Glossary

What Is Hardware Triggering in Multi-Camera Capture?

Hardware triggering uses a physical signal to start every camera’s exposure at once, so multi-camera and stereo streams capture the same instant.

Last reviewed: October 2026

Short answer

Hardware triggering is when a physical signal, such as a pulse on a sync line, starts each camera’s exposure so several cameras capture at the same instant. In multi-camera capture it reduces reliance on after-the-fact alignment, and it works alongside a shared clock such as PTP to keep timestamps comparable.

01

Why does hardware triggering matter for stereo and multi-camera robot data?

Stereo and multi-camera robot data only means something if every view describes the same moment. Cameras left to run freely stream asynchronously, and a depth-camera vendor’s multi-camera paper notes that each camera keeps its own clock, which can drift a few milliseconds over tens of minutes, so software has to correct for the offset. Hardware triggering attacks the problem earlier: a shared signal starts the exposures together, so frames from different cameras begin at the same instant instead of being matched afterwards. That matters most in egocentric manipulation, where the interesting motion is fast. As the cross-stream sync error entry puts it, if depth lags video by tens of milliseconds during a fast reach, the supervision is wrong exactly when the motion is interesting. A stereo pair is only a pair if both views were taken together.

02

How does Firsthand use hardware triggering?

Firsthand states that all streams are hardware-triggered onto a single PTP clock domain: a head-mounted global-shutter camera, two wrist cameras, depth, and 200 Hz 6-DoF IMUs. The PTP clock sync entry adds that Firsthand hardware-triggers exposures and then writes int64 nanosecond timestamps per stream, so each frame and sample carries a comparable timestamp. Head-mounted stereo, wrist views, depth, IMU and 3D hand pose all trigger on the same hardware clock within one episode. The site reports a measured residual of 1.12 ms median and 1.94 ms at the 99th percentile, with a 2.0 ms ceiling above which an episode is rejected.

03

What does a hardware trigger actually do?

A hardware trigger is a signal that arrives from outside a camera’s own software loop and starts an exposure. Teledyne’s application note on synchronized capture describes the pattern: one primary camera’s strobe, or an external hardware trigger such as a function generator, drives the other cameras, and the cameras start exposing each set of images within microseconds of each other. The same note says this also makes the secondary cameras’ frame rates follow the primary camera’s.

The depth-camera vendor’s paper describes the same arrangement with master and slave roles, or an externally generated sync signal with every camera set to slave, so the cameras capture at identical times and rates. The common thread is one signal driving many cameras.

04

How is hardware triggering different from software triggering?

The vendor documents cited here contrast hardware sync with free-running capture rather than naming software triggering, so the first table compares those two. When cameras free-run, each streams on its own schedule and alignment happens afterwards in software, using timestamps or frame counters. The depth-camera paper says that is valid, but you have to subtract per-camera offsets and take care not to read frames across a frame boundary. Firsthand’s PTP clock sync entry makes the matching point about timestamps: software timestamps drift and jitter.

In short, software alignment answers which frame is closest, while hardware triggering answers whether the exposures started together. The trade-off is physical. Hardware sync needs a signal path between cameras or from an external generator, and the vendor paper warns that without proper care, ESD or EMI events can reset frame counters even though streaming continues, and that an external generator must match the sensor’s real frame rate.

05

How does hardware triggering relate to PTP clock sync?

They solve different halves of the problem. The IEEE 1588 standard page describes PTP as a protocol for precise synchronization of clocks in packet-based networked systems, supporting the sub-microsecond range, with a grandmaster clock and defined timing domains. PTP aligns clocks, so timestamps from different devices are comparable. Hardware triggering aligns the moment of exposure.

A shared clock with free-running exposures gives comparable timestamps on frames taken at different moments. Triggers without a shared timebase give simultaneous exposures that other sensors cannot be placed against. Firsthand states both halves: one PTP clock domain, hardware-triggered exposures, then nanosecond timestamps per stream.

06

How is trigger sync measured?

Triggering is the method and sync error is the measurement, so a trigger does not remove the residual; it makes the residual small enough to measure. Firsthand measures the offset between streams against a periodic strobe visible to the cameras and detectable in the IMU, per episode, using the int64 nanosecond timestamps on the shared timebase. The sync-error guide reports the result as measured across delivered episodes, with a reject log recording any breach of the 2.0 ms ceiling.

You can check it yourself: the HDF5 and Rerun exports carry per-stream nanosecond timestamps, so the residual can be recomputed on your own copy. The depth-camera paper also suggests pointing cameras at a high-speed LED panel with a very short exposure to see whether they capture the same state.

