You cut the RS485 control cable between your controller and the camera head, spliced in extra length, and the head stopped answering. In almost every case like this the distance is innocent and the joint is guilty: a reversed pair, a lost termination, or a star-shaped junction. RS485 was built for exactly this job — a differential pair that spans a kilometre and more — so a sound extension should work.
This guide follows the failure in the order you will meet it: what the multimeter must read, which wire goes where, where the two 120 Ω resistors sit, how to make a splice that still behaves like cable, and the arithmetic that says when a repeater is the honest answer.
Why did the line stop working right after the splice?
Read 60 Ω across A and B with the power off: that one number separates the three usual culprits — reversed polarity at the joint, a missing 120 Ω termination, and a stub-shaped junction box — from any genuine distance problem.
RS485 sends every byte as a voltage difference between two conductors, A and B, so the receiver ignores ground shifts and needs only 200 mV of difference to decode. That property is why the standard reaches 1,200 m at all, and it is why a wiring fault kills the line instantly instead of degrading it slowly.
Distance failures look different: bytes arrive corrupted at high baud rates long before the line goes silent. Total silence right after a splice points at continuity, polarity, or power, in that order of probability. Prove it with the meter readings further down before you spend money on anything.
The same differential logic behind pin-to-ground and pair-to-pair limits governs every signalling pair in the cabinet, and the reasoning carries over to differential line input voltage limits: pin to ground or pair to pair.
How far can RS485 actually reach at control-line baud rates?
About 1,200 m at 100,000 baud or below on 24 AWG twisted pair, which is the classic TIA-485 envelope — so a run you just extended is already inside the limit unless you pushed well past a kilometre or raised the baud rate at the same time.
The trade is length against speed. The same pair that spans 1,200 m at control-line rates collapses to roughly 10–15 m at the 10 Mbps top rate, because edge settle time, not voltage, becomes the binding constraint. The standard behind both numbers is demanding: a compliant driver must hold at least 1.5 V of differential output into two 120 Ω terminations in parallel with 32 unit loads of receivers.
Your controller-to-camera traffic is the easy case. Preset calls, trigger commands and recipe selections are a few bytes each; at 9,600 baud a ten-byte command crosses the wire in about ten milliseconds. If the line worked at this baud rate before the splice, distance is not what broke it.
Which wires in the cable carry the RS485 signal?
Two conductors of one twisted pair — B, the positive or D+ line, and A, the negative or D− line, on the most common convention — plus a third wire for signal ground; a cable without that ground reference is the next most common reason a long line goes quiet.
Buy cable with two twisted pairs and an overall shield. One pair carries A and B; join the two conductors of the spare pair together at both ends and use them as the signal ground. Three wires for what everyone calls two-wire RS485 — and the missing ground shows up as random corruption once the run passes a few hundred metres.
Identify before you energise: mark A, B and ground at both ends, in ink or labels, the moment you strip the jacket. Vendor labelling is not consistent — D+ on one device is D− on another — so trust your meter and your marks, not the silkscreen.
On the controller side, this frame grabber I/O cable breaks a DB9F port out to open leads you can land directly on the pair, and the accessories finder filters the rest of the power and I/O cables by connector and length.
- Type
- Frame grabber I/O cables
- Length
- 6-8 m
Where do the two 120 Ω terminations belong?
One 120 Ω resistor across A and B at each end of the line, and nowhere else — the driver is rated to push 1.5 V through exactly that load, so a third resistor mid-line drags the level down and a missing one at the far end lets reflections corrupt bytes.
Two 120 Ω resistors in parallel measure 60 Ω, which is the number your meter should find across the pair with the power off. Many cameras, controller boards and light controllers ship with a termination jumper or DIP switch, so after any change to the line's shape, walk the devices and audit every one of them. A mid-line device left with its terminator switched on is a load the driver never budgeted for.
Termination belongs on the two devices at the extremes of the run. If you later split the line with a repeater, each segment becomes its own line with its own two terminations.
Does a splice in the middle of the run hurt anything?
One tidy splice costs nothing measurable at control-line baud rates if you keep the untwisted section under about 25 mm — the damage comes from topology: star branches, metre-long pigtails in a junction box, and spare conductors pressed into service as stubs.
