Twister3: a simple and fast microwire twister

In order to achieve straight TWPs, wire must be twisted together while under tension. In all other existing designs, a weight is hung from the wire bundle and a motor is loosely coupled to the weight in a way that does not constrain axial motion. This provides constant tension on the wire bundle due to gravity while allowing the helix to decrease in length as it is twisted. A variant of this method uses magnets hold the weight in place ⁷ , but, in our experience, this is unnecessary and is prone to causing wire breakage due to the nonlinear force/distance relationship of magnetic attraction.

A key design criterion for our device is that TWPs need to be turned very quickly. We aimed for < 10 seconds of twisting time (time when motor is in motion) per TWP. Assuming 100 total revolutions (upper estimate for tetrodes) this translates to an average turn rate of 600 RPM. Because previous methods rely on loose motor coupling, they were unsuitable to meet our speed requirements. For example, the SpikeGadgets machine, original Open Ephys Twister, and Neuralynx Spinner-2.0 are limited to approximately 60 RPM. The high centripetal forces involved in rapid turning inevitably leads to instability of the coupling mechanism, causing the bundle to vibrate wildly. Therefore, we sought to rigidly constrain the motor and wire bundle in the turning plane, but still maintain freedom in the axial direction to allow bundle tensioning and shortening during the turning process. To meet this goal, we made use of selective laser-sintered polyether block amide (PEBA) to create a monolithic, combined leaf spring/wire-retention mechanism (figure 1(B)). PEBA provides rubber-like mechanical qualities resulting in a spring-constant low enough for use with microwire (∼32 mN mm⁻¹ in the relevant range of motion (figure 1(C)).

To interface the microwire bundle with the motor, the bundle is clipped together using a standard, ferrous alligator clip that has been coated in insulating shrink wrap. The bundle is then drawn towards the motor while the leaf spring is lifted with with the user's free hand. While the leaf spring is under tension, the alligator clip is attached to a strong neodymium magnet on the leaf spring assembly, which provides the bundle-to-motor linkage (figures 1(A), (B)). During this process, a 3D-printed alignment jig on top of the magnet automatically guides the microwire bundle in line with the motor's axis of rotation. This alignment jig must be made from a material that can withstand the high temperature of the hot air during the fusing process and is smooth enough to ensure it does not snag the microwire as it guides it into position. We have found that 3D-printed rhodium-plated brass works very well for this purpose. The leaf spring is then slowly lowered until it reaches equilibrium with upward wire tension. At this point the motor can be turned.

Because of print tolerances and the fact that the leaf spring does not deform exactly vertically, the resting position of wire bundle will likely not be perfectly co-linear with the motor axle. To account for this, the motor mount consists of a two pieces: the base, which is rigidly concentric with the motor shaft, and an alignment plate. The alignment plate is friction fitted within the base and provides ∼2 mm of omnidirectional planar adjustment (figure 1(B)). This plate should be moved until the resting position of the wire bundle is in line with the motor axis.

This wire attachment mechanism has two advantages over existing designs. First it can be used rapidly because of magnetic coupling. Second, because the microwire bundle is automatically aligned with the motor's axis of rotation, high-turn rates do not result in oscillations or instability. We have found that this is a critical feature in order to produce straight, even-pitch, and consistent TWPs using very fast turn speeds.

To further increase the speed at which TWPs can be made, microwire needs to be drawn and attached to the motor quickly. Ideally, this process should occur with as few separate actions being taken by the operator as possible. To facilitate the rapid draw of wire from stock spools, we designed a torsion spring-based feeding assembly that allows wire to be quickly drawn from stock feeder bobbins (figure 1(C)). This mechanism applies enough friction to feeder bobbins to counter increased wire tension during a twist, transferring all slack compensation to the leaf spring (figure 1(B)). The holding (stiction) force of this mechanism is adjustable to account for the elastic deformation of different wire materials. We have found that, when working with standard tetrode wire ⁸ , the 2nd stiction setting (corresponding to a threshold of ∼11.5 N per spool), is adequate to counter wire tension and therefore prevent spools from improperly feeding during a twist. However, the proper stiction setting is dependent on the microwire material and will need to be adjusted for other wire types.

