Click here to go to the first RED TEAM post in this thread.   Thread: Red One genlock tests for 3D....

Reply to Thread
Page 1 of 10 12345 ... LastLast
Results 1 to 10 of 96
  1. #1 Red One genlock tests for 3D.... 
    Goal: To fully test weather the RED one will achieve a perfect lock at various frame rates/timebases.
    Main result: It never truly locks.
    In Reality: In most cases your eyes/brain will never perceive it. Which is why many have gone ahead and shot 3D with the reds with very little issues from all but the most trained eyes.

    If your still curious, read on.it might get boring.

    This all started sometime last year when I found myself involved in several stereoscopic Red one projects. One of them in involving lots of propellers and the other project had several high speed shots at different frame rates.

    I embarked on doing a thorough test of the cameras ability to lock.

    The first camera we were using for one of the projects was a Sony EX3. Which has a BNC genlock in port. We tested several methods of genlocking the cameras. Besides my visual sync test we had a Specsoft RAVE unit at our disposal. With help from specsoft we were able to use a command line utility that could actually measure the difference in phase/lock of the 2 cameras. It would display what line the sensor was scanning on one sensor and then the other and then give you the difference. It did this continuously. The program kinda looks like a continuous ping if your familiar with that. Very cool.

    2 out of the 3 tested methods resulted in PERFECT lock. 0 difference showing on a line by line analysis. Which was backed up by my visual tests. Thus proving my visual test method worked as a tool to accurately measure sync.

    My visual test setup. (sorry for lack of pics, Im on the road shooting)

    I setup a large household cooling fan. The blades were black, so I put 1 white gaff tape strip on one of the fan blades. I set the fan at its lowest speed. I also used a Deneke smart slate (with the digital LED readout) to record the current clips settings.

    Once recording, with this setup there was several visual ways of checking our frame lock of the cameras.

    #1 The famous slate clap.
    #2 At higher shutter speeds you could see the refresh of the LED redout of the slate. This turned out to be the most accurate method of checking sync. Since you would see half of a number in one eye and one quarter of that number in the other eye.(if they were out of sync that is)
    #3 The fan blade with the white strip. This provided a large visual representation of a fast moving object thru the frame and allow me to test how off the eyes might be if we shot high speed props on airplanes. Also I was able to learn how motion blur and shutter angle where hiding issues.

    Using the command line tool in the rave I established my "control" of my process. Which was the EX3s perfect lock and confirmation using every visual method I devised. (Congratulations Sony BTW.)

    Once I had this to verify my method and tests I embarked on testing the REDS.

    Cameras were placed side by side parallel, same setting, same lenses (red zooms) shooting the same tests aids I described above.

    One of the things I wanted to find out was when I placed the cameras in various high speed frame rates the genlock icon (small padlock on the screen) would get a little non-smoking type sign over it. Clearly indicating to me I had lost genlock. I wanted to find out if that was true. It did give me a good genlock icon at 60fps and 120fps BTW. But wanted to see how bad was the sync when I lost the sync according to the Icon. Basically if the icon was an accurate representation of genlock.

    For Genlock I used a AJA Gen10 with the switches at 1-2-7 ( also tried different switch positions giving me bad results so I stuck to that)

    So to sum my test parameters I tested,

    I ran the following tests @ 23.98
    24,48,60,80,105,120fps @ 1/48th shutter
    24,48,60,80,105,120fps @ 1/2000th shutter

    I ran the following tests @ 24
    24,48,60,80,105,120fps @ 1/48th shutter
    24,48,60,80,105,120fps @ 1/2000th shutter

    I re-ran those same tests parameters then using a Deneke TC/Sync gen device instead of the aja to see if that made a difference.

    No difference in results from AJA. (also AJA is much cheaper so I recommend the Gen10, BTW)

    At this point my results were already surprising. I discovered that technically speaking the RED One NEVER is truly genlock/phaselocked! Even when shooting at 24fps.

    To clarify, the slate clap will match at 24fps. But a more detailed analysis at a higher shutter speed reveals there is a clear difference in images captured between the eyes! This is when analyzing fast moving numbers of the LED TC display on the smart slate.

    For example. The slate clap would match but you will see half of a digital number 1 on the slate in one eye and a full number one in the other eye. Clearly demonstrating that each sensor is capturing images a slightly different time. Naturally this is an issue for 3D acquisition!

