime et lui envoie quelques manchons encore plus petits que les précédents.

Et peu de temps après, nous recevons des nouvelles… et des photos : Nathalie s’est mise à peindre, cela faisait tellement longtemps qu’elle en rêvait ! Et on doit dire qu’elle se débrouille plutôt très bien …  

Nathalie peint

…  et immédiatement après la livraison, conception d’une nouvelle extension articulée !!


Le porte fourchette

Maintenant que la base est construite, il devient aisé de concevoir d’autres extensions articulées et spécialisées.

En repartant du concept de porte fourchette décrit dans un article précédent ( pour réutiliser et améliorer le bloc “coinceur” de fourchette, imprimé en flexible. La réalisation de cette extension a été très rapide.

On retrouve les primitives de conception dans le Part Studio ‘fourchette_a_rotule‘ du même dossier team

A l’attention de tous les membres d’e-Nable France (Makers ou Demandeurs d’appareil)

Nos développements sont en open source, disponibles à tous pour être reproduits. L’adaptation de l’emboitement demande un peu plus de technicité qu’une simple compétence en impression. Mais, nous sommes là pour vous aider à acquérir cette compétence.

Soumettez-nous vos besoins et nous vous aiderons à réaliser votre MTH personnalisé. La seule petite contrainte, est que le design est trop complexe pour être réalisé avec le logiciel Openscad (surtout du fait de la forme non modélisable par simples primitives de l’emboitement).

Heureusement il existe une solution gratuite pour résoudre nos besoins, celle que nous maitrisons : l’application en ligne Son usage n’est pas plus compliqué qu’appréhender Openscad. Prenez quelques minutes pour lire notre article ‘’


Electrically Assisted Flexibone Hand

Electrically Assisted Flexibone Hand

About licensing

​ This work (Electrically Assisted Flexibone Hand) is provided under the terms of License Creative Commons Attribution 4.0 International.

This license concerns the whole related documentation, reports, CAD models, testing videos, etc.

The shape of the hand is based on Kwawu hand designed by Jacquin Buchanan.

This project is Open Source Hardware certified :  [OSHW] FR000008


This article presents a novel design for a low-cost, 3D printed, electronically assisted prosthesis for a transmetacarpal amputee. The design is the result of a recent collaboration between students from the University of Bath, England (Phil Barden, Thomas Eagland, Jay Pinion, Sevinç Şişman), and a member of Team Gre-Nable, located at Grenoble Institute of Technology, France (Philippe Marin). The team worked to design the prosthesis over a 5-month period, with the final design being the result of extensive force testing and mechanical analysis, rapid prototyping, and user testing. For a more in depth explanation of the final design and the project, please refer to the full report that providing all details. See at the end of this article for even more links and information, CAD model, Arduino code, video records of the tests, etc.

Much of what is presented here are proofs of concept, and it is our hope that by sharing these ideas with the rest of the e-Nable community, others will be inspired to build and improve upon them. Happy reading!

The mission

Most of “robotic hands” developed for amputees are designed with actuator motors integrated inside the palm, like this prototype we have seen on “”:

“your mission…”, Philippe said to the team, “is to design an electrically assisted prosthetic hand – I mean with motors, battery, mechanics… –  for a lady who lost ‘only’ her fingers. This means there is no room in the palm to put all this stuff”.

Nathalie’s residual hand

Background of the case


The project focused on a lady called Nathalie, a transmetacarpal amputee who approached team Gre-Nable back in 2018 saying she was unhappy with her €48,000 myoelectric prosthesis and asked whether the team could help designing a better alternative for her.

Her primary complaints were that it was far too heavy (weighting approximately 1.5kg, almost three times that of a human hand), and very difficult to control. The prosthesis consequently sat unused in its box.

The design presented below weighs approximately 500g, which is very close to that of a human hand. The approximate cost to recreate it is also close to €200, or less than 0.5% of the cost of her myoelectric prosthesis.


I like it more than my €48,000 prosthesis.