07

How do you get hardware-triggered egocentric data collected?

Request a custom dataset and describe the skills, environments and streams you need. Collection is fast, on demand and to a written spec, and the site states that hardware triggering is how the rig works rather than an optional add-on. Deliveries use RLDS, LeRobot format or WebDataset, with HDF5, zarr and Rerun exports carrying per-stream timestamps, under a buyer-owned commercial license.

If your program needs a specific trigger rate, signal type or sensor layout, raise it in the spec conversation: the site does not state those details, so they are confirmed per engagement.

09

Free-running capture versus hardware-synced capture, as the cited vendor documents state it

Free-running capture versus hardware-synced capture, as the cited vendor documents state it
AttributeFree-running (default)Hardware-synced
Capture timingCameras stream asynchronously, each on its own schedule (RealSense paper)Cameras capture at exactly the same time and rate (RealSense paper)
Start of exposureNot coordinated between camerasCameras begin exposing each set of images within microseconds of each other (Teledyne note)
Frame rateEach camera’s own rateSecondary cameras follow the primary camera’s rate (Teledyne note); an external generator must match the sensor’s real rate (RealSense paper)
Alignment methodTimestamps or frame counters in software; per-camera clocks can drift a few milliseconds over tens of minutes (RealSense paper)In principle no per-frame alignment is needed when buffering is minimal; larger buffers remove that guarantee (RealSense paper)
What it needsNothing between camerasA sync connection or an external trigger source; cabling quality matters (RealSense paper, Teledyne note)

10

What Firsthand states about hardware triggering and sync, and where it states it

What Firsthand states about hardware triggering and sync, and where it states it
ItemStatement on this siteWhere stated
All streamsHardware-triggered onto a single PTP clock domainHome page FAQ, “What hardware do you capture on?”
IMU (200 Hz, 6-DoF)Hardware-triggered onto the same PTP clock as the video and depth streamsIMU data glossary entry
Wrist camerasHardware-triggered onto the same clock as the head camera, depth and IMU streamsWrist camera glossary entry
Head cameraHardware-triggered onto the same clock as the depth, IMU and wrist-camera streamsHead-mounted camera glossary entry
Exposures and timestampsHardware-triggered exposures, then int64 nanosecond timestamps per streamPTP clock sync glossary entry
Median residual1.12 msCross-stream sync error entry; how-we-measure guide
p99 residual1.94 msCross-stream sync error entry; how-we-measure guide
Reject ceiling2.0 ms; any episode over it is rejectedCross-stream sync error entry; how-we-measure guide

11

What the site does not state about its triggering

What the site does not state about its triggering
DetailStatus
Trigger signal levels, cable type and trigger rateNot stated; confirmed per engagement
Numeric stereo baselineNot stated; the site describes a human interpupillary baseline only
Software-triggered capture optionNot stated; the site describes hardware triggering as how the rig works

12

Frequently asked questions

What is hardware triggering?
A physical signal, such as a pulse on a sync line, starts each camera’s exposure so several cameras capture at the same instant, instead of free-running and being aligned afterwards.
Is hardware triggering the same as PTP clock sync?
No. PTP (IEEE 1588) synchronizes clocks so timestamps are comparable; hardware triggering synchronizes the moment of exposure. Firsthand states both: one PTP clock domain with hardware-triggered exposures.
Does hardware triggering make sync error zero?
No. It makes the residual small enough to measure. Firsthand reports a 1.12 ms median and 1.94 ms p99 residual, with any episode above the 2.0 ms ceiling rejected.
Why not just align timestamps in software?
It works, but per the depth-camera vendor’s paper you must subtract per-camera offsets, and camera clocks can drift a few milliseconds over tens of minutes. Triggering reduces what software has to fix.
Which Firsthand streams are hardware-triggered?
The site states all streams: the head-mounted global-shutter camera, two wrist cameras, depth and 200 Hz 6-DoF IMUs, hardware-triggered onto a single PTP clock domain.
Can I verify sync on my own copy of the data?
Yes. The HDF5 and Rerun exports carry per-stream nanosecond timestamps, so you can recompute the residual yourself rather than trust the reported figure.
What trigger rate, voltage or cable does Firsthand use?
The site does not state those details, so they are confirmed per engagement in the written spec.
How do I order hardware-triggered egocentric data?
Request a custom dataset with your skills, environments and streams. Collection is fast, on demand and to a written spec.