Messages

Ted,

This is a very interesting and important Thread/Test!  Thanks for getting this going!

Have you tried this test with a medium focal length prime lens (like a 50mm or 100mm)?  I'm one of those cinematographers that believe primes are inherently sharper and produce less distortion than zoom lenses (not always true nowadays!).

In the test shot, what's the focal length?

What f-stop did you shoot at?

I'm going to set up my new Canon AF prime lenses and my old Nikon MF primes and run the same test criteria you've done here.

BTW, 2048x930 is 2.20:1 aspect ratio.  It takes me back to my days of working in 5-perf 65mm film (Super Panavision 70).  Gotta try it myself and get nostalgic!

Well RED MX, and all native 800 iso (like alexa, bmcc, c300) sensors tend to lose DR in highlights when changing iso to lower than 800 and in shadows in higher than 800.
Then you get not so nice rolloff in highlights and really messy noise in shadows. So in bright daylight you have to use heavy IR ND's.
All filmmakers want high native iso, because with lower iso rated sensors you need lots of lights on set to get good base exposure and then even more for lighting the actors to get contrast and separation from the background.

Now indoors with 800 iso (or ASA I should say) you can use more natural lighting without introducing lots of noise to the image, and even "light" actors faces with bounce boards etc.
So once again,  higher iso rated sensor = more money saved for producers, and low budget filmmakers.
7d sensor is 100-200 native so outdoors its all great, but indoors it can get messy. Even film stocks got to the point when you use 200 for outdoors and 500 for indoors.

All your points are valid.  But having spent 20+ years shooting 35mm and 65mm outdoors, I always liked to use whatever the current 50 ISO Daylight stock was available from Kodak (be it EXR, Vision I, II, etc.).  Typically it was the very best emulsion they made (i.e. finest grain, widest latitiude).  I also found it surprisingly forgiving in the shadows and good when pushed.   I don't know about using 200 ISO film out in the desert without again using a bazillion stops of ND with an 85 filter.

However, I would like to make one point without sounding argumentative, using a slower sensor in a 7D ML RAW camera (that I own) and having a slightly larger lighting package is way cheaper than purchasing or renting a RED or ALEXA or Panavision Genesis and using less lights.  Is that a fair conclusion?

Back to 7D ML RAW development, is there any word (or advisable procedure) for using an external HDMI monitor when shooting RAW?

[below]

I see in some photos using red cameras almost with iso 800, even if is sun/bright  ...i think should be their native iso and maybe they prefer to use filters or close the iris.

As a previous RED owner, the Original "M" sensor was Native 320 ISO.  The "MX" Sensor is Native 800 ISO. 

RED cameras do not like you to change ISO below the Native rating.  I can attest that the image starts getting crunched and less "filmic."  I hate having to put 13 stops of ND and a Hot Mirror (required for all RED shots) out there just to take a daylight shot at f4 (no joke!).  A serious flaw in the camera's design IMHO.

My experience so far with the 7D in RAW, the basic quality of the image doesn't seem to get wrecked if you vary the ISO up or down (within reason, of course).  I've done critical tests between 100 ISO and 3200 ISO and find staying within 100 ISO to 320 ISO is great.

My comments on several on-going issues that keep coming up:

Color Fringing:  I too have seen this ONLY when using DaVinci Resolve.  I believe 1% figures it's a debayering issue in Resolve.  Makes sense.  I do not see it using ACR/After Effects or Cineform/GoPro Studio.

Moire Issues:  The only "solution" to a line skipped sensor is to : A, not line skip aka crop mode or B, use the Mosaic Engineering VAF filter.  As I went over in this thread, I use the VAF filter and it works great and does not degrade the image.

The "Banding" Issue:  I only see these bands when the image ISO is pushed beyond 400 ISO.  All camera systems have limits and/or things you must avoid...Unfortunately for 7D owners, this may be one of them.

Absence of Motion "Blur":  This one is a puzzler, maybe I'm misunderstanding.  All the RAW footage I've shot has a fair amount of "Motion Blur" and not as much "Rolling Shutter" as it seems to have in h.264.  Maybe just another limitation we'll have to deal with.

Maybe, just maybe if we're real patient and lucky, the ML software gurus will solve all of this.  I'm confident.

[/img]
I am using the same 64GB 1000x Card and bitrom's new build and my write speeds are significantly slower.

Do you think I got a crap card or does this have to do with different versions of the 7d I bought mine new in 2009.
Other possibility: I'm doing it wrong.




Thanks for helping!

@Slink

I had the same problem at first--My speeds looked similar.  When the camera is in MOVIE Mode or PHOTO mode, speeds are WAY SLOWER.

Try this:

Turn Global Draw OFF.
Don't load MODULES
Switch to PHOTO mode on 7D.
Press the INFO Button until the screen is blank,

Go into ML and Run the Benchmark.

