System designHow-to

Wiring two cameras to one frame sync signal

Frame sync signal wiring for two cameras: one pulse output to every sync input, 5-24 V levels, drive-current math, pin ring-out and the timestamp check.

LumaSpec Vision11 min read Share on LinkedIn

On this page 10 sections
  1. 01What does a frame sync signal actually do?
  2. 02Which camera should be the master?
  3. 03What parts do you need to wire two cameras to one pulse?
  4. 04Which pins on the camera connector carry the sync signal?
  5. 05What voltage does a camera sync input need?
  6. 06How much current can one sync output spare per camera?
  7. 07How do you wire and verify the sync, step by step?
  8. 08How long can the sync cable run be?
  9. 09When does frame sync fail to align two cameras?
  10. 10Questions

Two free-running cameras drift apart within a frame or two, so a stereo pair or a dual-angle inspection station shows the part in a different place in each image. The fix is electrical, not software: generate one short pulse per frame, wire it to every camera's sync input, and let each sensor latch exposure on the same edge. You get the wiring order, the pin checks, the 5-24 V level and drive-current arithmetic, a cable-length calculation for your own run, and the timestamp test that proves the exposures start within microseconds of each other. It is written for engineers wiring GigE, USB3 or CoaXPress area scan cameras with opto-isolated I/O connectors, and the same wiring drives 3D sensors with trigger inputs.

What does a frame sync signal actually do?

A frame sync signal is one short electrical pulse, 10-100 µs wide, sent once per frame period — 50 ms at 20 fps — and every camera wired to it starts exposure on the pulse's rising edge, so both sensors sample the scene at the same instant.

One device owns the timing; everyone else listens. Slaves abandon their own pace and follow the pulse, so a 20 fps master holds every camera it drives at 20 fps. The pulse carries no image data and needs no protocol — it only marks an instant, which is why the wiring reduces to two cores: one conductor for the pulse, one for the return. Duty cycle is tiny — 10 µs on a 50 ms period is 0.02% — so the line sits idle almost all the time and the pair can share a small I/O cable with power cores without trouble. With global-shutter sensors the whole frame samples on that edge; the rolling-shutter catch has its own section near the end.

Which camera should be the master?

The master is whichever device's timing must lead the process — the camera that watches the trigger event, a photoelectric sensor with a 100 mA-class PNP output, or a PLC output — and every other camera is a slave that only listens.

When both cameras merely need to frame simultaneously with no external event, pick either camera as master and set its I/O line source to exposure-active so it emits one pulse per frame. The slave's pace then equals the master's and cannot drift. A frame grabber with two camera heads can instead fire both from one board-level output; that route is laid out in Trigger two cameras simultaneously from one frame grabber input. Whatever the source, keep the chain flat: one pulse source, every camera one hop away, no camera triggering the next.

What parts do you need to wire two cameras to one pulse?

You need three things before the first connection: a cable that brings each camera's I/O cores out to bare leads, a pulse source rated for the load, and a multimeter to ring out pins.

  • One power-and-I/O breakout cable per camera; the sync pair lands on its open ends.
  • A terminal block or soldered splices for the pulse core and the return.
  • A 24 V DC supply if the pulse comes from a sensor or PLC output; nothing extra if a camera GPIO drives a single TTL-capable slave.
  • A multimeter for the continuity check; a two-channel scope helps but stays optional.

Most cameras expose trigger and strobe lines on a round power-and-I/O connector, so the breakout cable is the part that does the job. The LS-CBL-0300 is an M8 8-pin power and I/O cable, male to open end, with eight cores over 7 m and a drag-chain-rated jacket — room for the sync pair plus power. For a camera with a 12-pin connector, the LS-CBL-0211 brings the whole pin set out over 5 m.

CablesCamera Power and I/O Cable, M8 8-pin connector, Male to Open end (8-core), 7 m, drag-chain use (LS-CBL-0300)LS-CBL-0300
Type
Power and I/O cables
Length
6-8 m
$79.00View partBuild a system with it
Cables5 m 12-Pin IO I/O and power cable (LS-CBL-0211)LS-CBL-0211$56.00View partBuild a system with it

The sync pair rides the I/O cable only; image data keeps to its own GigE or USB3 cable, which for long drag-chain runs should be a high-flex type from the camera cables range.

