uncompressed 14-bit RAW video recording

[1%]

Raw recording needs lv debug enabled... Also its still hardcoded to 5d3 for the 600D build.

Any chance to manually put the container on the raw files? I assume thats why they won't open.

[N.Mendes]

Can we translate your "technical words" by, "any chance to obtain a real 2k raw @24 fps on the 5d3"?

[1%]

Its not that good yet... although the 720P isn't far off from 1080P and can be stretched quite far.

You're getting like 2s at 24fps... maybe 1 on "2K" 1080 or zoom mode. CF card cameras will have more luck here, also depending on how much memory they have. 5d3 probably getting best frame rates so far with 6D close second and maybe 50D when it works will have enough memory to match or beat it since it has CF.

[N.Mendes]

The only thing i can say is, "good luck my friends", you're doing your best, and everyone knows that..

I'm french, and here in Paris, everybody is talking about what you're doing, about the Bmcc too, because your exploit could change the game..

Some are saying that the maximum you can obtain is a few seconds @ 24fps 2K raw, some others are saying that you can do that

Let's the time tell us who is right..

Again, Good luck from Paris

[1%]

Also for 600D how is the raw histogram looking... I'm not sure that its accurate. It looks squished and sometimes black level detects as 0.

[Alia5]

Raw recording needs lv debug enabled... Also its still hardcoded to 5d3 for the 600D build.

Any chance to manually put the container on the raw files? I assume thats why they won't open.

The RAWs are also corrupt in camera (when it saves them), i guess they really are broken, havent looked to deep into it though

[1%]

Most work ... I opened one 1740x1154 I think thats crop mode... also 1804x1020 works. I did not check with fps override yet... I basically compiled the bin and shot a few pics then went to sleep.

[Andy600]

I think there's a bug when changing pic size in the Canon menu (I'm using Tragic Lantern 2 on a 600d)



I can still change pic size in the ML menu but the Canon menu seems to be stuck like this now. Any ideas?

[1%]

Change to valid pic size, not MRAW/SRAW from ML menu. Also change won't accept until you switch modes so switch to raw only and flip between the 2 modes you were in. Menu should go back to normal.

Plus still other bugs, like bit rate stuff isn't all done.

[Andy600]

Change to valid pic size, not MRAW/SRAW from ML menu. Also change won't accept until you switch modes so switch to raw only and flip between the 2 modes you were in. Menu should go back to normal.

Plus still other bugs, like bit rate stuff isn't all done.

Thanks! That sorted it

[N/A]

Something I did on the 600D yesterday resulted in the camera never buffering out with yuv, got up to 400 or so frames and had to end up doing a batt pull lol. Think it had something to do with frame skip...

[Rush]

Something I did on the 600D yesterday resulted in the camera never buffering out with yuv, got up to 400 or so frames and had to end up doing a batt pull lol. Think it had something to do with frame skip...

You need a lot of frame skipping to make 600D to write continuously. It is sad, that 600D SD is limited to only 21 mb/s. So - only very short bursts (or timelapse) are usable.

[kgv5]

Is it possible to try YUV422 on 6d? How much those 600d yuv422 frames weights?

[N/A]

If each frame could be truncated in-camera to ~1 MB, would still be better than h.264.

Easier said than done I suppose.

[1%]

Yep, if they could be compressed to JPEG all problems would be solved.

[EOSHD]

Yep, if they could be compressed to JPEG all problems would be solved.

Is it a code problem or a CPU time issue? Is debayering and compressing to JPEG at 24fps just not possible?

[1%]

A1ex + G3ggo said it wouldn't work real time... nobody has tried to compress it in any way tho so who knows.

[N/A]

Random thought for music videos... Play the song back at half the bpm, have the singer perform along while recording raw/yuv at 12fps, and play it back in post at 24p. Feasible? Or just a choppy fuckin mess?

[shuggyrasklat]

this entire vid is comprised of 5x zoom recordings, 51 frames a piece, on the 5dmkii on a 60mb/s sandisk 16gb 2152x1076 comp size, all at 135mm.

http://www.youtube.com/watch?feature=player_embedded&v=e7Pu6m8GhdM

[P337]

this entire vid is comprised of 5x zoom recordings, 51 frames a piece, on the 5dmkii on a 60mb/s sandisk 16gb 2152x1076 comp size, all at 135mm.

http://www.youtube.com/watch?feature=player_embedded&v=e7Pu6m8GhdM

Hmm, still see some exposure hick-ups (or was that something actually in "the scene" or added in post processing during the 0:18-0:19 clip?) 

Was this filmed with the new "YUV/DNG recorder"?  I assume it was filmed with the older 14bit "silent burst pic" DNGs. 

Has anyone got the DNGs from the new recorder open yet? (and are the new DNGs still 14bit or just referenced from the 14bit raw image then compressed?  I don't understand yet why the DNGs from the new recorder won't open)

btw thanks for the test shuggyrasklat (looks sharp)

[P337]

Yep, if they could be compressed to JPEG all problems would be solved.

True for bandwidth, but from a purely image quality perspective wouldn't even the "Highest Quality" JPEG (8bit, 4:2:2, 2.6:1 compression) produce slightly lower quality than the YUV422 files or HDMI out?  If we had to resort to JPEG, is there even a point to use it over YUV422 or HDMI out? (for 5D3 users especially)

[minimimi]

I'm tried to aggregate FIO_WriteFile , like this ..
Because, I think contenious sequential access is most faster when we write file to filesystem.

Code: [Select]

@@ -416,8 +417,12 @@

         if(!skip_saving)
         {
-            FIO_WriteFile(save_file, UNCACHEABLE(&save_data->chunkData.chunkAddress[save_data->chunkData.chunkOffset]), avail);
+            memcpy(shmem+written,UNCACHEABLE(&save_data->chunkData.chunkAddress[save_data->chunkData.chunkOffset]),avail);
             written += avail;
+            if(written == length)
+            {
+                FIO_WriteFile(save_file, shmem , length);
+            }
+
         }

         length -= avail;
@@ -451,12 +456,20 @@
     strcpy(&data->fileSuffix[1], tmp);
 }

 void lv_rec_start()
 {
     int yPos = 3;
     lv_rec_data_t data;
     lv_rec_save_data_t save_data;

+    shmem = shoot_malloc(8000000);
+    if(shmem == NULL)
+    {
+        NotifyBox(3000,"8M shoot memory allocate failed");
+        msleep(3000);
+        return;
+    }
+
     /* set all values to zero */
     memset(&data, 0x00, sizeof(lv_rec_data_t));
     memset(&save_data, 0x00, sizeof(lv_rec_save_data_t));


But memcpy from DMA mem to shoot mem is really slow. Let me know more faster memcpy method, If you know.
Also , shoot mem is only allocated about 8MB. where located more fast and bigger memory?

Modified
5D3 has no dma_memcpy

[1%]

dma memcpy didnt' work? shoot_malloc(0) should be the big memory

[minimimi]

dma memcpy didnt' work? shoot_malloc(0) should be the big memory

Anybody didn't find it. and I did a actual test like this

shoot_malloc(4M)
shoot_malloc(8M)
shoot_malloc(16M)

and check these returns. Then , I can find only 8M to use from ML.

[1%]

Specific amount won't let you allocate the big chunk, its a new process.

dma mecpy is there, its called aj blt dma or something like that, just not renamed.