Tethers

A tether does the job a free-flying or wheeled vehicle cannot do on its own: it reaches a target that sunlight and line of sight do not, by carrying power, data and load-bearing force down a pit wall, into a crevasse, or across a rendezvous gap where propellant would otherwise have to do the work, an argument made at survey length for rappelling rovers, tethered antenna structures and ice-probe descent alike [2][6]. That reach comes at a structural cost the vehicle pays continuously rather than once. A tether on a planetary robot carries three loads at once, the vehicle’s weight during a rappel, electrical power, and data, and the mechanical property that governs it is not a single breaking force but the breaking force at the bend the tether is actually pulled over, at the temperature it is pulled at. The measured spread across those conditions is larger than the margin most sizing exercises carry, and the two numbers a designer would normally start from, the vendor’s rating and a design tool’s prediction, have both been checked against pulls and both were wrong.

The design space splits along what the tether is for. A rappelling rover’s tether is a structural member first and a conductor second: it must survive being dragged over rock edges under the vehicle’s full weight, so its cross-section is driven by abrasion and bend life rather than by conductor resistance [1][5]. A power tether is the reverse: its cross-section is driven by resistive loss over distance, and the resulting voltage is high enough that the design problem becomes power electronics rather than mechanics [4]. An orbital tether for momentum transfer or debris capture is neither: it is sized against tensile strength per unit mass and against micrometeoroid and debris impact risk over a mission of years, with no rock edges and no rover weight to carry [7]. All three converge on braided, jacketed, multi-layer construction, because no single material meets the strength, abrasion and flexibility requirements at once, and every one of them has had its vendor or design-tool prediction checked against a real pull and found off in one direction or the other.

The data below come from two JPL lines of work: the TYMPO lunar surface power tether, built to carry 1 kW at 1.5 kV over 1 km [1][4], and the Axel rappelling rover’s tether management and sensing hardware [2][3], plus a separate qualification of an Axel-family vertical-pit tether for the same vehicle class [5].

Breaking force is a property of the bend, not of the cable

The TYMPO electromechanical tether is four 22 AWG copper conductors in FEP insulation plus two radiation-hardened single-mode fibers, helixed on a central core inside a braid and an extruded FEP abrasion jacket [1]. It was pulled to failure over three bend fixtures at three temperatures. Every row below is one measured pull on one specimen [1]:

Bend	Temperature	Breaking force	Elongation
8x diameter	+22.5 C	4259 N	2.6 percent
2x diameter	+22.5 C	3080 N	1.6 percent
78 degree knife edge	+22.5 C	3083 N	1.5 percent
2x diameter	-190 C	7219 N	0.6 percent
78 degree knife edge	-190 C	3502 N	0.4 percent
78 degree knife edge	+200 C	990 N	1.3 percent

Source: [1]. Single specimen per condition, no replicates and no scatter.

The span is a factor of 7.3 [1]. Cold makes the tether stronger and less compliant: over a gentle bend at -190 C all layers failed together at 7219 N at 0.6 percent elongation. Hot over a sharp bend is the governing case at 990 N, and there the layers failed separately, the copper at 503 N and the fiber at 471 N. Neither temperature is a destination condition [1]. The cold case is liquid nitrogen rather than a measured permanently shadowed region temperature, and +200 C is above anything a tether on the lunar surface would see, so the pair brackets the range rather than reproducing it, and no specimen was thermally cycled.

The safe working load the authors recommend is 236 N. That is not a measurement: it is the worst case pull, the 990 N knife edge at +200 C, divided by a factor of safety of two [1].

Abrasion moves the same way as strength. The jacket was cycled under a suspended basalt rock through a bin of NU-LHT-2M simulant at 3.5e-5 Torr and failed at 3981 cycles at ambient temperature and at 1386 cycles at -120 C, a 65 percent drop in abrasion life [1]. The simulant is not regolith, the rock is terrestrial basalt, and 3.5e-5 Torr is many orders of magnitude above lunar surface vacuum, so the pair bounds nothing about vacuum welding or electrostatic dust adhesion.

