Full-resolution silent pictures (silent.mo)

Started by a1ex, July 01, 2014, 05:11:15 PM

Previous topic - Next topic

0 Members and 2 Guests are viewing this topic.

barepixels

Here is my test result from today.  mk11174 thank you so much for compiling for me and of course thnx A1lex for making the impossible, possible.



Manual mode, F2.8, 1/4sec, ISO 100, Tiffen variable ND filter, 15sec interval
361 frames, 2160p, 23.976 frame rate,  After Effects (best view fullscreen at 4K)

NOTE: 2 frames are completely black (frame 241 and 362 out of 363 were shot)
Don't know why or how to avoid.

BTW  silent.mo were the only module loaded.  and Global draw was set off.

Question:  would ETTR helps?
5D2 + nightly ML

vstrglv

What is min exposure time for raw full-res?
Canon 5D3,1.1.3; Canon EOS M,202,  CF-SanDisk Extreme PRO,160MB/s, 256GB, SD-SanDisk Extreme Pro, 170MB/s, 128GB.

Attero

Working on 600D! Thanks so much! Im gonna try this in the night with Intervalomat enabled.
For TimeLapse it would be realy nice if every sequenz is saved in a seperat .mlv or .raw or even a new .tls (TimeLapseSequenz) container.

U r the real MVPs!

X-RAY

This is so great ... after all this time alex is still in the game for great new developments!

I always wanted silent pics for timelapse, so my dream seems to come true.
Is it possible to get a nice configurable timelapse-section in ML?
www.frankenfotograf.com
instagram.com/frankenfotograf

garry23

Some testing feedback on the 5DIII.

If I try and use an ETTR setting, at 0%, the FF silent DNG is always about 1 stop over exposed. ETTR setting was 2s, so beyond FF silent min.

If aperture set fully open, ETTR seems to be OK, ie stopping down doesn't work.

Linking the FF-Silent with timelapse generated an ERR 70 and some crash logs: https://www.dropbox.com/sh/mnqjd20ejr6qowg/AAD0RXnnPGfEpsGe_xjdE8aCa

Will carry on testing and reporting.

budafilms

ASSERT: 0
at ./Param/AccumulateTime.c:53, task ShootCapture
lv:0 mode:3


Magic Lantern version : Nightly.2014Jul18.5D3123
Mercurial changeset   : e0a1e77aeb30+ (5D3-123)
Built on 2014-07-18 19:19:21 UTC by mathew@mathewubuntu64.
Free Memory  : 136K + 3731K

gcrook

EDIT:Exp override was off,just read the previous post.

MLV worked for me as well as ML raw video converter 1.9.1 with my 60D.

I did a simple exposure test primarily to find out how to expose.
I shot 1 photo mlv timelapses trying the next shutter step starting from 1'' exposure until 1/400.
Used f1.4 and iso 100 throughout and wrote down the milliseconds.
Extracted and imported to LR to measure (alas by eye and histogram) exposure and possible issues.

Here are my initial findings.

1''     -   1333ms
0.8''  -   1081
0.6' ' -    964
0.5   -    809
0.4''  -    640
0.3''  -    616
1/4   -    549
1/5   -    460
1/6   -    438
1/8   -    389
1/10 -    360
1/13 -    344
1/15 -    324
1/20 -    320
1/25 -    312
1/30 -    292
1/40 -    284
1/50 -    281
1/60 -    276 After this exposure bumped up and started getting random numbers from 269+ms to 280+ so i stopped the test
1/80 -    282 yep

So exposure changes are not very linear as far as i can see (or maybe i am wrong?),and after 1/30 there is no point in closing the shutter further.
Plus there is a strange top-corner vignetting after 1/30 which seems it exists anyway in the all pictures in every exporuse although it isnt always very noticeable.

EDIT:Exp override was off, just read the previous post.



gcrook

Quote from: gcrook on July 22, 2014, 04:33:02 PM
Thank you i will test this.
Cheers.

