where do you see a lean? If I look at the static shot and compare it to the panning shots i can't see any lean difference.
There Is No Softening When Moving the EX1
- Thread starter matthew77
- Start date
where do you see a lean? If I look at the static shot and compare it to the panning shots i can't see any lean difference.
RE1000
Well-known member
That could be the mystery of the res loss seen right there.
I agree - though Barry does not - he thinks there's more to it.
The thing I noticed about peaking in the EX1 is that is only looks for high frequencies horizontally. When you pan, the image blurs horizontally and the high frequencies disappear.
When you tilt, however, the peaking does not disappear, because the horizontal high frequencies remain.
Try it - peaking only disappears when panning, not tilting.
This alone should be enough to refute the softening assertion - no frame grabs necessary.
The thing I noticed about peaking in the EX1 is that is only looks for high frequencies horizontally. When you pan, the image blurs horizontally and the high frequencies disappear.
When you tilt, however, the peaking does not disappear, because the horizontal high frequencies remain.
Try it - peaking only disappears when panning, not tilting.
This alone should be enough to refute the softening assertion - no frame grabs necessary.
Barry_Green
Moderator
Look guys, I had one day with it, and a morning. I didn't get to test it as thoroughly as I'd like. I'll eventually get a chance to put it through a more thorough test. But I'm not the only one who saw what we saw, and it was on the HD-SDI output of the waveform monitor.
If I'm wrong, I'm wrong -- and I'll come back here and say "DOH! I blew it, there's no issue." But until I find out what it was, I'm sticking with what I saw. I also can produce no-res-loss on 1/2000th images with it. But on 1/2000th, there's no motion -- so that proves or disproves nothing. My conjecture was that it's somehow related to an aggressive noise reduction technique -- the same kind of thing that caused the "oil paint" effect on the V1. (of course, we also had guys claiming "there is no oil paint effect" on the V1 too.)
I am at a disadvantage here -- you guys have the EX1, I don't, I only had a brief time to check it, and I'm going to try to get to Vegas to get my hands on one this weekend to try some other stuff -- but that wouldn't be an opportunity to answer this particular question, so it'd still hang out there.
I am *not* referring to the peaking disappearing -- that's something that happens on the viewfinder/lcd, not something that happens on the HD-SDI output. And I'm *not* referring to MPEG-introduced compression softening; that would happen on record/playback and I'm talking about something that was clearly visible on the live output, and was verifiable on the waveform monitor.
I've allowed that it *could* be motion blur, but that it didn't seem so. It seemed like more than that.
I think the ultimate/ideal way to test would be to put a 1920x1080 CCD sensor on the same scope as the 1920x1080 CMOS EX1 and try to recreate the exact same pan speeds, etc, and verify whether there's an excess of softening or whether they're identical. The only 1920x1080 CCD I can think of that would be moderately accessible would be an HPX3000, and those don't exactly grow on trees. Maybe I can try talking Abel into letting me come in and test a 3000 against an EX1 on a scope just to verify if they have identical softening, which would then be attributable to motion blur. If they are identical, I'll take back everything I said. If they're not identical, and if the EX1 does show more softening, well, that's when things'll get interesting and this thread's title might have to change, yes?
If I'm wrong, I'm wrong -- and I'll come back here and say "DOH! I blew it, there's no issue." But until I find out what it was, I'm sticking with what I saw. I also can produce no-res-loss on 1/2000th images with it. But on 1/2000th, there's no motion -- so that proves or disproves nothing. My conjecture was that it's somehow related to an aggressive noise reduction technique -- the same kind of thing that caused the "oil paint" effect on the V1. (of course, we also had guys claiming "there is no oil paint effect" on the V1 too.)
I am at a disadvantage here -- you guys have the EX1, I don't, I only had a brief time to check it, and I'm going to try to get to Vegas to get my hands on one this weekend to try some other stuff -- but that wouldn't be an opportunity to answer this particular question, so it'd still hang out there.
I am *not* referring to the peaking disappearing -- that's something that happens on the viewfinder/lcd, not something that happens on the HD-SDI output. And I'm *not* referring to MPEG-introduced compression softening; that would happen on record/playback and I'm talking about something that was clearly visible on the live output, and was verifiable on the waveform monitor.
I've allowed that it *could* be motion blur, but that it didn't seem so. It seemed like more than that.
I think the ultimate/ideal way to test would be to put a 1920x1080 CCD sensor on the same scope as the 1920x1080 CMOS EX1 and try to recreate the exact same pan speeds, etc, and verify whether there's an excess of softening or whether they're identical. The only 1920x1080 CCD I can think of that would be moderately accessible would be an HPX3000, and those don't exactly grow on trees. Maybe I can try talking Abel into letting me come in and test a 3000 against an EX1 on a scope just to verify if they have identical softening, which would then be attributable to motion blur. If they are identical, I'll take back everything I said. If they're not identical, and if the EX1 does show more softening, well, that's when things'll get interesting and this thread's title might have to change, yes?
