VOE command 0x206 fail ("hal_voe_send2voe too long") on 2 boards, 2 SDK versions (4.1.0 stable + 4.1.1 pre-release), and the stock StreamRTSP/VideoOnly example

Arduino SDK
1

Board: AMB82-Mini, stock kit (Realtek RTL8735BDM), Arduino board package realtek:AmebaPro2 version 4.1.1-build20260417.

Camera module: the board’s own stock JXF37 module (confirmed against the official product listing). Physical module printed PA-FP0171-A-01 XCF. Camera Options JXF37, FCS Mode Disable — both defaults, confirmed via arduino-cli board details.

Symptom: RTSP server works completely normally (connects, DESCRIBE/PLAY succeed) but zero video frames ever flow. Boot log:

[video_pre_init_procedure] START
hal_voe_send2voe too long 51439 cmd 0x00000206 p1 0x00000000 p2 0x00000000
VOE command 0x206 fail ret 0xff

And at boot-loader stage, before any sketch code runs:

FCS KM_status 0x00002081 err 0x0000200a
It don't do the sensor initial process
RAM TM_STATUS 0x00bf1208 err 0x00001208

Troubleshooting done — seven independent full-boot-log captures, all byte-for-byte identical:

  • Ribbon cable reseated 3× (one with close visual inspection under bright light) — no change; timing stayed within ~170us across attempts (51439, 51448, 51433, 51278us), arguing against an intermittent connector fault.
  • USB cable and port swapped (COM14→COM13 confirmed) — ruling out marginal power delivery.
  • Confirmed physical module matches the documented stock JXF37 sensor and Camera Options/FCS Mode are already correct — not a sensor-selection mismatch (unlike some similar reports on this forum). Noted one nuance: a separate thread quotes Realtek staff saying the board actually ships with JXF37P and needs manual sensor.h editing — but that thread’s own resolution is unconfirmed, and our installed SDK has no SENSOR_F37P constant to select in the first place.
  • A second, completely different physical board (different MAC/sensor unit) — identical failure (51444).
  • Realtek’s own unmodified StreamRTSP/VideoOnly.ino example (not our code) — identical failure (51160).
  • The last stable 4.1.0-build20260213 release (not just the 4.1.1 pre-release/QC build we started on) — identical failure (51165), confirmed genuinely rebuilt via the changed boot-loader timestamp and VOE library version string.

All seven captures: cmd 0x00000206 fails within ~290 microseconds of each other (51439, 51448, 51433, 51278, 51444, 51160, 51165), and the boot-loader FCS status codes are identical every time.

Question: given this reproduces across 2 boards and 2 SDK versions including Realtek’s own unmodified example, is this a known issue, and does “It don’t do the sensor initial process” indicate something about how these units’ factory calibration data was provisioned? Happy to provide full logs or run any other diagnostic.

2

:waving_hand: Thanks for your post!

For documentation, SDK resources, FAQs, and community guidelines, please visit: here

Happy building with Ameba!

:waving_hand: 感谢您的发帖！

如需查阅官方文档、SDK 资源、常见问题（FAQ）及社区使用指南，请参考: 這裏

祝您使用 Ameba 开发愉快！

3

Hi @nkuecuek ,

Could you help to double check with the latest version for early release?
image

If possible could you also take a photo of camera and connection of the cable?

I am wondering if it is an HW failure. I have checked with same camera it is worked with video only demo.

BR