Hey can u provide the driver for the 4 lane imx415 module which can support 4k@60
For axon.
Okay, working on it.
Will update as check completes.
Could you send on which application you will run camera or send the command that you are going to use for your application !?
ffplay, gst-launch-1.0 etc.
I will using it with gstreamer and mediamtx
ffmpeg as well
More i will send later
run.sh (781 Bytes)
- Download in Axon.
- Give permission
sudo chmod +x ./run.sh sudo ./run.sh- You will find one overlays named as below :
Enable 8.3MP Camera (IMX415) on CSI0 4lane Alpha Axon V0.3
Let us know.
Thanks,
Actually im out of town so will test it next week.
In the mean time if u have any update of the current driver please share that as well
IMX415 @ CSI0 — 60 fps driver-mode report
Generated against: Vicharak Axon, kernel 6.1.75-axon (Ubuntu 24.04.4) Date: 2026-08-15 Audience: Vicharak support / kernel team Severity: Driver feature gap (not a hardware fault)
1. TL;DR
The imx415 driver shipped in 6.1.75-axon (v00.01.08) only exposes 2 sensor modes, and both are hard-capped at 30 fps. The IMX415 sensor itself supports 4K@60 (via 2-DOL HDR), 1080p@60 (via 2×2 binning), and other high-rate modes — these are missing from the driver. The fix is to backport the full mode table from the upstream Rockchip BSP imx415.c (or mainline drivers/media/i2c/imx415.c). The DT binding sony,imx415 does not need to change.
2. What the driver supports today (current state)
| Item | Value | Evidence |
|---|---|---|
| Sensor detected | IMX415, 4-lane MIPI | dmesg: imx415 1-0037: Detected IMX415 image sensor + detect imx415 lane 4 |
| I²C address | 0x37 on i2c-1 |
i2c_dev.txt |
| Driver version | 00.01.08 (builtin) |
dmesg: imx415 1-0037: driver version: 00.01.08 |
| DT node | &i2c0 (fea90000) → cam2-imx415@37 |
dmesg_imx.txt |
| MIPI route | sensor → rockchip-csi2-dphy0 → csi2-dphy0-hw@fedc0000 → rkcif-mipi-lvds2 → rkisp0-vir2 |
media_topology.txt |
| Link-freq menu | 5 entries: 297 / 446 / 743 / 891 / 1188 MHz per lane | v4l2_subdev_all.txt |
| Medibus codes | SGBRG10_1X10 (0x300e) + SGBRG12_1X12 (0x3010) |
v4l2_mbus_codes.txt |
| Frame sizes exposed (10bpp) | only 3864×2192 | v4l2_framesizes.txt |
| Frame sizes exposed (12bpp) | none (empty enum) | v4l2_framesizes.txt |
| Max framerate exposed | 30 fps (20 fps at 297 MHz) | v4l2_intervals.txt |
| Driver mode-table entries used | only cur_best_fit(0) under normal operation |
dmesg_modes_unique.txt |
| HDR modes exposed | none (hdr: 0 only) |
dmesg_modes_unique.txt |
| Binned / windowed modes | none | v4l2_framesizes.txt |
The full set of v4l2-ctl evidence is in v4l2_*.txt. The pipeline topology is in media_topology.txt.
3. What we need (target state)
The IMX415 datasheet supports these modes on a 4-lane MIPI CSI-2 D-PHY:
| # | Mode | Resolution | BPP | Path | Max FPS | Driver source for the mode table |
|---|---|---|---|---|---|---|
| 1 | Linear full | 3864×2192 | 10 / 12 | linear | 30 | ✓ already present (mode 0) |
| 2 | 2-DOL HDR full | 3864×2192 | 10 | hdr combine | 60 | ✗ missing — needs cur_best_fit(N) for hdr: 1 |
| 3 | 2×2 binned | 1932×1090 | 10 | binning | 60 | ✗ missing (driver only has 1944×1097@12b linear at 30) |
| 4 | 2×2 binned HDR | 1932×1090 | 10 | binning + hdr | 60 | ✗ missing |
| 5 | Windowed / cropped | 1920×1080, 1280×720, 1280×960, etc. | 10 | ROI | 60+ | ✗ missing |
| 6 | HDR windowed | 1920×1080 | 10 | ROI + hdr | 60 | ✗ missing |
The reference implementations are:
- Upstream Linux:
drivers/media/i2c/imx415.cat kernel.org — currently exposes 8 modes including all of the above. - Rockchip downstream BSP:
drivers/media/i2c/imx415.cinrk3588-linux— exposes the same mode table withmipi_freq_idxcorrectly mapped to the 5 link-freq menu already on the Axon.
4. Side-by-side: current vs. required mode coverage
| Resolution | BPP | Current driver | Needed (per IMX415 datasheet) | 4K60 path? |
|---|---|---|---|---|
| 3864×2192 | 10 |  30 fps (mode 0) |
30 fps linear | n/a (this IS the 4K frame) |
| 3864×2192 | 12 |  (empty enum) |
20 / 30 fps linear | n/a |
| 3864×2192 (HDR) | 10 | (no hdr: 1) |
60 fps 2-DOL HDR | this is the only “4K@60” path |
| 1932×1090 | 10 | (clamped to 3864×2192) |
60 fps 2×2 binned | n/a (1080p) |
| 1944×1097 | 12 | 30 fps (mode 7) |
only at 30 fps | n/a |
| 1920×1080 | 10 | (clamped) |
60 fps windowed | n/a (1080p) |
| 1280×720 | 10 | (clamped) |
60+ fps windowed | n/a |
| 1280×960, etc. | 10 | (clamped) |
60 fps windowed | n/a |
The 297/446/743/891/1188 MHz link-freq menu is already on the Axon and has enough bandwidth headroom (1188 MHz × 4 lanes = 4.752 Gbps) for 1080p@60 and 4K@60 in HDR. The driver simply never references any output configuration that would use that bandwidth at 60 fps.