RS485 is a bus: every device taps the same pair, in and out, device to device — a daisy chain. When you extended the line you may have created a T, where the old cable and the new cable both hang off one joint. That T is a stub, and a stub reflects the signal back into the bits behind it. Keep any unavoidable stub under about 0.5 m at 9,600 baud, and avoid stubs entirely at 115,200.
Making the joint: solder or crimp each conductor, stagger the two joints so they never sit side by side, cover each with its own heat-shrink, and restore the twist to within 25 mm of the joint. A joint made this way is electrically part of the cable at these baud rates. The routing that keeps the rest of the line alive is the same discipline that prevents robot camera cable failures: routing and fatigue checks — strain relief, bend radius, and no rubbing on structural edges.
Where does the shield get grounded?
At one end only — the controller end is the usual choice — and never as the signal return: grounding both ends turns any difference in earth potential between the cabinets into a current in the shield, which couples hum straight into the pair.
The shield and the signal ground do different jobs. The shield is a screen: foil or braid, landed on chassis at one end, its drain wire folded back and insulated everywhere else. The signal ground is the third conductor from the wiring section above, and it is what keeps both ends inside the standard's −7 V to +12 V common-mode window.
Keep the run away from motor and drive cables. Where a crossing is unavoidable, cross at 90 degrees; a long parallel share of a tray with a drive output injects noise that no termination cures — the same layout rule that protects you when extending a camera GPIO trigger cable without losing sync.
What should the multimeter read with the power off?
Under 2 Ω end to end on each of the three conductors, open from every conductor to the shield, and 60 Ω across A and B — three readings in five minutes, and the wiring is either cleared or condemned before you touch a single setting.
| Meter reading across A and B | What it means | What you do |
|---|---|---|
| About 60 Ω | both 120 Ω terminations present | wiring is sound; move to the polarity swap test |
| About 120 Ω | one termination missing | fit a 120 Ω at the open extreme end |
| Hundreds of Ω or open | no terminations; receivers alone load the pair (about 375 Ω with a full 32 standard receivers) | add both 120 Ω terminations |
| 0 Ω | A and B shorted together at the joint | remake the splice |
| Open on one conductor | broken wire, or the wrong conductor picked up | re-identify and re-mark the pair |
Then the polarity test. If all three checks pass and the head still stays silent, power down, swap A and B at one end only, and retry. Vendor silkscreens disagree about which line is D+, and this single swap resolves roughly half of all dead extended lines. If the swap changes nothing, confirm both ends share baud rate, parity, stop bits and device address before suspecting the copper again.
How do you budget a 1,400 m run at 9,600 or 115,200 baud?
At 9,600 baud the 1,400 m pair works — about 1.0 V arrives across the terminations and the edges settle inside an eighth of each bit — while at 115,200 baud the same wire fails, because its edges need 12.6 µs to settle into 8.7 µs bits.
The assumptions are all ordinary: 24 AWG copper at about 0.084 Ω per metre per conductor, pair capacitance of about 50 pF per metre typical of Cat5e-class cable, a 5 V driver, and both terminations in place giving the 60 Ω load. The DC arithmetic says the receiver sees 1.0 V, five times its 0.2 V threshold. The timing arithmetic says a 9,600-baud bit lasts 104 µs and the edge is settled within 12.6 µs, about 12 percent of the bit. Comfortable on both counts.
At 115,200 baud the bit lasts 8.7 µs and the edge is still settling when the next bit arrives, which is dead. Hold yourself to a quarter-bit settling allowance and the same arithmetic gives (0.25 × 8.7 µs) ÷ (3 × 60 Ω × 50 pF/m) ≈ 240 m of reach at 115,200 baud. This is the standard RC estimate rather than a cable datasheet curve, but it lands where field practice does: long RS485 runs live at low baud.
Splice, new cable, or repeater — which one do you choose?
Drop to 9,600 baud and keep your spliced but sound line when the traffic is commands; add one isolated repeater when you must stretch past about 1,200 m; buy one continuous cable when the original conductors are damaged the way a cut and re-jointed line usually leaves them.
Commands only, up to about 1,200 m: keep the direct pair, fit both 120 Ω terminations, run 9,600 baud, and stop. A preset or trigger command is milliseconds of traffic.
Longer than 1,200 m at low baud: one isolated repeater per additional 1,200 m, with automatic direction control so it needs no handshaking wire. The repeater also breaks the galvanic connection between segments, which stops a ground difference at the far building from eating the common-mode window, and each segment gets its own two terminations.