The final Prototype features…

  • Force Feedback
    • Pressure sensor in index finger
    • Haptic motors array in gauntlet
  • Attachment Method
    • Snowboard binding mechanism
  • Power supply
    • 7.4V Lithium ion rechargeable battery
    • 2200mAh
    • Providing several hours autonomy

Most of these elements are described in the following sections…

  • Mechanical Design
    • Flexibone finger design
    • Original fishing wire routing
    • ABS & Ninjaflex 3D printed parts
  • Actuation
    • Two servo motors
    • Whipple-tree pulleys
  • Controller
    • 2 axis Joystick
    • Arduino board




As for a patent description, we claim in this article that, as far as we know, the following elements are innovative in the context of low cost upper limb prosthetics:

  1. Finger kinematics made of a single flexible printed part that allow inter-phalanges bending, together with soft gripping pads in the fingers.
  2. Haptic feedback array in the gauntlet made of a series of vibration motors.
  3. Control by a two-axes joystick taking profit of a residual possible movement of the amputee limb.
  4. Twistable locking mechanism for attachment of the gauntlet on the forearm.

Features detailed Description

Flexibone Fingers



A novel finger model designed by team Gre-Nable and based on the Kwawu model, using a design with ‘bones’ passing through the centre of each finger and thumb, printed using a flexible 3D polymer called NinjaFlex. The bones are then covered by ABS shells representing the phalanx which give each finger three articulation axes and consistent rigidity. Like usual e-Nable designs (Phoenix hand), fishing wires are used to mimic tendons in all fingers to contract them.

This structure makes the “Flexibone Finger” very easy to manufacture by printing only one simple flexible part, and six rigid half phalanxes (made of ABS or PLA). Moreover, the flexible part realizes two functions:

  • the first one is making three pivot joints,
  • and the second one is integrating pads for a soft contact, that may be coated with rubber like painting (for example Plasti-DIP).

We believe this new Flexibone Finger concept is a major improvement for creating prosthesis, whatever the final type of hand is concerned. For the time being, our Kwawu based hand will benefit from the concept, in near future other type of hand prosthesis may integrate these Flexibone concept. Team Gre-Nable is in the process of developing a simple parametric standard finger based on the Flexibone concept.

The Flexibone Structure


This design was chosen through the means of testing and comparison against other common e-Nable models (namely Kwawu and Phoenix). Two tests were carried out:

  • a force test, to find the design that required the least amount of force to fully contract an index finger,
  • and a grip test, to find which model could grip the greatest range of objects, analysing both object sizes and maximum weight.

The Flexibone hand came second to the Phoenix hand in the force test, both needing considerably less force to fully contract a finger compared to the previous Kwawu design. However, the Flexibone design performed far better than both the Phoenix and Kwawu designs, with it being able to grip objects both larger and smaller, and objects of a much greater weight.

Test rig with servomotor, pulley, fishing line, force sensor, and sample finger in the vice.

Some of the printed series of parts for various parametric studies during the project.


The design works by having three areas of the bone which are thinner and are not constrained by the ABS plastic shells. The fishing wire runs up the centre of the flexibone, and when a load is applied to the wire, bending occurs at the three weak points hence contracting the finger. Due to the elastic nature of the flexible polymer, when the load is released from the fishing line the fingers return to their natural resting position.

This bar chart shows a sample result of a force test, comparing the maximum force required to completely bend a variety of fingers, for three different sizes (small, medium size and adult size). It appears that despite its very nice and ergonomic configuration the Kwawu requires more force in the tendon lines. On the other side, the Phoenix hand with dental elastics is still very good with this force criteria.

Fishing wire routing



The routes of the fishing wire (that represent tendon lines) travel from the tip of each finger to the actuation system located on the gauntlet. For most of the wrist actuated prosthetic hands, tendon lines pass over the wrist, and this routing strategy generates a direct link between the wrist flexion angle and the fingers bending position. As we don’t work on a wrist actuated but motor actuated solution, we had to separate the wrist movement from the fingers behaviour. The solution is to route the fishing lines through the wrist rotation axis. This allows the user to bend her wrist freely without altering the tension in the fishing wire.

After having tested fingers behavior, we know that the grip efficiency will be improved if the least force is lost between each actuator and its related finger. That is why we also want to minimise friction along the tendon lines. As expressed previously in this post, we felt that the wire routing strategy may have an impact on tendon line friction and we decided to assess the importance of this impact. This led us to make trajectories as smooth as possible, and to pass wires through PTFE pipes.  