    Furthermore, regardless of the padlock icon on screen there appears to be the same mismatch amongst all frame rates captured. So even with a valid genlock icon you will see a mismatch. Thereby making the icon irrelevant. So shoot at whatever frame rate you want since they are all off by the same amount.

    Having tested 2 different sync generators, multiple frame rates and time bases. I turned my attention to the phase adjustment of the camera. To see if adjusting the phasing would assist in bringing the cameras scanning of the frame closer in timing.

    Unfortunately the current phase adjustment at the time on build 21 was only 5 degree increments. Still I retested all the above listed parameters while adjusting the phasing of one camera -5, 0, +5.

    After a few hours of testing and analysis the results were disappointing. Any phase adjustment brought the cameras even father from locking.

    Fail.

    Being pressured from production to continue to look into this issue further I then drove down to RED HQ and had a chance to meet with and show Stuart English my test results on my laptop. First I wanted to know from him if my testing method was accurate and what his feelings were on the results..

    Stuart reviewed my tests and agreed the the cameras were not truly locked according to my results. He promised to modify the next build with 1 degree increments on the phase adjustment to see if that helped. Other than that he told me that the scarlet would have a special cable that would allow for TC lock and perfect genlock of the 2 cameras. Great for scarlet/epic owners but not so good for us RED one owners.

    So my thoughts after all this process?

    Well in most situations (24fps, 180 shutter) the viewer will never be able to detect any issues. Motion blur hides any difference at that point. Which is why you have never heard any complaints after watching any RED 3D content.

    The only time it * might * be noticeable is when shooting at higher framerates and narrower shutter angles. The narrower shutter angle reduces the motion blur allow your eyes/ brain to better record moving objects. Because of this when viewing 120fps material @ 24fps your eyes have more time to track objects moving thru the frame. Basically since you looking at the image longer your eyes will see the same object at a slightly different position during the same frame. Resulting in a bit of retinal rivalry and possible discomfort.

    Then again it happens so fast that most people still might not notice it. Naturally when shooting 3D you need everything to be as correct as possible and to me Genlock is vital and is the building block of anything 3D that comes after it.

    For me..step 1 get 2 cameras. Step 2 genlock them. Only then worry about a your rig etc

    I would love to fix this issue and be able to achieve the same results I get with other cameras.


    ..and yes, once the new build came out with the 1 degree increments of phase adjustment I did proceeded with a new round of testing.

    Unfortunately the results were not satisfying and did not address the issue. There continued to be a difference.

    I have held on to this information and posting this thread until the new build came out and I could test the new 1 degree increments. But now that it is out and it did not lock the cameras. I felt it was appropriate to finally bring this up to the community and possibly one again to REDs attention. See if anyone had any feedback. Especially considering that with the new MX sensor I believe RED one's will continue to see 3D duty for years to come.

    Anyone have any ideas? Might be a small annoyance to some but as I continue to shoot 3D RED productions one after another I just cant reconcile the fact that I still do not have a perfect lock.

    I want better.

    If sony and others can do it, Im sure RED can do it.

    If RED is serious about 3D then need to address this issue. Perfectly locked cameras are just mandatory for 3D if you ask me. But since its not that perceptible weve been skating by without it.

    So RED and others? What you think?
    ---------------------------------
    Pedro Guimaraes, SOC
    IATSE Local 600 - Operator

    Steadicam / Stereographer
    C-Motion Cvolution Package 3D-8 motor system for rent.
    Handsfree Segway for rent
    www.pgfilms.tv - Bio&Reel
    www.steadi3d.com - BLog
    www.cinemilled.com - DJI Ronin & other cinema Products
    Reply With Quote  
     

  2. #2  
    Senior Member mikeburton's Avatar
    Join Date
    Sep 2007
    Location
    Sacramento CA
    Posts
    1,673
    Hey Pedro,

