Messages

Yeah, it's clean in photoshop as well. Maybe it depends on what build of raw2dng I used at the time of posting that? I honestly can't remember whether it was one of your earlier builds or a.d.'s...

Yeah, the image I just posted was from the fresh build you linked. The stripes appear when opening the DNG with Adobe Camera Raw, but Raw Therapee (what I used for the image above) seems to read it fine.

The one from photoshop (no color corrections obviously): http://dl.dropboxusercontent.com/u/63799907/000350-3.jpg

update: Hang on, I think there's still stripes on the Raw Therapee version too. I was using the CLI earlier and might have used the wrong directory by accident, gonna doublecheck (edit: yeah, it turns out to have stripes too). The file opened with RT's GUI has stripes too. Sorry for the confusion.

a1ex: The stripes still appear under Adobe's camera raw tool, but Raw Therapee's output (what I've been using regularly) looks just fine to me: http://dl.dropboxusercontent.com/u/63799907/000350-2.jpg

@seanturco I know, but if a more elegant solution exists, I'd prefer to use that.
@a.d. Thanks! I can confirm that the May 31st build looks good.

I'm noticing vertical stripes in some test footage I just shot. Both the a.d. and ML bitbucket builds of raw2dng.exe (that was just updated) exhibit this anomaly. Unfortunately, I didn't keep the previous version around to see whether it's a problem with Raw2Dng, the latest ML build (6397e112ed41), or something else. Raw2Dng seems to be making a lot of corrections now whereas a lot of values were at 1.000 in the older version.

edit: JPEG for a quick look: http://dl.dropboxusercontent.com/u/63799907/000350.jpg
DNG: http://dl.dropboxusercontent.com/u/63799907/000350.dng

Hmmm, yeah I'm kinda stumped. Just updated to the latest build by a.d. (94190a95bbaf), recorded 6104 frames, merged them via the command line and now I'm currently extracting them. Both the raw2dng from the main ML bitbucket and the build from a.d. works. The a.d. build is more recent though and seems to do more with the vertical stripes correction (lots of the corrections are something other than 1.000).

Yeah, it may be a problem with the version of Raw2Dng. The command line you posted looks perfectly fine.

If you're using Windows, you can use the command line to merge the files without having to rename anything really. For Vista/7/8, hit the windows key to open up the search bar, type "cmd" (without quotes) and hit Enter to open up the command prompt.

Type the following to merge the files WITH quotes:

copy /b "C:\...\blah.RAW"+"C:\...\blah.R00"+"C:\...\blah.R01"+...+"C:\...\blah.RXX" "C:\...\blah-merged.RAW"

A couple slightly more advanced tips:

1) Open up a notepad file to type out the entire command first and then paste it into the command prompt. Press the windows key, type "notepad" and hit enter to open it up. Once you have everything written, hit Ctrl-A to select all the text, Ctrl-C to copy the text, then paste it into the command prompt window (AFAIK, you can only paste it by right-clicking the window and clicking "Paste").

2) You can quickly copy the path of each file by holding Shift and right-clicking on the file. A new option called "Copy as path" will appear. Click it and paste the path in notepad with Ctrl-V. It'll have quotes too.

edit: Oh, I forgot that some file extensions may be hidden by default and you may not see them unless you show them. The WinRAR files have the extension .R00, .R01, .R02, etc. The first file in the sequence is .RAW followed by .R00, .R01 etc.

The scripts dir comes from the regular nightly builds. Once you load the RAW recording module, the option to record RAW should appear last on the "Movie" tab.

edit: whoa, what's up with the post counts?

Thanks for the test.  I've used RawTherapee on Mac with some success -- it's free / open source so you can give that a try.  The nice thing is that is has numerous command line options, so if you're familiar with scripting (or willing to learn) you can setup a workflow that uses RawTherapee automatically without even opening the program.  Of course these will be automated settings so you will not have as much control as if you opened the app and made manual changes to each clip.

I've also been using Raw Therapee for Windows and the CLI is virtually a must for large scale conversions lol. The GUI is so unstable when dealing with thousands of images, I ended up learning how to avoid certain ridiculous crashes by using hotkeys and not touching the batch tab. It seems to crop 4 pixels on each side of the image for some reason too and I haven't been able to find any information on why that is. The great thing about it though is that it can output the raw with no color profile at 1.0 gamma, which is good for it8 calibration.

The file splitting implementation on the latest build corrupts frames after about 1600 or so, presumably where the split occurs.

Update: The latest build (from commit 9eaa84d...) by a.d. seems to have fixed the issue. I recorded till my 16 GB card maxed and it looks good all the way through.  Well, except for the last few dozen alternating frames, but it's always done that. Thank you!

Hm, I was not entirely following the 5D markII thread, what exactly is causing the write speed drop while using LV?

Just thought if LV mode is limiting the write speeds, maybe it will be also worth making a RAW video mode not involving the LV at all, for videos with static camera position that could mean a higher res possibilities.

