Persistent high latency (~500ms, 15-20 frames) on Depth stream over Ethernet with Femto Mega – no issue over USB-C

Technical Support & Q&A
1

Hi everyone,

I’m experiencing a consistent, severe latency issue with my Femto Mega when connected via Ethernet, specifically on the Depth stream. Would appreciate any insight from the team or community.

Setup:

  • Firmware version: 1.3.1
  • Connection: Direct Ethernet (Cat6 cable), no switch/router in between
  • NIC: Realtek Gaming 2.5GbE Family Controller
  • OS: Windows 11 (build 10.0.26200)
  • Tested in both OrbbecViewer and TouchDesigner — same behavior in both

Stream settings (from OrbbecViewer):

  • Color Stream Profile: 1280x720 30fps MJPG
  • Depth Stream Profile: 512x512 30fps Y16
  • Align Mode: D2C(HW)
  • Frame Synchronization: On
  • Frame Aggregate: Off
  • Decimation Filter Factor: 1

Symptoms:

  • Depth stream shows a consistent ~500ms (15-20 frame) delay from the moment the app starts. It doesn’t grow over time — it’s constant from the start.
  • Color stream is completely normal, no delay at all — same cable, same connection, at the same time.
  • PointCloud mode also lags (expected since it depends on Depth).
  • Switching to USB-C connection completely resolves the issue — full speed, no lag.
  • Viewer logs consistently show:
    [WARN] sensor: Depth fps too low. curStreamProfile fps: 30, fpsArrived_: ~8.8

What I’ve already tried (no effect on the lag):

  • Different Ethernet cables and ports, direct connection to PC (no switch)
  • Set NIC speed manually to both 100Mbps and 2.5Gbps Full Duplex — same lag in both
  • Enabled Jumbo Frames (9014 bytes)
  • Disabled NIC power management (“allow computer to turn off this device”)
  • Disabled Interrupt Moderation on the NIC
  • Set Ethernet interface metric to 1 (to rule out routing conflict with WiFi on the same subnet)
  • Disabled Windows KinectService
  • Fully power-cycled the camera (not just cable reconnect)

None of the above changed the behavior. The lag is isolated to the Depth stream specifically over Ethernet — Color stream on the exact same connection is unaffected.

Has anyone found a working fix for this? I noticed in GitHub issue #59 a similar report of dropped frames over Ethernet with no clear network bottleneck. Also curious if switching the Depth format away from Y16 (mentioned as reducing freeze issues in another thread) would help here — any guidance on how to do this in OrbbecViewer or via config would be appreciated.

Thanks in advance!