    I thought I was the only one having trouble with the Genlock on the RED's for 3D acquisition. I just recently was 1st AC / DIT on my first RED 3D job with the ET Rig. We were using the Transvideo monitor for 50/50 overlay and that rely's on the Genlock. It was such a pain to get the cameras to genlock and every time we shut down and powered back on we had to fight with the genlock and the monitor to produce the 50/50 overlay correctly. Not to mention, just like you have investigated and found, the RED's do NOT Genlock perfectly. Luckily, the shoot was all 23.98 at 180' Shutter and did not show off its inconsistency. However, this should be a no brainer! There is also a lot of criticism about the RED's 10 pixel offset which most stereographers find completely unacceptable. Makes for a lot of corner pinning / convergence fixes in post the frames. I'm sure Epic and Scarlet will address these issues but there is no reason RED should not work to solve these problems on the RED ONE. Probably nothing they can do now about the 10 pixel offset but the Genlock seems like a fix they should address.
    Attached Images
    Reply With Quote  
     

  3. #3  
    Senior Member
    Join Date
    Oct 2007
    Posts
    2,523
    Is there any likely hood of this 10 pixel offset problem being a fundemental issue or will this be cured by the MX sensor upgrade ? Im more concerned about Scarlet having the same issue as we wish to use it in a 3D rig. Any info welcome.
    Reply With Quote  
     

  4. #4  
    Senior Member
    Join Date
    Jan 2010
    Posts
    141
    What's the 10 pixel thing?
    Reply With Quote  
     

  5. #5  
    Senior Member
    Join Date
    Jan 2007
    Posts
    291
    Interesting write up Pedro. Thanks for the info.
    I very much agree with you that true accurate genlock is fundamental to 3D shooting.
    Do you have any idea as to precisely how much genlock is out between two RED 1's? Does it vary in anyway? And, out of interest, have you tried testing camera genlock with a 360 shutter setting using the same test methodology?

    Please excuse my ignorance but can someone explain "Red's 10 pixel offset" issue. What is it exactly?

    JohnF
    Reply With Quote  
     

  6. #6  
    Senior Member mikeburton's Avatar
    Join Date
    Sep 2007
    Location
    Sacramento CA
    Posts
    1,673
    Quote Originally Posted by Mike Gross View Post
    What's the 10 pixel thing?
    Here is how the Pixel Offset issue was explained to me. After seeing for myself in post it makes better sense. Basically, not every sensor is exactly the same and they can vary up to a certain number of pixel discrepancy in the final converged images because of this. In 3D this means, if there are two like images (left and right eye cameras) once converged they will not make an identical match and be offset from one another slightly making more work in post to correct. F900's for example have a 2 pixel offset (not bad), RED has up to a 10 pixel offset (not so good).

    From a Stereographer perspective this 10 pixel offset can be frustrating as their job is to make sure the 3D Convergence is as accurate as possible in camera so as not to make more work=money in post. With RED they have to fight the 10 pixel offset issue which is quite a bit more drastic than say a 2 pixel offset. All Digital cinema cameras have this Pixel offset issue but not all are as drastic (noticeable / acceptable vs unacceptable) as has been noted on the RED's.

    All that said, the images are amazing!
    Reply With Quote  
     

  7. #7  
    Senior Member
    Join Date
    Jan 2007
    Posts
    291
    RE pixel offset

    Is this an on-chip manufacturing issue or is it the chip mounting position that creates the position shift? And on what axis does this occur X,Y or both?

    I ask because I've seen many cameras whose imaging chips/block are not precisely centred to the lens axis (some quite drastically offset). Which, alongside getting precise genlock, poses yet another technical issue for the stereographer to watch out for!

    JohnF
    Reply With Quote  
     

  8. #8  
    Quote Originally Posted by JohnF View Post
    Interesting write up Pedro. Thanks for the info.
    I very much agree with you that true accurate genlock is fundamental to 3D shooting.
    Do you have any idea as to precisely how much genlock is out between two RED 1's? Does it vary in anyway? And, out of interest, have you tried testing camera genlock with a 360 shutter setting using the same test methodology?

    Please excuse my ignorance but can someone explain "Red's 10 pixel offset" issue. What is it exactly?

    JohnF
    Yes, John I tested various framerate + different shutter speeds/angles. More angle/speed more visible is the offset. In my observation this stayed consistent across framerates. (regardless of icon on the screen).

    And yes, every single camera I have ever worked with in 3D has the "sensor" in different positions. This might be a chip issue but more likley a "chip mounting in the body" issue or both combined.