Because the tested e-Nable prosthetic hands proved to feature a wide range of force to bend the fingers, and this could be due (among other reasons) to differences in tendons routing strategies, we decided to search for a routing that generates as little friction as possible. Several fishing line routes were tested to find the increase in contraction force due to friction, depending on the angle of the curves along the route (see figure below). Additionally, the results were compared to see how PTFE tubing affected this force. The test showed that if the route is lined with PTFE tubes then the friction force is reduced by half, and that the shallower the curves in the route the smaller the additional friction force.

Geometric parameters of tested trajectories for tendon path.

Arch and wrist axis, with tubing holes.

Final tendon path, as smooth as possible, and through PTFE tubings


For the fishing line actuation to work while allowing Nathalie to move her wrist, there should be no change in tension of the fishing line, therefore no change in its path length. To achieve this the five routes for the fishing line had to go through the axis of rotation of the wrist as the path length remains constant at this point. The smooth trajectories and PTFE tubing then ensure that the fingers still require a low force to contract as it has a lower coefficient of friction than 3D printed ABS.




Two servo motors, each driving a pulley, contract fishing lines that act as tendons in the fingers. The first servo motor drives both the thumb and index finger, and the second drives the remaining three fingers. A whippletree mechanism is built into the pulleys to allow for adaptive grip, as well as tensioner boxes similar to those found in other e-Nable hands.

Pulley CAD model

In case of unbalanced load fishing wire slips on Whippletree.


The pulleys are made up of two separate parts, the first being the pulley itself (blue) and the second being the pulley insert (red). Inside the pulley insert a Whippletree is located. The Whippletree allows two of the fingers to be attached to the same piece of fishing line. When one of these fingers is experiencing a greater load acting against it than the other, the imbalance in force causes the fishing wire to slip round the Whippletree, therefore ensuring that other finger can continue to contract without the motor overloading. The pulley insert can also be moved further into the pulley via a tensioner system. As a screw is rotated it pulls the insert into the pulley, therefore increasing the initial tension in the fishing line and tuning the initial position of the fingers.




 The user controls the prosthesis with their residual thumb joint using a joystick inside the palm.

Sensor types evaluation

After a bibliographic review of sensors classically used for prostheses control and their potential performance, an experimental study of a variety of interfaces have been performed in order to evaluate their usability in the context of this prosthetic hand. The tested sensors were low cost myoelectric sensors, pressure sensors (FSR), and flexion sensors.

Examples of sensors tests: (a) Two flex sensors. (b) One flex sensor. (c) One myoelectric sensor. (d) One flex and one myoelectric sensor. 

All sensors presented an element of inaccuracy, with the myoelectric sensor being particularly unreliable. This was primarily due to difficulty in finding optimal electrode placement.

After analysing the ergonomic capabilities of the recipient, the team finally converged on the opportunity of making use of the mobility of her residual thumb mobility.

A series of small joysticks have been tested, trying to find one easy enough to manipulate with low movement amplitude, and being also as compact as possible to be integrated in the palm thickness. 

First joystick attachment test rig.

Compact PSP joystick

Potential areas to place the joystick

Joystick connected on the palm interface, with sponge foam for elasticity.

Palm-joystick interface part, by offset of the scanned palm.

Joystick and interface integrated in the palm.

Finally, a very compact PSP joystick was found, saving space and avoiding too important modifications of the external prosthesis shape. Also sponge has been put in between the attachment and the prosthesis, to avoid unwanted movements and to ensure the joystick returns naturally to its neutral position.



Typically, electronically assisted prostheses use myoelectric sensors as a means of control, and this is a proven method that has seen much success in modern prostheses as a controller. Despite this, the users of such prosthetics often report them to be difficult to control and unpredictable. The myoelectric sensors often provide noisy signals that can worsen if sweat gets between the sensor and the users skin. A prosthetist is also usually required to find the optimal sensor locations which can be costly and time consuming. For a controller, predictability and reliability is key.

