Thread: Sony PMW-F3

Page 24 of 26 FirstFirst ... 1420212223242526 LastLast
Results 231 to 240 of 253
  1. Collapse Details
    Cinematography/Lighting Mod Ryan Patrick O'Hara's Avatar
    Join Date
    Jun 2006
    Location
    Los Angeles, CA
    Posts
    5,936
    Default
    Quote Originally Posted by D_and_G View Post
    I guess that's what we get for discussing the f35, in an f3 thread, and in the af100 forum. Blame it on Graeme :-)
    hahahaha. Yup!

    If cinematography wasn't infinite, I'm sure I would have found the end by now.


    Reply With Quote
     

  2. Collapse Details
    Junior Member
    Join Date
    Oct 2010
    Posts
    14
    Default
    I hope it is not posted yet:

    http://www.stargatestudios.net/channel/?p=948


    Reply With Quote
     

  3. Collapse Details
    Senior Member Duke M.'s Avatar
    Join Date
    Aug 2008
    Location
    CA & FL
    Posts
    2,736
    Default
    Quote Originally Posted by D_and_G View Post
    After I posted I thought "hey he might think I was referring to the F3 too". I guess that's what we get for discussing the f35, in an f3 thread, and in the af100 forum. Blame it on Graeme :-)

    I think the F35 discsussion came about since the early articles said the F3 sensor is like the F35 sensor so its not really off topic.

    The F3 thread in the AF100 section is purely a DVXuser issue. They closed the Sony XDCam F3 discussion in the Sony XDCam section.

    Quote Originally Posted by D_and_G View Post
    I'm not familiar with the Ex; but a 1080 sensor is about a 540 Nyquist limit. Funny, looking at the specs for the Ex they say it has a 1000 tv line horizontal resolution too - same as the F35.
    They say the F35 is 1000 tv lines, and at first I didn't believe Graeme that it had that artifact and resolution issue since it is Sony's flagship model. When I sought out some actual testing the F35 didn't test as well as expected. We can only hope the F3 has corrected that situation because it should test as well.

    Makes me appreciate my EX1 even more.


    Reply With Quote
     

  4. Collapse Details
    Red Team Graeme_Nattress's Avatar
    Join Date
    Apr 2005
    Location
    Ottawa, Canada
    Posts
    1,135
    Default
    If the F3 tests better than the F35, then Sony will have a lot of trouble on their hands explaining away the price of the F35. The issues with the F35 are solely down to Sony's choice of the RGB stripe pattern on the sensor. The F35 manual, pg 109, has some interesting precautions though: "Aliasing: when fine patterns, stripes, or lines are shot, they may appear jagged or flicker", and "Smear: when an extremely bright object, such as a strong spotlight or flashlight, is being shot, vertical tails may be produced on the screen, or the image may be distorted." Going CMOS for the F3 should rid them of the smear though.

    Graeme
    www.nattress.com - Film Effects and Standards Conversion for FCP
    www.red.com - RED - 4k Digital Cinema Camera


    Reply With Quote
     

  5. Collapse Details
    Senior Member Duke M.'s Avatar
    Join Date
    Aug 2008
    Location
    CA & FL
    Posts
    2,736
    Default
    Quote Originally Posted by Graeme_Nattress View Post
    If the F3 tests better than the F35, then Sony will have a lot of trouble on their hands explaining away the price of the F35.
    Not really. The F35 is still 4.4.4 out of the box. Smear has long been a concern with CCD sensors, but the CCD is much less subject to motion artifacts than a CMOS sensor (unless its a global shutter or reads really, really fast.)

    I read somewhere that the striped pattern is how they were able to extend the dynamic range too.

    Being introduced in 2008, the three year cycle for the F35 means there may be an update in the future for it too. (pure speculation)
    Last edited by Duke M.; 11-14-2010 at 08:24 AM.


