Messages

This new build (1222c61e8120) gets my write speeds up to 66.*MB/s, where as the last (d5a606e8c616) got me 63.*MB/s. At 1728x972 (25fps) thats got me up from around 700 to around 1200 frames before it hit's the wall.

Nice one chaps!

If its any help, vertical stripes are sometimes visible in cr2 files, especially if their pushed in post. Not sure if its the same thing.

It would be great if the .app included the original raw filename in the dng filenames, now we have usable names out of camera. This would make it easier to track footage from raw, through dng, to whatever offline codec, then back to the dngs for online.

If you use the 'New Comp from Selection' command in AE to produce compositions from image sequences, it will use the part of the filename common to all images in the sequence to name the composition. This would give you a relatively smooth process.

Drag a load of RAWs into raw2dng.app > use AE's Smart Import script to import the resulting folder full of dng sequences > do a quick one lamp grade in ACR as they import > select all the imported footage and right click for New Comp from Selection > select all the resulting comps and Add to Render Queue > If you have the right preset as default hit render

Not the prettiest workflow but manageable.

[In the zip folder is the command line:]

Does this fixed OSX user problem?
5D2_OSX_raw2dng.zip

In the zip folder is the command line:

1 cd destination_of_the_folder (drag&drop)
Terminal support drag and drop => don't need to type the path of the folder

2 /path_to_app/raw2dng /path_to_raw/JXX.RAW DNG_Name

Yep, works for me. Converted a 3GB and a couple of 8GB files no worries. So I'm now looking at a load of shaky, out of focus footage of kittens... Hmm. Nice tones though.

And the .app gui seems to work now too. So it calls the raw2dng command from /usr/bin?

@a.d.

Thanks fella. Likewise, loving your work. Your footer fix saved that little film, which in turn helped sell the location to the bbc, so ta!

I'll try this later cheers. We just got two kittens, so this shit just got less interesting... Ha ha

Thing move quickly on here, no flies on you chaps.

I had a shoot over the weekend, tides and moons for the bbc natural history unit. Mainly timelapse but they wanted some video too, and I foolishly decided to shoot ML RAW...

It was a little hairy at times, we lost one card full of video, totally corrupt, but that probably had nothing to do with ML. The results are pretty impressive. I've posted a few examples if anyones interested. It has a gritty quality that I'm not sure the NHU will appreciate but I really like, and the performance in some pretty extreme low light conditions was superb. Not sure I'd have got these images on anything else, certainly not on one of their varicams.



http://www.lightandtime.co.uk/media/redrocks/000287.dng



http://www.lightandtime.co.uk/media/redrocks/000000.dng



http://www.lightandtime.co.uk/media/redrocks/000105.dng

So thanks again to y'all, if they use any of the footage I'll donate my fee.

Particular thanks to a.d. who's builds I've been using all along. The new one is blisteringly fast for me but like others on here I am having problems converting longer clips. Getting the old 'aint lv_rec' error, seemingly for anything over 2GB. Been testing at 1880x800, will try more later if needs be, but I'm guessing I'm just being stupid and missing something. Updated raw2dng.app to 0.10. This runs it's own code right, it doesn't call the command line tool? Am I being a twat?

Is it true that 5dmk3 doesn't have that moire/aliasing issues 5dmk2 has?

The 5D3 may be slightly better but both cameras are going to suffer badly with aliasing, and no filter will really help without decimating the resolution. You're only recording from about 1 in 9 of the pixels on the chip. Thus almost 90% of the image is falling down the holes in between, and worse still the R,G and B source pixels being used for each output pixel will be some considerable distance apart.

So details will drop in and out of existence as they move across the sensor. Monochromatic edges will have serious colour fringing and bright details will cycle through the rainbow as they move. All very familiar to anyone who's used scientific high speed cameras with no low pass filter. That said, if you deal with this in post on a shot by shot basis, then you must be able to do a better job than Canon do in camera. ACRs colour noise reduction handles the colour artefacts remarkably well and I believe the moire reduction is pretty impressive for stills.

But this is all throwing some light on why DSLR video generally looks so terrible. Canon have really done a pretty amazing job getting rid of most of the colour fringing and detail flickering, in camera, in real time. Just wish they gave us the choice to turn that 'feature' off!

Heres some 5D2 test footage, shot in the wild at 1880x720 with the May 16th build.

Thanks to a.d. for the footer fix, and the new build. I shall give it a whirl.

https://vimeo.com/66442959

In fact I am getting the whole clips back with this fix, its just they have a load of extra garbage added on the end.

your last line
from

Code Select Expand

10 27 00 00 FB 19 00 00 10 27 00 00 A8 D1 FB FF
to

Code Select Expand

10 27 00 00 FB 19 00 00 10 27 00 00 57 81 FC FF
Update
The footer (the end part of the xy.RAW):

Code Select Expand

52 41 57 4d 58 07 d0 02 20 25 24 00 95 00 00 00
01 00 00 00 12 75 00 00 52 41 57 20 00 00 f3 5b
01 00 00 00 a4 28 3e 08 f4 04 00 00 f8 07 00 00
f2 0d 00 00 a8 12 45 00 0e 00 00 00 00 07 00 00
98 3a 00 00 00 00 00 00 00 00 00 00 58 07 00 00
e2 04 00 00 12 00 00 00 a0 00 00 00 f4 04 00 00
f8 07 00 00 00 00 00 00 00 00 00 00 00 01 01 02
01 00 00 00 6c 12 00 00 10 27 00 00 5b 02 00 00
10 27 00 00 c2 fc ff ff 10 27 00 00 8a e1 ff ff
10 27 00 00 72 3c 00 00 10 27 00 00 b0 09 00 00
10 27 00 00 28 fa ff ff 10 27 00 00 91 07 00 00
10 27 00 00 fb 19 00 00 10 27 00 00 57 81 fc ff

Replacing the whole 12 line footer as above seems to work as far as raw2dng is concerned. I've tried two of the previously broken files and they now convert and open in ACR. Nice one a.d.!

Not totally out of the woods though. Both of the files I've tried are fine for the first 4 secs or so, from then on every second frame is corrupt.

I'll try a few more.

Would love to test the new build of raw_rec but compiling is beyond me I think. Anyone got a binary for the 5D2?

API version mismatch is corrupted file (or worse, corrupted memory in the camera). I did not experience this, on what camera are you having trouble, and what changeset? (look in help menu)

5D2 with change set: 9c3ddbf5407

Files I've shot display a range of problems. Some fail in raw2dng with 'not lv_rec' or 'API Mismatch' errors. Some convert to DNG but then fail to open in ACR. A few open but come in bright green or pink. About 10% give me a usable image.

Make that about 10% success. Of the files that did convert to DNG, only four actually opened in AE, and one of these was bright green. So three out of thirty, so far...

Shame, it was a pretty sunset. Hope there's something salvageable in those files.

Getting the same errors as above, API mismatch or not lv_rec, from most clips. But some convert fine. 5D2 files at 1880x720, trying the conversion on Mac and Windows with about 30% success.

The Mac GUI seems a little fussy about file path. It likes obvious places, like desktop, movies, etc. Pop 'em deep in the directory structure and it says no such path.

Also, over half my clips wont convert, either on Mac or on Windows. I get either:

Error: This ain't a lv_rec RAW file
or
Error: API version mismatch: -2147483629

Seems random. All these clips are 1880x720 shot on the 5DmkII. Sometimes it's the big ones that work so doesn't seem to be about size.

Any ideas?

For those who like to put numbers on things...

I've put up an MTF resolution graph, ML RAW vs Canon H.264. Done on a 5DmkII.

There's jpegs of the test charts too.

http://blog.lightandtime.co.uk/post/50578814341/mtf-magic-lantern-raw-vs-canon-h-264-on-5d-mkii

Not an insignificant difference!