In the case of this project, the user is a trans-metacarpal amputee who, on her right hand, is missing all her fingers but retained some of her thumb joint. This joint has a large range of motion and is capable of precise movements, making it ideal for a means of control. To interface as closely as possible with the user’s central nervous system and to therefore implement an accurate controller, the team decided that utilising this joint would be preferable over myoelectric means or other types of sensors.

The joystick controller was successfully used throughout testing with Nathalie using a basic open-close or ‘bang-bang’ configuration. Moving the joystick towards the palm (abduction) closed the fingers, and away from the palm (adduction) opened them. It was proposed that variable speed could also be implemented, as well as being able to turn the system on and off and toggle between grip patterns, although sadly there was not time for this to be implemented or tested properly. A joystick is a very versatile tool that can be used as a complex controller and has been used as one for decades in devices like video game controllers. It’s therefore highly likely that, with training, a user could learn to give more complex commands to the prosthesis.

Force Feedback



A pressure sensor in the index fingertip indicates to the user the grip force being applied to an object by vibrating one of four haptic motors in the gauntlet. The pressure sensor is entirely concealed within the NinjaFlex fingerbone. The haptic motors vibrate on the surface of the user’s forearm, with a different motor vibrating depending on the force applied through the fingertip. The motor closest to the hand vibrates upon contact with an object, and the vibration moves further up the forearm as the force increases.

Left: Index finger and sensor location. Middle: partially disassembled index finger showing the conductive fabric route leading to the pressure sensor in the fingertip. Right: Haptic array in the gauntlet made up of four vibration motors.


Researchers have found that implementation of force feedback is of great use to an individual in terms of improving embodiment of a prosthesis and improving performance when grasping, particularly for delicate objects. The implementation of force feedback was therefore explored in this project using low cost materials. Conductive fabric was used in lieu of wire to avoid the wire fatigue that would inevitably occur in the finger under repeated flexion and extension. Although not done so here, pressure sensor(s) could also interface with the actuators using the Arduino as a means of closed-loop feedback (i.e. limiting the maximum force that can be applied to an object).

Attachment on the forearm


In place of a usual Velcro straps or similar, a more comfortable option has been investigated by the use of soft fabric that is tightened by a fishing wire and a locking mechanism inspired by BOA Fit system (often used for snowboard boots) and its use by Younes Zitouni in the community.


Parts of the twistable locking mechanism

Comfortable gauntlet tightening mechanism


In order to assess the maintaining potential of the solution, the wire strength was tested up to breaking force. Also different routings were tried, and it was seen that a double helix (cross hatched method) routing was a better option to have a uniform tightening and distribution of the load, together with the easiest releasing process.  Several types of fabric were compared to converge to “Tissus3D” which provides a comfortable feeling to the user, and, in addition, is often used by prosthesists for skin interface.


Based on the binding of a snowboard boot, this fastens the prosthesis to the user in a similar way to a shoe lace. This method was developed due to the user being unsatisfied with the Velcro straps on her existing prosthesis.

Optimizing finger configuration



It appeared during tests with Nathalie that she was often not able to grasp a bottle with our prototype. This was due to fingers contracting so that the distal joint rotated fully before the intermediate and proximal joints began to bend. This led to the distal bone making perpendicular contact as seen in figure Figure 16‑11, followed by the object being pushed away from the palm. To allow the prosthetic to grip the objects, a more natural order of contraction needed to be achieved.


Below is compared on the first line of images the bending process of the original Flexibone finger, and on the second line the bending process of the final one.


After a series of tests, this issue was solved by increasing the thickness of interim and distal flexible joints in the bone, by 25% and 50% respectively (see figure below). A drawback is that it increases the force needed to completely bend the finger (see curve below). Due to time constraints, a final global decrease of thicknesses in all three joints has not been investigated.

Grip patterns



Thanks to the Arduino code and the two-axis joystick, several functionalities have been implemented. And others could be developed if needed. The main grip patterns are described on the following figures: finger point grip, power grip, two-point tip pinch, lateral grip.


Final Gauntlet

Here is a summary of the final gauntlet layout and casing.