To obtain precise control over motor acceleration, speed, and position we to used a bipolar stepper motor to perform wire twisting. We have found that, due to their discretized motion, stepper motors can vibrate resonantly with taut microwire, resulting in irregular twists and wire damage. To overcome this issue, we drive our motor using an advanced microstepping driver (figure 2(A)); Trinamic TMC2130). In our case, we use 200 steps/revolution ($1.8^{\circ}$) motor. Microstep commands from the microcontroller are provided at 16 microsteps/step, which are further interpolated to 256 microsteps/step by internal driver circuitry. This results in a motor update resolution of 3 200 microsteps/revolution ($0.113^{\circ}$), and a motion discretization of 51 200 steps/revolution ($0.007^{\circ}$). Therefore, the motor operates approximately as smoothly as a continuous DC motor but with much improved motion control dynamics.

Zoom In Zoom Out Reset image size

We used an Arduino-compatible Teensy 3.2 ⁹ microcontroller module to perform step timing calculation. Its NXP MK20DX256VLH7 Cortex-M4 includes an integrated programmable interrupt timer (PIT) which is used to provide jitter-free step commands to the motor driver while acceleration calculations are performed ¹⁰ .

The control box (figure 2(B)) is powered using a 12 V DC center-positive barrel jack that supplies at least 1.5A. It has a single user input: a control knob consisting of a combined quadrature rotary encoder and tactile push-button. This knob permits the following user actions:

• Press: cycle through different settings (forward turns, backward turns, turn speed, turning mode)
• Turn: increment or decrement the selected setting depending on turn direction.
• Press and hold: execute the turn sequence using the current settings
• Press during motion: cancel the twist and stop the motor immediately.

The control box is used to perform two tasks: twisting electrodes (mode 0) and loading bobbins with microwire (mode 1). The turn mode is selected and changed using the dial on the controller. The selected turn mode is shown in the upper right corner of the liquid crystal display (LCD). After a mode is selected, all turning parameters (speed, forward and backwards turns) pertain to that mode only. All parameters are stored in non-volatile memory when a turn is started by pressing and holding the control knob. In the following sections, we detail how to use the mechanical components for making TWPs and loading bobbins with stock wire.

Before twisting electrodes, the bobbins on the wire feeder assembly must be loaded with microwire (figure 3(A)). The following paragraphs detail the bobbin loading procedure. A video of the bobbin loading process is available to aid these instructions (appendix A).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

First, the bobbins must be removed from the wire feeder assembly. To do this, the wire shield is removed by unscrewing its M6-retention screw as indicated in figure 6(6). Next, the bobbins are removed by unscrewing the M3 bolt that serves as the bobbin axle (figure 6(4)). The bobbins are then slipped off of each axle. If the bobbins have any left-over microwire, it should be removed and the wire grooves should be examined in to ensure they are clean of dirt and debris. They can be wiped with a cotton swab and water if they require cleaning.

Next, the 3D-printed leaf spring is removed from the rotor base. Because each bobbin has embedded magnets that mate with those on the rotor base, after the leaf-spring is removed, a bobbin can be placed onto to rotor base (figure 3(A1)). Once the bobbin is in place and reasonably centered using the adjustment plate (figure 8(5)), the wire guide should be positioned such that its tip points directly into the center of the wire groove on the bobbin. Properly loaded bobbins will have microwire tightly wound around their center groove. This relies on carefully adjusting the position of the wire guide such that it is tip is located about 1–2 mm from the center groove (figure 3(A1)). After this, wire is feed from from the stock spool through the wire guide and wrapped once around the bobbin in its center groove (figure 3(A2)).

Finally, to load wire onto the bobbin, the controller is set to mode 1 (figure 2(B)) and then the desired loading speed and number of revolutions are selected. We have found that 100 RPM works well for most wire (figure 2(B)). The circumference of the bobbin is ∼10 cm. Therefore, the length of wire loaded on the bobbin is turns × 10 cm. Including wastage, this results in a conservative estimate of 1 TWP per turn (table 1). After the desired parameters have been entered, loading is started by pressing and holding the controller button. Be careful not touch moving parts during this process: the microwire needs to have constant tension to ensure it is properly loaded on the bobbin. The loading processes should be monitored as it begins to ensure that microwire is being accepted by the bobbin. If there is an issue, pressing the knob on the control unit will halt the process so it can be corrected.

