Tethers

A tether on a planetary robot carries three loads at once: the vehicle’s weight during a rappel, electrical power, and data. 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 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].

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.

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].

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].