Potential improvements

Among many ideas to improve technical or functional aspects of the Flexibone assisted hand, here are the main priorities:

  • Thumb rotation plane: The orientation we made for the thumb does not really allow opposition with the forefinger (see power grip image above), which makes it difficult to grap some objects. Maybe we did not put the same way as the original Kwawu… This plane could be slightly modified, and the thumb movement amplitude also increased.
  • Haptic feedback : The position of the pressure sensor in the forefinger tip does not act always the same way depending on the contact configuration with the object. Could we find a better configuration, or a different sensor, so that the contact force would be properly detected in most situations? Also which haptic motors is activated seems difficult to figure out by the user. She says she feels vibration in the gauntlet but she feels difficult to know on which location.Reducing the contraction force of the fingers
  • Reducing the contraction force of the fingers. To improve finger contraction strategy we increased some of the Flexibone joints thicknesses, which finally increased the bending force that the fishing line has to provide. The drawback is that the greater the force that is lost through the tendons, friction, and bending of the fingers, the lower the force that remains for gripping objects. We should now investigate the possibility to decrease the flexion force by decreasing all joint thicknesses, keeping the relative flexion order and of course keeping a reasonable mechanical strength of the fingers.

A full open source project


Not only this article, but all data generated along this project is released as the open source “Flexibone Assisted Hand” under a licence Creative Commons Attribution (CC By).   We are happy to provide the community worldwide with:

  • Finally, if you just want to print the parts and build a copy of this, you can find STL files on Thingiverse
  • We will appreciate if you follow any works on the project with comments.


Thanks to the team Gre-Nable Philippe, Patrick, Fabien, Marie-Laure, for allowing us to work on this amazing project.

Thanks to Frédéric for his coaching in project management.

Thanks to all the Team of GINOVA, the University Fablab in Grenoble Institute of Technology for his technical support during this project.

Thanks to Patrick also for his help in publishing this article on the Team Gre-Nable blog.

And of course, many thanks to Nathalie (and her family) for her confidence in our work, and for being available for several tests along the project.

During the final meeting, from left to right: Tom, Phil, Nathalie, Sevinç, Jay. 

A knife holder for a child

A knife holder for a child

The context 

Lina is a 5 years old girl when we meet her for the first time. She was born with a malformed right hand equipped with two fingers. One of these is weak. She cannot flex both of her fingers, but only move them laterally as a kind of needle nose pliers. And she is very comfortable for most of everyday life activities… except for some actions that she cannot perform. And an example is holding a knife. Her parents are used to help her cutting meet, and she is used to push rice to the fork (which is handled by her left hand) with her two fingers.  

But now she has to have lunch at school, or sometimes diner in  a restaurant, and she probably feels not so comfortable in front of other children.  Thus Eric her father asks e-Nable community if someone can help in developing a kind holder for Lina.

Modeling of a fitting socket

After a first meeting with Lina and Eric, we consider that a standard e-Nable hand cannot fit the need of Lina, and we decide to develop a very dedicated knife holder.  It will be made of a socket in which Lina can insert her right hand, and some specific shapes where she can fix different kinds of knives.

Hand casting 

The first operation is to get a model of Lina’s hand. This is done by casting her hand in pink Alginate to get a positive plaster copy.  

Shape modification

Unfortunately she bent her wrist during casting operation. After the 3D scanning and reverse engineering process that leads to a clean digital model, the next operation to get a proper model usable to make a comfortable socket is then to unbend the CAD model. This is performed thanks to a specific “flexion” function available in SolidWorks. This allows to put the hand inline with the forearm, and also to put the two fingers closer to each other.

Design of the socket

The next step is to design the socket. We come back in our collaborative CAD software, Onshape, to draw a few sections and build a “loft” that approximately fits Lina’s hand. Then two offset operations of surfaces lead to the inner and outer shapes of the socket, with 3mm gap dedicated to put a comfortable 3D fabric that can absorb moisture and can be easily removed for washing. 

Designing the knife holder

The input information for designing a knife holder it to have an idea of the types of knives that will be used with this device. Eric, Lina’ farther, provided two CAD models (he drawn by himself) of knives available at home: the “knife to push” and the “knife to cut”.