Zoom In Zoom Out Reset image size

Once loading has completed, the process is repeated for the for the remaining bobbins. After each bobbin has be loaded with microwire, they are put back on their axle on the wire feeder assembly. Loose wire ends should point inward on both sides of the assembly (figure 3(B2)). Finally, the wire shield is replaced (figure 6(6)).

Zoom In Zoom Out Reset image size

The following steps detail TWP construction using Twister3. If you are interested in making stereotrodes instead of tetrodes, follow the same steps but only use two bobbins on diagonally opposing sides of the feeder. A video of the TWP construction process is available to aid these instructions (appendix A).

If it is not already in place, the leaf spring must be attached to the base rotor using its magnetic interface (figure 3(B1)). With the leaf spring in position, wire from each of the bobbins is grouped using a finger pitch. The bundle is then clamped using the alligator clip (figures 3(B2), 7(B)). The wire bundle does not need to be in a certain position within the clip or be tightly grouped Additionally, each wire does not need to begin at exactly the same length from the feeder because those with more slack will not pull on their bobbin until they it is equal in length to the shortest wire. At the point that all wires are drawing slack, they will be equidistant from the wire feeder.

Zoom In Zoom Out Reset image size

Once the wires are clamped in the alligator clip, it is flipped $180^{\circ}$ such that the wire bundle wraps around the bottom of the clip and exits its rear face (figure 3((B3), inset)). This will ensure that the wire does not slip out of the clip as it is draw from the feeder. Do not worry if the wire is not tightly bunched as the twist alignment jig will keep the bundle exactly concentric with the axis of rotation during twisting. Using your free hand, the twisting attachment's leaf spring is lifted until under slight tension, about 1.5 cm (figure 3(B3)). Then, with the leaf-spring raised, the alligator clip is drawn down to meet the magnet on the twisting attachment, feeding the bundle into the aignment jig (figure 3(B4)). After the clip is magnetically mated, the leaf spring is smoothly lowered until it is in equilibrium with the upward force produced by the wire. Each of the wires should be pulled straight. Do not let the spring snap back under its tension, as this will leave slack in the wires. If any wire has slack, its bobbin can be turned backwards slightly until it is taut. Finally, ensure all wires are guided through the center of the alignment jig (figure 3((B4), inset)) and the loose ends are not interfering with the taught portion of the wire. If so, they should be cut before performing a twist.

Before performing the first twist in a session, ensure that the alignment plate has been adjusted so that the point at which the wire exits the alignment jig is co-axial with the motor shaft (figure 8(5)). To do this, adjust the position of the plate until until the point at which the wires enter the alignment jig entry point does not 'wobble' with respect to a steady background object when the motor is manually turned back and forth. This should only need to be done once per session. After alignment, the controller is set to mode 0 and the desired number of turns and turn speed are selected (figure 2(B)). We use 900 RRM for our wire (table 1). These settings only need to be entered once because they are saved to non-volatile memory by the controller every time they are changed. The button is then pressed and held to start a twist.

Zoom In Zoom Out Reset image size

After the twist is complete, the wires are fused using hot air. The exact parameters of the hot air gun will dependent on the melting point of microwire insulation material. For polyimide, we have found that 480 ∼ $480^{\circ} {\rm C}$ is a good temperature. We have found that the air flow is not critical, but should not be so high that it deforms the taught wire during the fusing process. Settings for our hot air gun (table B3) lab are provided in table 1.

Once the air gun has reached a stable temperature, the nozzle is held ∼5 mm from the from the point at which the twisted microwires separate towards the feeder bobbins (figure 3(B5)). It is then moved downward slowly and smoothly (∼2 cm s⁻¹) until it reaches the wire alignment jig. The direction is then reversed and it is raised until the nozzle returns to its initial position at the wire separation point. The nozzle is then moved away from the wire. Starting from the top and fusing downward is important: we have found that fusing from the bottom and moving upwards will cause the lower portion of the TWP to 'absorb' slack from above resulting in a very fine twist pitch and a TWP that is shorter than intended.

