Page 8 of 26

Re: Getting RAW Video

PostPosted: Sat Apr 18, 2015 4:26 pm
by leegong
@midairtoss ,
Nikon SDK is greatly helpful for hacking , i frequently search for data description in SDK document .
Thank you very much for giving me help , don't hesitate to post any info for hacking , thanks again .

Re: Getting RAW Video

PostPosted: Sun Apr 19, 2015 1:11 am
by leegong
Just check IMX071 settings in MV defect shooting mode , it is same as sensor setting in Liveview 24 FPS mode ,
i can safely say that MV defect RAW is same as LV RAW
, according to decoding of sub "rebuild_MvDefectTable_xx" spotted by Simeon ,
the total number of pixels of MV defect RAW should be 0x1BCD80 , decimal 1822080 , meaning size of 1664 x 1095 ,
according to sub "rebuild_MvDefectTable_xx" , bit depth of MVdefect RAW is 8 bits , not sure about it at all .

Re: Getting RAW Video

PostPosted: Sun Apr 19, 2015 1:33 am
by Fred
Does it mean that only 8 bit video is available with LW raw?

Re: Getting RAW Video

PostPosted: Sun Apr 19, 2015 1:40 am
by leegong
Fred wrote:Does it mean that only 8 bit video is available with LW raw?

Based on current decoding , yes .
However it's possible that IMX071 sends more bits(10bits , 12bits or even more ) over , but chip4000 substract images bits ,
we have to observe LVDS transfering and take MV defect shooting to verify it .
EDIT : Since 8bits YUV output for later Liveview displaying and H.264 encoding ,
8bits RAW is the right and economic system design , unless IMX071 doesn't support 8bits output .

Re: Getting RAW Video

PostPosted: Tue Apr 21, 2015 7:44 am
by leegong
When looked at the firmware yesterday , i suddenly realized that no need to patch the firmware for MV defect ,
the camera itself is able to shoot MV defect without any patch , this is why my patch for MV defect RAW always failed.
then i traced lots of codes related to MV defect data , i understand that it's really not diffciult to shoot MV defect ,
finally i successfully capture RAW with Nikon offical firmware 0101 , just inject the patch which stores struc_79
into SD card .
I take a quick look at it , it looks like a still dark frame , see the attachment .
EDIT : It looks like a 12bits or 10bits dark frame ( i kept the lens attached ) .

Re: Getting RAW Video

PostPosted: Tue Apr 21, 2015 10:40 am
by stvnms
Simply make a great effort and share your knowledge, is reason enough to thank you and everyone working on this project. Thank you.

Re: Getting RAW Video

PostPosted: Tue Apr 21, 2015 12:38 pm
by a1ex
The image looks like 16-bit data, high 8 bits being zero (probably because it's a dark frame).

From FFT analysis (probably biased by the horizontal lines in attached picture), possible values for image pitch (in bytes) are 4992, 3328, 2496. Assuming a 3:2 aspect ratio, the image is most likely 1248x840. Do these numbers make any sense for you?

Re: Getting RAW Video

PostPosted: Tue Apr 21, 2015 1:48 pm
by Macman
Does this mean the patch will be ready to test soon? Because this is really good work been really looking forward to Nikon itself to set up it's game in the video area for a while now, but you guys have made it possible to stay with Nikon. Thank you! Awesome work!

Re: Getting RAW Video

PostPosted: Tue Apr 21, 2015 7:38 pm
by Simeon
a1ex wrote:The image looks like 16-bit data, high 8 bits being zero (probably because it's a dark frame).

From FFT analysis (probably biased by the horizontal lines in attached picture), possible values for image pitch (in bytes) are 4992, 3328, 2496. Assuming a 3:2 aspect ratio, the image is most likely 1248x840. Do these numbers make any sense for you?


3328 is double the 1664 x 1095 value leegong quoted above, which makes sense in its recorded in int16

Re: Getting RAW Video

PostPosted: Wed Apr 22, 2015 6:47 am
by leegong
This time i modified patching location in the firmware , set capturing size = 0x200000 , the captured RAW
is interesting and weird :
1) 0x1BCD80 (decimal 1664 x1095 ) bytes from the beginning of captured RAW look like dark frame , 8bits depth ?? 16bits depth ??
2) Most of value are 0x00 , meaning very low standard deviation or standard mean variance compared to previous
RAW captured yesterday , does it hint this is due to 3x3 addition mode meaning random noise is decreased to 1/9 ? 9 times
3) i tend to think the RAW i captured yesterday is a still photo defect dark frame ,
probably is same as normal size of nikon NEF RAW , because i found the patching location yesterday was a common entry ,
it may be triggered by another unknown RAW routine also , this time i'm very sure the following attachment is
only for building MV defect table .
4) value in offset greater than 0x1BCD80 aera are quite weird .