Then we started with the idea of fixing the knife handle in a flexible interchangeable part that should fit both types of knives, and to add a simple slot to center the blade. In the first prototype the blade has been tied with adaptive strap made of Velcro type band (ID-Scratch).

After validation of the orientation and position of the knife by Lina, for the second version the strap was replaced by a simple neodymium magnet. You will find below a photo of the first version in test (pushing knife), and several CAD views of the last version.

And then…

Lina is happy, she can eat without asking for help to her classmates;

Eric is happy, Lina eats now without pushing noodles with fingers.

Eric asked for advice because (and that’s also good news!)

  • he is in the process to learn how to design with Onshape. He already made a new insert (the blue part on screen shots above) that will fit another knife handle.
  • And he is in the process of buying a 3D printer and we hope him to become a new maker within e-Nable France community 🙂


The CAD models are available for inspiration, adaptation to other cases, and hopefully for improvements under, if you have an account (free for non profit activities and public models), you just have to search for “Team Gre-Nable : knife_holder” among the public models. 

Please just let us know if you design an adaptation of this!


Terminal Socket for Keyboard Typing & Smartphone Use

Terminal Socket for Keyboard Typing & Smartphone Use

Open source project:

This design, like all those made by Team Gre-Nable, is open source. We explain in these articles why and how they were done, the reasons for our technical choices like the means and tools we used. All 3D models are available under the Onshape environment (free access for public models). This design is accessible on Onshape by using the Search tool with team : manchon Jean“.

Jean’s Situation …

Jean is an adult suffering from agenesis of the right arm. He has his elbow and forearm only 9cm long, with very conical form and ending with a small “bud” … he has become accustomed to use among others to typing his MacBook’s. And we must admit that it is impressive with his “dexterity”. But the more and more intensive use of his computer for his work ends up hurting him and generating pain from his little “finger”. He would like to protect his skin while maintaining his dexterity he has acquired in the usage of the keyboard. He has already consulted two or three professionals who made him special appliances, which ultimately did not suit him. He then contacts Team Gre-Nable.


Molding & Casting

An alginate molding session provides a plaster model of Jean’s forearm and elbow. Note on the image below many bubbles that have been trapped in the hair of his arm during molding, which generates these small “balls” distributed on the surface of the plaster model. These were removed very easily with a cutter blade on the plaster. We can distinguish very well in the upper part the “bud finger” Jean used until now to type on the keyboard of his computer. In the lower part of this molding, the slight restriction of section is due to the presence of an elastic band we place intendely to clearly mark the location of his elbow during molding. This restriction will be removed (smoothed) between the 3D scanning operation and the CAD volume reconstruction.

Plaster model cast from the alginate mold.

Comparison of 2 brand/types of 3D scanners.

This plaster model is digitized with a 3D scanner. We took this opportunity to compare the iSense (see also  with a high-end HandyScan 700 professional scanner. HandyScan 700 .

The Handyscan will be considered as the reference tool, with a reported accuracy well below 0.05mm on this type of object.

Results of scan shows the iSense giving efficient results for this application (accuracy less than 0.7mm, average around 0.5mm) in most areas with few variation of curvature, however, in more rugged areas, errors can reach nearly 2mm. But since we foresee the insertion of a comfort glove with thickness of 3mm (blue 3D fabric visible on the featured image near the Jean’s elbow: the process of designing this comfort glove is fully described in his post), and that we can also count on the adaptability of the flesh in contact with the socket, it seems that the quality of the iSense will be sufficient to digitize this model in plaster.

Please note that a direct digitization of the arm would probably have generated larger dimensional variations. We will keep the accurate scan made with the Handyscan as we have it.

A CAD model will be created based on this 3D scanning (obviously a mesh file), and exported in STEP format for use by Onshape online webapp.


Socket Design

Two new models are generated thanks to VXelements application (associated app with Handyscan scanner of Creaform)), with offsets of surfaces of 3mm then of 5mm. We wanted to test these features in VXelements (and we were very satisfied with the results) nevertheless these surface offsets and the generation of the new volumes could have been done with other modeling tools, either on the STL model (with Meshmixer for example) , or on the reconstructed STEP model (Fusion360, Onshape, etc …). You can read our post “Création d’un Multi-tool holder“for an example of using the “Surface Offset” function in Onshape.