A separate qualification campaign for the same rappelling-rover class of tether reached the opposite conclusion about how to build the strength member. That campaign tested a tether meant to lower a rover down a lunar pit wall and found that a conventional central strength member is the wrong architecture for that duty: when the tether presses against rock under tension, the core cuts into the surrounding conductors from the inside. The fix was to move the strength member outward, to an over-braided Vectran layer outside the conductors, under an extruded Tefzel abrasion jacket. That construction choice, not a stronger material, is what raised abrasion life: extruded jackets outlasted braided ones by roughly an order of magnitude against both basalt and JSC-1A simulant (12,987 and 2,733 cycles against 1,557 and 1,952) [5]. Tensile strength on that tether held to a modest reduction below its 5.3 kN rating at +150 C and roughly doubled at -50 C, the same cold-strengthens-and-stiffens direction TYMPO measured [1][5]. A 90 degree knife-edge bend under tension broke wires only above 1105 N against a 200 N requirement, a margin the authors describe against their own conservative abrasion model rather than against a measured pit-descent duty cycle. The same tether carried 3 Mb/s RS-485 data over 477 m alongside 300 VDC power on an adjacent conductor pair with negligible jitter, evidence that data and power can share one jacket at rover scale without the power line degrading the link [5].

The vendor rating and the design tool are wrong in opposite directions

The same tether was measured against both the vendor’s numbers and the predictions of TetherCAD, the JPL design library used to lay it out.

Quantity	Measured	TetherCAD	Vendor
Minimum bend radius	12 mm	47.9 mm	109 mm
Breaking force	2524 N mean	8027 N	not stated
Linear mass	58.2 g/m	67.1 g/m	45.5 g/m
Outer diameter	6.0 mm	5.99 mm	not stated

Source: [1]. The measured breaking force is the mean of the three knife-edge pulls at +22.5, -190 and +200 C; the TetherCAD figure averages its straight-run and un-helix analyses.

The vendor’s minimum bend radius is about nine times more conservative than the tether tolerated, which costs spool and capstan diameter directly. The tool’s strength model is roughly a factor of three high, which is unconservative, and it is high for a structural reason: TetherCAD models yield from an elastic modulus at a single temperature, using ambient-temperature datasheet entries, so it cannot predict the cryogenic and hot behavior the same campaign measured [1]. The specimen was also built from wires slightly out of specification, so the article tested is not exactly the article designed.

Managing the tether costs about a tenth of the vehicle

Axel’s passive tether management module masses 6 kg, about 10 percent of rover mass [2]. The driver is the cable rather than the winch: a 50 mm minimum bend radius forces capstans of at least 100 mm diameter, a constraint the authors themselves call overly conservative [2]. The TYMPO result above is what that conservatism looks like when it is checked.

Measured performance of the series elastic tether management module, all on a benchtop rig and none of it integrated with a rover [2]:

Quantity	Value
Tension control bandwidth	5 Hz
Resonant peak and phase reversal	7 Hz
Step response, 0 to 500 N	about 0.2 s
Disturbance rejection error, to 1 Hz	plus or minus 50 N
Disturbance rejection error, 0.1 to 2 Hz	plus or minus 120 N

Source: [2].

Those figures have no requirement to be judged against: the authors state it is unclear what performance would satisfy Axel’s mobility needs. They also report an unexplained disagreement between the tension inferred from the series elastic element and the tension read by the load cell across the whole frequency range, so the module’s true output tension accuracy is unknown [2]. The series elastic mount raises the admissible fall factor by nearly three times over a rigid attachment, by analysis plus drop simulation at 0.5 m/s, with the benefit largest below 5 m of free tether [2].

What a tether sees in the field

One field run bounds the loads the numbers above have to survive. During a descent in the JPL Mars Yard, a 50 m fiber-optic sensing tether snagged on a fiberglass rock at 11 m while the rover’s wheels were bogged in sand [3]. Distributed strain sensing read 700 N at the snag and 500 N along the span back to the anchor. Both readings are above the tether’s own 250 N calibrated sensing range, so they are measurements outside calibration rather than inside it [3].

That instrument cannot see the worst case. A scan takes 35 ms and about 60 s to process, so the tether must be static or near-static, transient loads during a jerk are not captured, and the load at which the cable eventually broke was never recorded [3]. Its curvature channel is known to be miscalibrated: wound on a 4.2 cm hub it reported a curvature implying a 12.5 cm bend radius, so curvature from it is qualitative only [3].

The same work reports that the tether stiffens under high stress and stays semi-rigid until flexed again at low tension, attributed to woven layers tightening around the central copper helix. That behavior is undocumented elsewhere and unquantified [2].

Flight hardware has failed at the mechanism rather than the cable. On Apollo 15, both crew retractable tethers, spring-loaded reels used to secure hand tools, failed during surface operations: the Commander’s cord snarled and jammed when a slack release let the reel spring expand off its spool, then broke against the spring’s sharp edge on an attempt to re-extend it, and the Lunar Module Pilot’s tool clamp came off after its bowline and figure-eight knot untied and let the cord retract into the housing [11]. The mission report calls the jam-and-break mode repeatable and closes the anomaly with a knot change and a crew training note rather than a redesign [11]. It is a small mechanism next to the load-bearing tethers above, but it is the only case in this record of an as-flown tether failure, and both failure modes are about attachment and retraction hardware, not the cord’s breaking strength.