On your Benchmark Info Screen, Near the top it should now say MODE: playback, Global Draw OFF.


Hope this helps.

@Britom

New build seems to work great!  Using Komputerbay 64GB 1000x card, the first bar when recording RAW barely flickers.  I prefer the RAW instead of MLV (for now) because of one too many steps.

My card speed improved slightly, but continues to be very good.



For ACR users:  I'm not seeing any of the "flickering issues" that others report when using the Highlights Recovery slider.  I'm going to update my DaVinci Resolve and test my footage using it.  So far I've used 2 methods of conversion to AVI:  1. ACR grading then to AE render to Cineform 444 1080p23.976 and  2. Cineform converter to 444 1080p23.976 and grading using Metadata thru GoPro Cineform Studio Premium.  So far #1 produces better results in a shorter timeframe.

Question for anyone who might know these details:

For some RAW shots I make at different ISO's, when I make the .DNG's using Raw2Dng displays this message while rendering each frame:

"Interlace pattern not supported.  ISO Pattern: BdBd"

Funny thing is, the .DNGs are fine and the shots come out normal looking.  Odd???  Am I doing something wrong?  Are there ISO's we shouldn't use?

I am experimenting with 7D raw as well. Do you have the same problems with vertical lines when shooting in low-iso performance?

I've seen this in my experiments with High ISO.  Not unusual, the RED ONE does the exact same thing!

Just ran Britom's new modules.  Live View is choppy, doesn't freeze, but my buffers immediately fill and after about 100 frames (at 1472x626) it stops and alerts me to "corruption in frames"  So I went back to the previous version and everything is as was before (I use the "greenscreen" checker for monitoring--great when shooting in low light).

Too bad it didn't work, but I definitely think you're on the right track!  Keep at that code...it's in there somewhere!

Man, I'd love to have a 5D3, but then my Tokina 11-16 2.8 would be useless...

I've been with CineForm from nearly the beginning (I had a JVC GYHD100 and the only way you could transcode the files to PC was with CineForm).  It's a great codec and a great product with patient, competent support folks.  The 444 CineForm avi is 12-bit, so there's only a little loss in going from a 14-bit original (a huge PLUS).  The 422 CineForm avi is 10-bit and in most instances it's indistinguishable from the original.  File sizes are very good and you can choose from 5 quality settings (Low, Medium, High, FilmScan 1 and FilmScan 2) I tend to use FilmScan 1, the perfect balance of pristine quality and manageable file size. (sorry, sounds like a sales pitch, trust me, I'm not on their payroll, just a very pleased customer!)

With the VAF, yes it's true you can't get something for nothing, if the lines are skipped, they're skipped,  but it really doesn't appear to have a "blurred" quality to me.  Overall, very little trade-off, way more up-side.  BTW, In Canon's native 1080p24 H.264 format, you can resolve about 760-800 lines of resolution.  In ML RAW with a smaller frame 1472x626 I'm getting at least that performance already....Interesting.

Shot some more today.  I just wish the Live View wouldn't freeze when you start recording, probably an easy bug to work out.  The image is just plain VIVID compared to the Red.  Red One images were very dull and lifeless out of the camera.  You really had to finesse the life back into a Red shot.  And let's be honest here, the Red has some serious aliasing issues of it's own.  A moving/panning chain-link fence shot in 4K, downsampled to 1080p with Redcine-X was NASTY looking, distracting and unnatural.

I'm not positive about this yet, but everything I've shot and seen in the 7D ML RAW format seems to have less rolling-shutter "tearing" than the H.264 7D footage.  More testing by myself and others needed though!  It could be that it's exactly the same, but the H.264 compression makes those rolling-shutter artifacts more pronounced.  Either way it's a potential plus.  Anyone noticed this on the 5D3 or other cameras using ML RAW?

Working on getting some of these tests up.  I just hate truncating them to H.264 for Vimeo or YouTube, doesn't really do a ML RAW shot justice.  I'd be happy to send interested ML'ers the CineForm files, but they're 1GB+ so that's not realistic.

Nice!! Was it shot at 2.5k?

I have a question.. WHY do we have a lot of alisasing and moire on 7D, same as 5D2 and others... I mean, the 7D when shooting H264 does NOT have that HUGE amount of moire.. it just has a tiny biy, which makes it completely usable!! But the raw.. it just makes the footage not usable at all if you do a wide-shot..

I'm thinking.. if it makes the same extact line-skipping to get to 1920x1080 as the H264 does... then, why?? And, it could be possible to fix it right??

Thanks!!

I did my first RAW tests with my 7D over the weekend and let me tell you as a Former Red One Owner, the 7D sensor is miles ahead of the original Red M Sensor.  I shot using the 2.35:1 - 1472x626 (which is slightly more total pixels than traditional 720p).  It blew my mind.

