khangly commited on
Commit
648cceb
·
verified ·
1 Parent(s): 15ec1f0

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +30 -0
README.md ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ pretty_name: Nikon Z 30 dark frames
4
+ tags:
5
+ - raw
6
+ - nikon
7
+ - sensor-calibration
8
+ - dark-frames
9
+ size_categories:
10
+ - n<1K
11
+ ---
12
+
13
+ # Nikon Z 30 dark frames, October 28, 2026
14
+
15
+ Lens-cap frames from a Nikon Z 30 (firmware 1.00), shot over USB from a script, for measuring the sensor's black level, read noise and the raw spatial filtering Nikon applies at long exposures. 14-bit lossless NEF.
16
+
17
+ ## Contents
18
+
19
+ - `t78/dark_iso{100,1000}_{t}s.nef`: ISO 100 and ISO 1000 at 1/15, 1/8, 1/6, 1/4, 1/2, 1, 2, 4, 8 and 30 s (`0p25` reads 0.25 s), long-exposure noise reduction off.
20
+ - `t78/cap_iso{100,400,1600}_1s.nef`: 1 s at three ISOs.
21
+ - `t0/`: the lens-cap check frames (small JPEGs) made before the series.
22
+
23
+ ## What was found
24
+
25
+ - Black level 1008 on every channel at every ISO; white level 16,383; no PDAF-row signature (the blue-pixel row variance is the same on the rows said to carry phase-detection pixels and elsewhere).
26
+ - From 1/4 s upwards the NEF is spatially filtered even with long-exposure NR off: a pixel more than about 50 ADU above the maximum of its 5 × 5 neighbours is clipped to that level, so hot pixels and single-pixel stars vanish at 1/4 s and longer. Frames of 1/6 s and shorter are untouched. The analysis follows Jim Kasson's row-spectrum test and Shelley's neighbour-maximum plot.
27
+
28
+ The analysis scripts (`analyze_dark.py`, `analyze_dark_shelley.py`) and the per-frame numbers are in the camera repository the dataset belongs to.
29
+
30
+ Licence: CC BY 4.0. Frames by Khang Ly.