    So this is nothing new, just one more thing to correct for in post. Although with the new Angeniuex rouges you can correct for this on camera with the allen screws that allow you to adjust the projection cone over the sensor. Keep in mind this does affect other things as well. Proceed with caution. This is yet another reason Angeniuex rouges are the best lenses for 3D so far.

    The Cook 5i's aupposed match from the factory (they had 3D in mind allegedly)

    http://www.cookeoptics.com/cooke.nsf...s/cooke3d.html

    But still I bet they differ, just the nature of the beast. So I'll side with angeniuex's concept of allow us to adjust on camera which is just awesome.


    I will try to post frame grabs once I stop working and get a chance to load up the test footage (I hope I didn't delete it)
    ---------------------------------
    Pedro Guimaraes, SOC
    IATSE Local 600 - Operator

    Steadicam / Stereographer
    C-Motion Cvolution Package 3D-8 motor system for rent.
    Handsfree Segway for rent
    www.pgfilms.tv - Bio&Reel
    www.steadi3d.com - BLog
    www.cinemilled.com - DJI Ronin & other cinema Products
    Reply With Quote  
     

  9. #9  
    Quote Originally Posted by mikeburton View Post
    Hey Pedro,

    I thought I was the only one having trouble with the Genlock on the RED's for 3D acquisition. I just recently was 1st AC / DIT on my first RED 3D job with the ET Rig. We were using the Transvideo monitor for 50/50 overlay and that rely's on the Genlock.
    Yes the transvideo.

    Well, "genlocked" output of the RED is a recent feature (b21) and not sure what the deal is with it.

    When I shoot 3D with the sony ex3's I NEVER have any issues with the transvideo. Which makes sense since I know they lock 100%.

    When using that monitor on with the reds whenever you enter playback or connect the monitor you will see a double image (sort of over/under) to get rid of this simply cycle the power of the aja a few times until you see a clean 3D overlay image. A few times it acted up and refused to lock. Then I simply plugged into a different output on the aja (there are 6) and that seamed to fix it.

    The fact that the transvideo doesn't like the "locked" signal from the red is probably another sign that the camera is not truely locked. But that might be a whole other circuit or something. Since output lock was added later.

    I've now completed 2 3D features (one destined to IMAX) and 6 smaller 3D projects with the RED one. Also countless "3D tests" with potential clients of companies I'm working with. To think that all that footage is not 100% in sync makes me cringe as a stereographer. Regardless if you can tell or not.

    This needs to be fixed IMHO.
    ---------------------------------
    Pedro Guimaraes, SOC
    IATSE Local 600 - Operator

    Steadicam / Stereographer
    C-Motion Cvolution Package 3D-8 motor system for rent.
    Handsfree Segway for rent
    www.pgfilms.tv - Bio&Reel
    www.steadi3d.com - BLog
    www.cinemilled.com - DJI Ronin & other cinema Products
    Reply With Quote  
     

  10. #10  
    Senior Member mikeburton's Avatar
    Join Date
    Sep 2007
    Location
    Sacramento CA
    Posts
    1,673
    Quote Originally Posted by Pedro Guimaraes View Post
    Yes the transvideo.

    Well, "genlocked" output of the RED is a recent feature (b21) and not sure what the deal is with it.

    When I shoot 3D with the sony ex3's I NEVER have any issues with the transvideo. Which makes sense since I know they lock 100%.

    When using that monitor on with the reds whenever you enter playback or connect the monitor you will see a double image (sort of over/under) to get rid of this simply cycle the power of the aja a few times until you see a clean 3D overlay image. A few times it acted up and refused to lock. Then I simply plugged into a different output on the aja (there are 6) and that seamed to fix it.

    The fact that the transvideo doesn't like the "locked" signal from the red is probably another sign that the camera is not truely locked. But that might be a whole other circuit or something. Since output lock was added later.

    I've now completed 2 3D features (one destined to IMAX) and 6 smaller 3D projects with the RED one. Also countless "3D tests" with potential clients of companies I'm working with. To think that all that footage is not 100% in sync makes me cringe as a stereographer. Regardless if you can tell or not.

    This needs to be fixed IMHO.
    Agreed! I realized quickly that unplugging and replugging worked but it was never consistent and I couldn't tell if it was the Transvideo Monitor or the cameras not Genlocking 100% or a combination of both.
    Reply With Quote  
     

Tags for this Thread

View Tag Cloud

Posting Permissions
  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts