Machine Builders Buying Servo Cable From China: Where the Risk Actually Is

A machine builder sourcing servo cable from China usually starts the exercise in the wrong place. The concern goes to the assembly: are the crimps sound, is the overmould neat, will the connector mate properly with the drive. Those are answerable questions, and a competent harness shop answers them well. Meanwhile the thing that will actually take a machine down at a customer site in eighteen months is a decision nobody discussed — which spool the shop pulled the cable off.

Here is the uncomfortable version. Every bench test in a harness shop's outgoing QC — continuity, insulation resistance, withstand voltage, terminal pull force, cross-section analysis — will pass a cable that is going to die in a moving drag chain at four hundred thousand cycles. There is no correlation between those tests and flex life. A shop can build a genuinely excellent assembly on the wrong cable, ship it with a clean test report, and you will find out during your customer's second production year.

So the risk in this category is not workmanship. It is material selection, and it hides behind a single ambiguous word.

"Flexible" and "continuous-flex" are not the same word

Ask a cable catalogue for a flexible cable and you will get one. Flexibility in that sense is a conductor property: the international conductor standard IEC 60228 defines stranding classes, and Class 5 designates a flexible conductor — many fine strands rather than a few thick ones — while Class 6 designates a finer, more flexible one still. A Class 5 cable bends easily around a corner during installation. That is what the classification means. It says nothing about whether the cable survives being bent back and forth ten million times.

A cable designed for continuous motion is a different object underneath a jacket that looks identical. The construction differences that matter are all invisible from the outside:

• Bunched fine strands, then short-lay bundle stranding. The cores are laid up in concentric layers around a central element with a deliberately short lay length, so that in a bend the cores move along a helix instead of stretching on the outside of the curve and compressing on the inside. Long-lay or randomly laid construction lets cores migrate; migration is what breaks conductors.
• A separator or fleece wrap between layers, so the cores cannot rub against one another or against the jacket during motion.
• A pressure-extruded jacket, not a tube-extruded one. A pressure extrusion fills the interstices and grips the bundle so it moves as one body; a tube extrusion leaves the jacket loose over the cores, which is fine for a static installation and fatal in a chain, where the cores work their way toward the neutral axis and the cable corkscrews.
• A braid shield with a specified optical coverage and braid angle, chosen so the braid can articulate through millions of cycles without work-hardening and opening up. Foil shields alone crack under motion; that is why a moving feedback cable uses braid, or braid over foil, and not foil over a drain wire.

None of that is visible at incoming inspection. You cannot tell a drag-chain cable from a merely flexible one by squeezing it, and the price difference is large enough that the substitution is always tempting to somebody in the chain.

What a real flex specification contains

The correct response to a supplier saying "yes, it is flexible" is to ask for the number. Then ask for the four conditions attached to it, because a bend-cycle count without them is not a specification.

Two consequences follow that catch people out.

The first is that a torsional-flex rating is not a substitute for a chain rating. A cable that survives a robot's wrist twisting is qualified against a different load case than one running back and forth on a linear axis. Suppliers quote whichever number they have.

The second is that the bend radius is doing more work in that table than the cycle count. Halving the radius does not halve the life. Machine builders who install a cable at 6×d because that is what fitted in the chain, when the data sheet's cycle count was measured at 12.5×d, have quietly discarded the specification they paid for.

Worth knowing as well: an installed chain should not be packed. Common practice keeps cables to roughly 60 per cent of the chain's usable cross-section, separated by dividers so power and feedback cables cannot lie on top of each other and abrade, with each cable strain-relieved at both the fixed and the moving end so no axial load transfers into the chain run. A perfect cable installed in a full, undivided chain fails on schedule anyway — and the failure will be blamed on the cable, and then on the country it came from.

Where the risk actually sits: the bulk cable, not the crimp

Put the two previous sections together and the sourcing conclusion writes itself.

A harness shop is being bought for cutting, stripping, crimping, shield termination, overmoulding, labelling and testing. Those are real capabilities and they are the shop's contribution. Cable qualification is not, and it is unreasonable to expect a small assembly house to run a chain-life test programme. What the shop does is buy bulk cable on its domestic market against whatever description was written on your purchase order. If that description said "flexible shielded 4×1.5 + 2×0.75", the shop will supply exactly that, at a good price, and be entirely within its brief.

So the control point moves upstream, and you have three workable options:

1. Nominate the cable. Name the bulk cable by manufacturer and type designation on your drawing, require the cable manufacturer's data sheet with the first article, and require per-lot traceability — a lot number on the assembly's label or its certificate of conformity that maps back to a spool. This is the normal arrangement and the one most shops can meet.
2. Free-issue the cable. Buy the bulk cable yourself from a European or Japanese supplier, ship it to the assembly shop, and buy assembly only. Costs you inventory, freight and customs handling; removes the substitution risk completely. Sensible for the moving axes only, not for the whole machine.
3. Accept the shop's cable and qualify it yourself. Buy sample assemblies, run them in a chain at your own axis speed and radius until they fail, and decide with data. Slow, but it is the only route that produces a number you own — and if you are moving a high-volume machine platform, it is worth the fixture.

A fourth answer is also legitimate: for the high-cycle moving axes on a flagship machine, keep buying finished cables from an established European or Japanese cable brand, and split the bill of materials deliberately. The economic case for the Chinese assembly sits in the static harnesses, the panel wiring, the fixed motor leads and the lower-duty axes — which on most machines is the majority of the copper by value anyway.

The mistake is doing none of the four and treating a low quotation as a like-for-like price. When quotations for the same drawing come back with a wide spread, the spread is usually construction, not margin — which is the general pattern the CMH piece on why China quotes for the same spec differ walks through in detail.