2nd sloppy test with the 60D with the modules brapodam linked.
Exposure simulation, and override, (if the latter matters) ,was ON.
f1.4 iso 100 exposures starting from 5'' until 1/250 exported using mlv converter and imported into LR.
Bug?next shutter (ie to go from 1/50 to 1/60) wouldn't change unless i went 2 steps ahead and then 1 back,
(had to go from  1/50 to 1/80 and then go back to 1/60 for it to engage and show in the screen).Maybe exp override does that?
So:

5''     -  5686ms  f-stop reports incorrectly to LR (shutter reports 1.0s)
4''     -  4442      same as above
3.2''  -  3490      same as above
2.5''  -  2980      same as above
1.6''  -  1887      same as above
1.3''  -  1631      same as above
1''     -  1333      same as above
0.8''  -  1091      same as above
0.6' ' -   954       same as above
0.5   -   799       reports correctly as 0.5''
0.4''  -   680      ----reports as 0.5''---
0.3''  -   616     
1/4   -   539
1/5   -   460
1/6   -   438
1/8   -   399
1/10 -   360 After this step exposure changed very little and top corners vignette became more and more noticeable
1/13 -   354
1/15 -   324
1/20 -   310
1/25 -   301
1/30 -   302  ?
1/40 -   284
1/50 -   291 
1/60 -   276 
1/80 -   271
1/100-  270
1/125-  268
1/160-  266
1/200-  275
1/250-  264

Dont know if all these mean much, if there are other means to test this further and provide useful info, let me know.

A few random pics with shutter 1/3, 1/8, 1/30, 1/125.

http://postimg.org/image/3uxdr5bdv/
http://postimg.org/image/yb3ap9wwz/
http://postimg.org/image/t66ole9cz/
http://postimg.org/image/eygzwqwo3/

dsManning

A few tests on the 50D show different ms. Haven't gone through the whole range, but matched ISO100 and f/1.4

Shooting at 5s gave 5607ms
Shooting at 1/10s    310-312ms (this too seems to be the sweet spot for no darkness at top of frame, with least amount of motion)

a1ex

Shutter speed is set from a table, in scsDummyReadoutDone, first call right after NormalCapture; on 60D, the table is at 0xFF5B4DA8; it was first mentioned here.

These numbers are plugged into a routine which prints something about VSizeSetting, and does some math, which transforms the time value in units for FPS timer B.

On 60D, that table looks like this:


raw_shutter  printed_time   time_ms     fps_timer_B_units
      16        32"    =>   32292 ms    2131 + 132*5004
      17        32"    =>   30480 ms    5000 + 124*5004
      18        32"    =>   27951 ms    3176 + 114*5004
      19        25"    =>   25632 ms    655  + 105*5004
      20        20"    =>   23505 ms    2069 +  96*5004
      21        20"    =>   21555 ms    2110 +  88*5004
      22        20"    =>   19767 ms    471  +  81*5004
      23        20"    =>   18127 ms    1864 +  74*5004
      24        16"    =>   16623 ms    1045 +  68*5004
      25        16"    =>   15244 ms    2786 +  62*5004
      26        16"    =>   13980 ms    1885 +  57*5004
      27      12.5"    =>   12820 ms    3114 +  52*5004
      28        10"    =>   11757 ms    1331 +  48*5004
      29        10"    =>   10782 ms    1352 +  44*5004
      30        10"    =>    9888 ms    3032 +  40*5004
      31        10"    =>    9068 ms    1229 +  37*5004
      32         8"    =>    8316 ms    819  +  34*5004
      33         8"    =>    7626 ms    1680 +  31*5004
      34         8"    =>    6994 ms    3729 +  28*5004
      35         6"    =>    6414 ms    1844 +  26*5004
      36         6"    =>    5883 ms    963  +  24*5004
      37         5"    =>    5395 ms    963  +  22*5004
      38         5"    =>    4948 ms    1803 +  20*5004
      39         5"    =>    4538 ms    3401 +  18*5004
      40         4"    =>    4162 ms    696  +  17*5004
      41         4"    =>    3817 ms    3627 +  15*5004
      42         4"    =>    3501 ms    2151 +  14*5004
      43       3.2"    =>    3211 ms    1209 +  13*5004
      44         3"    =>    2946 ms    778  +  12*5004
      45       2.5"    =>    2702 ms    778  +  11*5004
      46       2.5"    =>    2478 ms    1188 +  10*5004
      47       2.5"    =>    2273 ms    1987 +   9*5004
      48         2"    =>    2085 ms    3135 +   8*5004
      49         2"    =>    1913 ms    4610 +   7*5004
      50         2"    =>    1755 ms    1372 +   7*5004
      51       1.6"    =>    1610 ms    3401 +   6*5004
      52       1.5"    =>    1477 ms    676  +   6*5004
      53       1.3"    =>    1355 ms    3176 +   5*5004
      54       1.3"    =>    1243 ms    881  +   5*5004
      55       1.3"    =>    1141 ms    3790 +   4*5004
      56         1"    =>    1047 ms    1864 +   4*5004
      57         1"    =>     961 ms    5102 +   3*5004
      58         1"    =>     882 ms    3483 +   3*5004
      59       0.8"    =>     809 ms    1987 +   3*5004
      60       0.7"    =>     743 ms    635  +   3*5004
      61       0.6"    =>     682 ms    4385 +   2*5004
      62       0.6"    =>     626 ms    3237 +   2*5004
      63       0.6"    =>     575 ms    2192 +   2*5004
      64       0.5"    =>     528 ms    1229 +   2*5004
      65       0.5"    =>     485 ms    5348 +   1*5004
      66       0.5"    =>     445 ms    4528 +   1*5004
      67       0.4"    =>     409 ms    3790 +   1*5004
      68       0.3"    =>     376 ms    3114 +   1*5004
      69       0.3"    =>     345 ms    2479 +   1*5004
      70       1/3     =>     317 ms    1905 +   1*5004
      71       1/3     =>     292 ms    1393 +   1*5004
      72       1/4     =>     268 ms    901  +   1*5004
      73       1/4     =>     247 ms    471  +   1*5004
      74       1/4     =>     227 ms    4651
      75       1/5     =>     209 ms    4282
      76       1/6     =>     192 ms    3934
      77       1/6     =>     177 ms    3627
      78       1/6     =>     163 ms    3340
      79       1/6     =>     150 ms    3073
      80       1/8     =>     138 ms    2827
      81       1/8     =>     128 ms    2622
      82       1/8     =>     118 ms    2418
      83      1/10     =>     109 ms    2233
      84      1/10     =>     100 ms    2049
      85      1/13     =>      93 ms    1905
      86      1/13     =>      86 ms    1762
      87      1/13     =>      79 ms    1618
      88      1/15     =>      73 ms    1495
      89      1/15     =>      68 ms    1393
      90      1/15     =>      63 ms    1290
      91      1/20     =>      59 ms    1209
      92      1/20     =>      54 ms    1106
      93      1/25     =>      51 ms    1045
      94      1/27     =>      47 ms    963
      95      1/28     =>      44 ms    901
      96      1/30     =>      41 ms    840
      97      1/35     =>      38 ms    778
      98      1/38     =>      36 ms    737
      99      1/40     =>      33 ms    676
     100      1/45     =>      31 ms    635
     101      1/50     =>      30 ms    614
     102      1/55     =>      28 ms    573
     103      1/58     =>      26 ms    532
     104      1/60     =>      25 ms    512
     105      1/70     =>      23 ms    471
     106      1/80     =>      22 ms    450
     107      1/80     =>      21 ms    430
     108      1/90     =>      20 ms    409
     109     1/100     =>      19 ms    389
     110     1/110     =>      18 ms    368
     111     1/120     =>      17 ms    348
     112     1/125     =>      17 ms    348
     113     1/140     =>      16 ms    327
     114     1/150     =>      15 ms    307
     115     1/160     =>      15 ms    307
     116     1/180     =>      14 ms    286
     117     1/200     =>      14 ms    286
     118     1/215     =>      13 ms    266
     119     1/235     =>      13 ms    266
     120     1/250     =>      13 ms    266
     121     1/280     =>      12 ms    245
     122     1/300     =>      12 ms    245
     123     1/320     =>      12 ms    245
     124     1/350     =>      11 ms    225
     125     1/400     =>      11 ms    225
     126     1/430     =>      11 ms    225
     127     1/470     =>      11 ms    225
     128     1/500     =>      10 ms    204
     129     1/560     =>      10 ms    204
     130     1/600     =>      10 ms    204
     131     1/640     =>      10 ms    204
     132     1/750     =>      10 ms    204
     133     1/800     =>      10 ms    204
     134     1/850     =>      10 ms    204
     135     1/900     =>      10 ms    204
     136    1/1000     =>      10 ms    204
     137    1/1100     =>      10 ms    204
     138    1/1200     =>      10 ms    204
     139    1/1250     =>      10 ms    204
     140    1/1500     =>      10 ms    204
     141    1/1600     =>      10 ms    204
     142    1/1700     =>      10 ms    204
     143    1/1900     =>      10 ms    204
     144    1/2000     =>      10 ms    204
     145    1/2300     =>      10 ms    204
     146    1/2400     =>      10 ms    204
     147    1/2500     =>      10 ms    204
     148    1/3000     =>      10 ms    204
     149    1/3200     =>      10 ms    204
     150    1/3500     =>      10 ms    204
     151    1/3800     =>      10 ms    204
     152    1/4000     =>      10 ms    204
     153    1/4500     =>      10 ms    204
     154    1/4800     =>      10 ms    204
     155    1/5000     =>      10 ms    204
     156    1/6000     =>      10 ms    204
     157    1/6400     =>      10 ms    204
     158    1/7200     =>      10 ms    204
     159    1/7800     =>      10 ms    204
     160    1/8000     =>      10 ms    204