    Reply With Quote
     

  6. Collapse Details
    Senior Member
    Join Date
    Mar 2005
    Location
    toronto, Canada
    Posts
    839
    Default
    They could have been rid of the smear with a mechanical shutter, and while your at it throw in an ovf. I wouldn't worry too much about smear though because if it's close to the F23 it won't be that much of a problem. Heck, if it does occur just tell them it's a new type of anamorphic artifact, except vertical :-) The two cameras are in different leagues (or at least should be). Besides what Duke mentioned the F35 is supersampled.

    If the F3 gets close to the f23 it'll be a smashing technological success (for the price point), and a real contender if shot with the right conditions. The only question for me is the codec.
    Quote Originally Posted by Graeme_Nattress View Post
    If the F3 tests better than the F35, then Sony will have a lot of trouble on their hands explaining away the price of the F35. The issues with the F35 are solely down to Sony's choice of the RGB stripe pattern on the sensor. The F35 manual, pg 109, has some interesting precautions though: "Aliasing: when fine patterns, stripes, or lines are shot, they may appear jagged or flicker", and "Smear: when an extremely bright object, such as a strong spotlight or flashlight, is being shot, vertical tails may be produced on the screen, or the image may be distorted." Going CMOS for the F3 should rid them of the smear though.

    Graeme


    Reply With Quote
     

  7. Collapse Details
    Red Team Graeme_Nattress's Avatar
    Join Date
    Apr 2005
    Location
    Ottawa, Canada
    Posts
    1,135
    Default
    Other than sensor size effect, I think the F23 image is superior to that of the F35, showing less chroma moire. The whole confusion and discussion here exists only because Sony are not forthcoming on their new sensor details, only that the area of the photosite is 4x that of a typical DSLR pixel, or words to that effect. On a S35 sized sensor, what does that mean for pixel dimension? If they're using their 45 degree rotated Bayer Pattern with 1920x1080 on the greens, that leaves the chroma sub-sampled and makes their 4:4:4 option somewhat pointless, for instance. Nothing quite joins all the "known" pieces of information together. It will certainly be interesting to see what they come up with, and how the image quality positions itself between their small sensor HD and large sensor F35 HD cameras. Speculation is fun, but it's not where near as good as actually measuring and viewing images!

    I find it hard to describe the F35 as supersampled though, any more than a 3chip+prism camera being super-sampled. The F35 sensor arrangement equates to 3 chips translated onto a single sensor, interlacing the columns to get them all to fit, causing a spatial mis-alignment of the channels that is not accounted for in the image processing. The vertical use of two photosites per pixel, rather than one could indeed be used to over-sample the image, but the excessive amounts of luma aliasing vertically is enough to show we do not get the benefits of oversampling vertically. From what I can tell, the two photosites are binned, which can help with DR, but not with aliasing.

    So, we could end up with a situation where the F3 is a CMOS F35 with stripe, or it's a rotated Bayer CMOS, or even a normal Bayer CMOS at HD resolution, or normal Bayer at a higher resolution like Alexa. Or even something different to all of the above. Each decision is a compromise, each decision has a different set of quality factors.

    The question of the codec is a simple one - you're meant to record to SR and the codec is purely for offline.

    Graeme
    www.nattress.com - Film Effects and Standards Conversion for FCP
    www.red.com - RED - 4k Digital Cinema Camera


    Reply With Quote
     

  8. Collapse Details
    Senior Member Duke M.'s Avatar
    Join Date
    Aug 2008
    Location
    CA & FL
    Posts
    2,736
    Default
    The F35 sensor arrangement equates to 3 chips translated onto a single sensor, interlacing the columns to get them all to fit, causing a spatial mis-alignment of the channels that is not accounted for in the image processing.
    That's what I was getting at. My wish list, and its perfectly possible, their next generation will account for the mis-alignment with their software. (All the camera companies need to pass out more information.)