I rendered out these images from AE CS6 to CineForm 444 1080p24 avi's and compared to similarly transcoded Red files.  Here's the Truth:  The Canon 7D footage is deeper, richer and definitely more Film-like than the Red One (and I don't say this lightly: I'm one of those guys that shot in 65mm Super Panavision and did own the Red).  In the end, the Red One is too bulky, gets WAY TOO HOT and the RAW images are just way too...RAW (takes too long to grade a Red Image).  And with Magic Lantern RAW for the 7D, the Red is WAY TOO EXPENSIVE.  I can't believe how little tweaking the 7D RAW footage needed to look so....RICH!

Magic Lantern is really doing something truly special here.  My hat's off to all involved.  This really does change the game.  These great software guru's have proved to me that at this level Bit depth becomes truly more relevant than pixels.

Man, I wish I could write code!  I know the 7D is a struggle because of it's hardware architecture, but let's hope the ML folks stay at it!  It's definitely worth it!

BTW, I use the VAF filter on my 7D and there's ZERO ALIASING.  Get a VAF filter and forget about it.  You won't regret it.  It's a wondrous optical device.

Thanks for the words of advice.  Got it up and running with my 16gb 500x.

I think my Komputerbay 1000x 64gb may be one of the "duds."  When I benchmarked it, the Write speeds were 45 to 61 MB/sec and the Read speeds were 41 to 67 MB/s.  Seems to be WAY OFF what others here get with the same CF card!

Is it possible to run the dual iso, raw burst and raw modules on the 7D without the "risky" autoboot?  Maybe I'm just plain stupid, but I've read these 7D threads up and down, followed various instructions and can't get the "Modules Tab" to appear. 

And I can't seem to get (or find) that nice Benchmark Test that everyone posts their results with (mine only does two or three lines and that's it! FYI using a Komputerbay 1000x 64gb).

Am I missing something here...?  Pathetic, I know, but if someone could help, I would be eternally grateful!  Love ML and my 7D, just can't seem to get the latest and greatest up and running!

Here's what I've found.

With the ML booted, 1080p24 video with the Marshall monitor connected through HDMI, the problem occurs on BOTH Live View and Record.  Pan the camera fast, back and forth and there's a periodic "hitch" or "freeze."

When I turn OFF "GLOBAL DRAW," everything works Normal.

So I re-enabled GLOBAL DRAW and started turning off items in the OVERLAY menu.  What I've found is that if you turn OFF HISTOGRAM and WAVEFORM, the problem goes away totally.

If you enable HISTOGRAM, it works longer between hitches.

If the MAGIC ZOOM is red dotted (can't use under 1080i), it makes the problem worse.  Disabling it (no dot) makes things a little better.

AUDIO METERS and the SPOTMETER have no effect, they are fine enabled or disabled.

@chroma:
np.
can you do this please? thanks.

Ran 2 different Benchmarks.  First, clean ML boot with No Monitor Attached:
https://skydrive.live.com/redir?resid=E642B6CE19214075!132&authkey=!AOZD8t2Pd1dMjkA

2nd, clean ML boot with the Marshall Monitor Attached:
https://skydrive.live.com/redir?resid=E642B6CE19214075!133&authkey=!AItmv5ki8o_HASQ

Praying you can download these files!

I've had such reports before, and so far, in all cases, the computer / camera were not fast enough to play the video.

Not wanting to step on any toes here!

The PC I use for video editing is fast enough to play back RED 4k files at 24fps, and the 7D video files are copied from the CF card to the PC's hard drive.  I convert to CineForm DI for editing.  I've also gone frame by frame and you can see the jump (in both mp4 orginals and CF avi DIs).

The "dropped frames" situation only seems to occur when the camera is in ML w/ the Marshall monitor hooked through the HDMI...also the 7D is in Manual mode when I shoot video.

New Member, First Post.

Thanks for developing the 7D ML software.  As a former RED ONE owner/operator, I'm very used to histograms and RGB overexposure indicators...So ML is GREAT!

I am noticing a strange behavior with my 7D using the new ML alpha, so I wanted to run it by you.

When I record a video 24p1080 with Canon 2.0.3 firmware, video file looks normal.

When I record a video 24p1080 with Canon 2.0.3 firmware with Marshall 5" Monitor connected via HDMI, video file looks normal.

When I record a video 24p1080 with ML firmware running, video file looks normal.

When I record a video 24p1080 with ML fimware running with Marshall 5" Monitor connected via HDMI, the video file recorded to the CF card has a periodic "stutter" or "dropped frames" throughout.

The video file shows the "dropped frames" when played back by the 7D, through the external HDMI or if the video file is copied and played back through a computer.

I am wondering if there's some kind of data "bottleneck" under those 3 specific conditions:  1, running  ML, 2, using external HDMI, 3, recording video.  Could this be happening because the 7D ML "alpha" is a CF card driven firmware, not flashed or installed?  Sorry if this is a bone-headed hypothesis!