I think a1ex implemented that at one point, but I don't think it made a difference in the end. No clue what's limiting the 5D2, maybe it's just the hardware limit, but we have been seeing gradual increases throughout the weeks. I've gained about +4-6 MB/s of write speed compared to when I first started playing with this, with the same card.

What kind of corruption?

Lots of pink noise on some sections, noise/discoloration and vertical lines in others. I'll get some examples up in a bit. Gotta re-join/extract the frames again.

The file splitting implementation on the latest build corrupts frames after about 1600 or so, presumably where the split occurs. edit: oops, meant to post this in the 5D2 builds thread.

Can a video camera ever really resolve "true" 1080p? I recall there being a theorem (Nyquist something or other) that states that when converting things to digital, you need to acquire the source at twice the target resolution, then downscale it to maximize the quality at the desired resolution. This happens with things like flatbed scanners too. Scanning at 600 dpi looks worse than scanning at 1200 dpi and downscaling it to 600 dpi. Resizing algorithms can preserve the detail that was impossible to acquire at the native target resolution.

Can this be right? Transcend should be capable of around 100Mb/s? Has the card been through low level format using the Transcend format programm? I heard it may rise the write speed.

Though the 5D2 can't reach those numbers, this benchmark is still lower than what it's capable of. The highest values I've seen so far were in the high 60s, almost 70 MB/s. Heck, with this new build, my 600x Sandisk card can record at a stable 59.5+ MB/s. Even though it's just below the speed required, I can finally do 1880x800 continuously at least until my 16GB card fills up (love the huge buffer we have )

lol @ the pissing ape

Great video. It seems to be one of the few I've seen that haven't been mangled *after* being captured in post. Lots of high frequency detail preserved, particularly noticeable on the animal textures (fur/skin) and the girl's striped cap (your daughter I presume?) It really shows the difference between the 5D3 and 5D2 too, even though my jaw still constantly drops by how good the 5D2 looks in RAW.

Well, it's been about 2 weeks since I started fooling around with RAW and I'm still constantly amazed by the quality. My brain keeps telling me "this shady region should be clipped to pitch blackness so don't expect any details" but I still see everything and get surprised lol.

RAW is not just about the sheer quality and resolution increase either. I've been fooling around with other technical stuff, like IT8 targeting for increased color accuracy which is more flexible with RAW data. I still have a reflective target used for flatbed scanners. so it's not optimal and a hassle to avoid reflections. However, the results have been pretty darn good I think the colors are fairly representative with what i see with my eyes. I've also learned about color-managed video. All my frames are converted to Rec.709 (using this profile: http://www.color.org/rec709.xalter) instead of sRGB and encoded with the Rec.709 flags on in x264. Playback in MPC-HC with color management on makes everything look exactly as it should (at least, I think it does). Fun times with the new opportunities RAW gives us!

I wanted to test how well this worked on the 5D2, and as others have reported 1880x840 is maximum. I am able to record from 200-360 frames @23.976fps, and then it stops.

So the 5D2 skips frames at 1880x840 even though the benchmark results are well above the write speed requirements? 69 MB/s should be plenty... Were you recording with global draw on at the time? If so, did you try recording without it?

Thanks for the results btw, I lets us 5D2 owners get an idea of what our limits are and what to look out for in terms of future upgrades.

Very nice indeed! 1880x800 lasted almost 2200 frames and 5.37 GB on my Sandisk 16GB 600x card (averaged above 58 MB/s which is a little better than it was before). I haven't checked out the frames yet though. Great job and thanks!

Just recorded a 14.8 GB clip at (almost) the usual stable resolution, which maxed out the card. Gonna see how continuous recording looks, though it'll take a while lol.

Just to clarify, full frame is 3:2 ratio and anamorphic is a 4:3 ratio (unstretched), right? You can record at something like 1440x1080 (the 5D2 can barely handle that) and stretch it to 2538x1080. That's well over that full HD can handle (1920x1080) so it even gives you a bit of room to downsample to 1920x817 or some mod2 resolution.

Full frame video resolution is 1880x1250 for the 5D2. Kinda funky, but it was designed this way. Unless we can get 94+ MB/s of write speed, it's kind of a pipe dream to get full frame resolution continuously, which in turn would've provided even greater resolution at 2.35:1 with an anamorphic lens (effectively 2938x1250 resolution when stretched horizontally). I'd be happy enough with just ~1880x800 which the 5D2 seems just barely able to pull.

Yeah, I think it's related to this change from a while ago:

ed94a01ee362 | alex: raw_rec: added live raw preview for x5 zoom mode (so you can frame in crop mode); added option for panning in menu (dolly mode)

I think that BW image is a simulation of the coverage of the raw recording feature. Unlike in full frame mode, where you can just show a white box to indicate the crop areas, the zoom mode by itself can't show you the full extent of your crop zone so it has to show a distorted BW image to illustrate the full extent of your frame.

For me, the audio loses sync continuously and maybe not even smoothly. It starts off close to sync, but gradually shifts out of sync. I don't know if it's the fault of the program I use (Avisynth with the ffms2 plugin), but audio remained in sync with the same program in the "vanilla" recording mode.