Messages

http://pastebin.com/GrJxcTcC -- 2hr+ on Delkin Devices 1050x Compact Flash card

As in angle of camera for example pointing 10 degrees down.

ML devs, any idea how accurate the level is in-camera?

good suggestions noisyboy.  I would add "Reel" to come before "Scene" and change "Camera Height" to "Lens Height".  I'm not sure what it means by "Angle"

Any insight on the actual shooting performance of the Delkin 1000x cards?

I could not get above 1920x720 @23.976fps sustained on either the Delkin 1000x 32GB or 64GB CF cards.  Average datarate before dropping frames was 64-66MB/s.  pretty lousy performance for a 1000x UDMA7.  does the 5D3 support UDMA protocol anyway?

You're talking about software denoising right?  That pattern noise is the noise floor of the camera (a hardware problem).

that's correct, software noise removal specific to the 5DmkIII's sensor and any others that can be extrapolated from that data

http://en.wikipedia.org/wiki/Fixed-pattern_noise
This is how your sensor looks +3 stops @ 160 ISO. It's got nothing to do with ML except for the fact it's no longer denoised and compressed in camera.

Dan Hudgins http://www.dancad3d.com/ is currently working on eliminating this.  He's making great headway and, as with ML, it's only a matter of time.

apologies I misunderstood.  I see now from the nametag you are shooting on a 600D, I assumed 5Dmk2 or 3.  Still, the fact that a Canon DSLR can compare to a camera from RED because of the efforts of volunteers is astonishing.  The fact they got it to shoot a picture with more color resolution than an Alexa is ridiculous.  I see your point on in-camera processing, just don't agree it's necessary.

Well, I'd rather say that if possible - we should do something with uncompressed raw - cause 25 minutes at 128 GB card is just HUGE and unreal.

The benefit of what Magic Lantern is doing here is pristine data straight from the sensor.  Compression ruins that and makes it just Lantern.  Voluminous data is a consequence of shooting raw, no getting around that.

The other question is - if it's feasible to process the full uncompressed RAW stream with our DIGIC,

the DIGIC processor is not processing the raw datastream.  The beauty of all of this is ML devs were able to bypass all of Canon's image processing and deliver a stream of image data that is 14bits deep and glorious.

A lot of CF cards have been tested already: http://www.magiclantern.fm/forum/index.php?topic=5471.msg37692#msg37692

As you can see in A1ex graph (which is not up to date), the Delkin cards habitually underperform.  In my testing, I have found the Delkin 1000x to be >40MB/s slower than Lexar 1000x and >20MB/s slower than SanDisk Extreme Pro CF cards.

Could the raw_rec module offer options to strip the 14bits depth down to 12, 10 and even 8 bits.

I think all the raw-rec module is doing is saving a datastream the sensor is outputting anyway.  There is no processing happening which is why the camera will not overheat and the sensor will not be damaged.  Also why post workflow has not been solidified.  So no, it would not be possible to convert the 14bit datastream into a lower bitrate without processing in-camera, which defeats the purpose of this modification.

for the graph a1ex: http://pastebin.com/974H86ey -- 2+hrs on 5DmkIII Delkin Devices 64GB 1000x UDMA7 CompactFlash card

can some of you with a LEXAR 1000x post the VRAMx.BMP that is being saved with this plugin?

here's one from a 32GB CF card g3ggo: https://docs.google.com/file/d/0B-UavK-PpYzwQk9xc0F2TUZTMWc/edit?usp=sharing

@N/A,
   Dan Hudgins http://www.dancad3d.com/ is currently working on fixed-pattern noise reduction and figuring out what exactly is going on with Canon's 14bit buffer...I'll post back more info in the next couple days as to his findings.

thanks Audionut!

http://pastebin.com/fKzZ8ee5 -- 4+hr benchmark of SanDisk Extreme Pro 100MB/s

http://pastebin.com/jU0Jv3DH -- 2+hr benchmark of Lexar Professional 1000x

confirmed, lourenco helped me out and the build works great.

lourenco,
  I get an error message when trying to load the modules:

Code Select Expand

[i] proprs    at:  0x00000000
[i] cbr       at:  0x00A97628
------------------------------------------


These long tests are really nice. Keep them coming, please!