115,200 baud over a kilometre: do not chase it with copper. Quarter-bit settling wants segments under about 240 m, so a repeater chain costs more than the speed returns. Drop the baud, or convert to fiber or Ethernet for the long haul and keep RS485 at the two ends.
More than 32 devices: a repeater restarts the unit-load count on its far segment, or fit transceivers that present one-eighth of a unit load and take the bus to 256 receivers.
Damaged cable, as after a bad cut: replace the whole run with one continuous length. Every splice is a future failure point on a moving or vibrating machine, and a drag-chain-rated cable survives flexing that a joint never will.
- Type
- Camera data cables
- Length
- 15 m
- Type
- Power and I/O cables
- Length
- 6-8 m
Serial control is everyday business here: the LS-CTL-0487 is a 2-channel 24 V digital LED light controller with an RS-232 port on board, and getting the first reply out of a port like that is its own checklist in RS232 serial control of a vision controller: get the first reply. When the controller must sit far from the head, wiring an RS232 device to RS485 with a converter works the same for a control box as for a scanner, and the light controllers range carries the dimming and strobing end of the job.
Bring-up sequence for the re-extended line:
- Isolate the cabinet; confirm under 2 Ω end to end on A, B and ground.
- Confirm 60 Ω across A and B — both terminations in, none mid-line.
- Confirm the shield lands on chassis at the controller end only.
- Set both ends to 9,600 baud, 8 data bits, no parity, one stop bit, matching addresses.
- Power up and send one command; if silent, power down, swap A and B at one end, retry.
- Only after the head answers, step the baud up within the 240 m quarter-bit rule for 115,200.
- Label A, B and ground at both ends before you close anything.
If the numbers leave you unsure which cable or converter the line needs, contact us with the baud rate, the one-way length and the connector at each end. The neighbouring wiring guides sit on the systems topic page, and you can start the parts list at the cable step of the configurator.
The configurator matches a camera, lens, light, light controller and cables to each other, step by step, and lists only parts that fit what you already picked.
Questions
Can I run the RS485 cable in the same tray as mains and motor cables?
Keep them apart. A control pair strapped along a mains or drive cable picks up coupled noise that eats into the 200 mV receiver margin. As a working minimum leave 300 mm of separation, cross power runs at 90 degrees, and let the shield, grounded at the controller end only, do the rest. Where a tray forces a long parallel share, move the pair into conduit of its own on the far side.
Can I use Cat5e Ethernet cable for an RS485 run?
Yes, and it is a good fit. One twisted pair carries A and B, a second pair has its two conductors joined at both ends to serve as the signal ground, and the overall shield lands on the controller chassis. Cat5e is 24 AWG at roughly 50 pF per metre, which is the cable the 1,200 m low-baud figure assumes. Keep the pair twisted to within 25 mm of every joint.
How many devices can share one RS485 line?
Thirty-two unit loads on one terminated pair, per TIA-485. A unit load is the standard's 12 kΩ receiver input, and most modern transceivers present one-eighth of it, lifting the practical ceiling to 256 receivers. Count every repeater as a load on each segment it touches, because the driver must still hold 1.5 V differential across the whole set.
Do I need 120 Ω terminations on a short line?
Electively, below about 170 m at 115,200 baud and about 2 km at 9,600 baud: at twisted pair's typical 5 ns per metre, reflections arriving within a tenth of a bit time die before the receiver samples. Fit the terminations anyway, because they cost nothing, they make the 60 Ω power-off check meaningful, and the driver is rated for exactly two of them.
Should the shield double as the signal ground?
Never. The shield is a screen grounded at one end; the signal ground is a dedicated third conductor carrying the reference between the two ends. Use the shield as the return and ground it at both ends, and any earth-potential difference between the cabinets drives loop current through it, coupling mains hum directly into the pair. Keep the two jobs on separate wires.
What baud rate should both ends use after a long extension?
Set controller and head both to 9,600 baud. On a 1,400 m pair that leaves five times the receiver's 200 mV threshold in hand, while 115,200 baud fails the settling arithmetic on the same copper. Confirm with one command, then step the rate up only if speed earns it: preset calls and trigger strings are a few bytes, milliseconds each at 9,600 baud.