Feedback cables fail first, and it looks like a drive fault

There is a diagnostic pattern worth recognising, because it wastes weeks of service time.

In a mixed chain, the encoder or feedback cable degrades before the motor power cable. The conductors are finer, the shield is doing electrically critical work, and the whole assembly is less forgiving of construction faults. As the braid's fine wires begin to break, optical coverage drops. Shielding effectiveness falls before continuity does — which means the cable still passes every electrical test you can run on it while the machine starts throwing intermittent following-error or position-feedback faults under load, more often when the axis is at one end of its travel, more often when the spindle drive next to it is switching.

The service engineer replaces the drive. Then the motor. Then, eventually, the cable, and the fault clears.

Three construction decisions determine whether you meet that failure at 400,000 cycles or 10 million:

• Shield termination geometry. The construction and shielding of a servo encoder feedback cable is a subject in itself, and the termination is where most of the performance is won or lost. A braid terminated 360 degrees into a conductive backshell or an EMC gland behaves completely differently from a braid twisted into a pigtail and run to a terminal. The pigtail is an inductor; at the frequencies that matter it undoes the shield. If a supplier's standard practice is pigtailing, that is a specification decision you need to override in writing.
• Which end the shield is bonded at. Follow the drive manufacturer's manual rather than a general rule, state it on the drawing, and make sure both the assembly shop and your own panel builder are working from the same instruction.
• Separation. Motor power and feedback in the same chain, undivided, is a standing invitation to conducted and coupled noise, and to mechanical abrasion between the two. Divide the chain.

Qualifying a Chinese servo cable assembly without betting a machine on it

A staged path, roughly in the order a machine builder can actually execute it:

1. Start with the static parts. Panel-internal harnesses, fixed-installation motor power cables, non-moving encoder extensions and cabinet control looms carry far lower risk than chain cables. Move these first; they are ordinary build-to-drawing harness work and a competent shop will do them well.
2. Issue a complete drawing. Wiring diagram, from-to list, wire list with cross-section, insulation material, colour and marking, connector specification per end, shield termination method and bonding end, overall length with an explicit tolerance convention, and the conductor-identification rules you work to. For European machine electrical equipment, EN 60204-1 reserves green-and-yellow for the protective conductor and light blue for the neutral, and expects consistent identification for control conductors — a shop working from a US-derived convention will not guess this, and the wire-numbering and ferrule conventions that go with it need stating too.
3. Specify the end-of-line test. Continuity plus insulation resistance and a withstand-voltage test at a stated voltage and duration on 100 per cent of assemblies, with results retained by lot.
4. Approve a first article properly. Measure the length, section a crimp, examine the shield termination, and mate it with the actual drive and motor on your bench — not with a mating half the supplier chose.
5. Then, and only then, move a moving axis. With cable nominated or free-issued, a chain endurance sample running on your own rig, and a limited first production quantity while the sample accumulates cycles.
6. Fold cable and harness items into the acceptance test. If you run a formal factory acceptance test on machines — the CMH FAT guide for machinery from China covers the structure — put an axis-cycling block in it, because that is the only stage where a flex problem can surface before shipment.

Common questions

Can a small Shenzhen harness shop build feedback cables at all?

Building them, yes — servo power and feedback cables, encoder lines and network-style signal cables are ordinary work for a shop with controlled crimping and a test bench. The distinction to hold onto is between building and qualifying. Ask what the shop builds today, ask to see a sectioned crimp on a fine conductor, and ask specifically how it terminates shields. Then hold the cable-construction decision yourself.

What order quantity makes this worth doing?

Harness shops typically quote four-figure per-item minimums, which is awkward for a machine builder buying eight cables per machine. Huisheng Electronics, for example, states a 1,000-piece minimum with samples dispatched in 7–10 days and 15–25 days to warehouse on volume — supplier-stated figures to confirm against your own drawing. The workable structure is a blanket order against an annual quantity with scheduled call-offs, and standardising lengths across machine variants so a small number of part numbers carries the volume.

How do I stop a cable substitution after the first order?

Write it into the purchase order rather than the drawing alone: no change of bulk cable manufacturer, type designation, conductor class, shield construction or jacket material without written approval, and a requirement to state the cable lot on each shipment's certificate of conformity. Then check it at least once, unannounced, against a section of the delivered cable. A condition nobody audits becomes optional.

What to ask the supplier next

The six questions that sort suppliers in this category:

• Do you stock or source continuous-flex, drag-chain-rated cable? From which cable manufacturer, and what type designation?
• What bend-cycle figure does that cable's data sheet state, at what bend radius, travel speed and temperature — and can you send me the data sheet?
• How do you terminate braid shields as standard: 360-degree through a backshell or gland, or pigtail?
• Can you give me lot traceability from each finished assembly back to a bulk-cable spool?
• What is your end-of-line electrical test on a feedback cable, at what voltage, on what percentage of units?
• Which items on my drawing would you propose to substitute, and what would change?

Huisheng Electronics in Longhua, Shenzhen is a plausible shape of supplier for the static and panel side of this work: servo power and feedback cables, encoder lines and network-style signal cables sit among its six listed lines alongside multi-branch machinery and control-cabinet looms built to the buyer's drawing, with automatic cut-strip-crimp machines and servo crimping presses on its published equipment roster and cross-section, tensile, withstand-voltage, insulation and continuity checks on its test bench. Its catalogue points at fans, power supplies and motor drives, and it lists buyers in Europe among its export destinations. Whether drag-chain-rated cable and flex-cycle data are available is a question to put in writing before anything moves in a chain — and stated capacity from any small Shenzhen shop is worth confirming in a live walkthrough before you commit a build schedule to it. A well-structured specification document is what makes both of those conversations short.