I have the results from 4hrs of SanDisk Extreme Pro 128GB 100MB/s CompactFlash card and 2.1hrs of Lexar Professional 32GB 1000x CompactFlash card, but it exceeds the maximum limit of characters for the forum.  A1ex, send me instructions to email to you if you wish.

The latest 5D3 modules build from lourenco does not work for me.  After loading the raw_rec module, its status is "No_Info" instead of the usual "OK" and the "Raw Video" ("Video" being a misnomer if de-bayering is not happening in-camera) option is absent from the ML video menu.  Is anyone else experiencing this?

There should be videos on quick movements, actual real life shooting.

https://vimeo.com/66296381
https://vimeo.com/66299918
https://vimeo.com/66304472
https://vimeo.com/66310082
https://vimeo.com/66296375
https://vimeo.com/66294984

Lexar Professional 1000x 32GB CompactFlash

Code Select Expand

Buffer size experiment
ML v2.3.NEXT.2013May17.5D3113, 3046238719b2+ (unified) tip
Mode: playback, Global Draw: OFF
CF LEXAR ATA FLASH CARD LEXAR ATA FLASH CARD                   
10312704 1152
15885312 1162
80896 296
8949760 1142
28661760 1164
16264192 930
24521728 1080
16424960 1165
13148160 1163
15619072 1079
1658880 1032
21001216 1171
8548352 1053
19760128 1165
5965824 1110
24906752 1166
13968384 1132
16770048 1166
4542464 1121
13063168 1159
33263616 1174
30312448 1172
16713728 1166
3887104 1112
21841920 1168
12846080 1159
16192512 1166
3333120 1102
6719488 1141
17313792 1164
23374848 1168
2865152 1091
32488448 1169
18588672 1166
4186112 1093
821248 910
3661824 1108
18359296 1167
19875840 1166
22754304 1165
13665280 1159
24957952 1161
28555264 1173
33442816 1174
10708992 1156
20242432 1168
8774656 1151
11216896 1155
23156736 1167
12750848 1159
12369920 1158
2157568 1042
15279104 1164
29943808 1196
30321664 1165
19004416 1170
268288 611
23778304 1169
467968 730
1905664 1048
32368640 1172
19829760 1167
26281984 1172
24153088 1173
11364352 1151
32160768 1173
5944320 1136
24118272 1170
28511232 1174
31526912 1170
15956992 1163
1732608 1032
13881344 1160
6106112 1133
17367040 1165
21693440 1166
30835712 1166
1139712 958
13698048 1077
19388416 1086
1532928 1019
16333824 1163
15963136 1164
23540736 1168
23770112 1169
7317504 1143
6019072 1135
13303808 1136
32039936 1171
13137920 1132
24311808 1140
25442304 1168
8986624 1125
25421824 1197
2720768 1061
18526208 1165
12448768 1155
6514688 1140
6632448 1063
17897472 1166
26569728 1174
22568960 1166
11912192 1158
22007808 1170
30651392 1176
29615104 1173
18367488 1166
21049344 1169
15778816 1167
24951808 1168
5359616 1133
23476224 1169
17125376 1167
3318784 1101
12782592 1164
30965760 1163
1974272 1054
32021504 1059
24463360 1080
28246016 1054
23352320 1146
27161600 1177
25191424 1068


I'll run a test on Sandisk Extreme Pro CF for a few hours and post back later

for those who are interested, I posted a blog yesterday with basics on what ML is and currently does, and a straightforward workflow for processing .raw files: http://shop.texasmediasystems.com/and-Canon-will-now-sell-a-grip-of-5DmkIIIs_b_5.html

for this forum's members, a lot will be redundant knowledge, but still worth a look.

beautiful "test" here: https://vimeo.com/66296381

If we are formatting the cards in the camera before using them wouldn't the camera dictate the format of the card?

I haven't found that to be the case, rather the firmware of the card dictates its native formatting schema and size allotment, etc.

also, 1080p23.98 14bit raw footage: https://vimeo.com/66268193  with more to follow

just a guess, but it may be that the 128GB cards are formatted as exFAT, while the <64GB cards are FAT32 and the Canon's firmware is optimized for FAT32..somewhat analogous to the possible 32bit v. 64bit problem with raw2dng.

that's interesting...essentially, what is raw2dng doing to the .raw file?  does it just parse it into individual frames, or does it also de-bayer, or something totally different?