Power over a tether

The tether is also a conductor, and the voltage is set by conductor mass rather than by the load. Sending 1 kW over 1 km of 22 AWG at 5 percent transmission loss requires 1.5 kV at the sending end [4]. That is a design calculation, not a measurement, and it is what fixes the TYMPO operating voltage: raising voltage is how conductor mass comes down for a given loss. With an 80 percent end-to-end efficiency requirement, it imposes 92.5 percent minimum efficiency on the conversion chain [4].

A GaN multilevel DCX converter breadboard measured 99.2 percent peak and 96.2 percent full-load efficiency at a 150 kHz resonant frequency [4]. That was a scaled-down breadboard, not the flight design, and not at the full 1 kW and 1.5 kV rating. The 91 percent end-to-end system figure quoted alongside it is modeled, combining a 98 percent modeled converter pair with 5 percent tether loss, and has not been measured [4].

Tethered power and data delivery is not limited to surface rovers descending pits. jaynes2024design describes a ground testbed at NASA Marshall, an epoxy floor supporting air-bearing spacecraft simulators, built to develop a tethered “space tug” concept for in-space servicing and assembly tasks such as structural mating and laser welding. The tether there stands in for the umbilical a small servicing vehicle would use to stay powered and commanded while docked to a work platform, and the reported work is a design and test description of the simulator rather than a measured tether performance result [10].

Tethers off the surface

Everything above is a tether anchored to a vehicle that also has wheels, legs or a lander under it. Orbital tethers carry no such vehicle and answer a different problem: momentum and attitude control without propellant. carroll1985guidebook, a 1985 NASA guidebook that predates the surface and orbital work above by decades, gives the first-order sizing relations, a characteristic velocity and characteristic length derived from a material’s specific strength, tabulated for steel, titanium, Kevlar and graphite, plus a debris-impact scaling law calibrated against one hypervelocity test series at 6.5 km/s. The author is explicit that the tool is broad rather than deep: the debris-risk estimate for a kilometer-scale tether is built from inputs marked “ESTIMATED” on the source figure, and the impact-velocity calibration is below both the assumed micrometeoroid impact speed and the axial sound speed in Kevlar, so the scaling law is stated as an assumption that may not hold for braided fiber, not a validated result [7].

Tethers reappear in orbital debris removal for a different function again: as the connection between a servicing spacecraft and a captured, tumbling target rather than as a load path to a fixed anchor. wang2015coordinated models a tethered space manipulator holding a tumbling target after capture, using tether tension itself as a pitch and yaw control input alongside thrusters and an arm, with results from simulation rather than hardware. bigdeli2025mechanics, a broader review of debris-removal mechanics written from a contact-and-tribology perspective rather than a guidance-and-control one, places tethered capture alongside nets, harpoons and electrodynamic tethers used as a contactless deorbit method, and gives net-closure timing of roughly 26 to 30 seconds after contact from its own simulation library [8]. Neither source shares a material, a construction, or an anchor point with the planetary rover tethers above; what they share is that the same word covers a load-bearing cable, a control actuator and a deorbit device, and a design decision made for one function does not transfer to another.

What is not established

No test above combines abrasion, bend and cold or hot temperature in a single specimen, so the effect of a rock edge on a tether that is also cryogenic has not been measured, only bracketed separately at each extreme [1][5]. TYMPO’s liquid-nitrogen cold case and 200 C hot case are not surface temperatures actually measured in a permanently shadowed region, and no specimen was thermally cycled before the strength pull [1]. Vacuum abrasion testing used simulant and terrestrial basalt at a pressure many orders of magnitude above lunar surface vacuum, so it bounds nothing about vacuum cold welding or dust adhesion [1]. The Axel tether management module’s control figures were measured on a benchtop rig with no rover integration and against no stated mobility requirement, and its tension sensor disagrees with its own load cell across the measured frequency range without explanation [2]. The one field measurement of tether load under a real snag saturated the sensor that recorded it, cannot resolve transient jerk loads, and never captured the load at which any tether has actually broken in the field, apart from the mechanism failures on Apollo 15, which were never load related [3] [11]. TetherCAD’s strength model has been shown wrong by a factor of three in one direction while the vendor’s bend-radius rating is wrong by a factor of nine in the other, and no third, independent prediction has been checked against the same pulls [1]. The orbital tether sizing relations in carroll1985guidebook rest on 1978 commercial material data and a debris-impact model the author himself flags as built from order-of-magnitude, unvalidated inputs, and neither tethered-capture control scheme described here has been demonstrated on hardware [7][9].