5. Live measurement (sensor under 30 fps linear)
| Probe | Result | Interpretation |
|---|---|---|
v4l2-ctl -d /dev/video11 --stream-mmap=4 --stream-count=90 |
30.00 fps (see earlier run in conversation) |
Sensor delivers exactly 30 fps in linear mode, as expected |
v4l2-ctl --set-ctrl link_frequency=4 (max 1188 MHz) |
Still 30 fps reported in ENUM | Mode table cap, not bandwidth |
v4l2-ctl --set-ctrl vertical_blanking=46 (min) |
Still 30 fps reported in ENUM | Same — min VTS still > required VTS for 60 fps |
| Force 1944×1097@12bpp + max lf + min vblanking | Still 30 fps in ENUM | Mode table is the cap, not the sensor |
6. Repro steps for the support team
# 1) Show sensor detection (4 lanes confirmed)
dmesg | grep imx415
# 2) Show only 1 frame size is exposed
v4l2-ctl -d /dev/v4l-subdev2 --list-subdev-framesizes pad=0,stream=0,code=0x300e
# -> "Size Range: 3864x2192 - 3864x2192" (4 times, one per link-freq slot)
# 3) Show only 20/30 fps are exposed, never 60
v4l2-ctl -d /dev/v4l-subdev2 --list-subdev-frameintervals \
pad=0,stream=0,width=3864,height=2192,code=0x300e
# -> "Interval: 0.033s (30.000 fps)" and "0.050s (20.000 fps)" only
# 4) Show only 1 driver mode (cur_best_fit(0)) is ever used
# (after letting a few seconds of operation pass)
dmesg | grep cur_best_fit | sort -u
# -> only "cur_best_fit(0)" — never 1, 2, 3, 4, 5, 6, 7
7. Why this is a driver bug, not a hardware issue
- The part is detected as a genuine IMX415 on 4 lanes (
dmesg: Detected IMX415 image sensor+detect imx415 lane 4). - The Sony IMX415 datasheet supports 4K@60 (2-DOL HDR) and 1080p@60 (2×2 binning / windowed) — the silicon can do it.
- The driver exposes a 5-step
link_frequencymenu (up to 1188 MHz/lane = 4.752 Gbps total), which is enough MIPI bandwidth for both 1080p@60 and 4K@60 in HDR. - The driver’s own
pixel_ratecontrol (read-only 356.8 MHz for mode 0) is computed from the mode table, not from the sensor — so the “max 30 fps” ceiling is a driver-side decision, not a sensor-side limit. - Manual attempts to force 60 fps (max link-freq, min vblanking, mode change to 1944×1097@12b) all return 30 fps in
ENUM_FRAMEINTERVAL— the driver itself rejects them.
8. Suggested fix
Two options, in order of preference:
Option A — Vendor backport (recommended): Pull the latest imx415.c from Rockchip’s rk3588-linux BSP. It exposes the same DT binding (sony,imx415) and the same link_frequency menu already on the Axon, but with the full mode table including 2-DOL HDR, 2×2 binning, and windowed modes. No DT changes needed.
Option B — Upstream backport: Replace with mainline drivers/media/i2c/imx415.c from kernel.org. Same binding, slightly different control IDs (exposure / analogue_gain / link_frequency) — fully backward compatible with the existing DT.
Option C — Minimal patch: If neither full backport is feasible right now, the minimum patch to add support for 1080p@60 is ~20 lines: add 2 entries to the driver’s supported_modes[] table for 1932×1090 @ 10bpp linear and 1920×1080 @ 10bpp linear, both with mipi_freq_idx = 4 (1188 MHz). A reference diff is available on request.
9. Workaround until the fix lands
We’re using the 3864×2192 @ 30 fps linear mode, which is the only mode the driver exposes. This is confirmed working at 30.00 fps via:
v4l2-ctl -d /dev/video11 --set-fmt-video=width=1920,height=1080,pixelformat=NV12 \
--stream-mmap=4 --stream-count=90 --stream-poll
# <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< 30.00 fps
For higher framerates we’d need a kernel rebuild with the upstream imx415.c.
10. Files in this attachment bundle
| File | What it contains |
|---|---|
IMX415_60fps_Driver_Report.md |
This document |
system_info.txt |
uname -a and /etc/os-release |
dmesg_imx.txt |
`dmesg |
dmesg_modes_unique.txt |
Unique cur_best_fit entries the driver ever logs |
dmesg_full.txt |
First 100 lines of dmesg for context |
media_topology.txt |
media-ctl -d /dev/media0 -p (full pipeline graph) |
v4l2_subdev_all.txt |
v4l2-ctl -d /dev/v4l-subdev2 --all (full control list) |
v4l2_framesizes.txt |
ENUM_FRAME_SIZE for 10bpp + 12bpp |
v4l2_intervals.txt |
ENUM_FRAME_INTERVAL for 3864×2192 |
v4l2_mbus_codes.txt |
ENUM_MBUS_CODE |
v4l2_devices.txt |
All V4L2 devices on the system |
v4l2_video11.txt |
ISP mainpath format list + current fmt |
v4l2_video0.txt |
CIF raw path format list + current fmt |
v4l2_fps_measured.txt |
Live FPS measurement (90 frames) |
modinfo_imx.txt |
modinfo imx415 (driver build info) |
i2c_dev.txt |
I²C device tree for the sensor |
imx415_60fps_evidence.zip (11.2 KB)