These numbers are the exposure times for the very first image row (which is actually in the optical black border). The gradient effect was already discussed (if you have trouble understanding it, read this first), and the last number from this table should match the timer B setting before the final readout. On 5D2, the last number from the same table is 14ms = 224 timer B units, which is 0xdf + 1 and matches the previous results.

When comparing the exposure times from the table above, with the run times of the capture function (as posted by gcrook), keep in mind there are other delays in the process, not exactly known (the capture code is quite complex), and most cameras (except 500D and 1100D) use a dummy capture step (which adds a constant delay before the actual capture).

So, these exposure times are approximate and slightly larger than the actual exposure; the exact exposure time is done (or expected to be done) by mechanical shutter.

Reminder: these are some test results I'm looking for, to complete the timing analysis. The test script is not present in the pre-compiled builds, so it's for those of you who compiled ML yourselves.

Quote from: a1ex on July 14, 2014, 11:32:29 PM
take the two test images with this script, then get the ADTG screenshot by following these steps.

escho

600D full-res silent pics:

Manual mode and AutoISO gives an error:
ASSERT: 0
at Param\HivshdParam.c:115, task ShootCapture
lv:0 mode:3


Magic Lantern version : Nightly.2014Jul22.600D102
Mercurial changeset   : 4493f6e04e5f (fullres-silent-pics)
Built on 2014-07-22 19:10:51 UTC by edgar@linux-indh
Free Memory  : 232K + 947K.


Manual mode and fixed ISO: no error

Edgar
https://sternenkarten.com/
600D, 6D, openSUSE Tumbleweed

Audionut

Ok, lets get this thread back on track.    :D

This thread is for full resolution silent picture development discussion.  Before posting here, please stop to consider your post.  Will it help development?

For discussion about capturing silent pictures using raw video formats, please see this thread by dmilligan:  http://www.magiclantern.fm/forum/index.php?topic=12733.0

Want to request a build to help with development?  There is already a thread created here:  http://www.magiclantern.fm/forum/index.php?topic=12608.0
Note:  Other members who can compile builds are more then welcome to fill requests in that thread.


