A two-camera calibration that times out on the first synchronized capture is rarely a broken camera. The error means the capture group waited for a timestamp match that one device never delivered: somewhere in the chain, an armed camera saw no sync pulse. This guide walks the fault in the order it actually happens — the cameras you powered, the wiring between sync-out and sync-in, the master and subordinate modes — and puts firmware last, where it belongs. Work through it with a spare sync lead on the bench and you will have captures passing within the hour.
What the timeout actually means
Your program opens each camera, assigns master and subordinate roles, arms synchronized capture, then waits for a capture from every device whose timestamps agree within the SDK's tolerance. A subordinate that receives no pulse on its sync-in input never produces a capture, the wait window expires, and the SDK raises the timeout.
That leaves only three ways the pulse goes missing:
- A camera is powered and enumerated but is not wired into the chain.
- The cable, a connector or the sync port does not carry the signal.
- The modes are wrong, so no device drives sync-out.
Firmware, SDK builds and host USB controllers fail far less often and are worth touching only after these three are cleared.
Fix the count first: power only the cameras in your chain
The most common cause is a camera you forgot was on the desk. A typical bench failure: three depth cameras powered and cabled, a two-camera capture that times out on every run. Disconnecting the third camera cleared it on the next run, with no firmware change and no SDK reinstall — the extra camera was enumerated but undriven, and the group waited forever for its capture.
So before anything else, power down every camera that is not in the capture plan. Check the device-count argument in your capture call and make it equal the number of cameras you actually powered, then run again. If the capture passes, add cameras back one at a time, wiring and mode-setting each one before it powers up.
A camera you intend to use later counts too: an idle-but-enumerated device stalls the whole group. Unplugged means unplugged. When captures fail only in a multi-sensor set-up, the isolation method in camera open fails only with multiple sensors narrows the fault the same way.
Wire the chain: sync-out to sync-in, one hop at a time
The chain is a daisy. The master's sync-out feeds the first subordinate's sync-in, and that subordinate's sync-out feeds the next camera's sync-in. Depth cameras usually take a shielded 3.5 mm stereo-style lead for this, and a half-seated plug is the classic silent failure — it looks mated from above and carries nothing.
Check three things on every hop:
- Orientation: every lead runs out into in. Two sync-outs joined together, or a lead looped from a camera back into itself, time out exactly like a missing cable.
- Seating: push each plug home and give it a light tug.
- Cable type: shielded, and the type the camera documentation names for sync.
The connection conventions between two cameras are set out in wiring two cameras to one frame sync signal, and the camera cables range together with the accessories finder lets you filter leads by length and use.
Set the modes: master drives, subordinates follow
Three modes matter, and one is a trap:
- Standalone (the default): the camera captures on its own clock and drives nothing. A master left in standalone mode arms happily and pulses no sync-out, so every subordinate waits forever.
- Master: the camera captures on its own clock and pulses sync-out for the chain.
- Subordinate: the camera waits for a pulse on sync-in and captures only when one arrives.
Only subordinates take a delay value. The delay shifts each subordinate's exposure start by its own trigger-to-exposure latency, so its depth and color windows overlap the master's; the first-frame behavior of subordinate mode is covered in subordinate sync mode: first frame timing and rolling shutter.
Chains that mix device families, or run one sensor from a shared pulse, follow the same discipline; sync a depth camera and a third-party camera on one external pulse walks that case.
Isolate the failing camera one pair at a time
Work in pairs and the fault can hide nowhere:
- Run two cameras only: the intended master and the first subordinate, one direct lead, nothing else powered.
- Confirm the subordinate delivers captures. If yes, the chain is proven to that point.
- Power the next camera, lead it from the previous camera's sync-out, set it subordinate, run again.
- When a camera fails at step 3, swap the hop lead first, then reseat both plugs, then re-read its mode.
- A camera that captures as a master but not as a subordinate has a sync-in problem: lead, seating, port or mode, in that order.
| Symptom | Most likely cause | Test | Fix |
|---|---|---|---|
| Timeout with every camera powered | Extra device outside the capture plan | Power down unchained cameras, rerun | Leave it unplugged or wire it in properly |
| One subordinate never captures | No pulse on its sync-in | Swap the hop lead | Reseat or replace the lead, check out-to-in orientation |
| No camera captures anything | Master left in standalone | Read the mode each device reports | Set master mode on the first camera |
| Captures arrive but timestamps disagree | Subordinate delay unset | Compare the frame timestamps | Set each subordinate's delay from its data-sheet latency |
A timeout that survives a two-camera run on a known-good lead points at modes, and modes cost nothing to fix.
Worked example: flight time of a sync pulse on a 3 m cable
The inputs, and where each comes from:
- Hop length L = 3 m — a standard stocked length for camera cables and a typical bench hop.
- Signal speed in the cable v ≈ 2×10^8 m/s — about two-thirds of light speed in vacuum; insulated cable commonly propagates at 0.6 to 0.8 of that speed.
- Subordinate delay values are set in whole microseconds.