The three volume models obtained, which we will call “ arm “, “ arm + 3 ” and “arm + 5” are imported into Onshape in STEP format. A boolean (volum) subtraction operation between “arm + 3” and “arm + 5” makes it possible to obtain a socket with a thickness of 2 mm thick, 3 mm away from the arm.


ARM and initial socket


Clearance around pinky

An additional offset of the surface, followed by some cuts and re-assemblies of volumes allow us at this stage to release a significant clearance at the end of socket, which will avoid the contact between the Jean’s finger and the socket. This protection is the main element of his design “specifications” transmitted to us, so we pay special care for this!

Then we add an artificial “finger”, which will be equipped with a flexible tip ( made of NinjaFlex) to allow a soft touch with keyboard’s keys.

Socket and finger, external view.

Preparing 3D fabric, and clearance provided around Jean’s bud to prevent any injury.

Socket and finger, section view on.

3D fabric : finger clearance.

Adaptation for touch screen

Like most of us Jean uses a smartphone or tablet and thus a touch screen.Could he take advantage of this artificial finger to manipulate the apps on this type of screen ?

I had studied this issue few months ago, and I came to the conclusion that it should be possible. Indeed most of the current touch screens are capacitive, and speaking without too technical terms (which I would not control for that matter!) the touch screen detects a variation of potential generated by a slight leak of electrons when the finger touches the sensitive surface. It “would be enough” therefore that the end of our plastic finger (in this case the foam cap) would be connected to a sufficient electrical mass to provoke a slight electrically discharge during contact with the screen. The solution is …

  • drilling a small channel inside the artificial finger to insert a flexible electrical wire,
  • to replace the current printed tip (made of Ninjaflex) with an conductive tip retrieved from a smartphone stylus,
  • and to connect the wire to a metal mass, and possibly touching the skin of the user.

Channel to end of the tip for the wire

Wire exiting toward conducting tip

Former stylus tip placed on the finger

Electric mass made of aluminium foil (kitchen product)

Electric wire as an alternate solution

The smartphone stylus tip was taken on a stylus of this type.  Tests show that direct contact with the user’s skin is not necessary and good news as well ! What’s more, aluminum foil is not essential either. The system works very well with just a turn of wire in the socket, wire which is located about 3mm (the thickness of foam “3D fabrics” of comfort) of Jean’s arm.

Results of first tests

The second meeting with Jean (after the molding) was really satisfying.
remaining point was to imagine a way to maintain the socket well in place around his forearm. The first tests were done by placing two pieces of adhesive Velcro on the edge of the socket, and placing another strip of Velcro around his arm. The following videos taken during the first minute of use in each context (keyboard and touch screen) show that Jean will undoubtedly succeed in appropriating this new tool … if it is already done from this moment.


Very first try, Jean typing on his Mac.

Jean’s first test on his touch screen. See how smoothly he zooms with his left thumb and right “index”!

Holding the socket on the arm

The very conical shape of the Jean’s forearm does not allow effective hold by simply clamping in this area. The first tests showed the feasibility of maintaining a strap around the arm, but we are looking for a solution that would be easier to handle, and that would avoid a localized tightening certainly not comfortable.

It seems then that the proposal made by Dominick Scalise to use a fabric sleeve for some prostheses would fit here.
A first test is done by cutting a sock … it seems to work well, and we propose to Jean, by email, to test himself this solution with the socket it uses.  

Meanwhile, we also offer a socket version printed in TPU (semi-flexible material) rather than PLA, to further improve comfort. Let’s not forget that John wants to use this device every day for several hours. 

In our third meeting, Jean introduced us to the fixation solution he found : he replaced the idea of ​​the sock with an ankle (usually used in the case of a sprain) that provides homogeneous and very efficient clamping. Just find the right dimension, and he will use or not the complementary elastic straps that come with it.

The TPU version seems to seduce him on the comfort side, and he plans to reduce the length, or even cut a slot on the side … what we do on the spot.

Printed release with TPU with slots. Please note the “glove” made of 3D fabric.

Hold with ankle (it may be necessary to use the size below)