    Reply With Quote
     

  9. Collapse Details
    Red Team Graeme_Nattress's Avatar
    Join Date
    Apr 2005
    Location
    Ottawa, Canada
    Posts
    1,135
    Default
    Well, if you took an uncompressed feed off the camera, you'd be able to apply whatever post-filtration you desire to the data to correct for the 1/3 pixel alignment offset on each of the two channels that are out of alignment with the third. However, you'd have to ensure that no colour correcting matrix had been applied to the data in-camera, or the co-mingling of the channels would ruin your ability to correctly correct for offset.

    Graeme
    www.nattress.com - Film Effects and Standards Conversion for FCP
    www.red.com - RED - 4k Digital Cinema Camera


    Reply With Quote
     

  10. Collapse Details
    Senior Member
    Join Date
    Mar 2005
    Location
    toronto, Canada
    Posts
    839
    Default
    Quote Originally Posted by Graeme_Nattress View Post
    Other than sensor size effect, I think the F23 image is superior to that of the F35, showing less chroma moire. The whole confusion and discussion here exists only because Sony are not forthcoming on their new sensor details, only that the area of the photosite is 4x that of a typical DSLR pixel, or words to that effect. On a S35 sized sensor, what does that mean for pixel dimension? If they're using their 45 degree rotated Bayer Pattern with 1920x1080 on the greens, that leaves the chroma sub-sampled and makes their 4:4:4 option somewhat pointless, for instance. Nothing quite joins all the "known" pieces of information together. It will certainly be interesting to see what they come up with, and how the image quality positions itself between their small sensor HD and large sensor F35 HD cameras. Speculation is fun, but it's not where near as good as actually measuring and viewing images!
    It all depends on a number of factors, like orientation, how they read the photosites, and the entire image system. That's what I try and impress upon people is that it's the image system along with it's implementation that are vastly underrated in all our talk about specs etc... Having all these new cameras to test will be fun :-) (for those who have the $ to do it)

    I find it hard to describe the F35 as supersampled though, any more than a 3chip+prism camera being super-sampled. The F35 sensor arrangement equates to 3 chips translated onto a single sensor, interlacing the columns to get them all to fit, causing a spatial mis-alignment of the channels that is not accounted for in the image processing. The vertical use of two photosites per pixel, rather than one could indeed be used to over-sample the image, but the excessive amounts of luma aliasing vertically is enough to show we do not get the benefits of oversampling vertically. From what I can tell, the two photosites are binned, which can help with DR, but not with aliasing.
    I say supersampled more as a general term. When you take a 5K chip, subsample to 2K; basically smaller photosites are added together to make larger ones. Doing that you get pretty clean signals because of the amplification. Yes, pixel binning can sacrifice spatial resolution (especially in older model ccds) but the new ccd architecture has an improved floating diffusion sense node which helps keep the varied potential levels linear. That's why I don't think this is an alignment issue. It could be, but i'm leaning more towards that what you're seeing is high frequency detail in the chart/plate at or near the carrier wave (900- 1050 tvl/ph depending on the cam) generating a reciprocal low frequency response in the signal that is hard to get out through the dsp (because it's actually mixed in the signal) - thus you have aliasing or moire .

    We could look at the prefilter olpf, macro and unit cell alignment, the optical sampling lattice, band limiting electrical filter, a/d converters, dsp etc..., but only look there if you see aliasing/moire and are well within the Nyquist limit that you're sampling.

    If that was the case then there are a number of things you can do, especially in the dsp. However, if it isn't the case, then what you're seeing is the natural limits of a single chip 1080 cam, which you still won't really notice on a film out, imho



    Each decision is a compromise, each decision has a different set of quality factors.
    Absolutely, every sensor has its compromises and things it does better than others. The more choices the better. :-)

    The question of the codec is a simple one - you're meant to record to SR and the codec is purely for offline.

    Graeme
    You have some $ you're not using Graeme :-) I wouldn't mind recording to Sr. At the end of my film I'll give you the credit "codec supplied by the beneficence of Graeme".
    Last edited by D_and_G; 11-17-2010 at 01:20 PM.


    Reply With Quote
     

Page 24 of 26 FirstFirst ... 1420212223242526 LastLast

Posting Permissions

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