Which pins on the camera connector carry the sync signal?

The pulse rides on the two cores the camera's own I/O table calls the input line and its return — the IN+ and COM cores of an opto-isolated input — and the pin numbers come from that table, never from a similar-looking drawing of a different plug.

Round connectors are where sync wiring goes wrong. Male and female pin diagrams are drawn from opposite faces, so the same physical position carries different numbers at the two ends of one cable; a core that reads pin 3 at the male moulding can land in pin 7 territory at the female panel connector. Cable makers also change wire colours between batches. Guessing is the expensive step.

Ring every core out instead. Put one multimeter probe on the numbered pin at the camera end, sweep the bare leads at the other end, and write down what conducts. Compare the result with the camera's I/O table, then label both cable ends. The same discipline applies to I/O breakout boards, as Camera IO board pin alignment: verify the pinout before you mount shows, and to assembling the open-ended plug itself, which Open end trigger cable wiring for optocoupler trigger and strobe walks core by core.

What voltage does a camera sync input need?

An opto-isolated camera input switches reliably on a 5-24 V pulse once roughly 7-10 mA flows through its internal LED, so drive it at 12-24 V when you have the choice, and never put more than 5 V on a line the I/O table marks as TTL or non-isolated.

Two input families exist and they fail in opposite directions. The opto-isolated input is an LED behind series resistance: it needs voltage and a few milliamps. With V_drive as the pulse voltage, V_LED the opto LED forward drop near 1.2 V, and R_series the input's series resistance, the switch current is:

I = (V_drive - V_LED) / R_series

At 24 V into a typical 2200 Ω input, I = (24 - 1.2) / 2200 Ω ≈ 10.4 mA — a solid switch. Feed the same input 5 V from a GPIO and I = (5 - 1.2) / 2200 Ω ≈ 1.7 mA, below the ~7 mA most opto inputs ask for, which is exactly the intermittent no-trigger fault. So 24 V from a sensor or PLC output is the dependable default; use a camera's 5 V GPIO output only where the receiving input is listed as TTL-capable. The TTL case, including pull-ups, is wired out line by line in Wire a GPIO output to a camera trigger input: TTL levels and pull-ups.

How much current can one sync output spare per camera?

Count every camera input as a 10 mA load: a 100 mA-class PNP sensor output drives six of them with headroom, while a camera GPIO pin should feed one slave directly and hand two or more slaves to a buffer or a sensor output.

The loads add up, one input chain per camera. Two slaves at 10.4 mA draw 20.8 mA; six draw 62.4 mA, about 62% of a 100 mA output rating, which leaves margin for cold corners and cable losses. A camera GPIO is different: its pin rating sits in the low tens of milliamps on most models, and one opto input consumes that, so the rule is one GPIO to one slave. For a third camera, move the pulse to a 100 mA-class PNP output or insert a transistor or line-driver buffer stage. The pin's exact rating is in the camera's I/O table; if the planned load exceeds it, buffer.

How do you wire and verify the sync, step by step?

Wire the master's output core to every slave's input core, join all returns at the source ground, then set the trigger features on each slave and prove alignment with frame timestamps — the whole job fits inside an hour once the pins are known.

  1. Power everything down. Ring out the breakout cables against the camera's I/O table and label both ends.
  2. Land the master's output core and each slave's input core on adjacent terminals of the terminal block.
  3. Land the returns: every slave input's COM core joins the source's ground or COM, one common return through the cable.
  4. Match voltage families: a 24 V source to opto-isolated inputs; a 5 V GPIO only to an input listed as TTL-capable.
  5. On each slave, set the wired line to Input mode, then enable the trigger with the GenICam features: TriggerMode to On, TriggerSource to the wired line, TriggerActivation to RisingEdge. Keep exposure under programmed control (edge mode) unless you want the pulse width itself to set exposure.
  6. On the master, set the output line's source to exposure-active so it emits one pulse per frame period, and set the frame rate every camera must share.
  7. Power up and measure: a scope shows one pulse per period — 50 ms at 20 fps — and a multimeter on frequency range reads about 20 Hz on the same line.
  8. Capture from both cameras with timestamp chunk data enabled and compare exposure-start times. Expect agreement in the tens of microseconds; anything near a frame period means a slave is still free-running. Chunk data timestamp: camera time at exposure end and how to latch it covers the latching detail, and Verify hardware sync between two cameras actually works gives the bench test that separates a wired pair from a coincidentally aligned one.