After wire fusing, both hands are used to simultaneously roll each of the bobbins forward until the leaf spring relaxes and there is no tension on the microwire (figure 3(B6)). The loose wire above the fused TWP is then cut using scissors (figure 3(B7)). When performing this cut, make sure to leave enough free wire for connectorization. Finally, the alligator clip is removed from the magnet in the wire alignment jig and the finished TWP is cut into a storage box (figure 3(B8)). The process can then be repeated for the next TWP.

Choosing twisting parameters will require some experimentation in order to produce TWPs with the desired geometric and mechanical properties given the user's choice of wire, implant type, and animal model. In our labs, we use Twister3 to make tetrodes for microdrive implants in both mice and rats. The operation settings that we use are shown in table 1.

Two settings of note are the height of the wire feeder above the motor assembly, which determines the probe length and microwire twist pitch and the bobbin stiction threshold which determines the wire tension during twisting and drawing. Figure 4 shows the effect of changing the wire feeder height on resulting tetrode characteristics for 12.7 µm diameter, polyimide coated nichrome and 0.6 µm, polyimide coated nichrome mircowire types. For both wire types, lowering the feeder closer to the motor, and therefore increasing the angle of wire divergence from the axis of rotation does decrease twist pitch and probe length for a given number of turns. However, the buckling point (figure 4) and stiffness (figure 4) of tetrodes across feeder heights for each material are remarkably stable. Probe stiffness across materials and wire diameters is very different, as expected.

To test probe functionality, we produced tetrodes using the device settings in table 1. Using procedures that were approved by the Committee on Animal Care of Massachusetts Institute of Technology and followed the ethical guidelines of the US National Institutes of Health, these tetrodes were gold plated [15] and used in combination with microdrive assemblies [5, 16] to obtain recordings in the pyramidal cell layer of CA1 of mice and rats. As expected, these tetrodes reliably produced characteristic LFP and multiple, well-isolated units in both mice (figure 4(D)) and rats (figure 4(E)).

To quantify Twister3's speed of operation, we measured the tetrode construction time of three users. All users had ∼1 hour of experience with the device at the time of testing (figure 5). We divided speed measurements into three steps: (1) wire clipping/drawing, (2) wire twisting/fusing, and (3) tetrode removal and storage. Time trials were performed using the device parameters shown the 'Mouse' column of table 1. The motor-in-motion time was constant (9.1 s at 1000 RPM max speed) and therefore incurred a constant offset on step (2), as indicated by the dashed line in figure 5. All users were able to create tetrodes at a pace exceeding 1 tetrode/minute. However, each user had slightly different strategies when using the device. For instance, user 3, who has large amounts of experience with tasks requiring fine motor skills, was relatively quick for steps (1) and (3) and was relatively slow for the fusing step (2). User 1 had occasional difficulty clipping and the wire during step (1), increase their average time. These discrepancies indicate that there is room for future improvement and automation, especially with respect to wire clipping and fusing.

The mechanical portion of Twister3 consists of common hardware, standard optomechanical components, and 3D-printed parts. Wherever possible, we used standard (and easy to replace) parts. The mechanical bill of materials is shown in table B1. In addition to these standard mechanical components, several 3D-printed parts are required (table B2). These parts are available for direct purchase from third-party 3D-printing services via the links provide in the table. After obtaining the required components, the wire feeder, wire-guide, and stock-spool assemblies can be constructed by following steps detailed figures 6 and 7(A)–(C), respectively. After each of these modules is complete, they are combined into the complete device by following the steps presented in figure 8.

The control board is comprised of the following blocks: power regulation, motor-driver, microcontroller, and user interface (figure 2(A)). Wherever possible, we used pre-assembled modules (microcontroller, motor driver, and LCD display). The bill of materials for this board is shown in table B3. Printed circuit board designs and Gerber files are available on the Twister3 repository ¹² .