The rule in words: flight time equals hop length divided by signal speed.
The arithmetic:
t = 3 m ÷ 2×10^8 m/s = 1.5×10^-8 s = 15 ns = 0.015 µs
A long hop barely moves the answer: 20 m ÷ 2×10^8 m/s = 1×10^-7 s = 0.1 µs.
The result: 15 ns of flight time on the 3 m hop and 0.1 µs on a 20 m hop — respectively 67 and 10 times smaller than the smallest whole-microsecond delay step you can set.
The decision it forces: flight time cannot explain a missing capture. If a subordinate records zero captures, the fault is an open circuit or a wrong mode, not the delay budget. Set the subordinate delay from the trigger-to-exposure latency printed in the camera's data sheet, and revisit cable length only when a hop grows past tens of meters.
Firmware: read it, then update it last
The bench case above passed without any firmware change, and most do — but firmware is a real cause, just a rare one. Some releases fix multi-device synchronized capture, and two cameras on different firmware versions can behave exactly like a wiring fault: one arms, one stalls, the group times out.
Read instead of guess. Every depth camera vendor ships a firmware utility whose query option prints the installed version; compare each camera against the minimum the vendor publishes for synchronized capture, and update only cameras below it, from the vendor's own package. Do it after the count, the wiring and the modes, which fail dozens of times for every genuine firmware fault.
When the mismatch surfaces as an SDK error rather than a timeout, the alignment procedure is separate: 3D camera header version mismatch: fix firmware and SDK alignment. The 3D camera topic page gathers the capture and streaming fixes around this one.
Prevent the next timeout
A chain that passed once can fail after the rig moves. The prevention list is short:
- Label every sync lead at both ends: out on the driver, in on the driven camera.
- Power only the cameras in the capture plan; unused cameras stay unplugged, not idle.
- Prove the chain with a two-camera capture before mechanical assembly, using the checks in verify hardware sync between two cameras actually works.
- Keep each camera rigid through capture; a loose mount shows up as skewed depth rather than timeouts — the mounting rules are in 3D camera mounting: stillness during capture decides image quality.
- Treat a flaky data link as a suspect on its own: the USB fault that appears only while the depth stream runs has its own test, in USB extension cable fails only when you record the depth stream.
Specifying a rig that syncs on the first power-up
If you are building the rig rather than repairing one, three buying decisions prevent this error class: buy the depth cameras as a matched set so firmware and modes line up across the chain, fix the master and subordinate roles on the drawing before any lead is cut, and order the sync and data leads with the cameras so every hop is a known length of the right type.
The 3D camera types most chains are built from, all in the 3D sensors range and filterable in the camera finder:
- Type
- Stereo / structured light
- Type
- Stereo / structured light
- Type
- Depth / ToF camera
On a moving rig the data leads flex with every cycle, so specify drag-chain-rated cable rather than office-grade stock:
When the parts are chosen, the configurator walks the trigger and cabling steps with you: configure the trigger and cabling. If you would rather have the chain checked before ordering, send the rig drawing through the contact page and we will review the sync topology with you.
Questions
Can I synchronize two depth cameras without a sync cable?
Only if your application tolerates a few frames of skew. Starting both cameras in software leaves each on its own clock, so capture times drift apart by milliseconds and a timestamp match fails or wobbles. One hardware pulse from the master's sync-out into both sync-in inputs, or from a pulse generator into both, lands captures within microseconds. The wiring takes minutes; chasing software sync afterwards costs days.
Does the master camera need a delay setting?
No. The master captures on its own clock and drives sync-out, so it has nothing to wait for. Only subordinate cameras take a delay value: the delay shifts each subordinate's exposure start by its own trigger-to-exposure latency, so its depth and color windows overlap the master's. Set the master to master mode with no delay, and put the data-sheet delay on every subordinate in the chain.
How long can a sync cable between cameras be?
Longer than you need. At about 2×10^8 m/s, a pulse crosses 20 m of cable in roughly 0.1 µs, so propagation never sets the limit. Noise pickup and connectors do: keep each hop a few meters, use shielded cable of the type the camera names for sync, seat every plug fully, and route the leads away from motor and drive wiring. A misbehaving long hop is fixed with a shorter cable, not a bigger delay.
Why did a third camera break a two-camera capture?
The wait needs a synchronized capture from every device the program opens, and a powered camera that is not wired into the chain never delivers one. The group waits for its frame until the window expires, then reports the timeout. Disconnect or power down every camera outside the capture plan, set the device count to the cameras you actually run, and add cameras back one at a time, wired and mode-set before each powers up.
Should I flash new firmware before debugging a sync timeout?
Treat firmware as the last check. Wrong modes and loose sync leads produce the identical error in seconds and cost nothing; a firmware flash before the chain is proven can waste an afternoon and leaves cameras on mismatched versions if it goes wrong. Read each camera's version with the vendor's firmware utility, compare it against the vendor's documented minimum for synchronized capture, and update only cameras below that version.



