Camera settings#

Where camera settings live#

Each configured camera gets its own card in the Video settings tab (CameraCardW), with five sub-tabs: Image, Exposure, Gain, Advanced, and HW Trigger. Every field is backed by one member of CameraParameters (src/core/settings.hpp) and is persisted per-profile in that user’s settings.json under video.cameras[i].

Most fields live-apply: editing them pushes the new value straight to the already-open camera (debounced ~150 ms) without interrupting preview or recording. A few structural changes — resolution, pixel format, frame rate, and the HW Trigger tab’s fields — instead trigger a full close-and-reopen of every configured camera, since those can’t be changed on an already-streaming GigE device. Both paths go through the same settings.json, so there’s no separate “apply” step to remember.

Image tab#

Field

Default

Notes

Width / Height

1920 / 1080 px

Sensor ROI size. Structural — triggers a reopen.

Offset X / Y

0 / 0 px

ROI top-left corner. Structural.

Reverse X / Y

off

Horizontal/vertical flip.

Pixel format

BGR8

One of BGR8, RGB8, Mono8, Mono12, BayerRG8, BayerBG8. Structural.

Specify frame rate

on

If off, the camera free-runs at whatever rate its current exposure/ROI/bandwidth allows.

Frame rate

25 fps

Matches the acA1920-25gc’s real sustained maximum, not the generic “30 fps” a lot of camera UIs default to. Structural — see good default values below for why requesting more than a camera can sustain just under-delivers silently rather than erroring.

Exposure tab#

Field

Default

Notes

Auto mode

Once

Off | Once | Continuous — see auto modes explained below.

Exposure time

10000 µs

Manual value, only used when Auto mode is Off.

Auto range (lower/upper)

100 / 50000 µs

Clamps what Once/Continuous is allowed to pick.

Gain tab#

Field

Default

Notes

Auto mode

Once

Same three options as Exposure. On the ``acA1920-25gc``, manual (``Off``) gain does not currently work — see the callout below.

Gain

0.0 dB

Manual value; only meaningful if the camera actually has a working manual-gain node (see below).

Auto range (lower/upper)

0.0 / 24.0 dB

Clamps what Once/Continuous is allowed to pick.

Note

Why you may see “Skipping ‘Gain’: … No node attached.” — this camera generation only exposes the older SFNC 1.x GainRaw integer node, not the modern SFNC 2.0 Gain (dB) float node Mosaic writes to when Auto mode is Off. The write is skipped safely (this is not a crash or a real error — see VideoGrabber::apply_image_params() in video_grabber.cpp), but it also means manual gain is a silent no-op on this hardware today. This is exactly why the default is Once rather than Off — with manual gain effectively unusable, Off was leaving cameras at whatever gain they happened to power on with. If you need a working manual gain control, that would require implementing a GainRaw fallback path (the same pattern already used for BlackLevel’s SFNC 1.x fallback) — not done yet.

Advanced tab#

Field

Default

Notes

Gamma

1.0

Black level

0.0

Falls back to the SFNC 1.x BlackLevelRaw node on this camera generation (the SFNC 2.0 float node isn’t present) — this one does have a working fallback, unlike Gain above.

Saturation / Contrast / Brightness

1.0 / 1.0 / 0.5

No GenICam node on the acA1920-25gc at all (no on-camera ISP for these) — saved with the session regardless, but only take visible effect on a camera model that exposes the matching node.

Auto target brightness

0.5

Digital shift

0 bits

For 12/16-bit → 8-bit conversions.

White balance auto mode

Once

Off | Once | Continuous, same semantics as Exposure/Gain.

Test pattern

Off

Simulated pattern shown in the monitor when no real camera is connected — Off | ColorBars | Horizontal | Vertical.

HW Trigger tab#

Covered in full in Synchronization, including what “Action1” actually does and the live per-camera “Action-command support” readout. Quick reference:

Field

Default

Notes

Enable hardware trigger

on

Structural — reopens every configured camera.

Trigger source

Action1

Line1 | Software | Action1. Action1 needs no physical cable — it’s a GigE Vision Action Command broadcast over the existing camera network, fired once per frame for the whole session.

Trigger delay

0 µs

Only meaningful with a physical trigger source (Line1); not used by Action1.

Auto modes explained (Exposure, Gain, White Balance)#

All three share the same three-value semantics:

  • Off — fixed, manual value you set yourself. Never changes on its own.

  • Once — the camera auto-calibrates a single time (when preview starts), then locks that value for the rest of the session, including through recording.

  • Continuous — the camera keeps re-adjusting for as long as it’s running, including during recording.

Continuous tends to look the best in a live preview, since the camera is always compensating for whatever’s currently in front of it. The tradeoff: brightness/color can visibly shift mid-recording, which is a real risk if anything downstream assumes consistent lighting frame-to-frame (facial expression confidence, luminance-based measures, anything comparing frames across time). Once is the better default for a research recording tool for exactly that reason — you get a scene-matched image without the mid-recording drift risk. Off gives full manual control but requires you to already know a good value, and — for Gain specifically on this camera generation — currently doesn’t work at all (see the Gain tab’s note above).

Default values, and where they come from#

Every field’s default lives in exactly one place: CameraParameters’s own member-initializers in src/core/settings.hpp. Three code paths all read from that single source of truth, so there’s nowhere else a “default” can silently diverge:

  1. A brand-new profile gets seeded with room 11’s real 6 camera serials (default_room11_cameras() in src/core/application.cpp) — every other field is left at CameraParameters’s defaults.

  2. “+ Add camera” in the Video tab (VideoSettingsW::add_camera()) constructs a plain default-initialized CameraParameters.

  3. “Discover cameras” fills in only the serial number and model name it found on the network; everything else is, again, left at the defaults.

To change a default for every future camera (not ones already saved), edit the member-initializer in CameraParameters (settings.hpp) and rebuild — see the Gain/Exposure/White-balance defaults above for a worked example ("Off" → "Once", changed 2026-07-27 for exactly the reasons described in this page).

How to change settings for a camera that already exists#

Existing, already-configured cameras do not pick up a new code default automatically. CameraParameters::from_json() only falls back to the struct’s default when a key is missing from settings.json — for a camera that’s already been saved once, every field is present, “Off” included, so a code-level default change has no effect on it. Two ways to actually change an existing camera’s settings:

  • Through the UI (recommended for a one-off/manual change): open the relevant tab in that camera’s card and change the control directly. Most fields live-apply within ~150 ms; HW Trigger and Image-tab structural fields reopen the camera.

  • Editing ``settings.json`` directly (useful for changing several cameras at once, or scripting a rollout): find the file at %LOCALAPPDATA%\CSRU\MOSAIC\settings.json (or profiles\<username>\settings.json for a named profile — see Research-group profiles), and edit the relevant key inside video.cameras[i], e.g.:

    {
  "video": {
    "cameras": [
      {
        "serial": "24925616",
        "exposure_auto": "Once",
        "gain_auto": "Once",
        "balance_white_auto": "Once"
      }
    ]
  }
}

    The app must be closed while you edit — it only reads this file at startup and rewrites it in full on shutdown, so any change made while it’s running will be silently overwritten when it next saves.

See also#