Three faults cause most failed first attempts: returns left floating so no current ever flows, a pulse rate above the slave's maximum frame rate so the slave drops frames, and 5 V delivered to an input that wanted 24 V. The same timestamp check applies to colour rigs; Multi-camera RGB synchronization: verify exposure timing lists what to compare when exposures look aligned but brightness pulses differ.

How long can the sync cable run be?

Size it from cable capacitance, and at frame-sync pulse rates the answer is generous: a 15 m twisted-pair run driven at 100 Ω shows a rise time near 165 ns, under 2% of a 10 µs sync pulse, so runs of tens of metres work as drawn.

C = cable length x capacitance per metre rise = 2.2 x R_drive x C

Worked example, 15 m of shielded twisted pair at a typical 50 pF/m: C = 15 m x 50 pF/m = 750 pF. With a 100 Ω drive resistance, rise ≈ 2.2 x 100 Ω x 750 pF = 165 ns. The skew between a 3 m slave and a 15 m slave on the same pulse is ΔC = 12 m x 50 pF/m = 600 pF, so Δrise ≈ 2.2 x 100 Ω x 600 pF ≈ 132 ns — about 0.0003% of a 50 ms frame period and invisible next to sensor timing. Keep the pair twisted, cross motor and drive cables at right angles, and never run it parallel alongside a variable-frequency drive output: induced edges at these rise times can fake a trigger.

When does frame sync fail to align two cameras?

Frame sync aligns exposure start and nothing else, so it cannot rescue a rolling-shutter pair on fast-moving parts, cannot make two different sensor models latency-match by itself, and cannot fix a pulse that never leaves software.

Rolling shutter reads the frame line by line, and the scan takes 10-30 ms on a 12 MP-class sensor, so a part moving 300 mm/s shifts 3-9 mm between first and last line even with a perfect pulse. Global-shutter cameras remove the skew entirely; for a synced pair on a moving line, shutter type matters more than sync wiring. Latency between different camera models adds a few microseconds of offset between trigger edge and exposure start — keep the pulse shared so the residual stays in the low microseconds, and measure it once on the bench. And the pulse must be real: two cameras started from the same software loop carry millisecond scatter, which the timestamp check in the step section exposes immediately. If the station mixes acquisition with robot control, the systems topic page collects the integration guides for both sides.

To pick the cameras and cables for a sync pair, the camera finder filters by interface, shutter type and resolution, the camera configurator pairs a camera with its I/O cables and power in one pass, and the applications team answers wiring questions through the contact page.

Questions

Can a GigE camera and a USB3 camera share the same sync pulse?

Yes. The pulse is purely electrical and the data interface never touches exposure timing. Wire both cameras' opto-isolated inputs to the same output core and common return, set the same trigger features on each, and both sensors latch on the same edge. Only the image transfer path, and its cable limits, differ.

Why do synced cameras still return images with different brightness?

Sync aligns exposure start, not exposure length or gain. If one camera still runs auto exposure, or the two ExposureTime values differ, brightness differs frame by frame. Set identical ExposureTime and gain on both cameras, or switch the inputs to width mode so one pulse's duration sets the exposure everywhere.

Can I daisy-chain the pulse from camera to camera instead of a star split?

Some cameras can route an input line back out through an output line in their GPIO matrix, and that works, but every extra optocoupler hop adds a few microseconds and one more wiring fault to find. A star split keeps every slave one hop from the source and equally delayed.

Can the pulse come from a photoelectric sensor or PLC instead of a camera?

Yes, and for event-driven stations it is the better master. A PNP output rated around 100 mA at 24 V drives several opto inputs directly, and the pulse then marks the part's actual arrival rather than another camera's frame decision. The trade is one more device in the timing chain.

What frame rate do the cameras run once they are synced?

The slaves run at the pulse rate, whatever their own frame-rate settings say, so set the pulse rate at or below the slowest camera's maximum or the slow camera drops frames. A 20 fps pulse holds a 20 fps master and its slaves there; nothing drifts because nothing free-runs.