Thankyou all for your understanding.

marekk

what is the difference between battery consumption with full res silent pics timelapse and standard timelapse ?

a1ex

It takes about 10 minutes to find out. Look in Debug menu - once the percentage decreases by 2 units under a given workload, you can write down the estimated drain rate.

Of course, if you want better accuracy, you can do a few tests and average them - that will take some more time :P

a1ex

Everybody, please double-check the skip offsets in raw.c and post the confirmation screenshots. The ones from 600D are wrong.

Quote from: a1ex on July 01, 2014, 05:11:15 PM
download raw_diag.mo, select the "OB zones" option, and adjust the offsets until you get the following picture (note the cyan box being just inside the noisy borders):



Greg


escho

Quote from: a1ex on July 25, 2014, 03:36:55 PM
Everybody, please double-check the skip offsets in raw.c and post the confirmation screenshots. The ones from 600D are wrong.

600D, raw_diag.mo loading

tcc:error: undefinied symbol 'log2f'
failed to link modules


Edgar
https://sternenkarten.com/
600D, 6D, openSUSE Tumbleweed

a1ex

It's there, just checked the current nightly.

00cd1644 log2f

If it's not present in your custom build, you can add a call to it in builtin-enforcing.c.

mk11174

Quote from: a1ex on July 25, 2014, 03:36:55 PM
Everybody, please double-check the skip offsets in raw.c and post the confirmation screenshots. The ones from 600D are wrong.
I did post a screenshot the night I added the edmac and before you accepted it you said the skips were correct, so how are they wrong now???

Update: Just saw you commented me on bitbucket, but the screenshot was after edmac change because I got error if no change was made and tried full res silent picture.
500D/T1i  550D/T2i  600D/T3i  700D/T5i

a1ex


mk11174

Quote from: a1ex on July 25, 2014, 07:19:57 PM
http://www.magiclantern.fm/forum/index.php?topic=7139.msg123281#msg123281

Your screenshot was most likely done before the EDMAC changes.
Maybe something wrong in my source, I can try raw diag using the nightly and see what we get unless someone already got you one, but trust me, there was no way for me to get a screen before the edmac change, so I don't know what is wrong now?
500D/T1i  550D/T2i  600D/T3i  700D/T5i

a1ex

Hm, the raw_diag screenshot was directly on a silent picture? Can you upload a DNG?

edit: I get it now, you got the screenshot from LiveView, not from photo mode...

You can get the raw_diag screenshot from regular photo mode, outside LV. If you really want to get it from a silent picture, use the intervalometer, not half-shutter. In both cases, the offsets should be equal, because it's the same Canon code that writes the raw data.

mk11174

Quote from: a1ex on July 25, 2014, 07:35:22 PM
Hm, the raw_diag screenshot was directly on a silent picture? Can you upload a DNG?
Well, I turned on the fullres on the silent module, then I turned on Raw Diag with the zone option, then I did a half shutter press and held it for a second till I saw the Raw Diag text so I guess it was a silent picture it took cause the shutter did not activate.

Let me get her camera so I can do more tests with the tool and send you the DNG too.

Crap, just remembered the fullres is not in nightly yet.

So Ill create a new local repo and build and try, but do you want one with nightly too without fullres?
500D/T1i  550D/T2i  600D/T3i  700D/T5i

mk11174

Quote from: a1ex on July 25, 2014, 07:35:22 PM
Hm, the raw_diag screenshot was directly on a silent picture? Can you upload a DNG?

edit: I get it now, you got the screenshot from LiveView, not from photo mode...
Ok, so just tell me exactly what you want and Ill grab it for you, I have camera now.

Gotcha, sorry, I thought it had to be a regular silent picture, I better check my 700d now to then!

Here is the screen from photomode and a regular photo take without silent picture turned on.

Looks real off doesnt it?

So any idea what the offsets should be now to start with for correction?

500D/T1i  550D/T2i  600D/T3i  700D/T5i

a1ex

Look for CONFIG_600D in raw.c, and the last block has the offsets from photo mode. You can start from the values I've linked above, and maybe fine-tune a few pixels from there.