Luis Caffesse
Admin
I see it flipping back and forth between the shots, very clearly -
After looking at them in 'preview' on the Mac, I took them into photoshop,
layred them on top of eachother, lined up the top edge to make sure they're aligned.
Then you can just turn the one layer on and off and the lean is pretty evident.
I just did the same thing with the static and the 1/60 pan and I can see no lean. There is motion blur but no lean I can see.
Luis Caffesse
Admin
Well, here is a quick animated GIF to show you what I was talking about.
I'm not sure how controlled this test was - but after lining up the charts as closely as possible
(using the scale which reads: 1 3 5 7 9 10 8 6 4 2 (in the center, lower half))
This is what it looks like:
I've left the Photoshop guides in there as reference to show the 90 degree angle.
You can see the scales on the left and right slant while the scale in the lower center stays fairly horizontal.
I never said it was a huge lean - only that I saw one.
Maybe something more controlled would need to be done to test the effects of the rolling shutter - as this wasn't the intent of this test.
Regardless, I see one there pretty clearly.
Whether or not it's 'significant' is really a subjective call.
I'm not sure how controlled this test was - but after lining up the charts as closely as possible
(using the scale which reads: 1 3 5 7 9 10 8 6 4 2 (in the center, lower half))
This is what it looks like:
I've left the Photoshop guides in there as reference to show the 90 degree angle.
You can see the scales on the left and right slant while the scale in the lower center stays fairly horizontal.
I never said it was a huge lean - only that I saw one.
Maybe something more controlled would need to be done to test the effects of the rolling shutter - as this wasn't the intent of this test.
Regardless, I see one there pretty clearly.
Whether or not it's 'significant' is really a subjective call.
gaffney film
Active member
the gif animation only shows the plane of the chart changing relative to the plane of the sensor. as if the upper left corner leans away. if this was demonstrating skew, which it's not, the information in the frame would not change - only it's position would change.
v
v
Last edited:
Luis Caffesse
Admin
To be fair we should say when you pan on a camera using a rolling shutter, things lean.
I just don't want anyone to think this is an EX1 specific phenomenon.
It's a consequence of the technology used - not a specific Sony issue.
NorthernFilmMaker
Active member
Matthew, I totally agree with your statement...."WHO CARES" .... well obviously people who are trying to find any type of flaw in this camera!. Its quite funny I think the EX1 has been the most criticized camera that has come into this market. Every camera to come out had its issues and the attitude towards those cameras where ok, it has problems... it's expected, well for its price point. lets see what we can do to find a way to help people to minimize the problems that are seen. But with the EX1 its been.... ok it has this problem and this problem, I wouldn't get the camera and certainly wouldn't recommend it... Ok, the EX1 has issues, but are they issues that can be worked around? In regard to the rolling shutter skewing issue.... I want someone to post footage with skewing or wobble that looks so bad that it's not usable! I mean practical footage.... i challenge someone to take the EX1 out and shoot on tripod and handheld and do some quick pans of buildings and the like, and then we can see some solid evidence that the footage would not be usable in practical terms. Otherwise please stop with the...Check out these graphs and charts with the terrible skewing.... I want some real world footage!!
Anyways, I am at the moment researching and doing the homework to upgrade to a good HD Camera in this form factor. And to be honest i have recently like others, as a result of certain posts and comments about the EX1 been a little bit weary about getting the camera... On the other hand after looking at Philip Bloom's stuff and other posts of some really awesome footage this little camera produces.... I was blown away. i liked the richness of the color, the detail and the versatility of the EX1, i know Phil uses Magic Bullet looks and that it really contributes to the color, but again who cares, what a great tool. Magic Bullet really is an awesome companion to the EX1. So, i think I am coming to the conclusion that the EX1 is the right camera for me, unless someone can post some footage that shows the Eifle Tower doing the Limbo... I don't think the skewing is an issue.
I would also just like to commend guys like Philip Bloom on his professionalism... i mean, if you read his posts you will find that he is quite an honorable guy....he doesn't criticize one camera over another.... he just tells it like it is.... now, if we could share his attitude about the tools of our trade i think this forum and our Individual businesses would be better off for it. I like that he said that in his opinion that the hvx200 and the EX1 are crap! but you just got to make due with the one that best suites your needs and that gives you a look and feel that you and your clients are satisfied with!
I meant "who cares" in the context of this thread.
Of course I care about rolling shutter issues. But that's not the purpose behind my shooting those charts. If I were doing that I would used a grid and paid very careful attention to the verticality.
Sorry that you misunderstood.
Of course I care about rolling shutter issues. But that's not the purpose behind my shooting those charts. If I were doing that I would used a grid and paid very careful attention to the verticality.
Sorry that you misunderstood.
Barry_Green
Moderator
People who will be doing match-moving and compositing. Anyone using SynthEyes, for example, which is why its maker went on a rant about rolling shutter cameras. Skewing and wobbling make those type of processes extraordinarily complex, if not impossible.
Anyone else? Not so much. Those who care are those who are going to be affected by it. (well, maybe sports shooters who have to pan back and forth quickly.)
I'll look at 'em. My recollection from the tests we were doing is that it was clearly resolving 800 vertical lines of horizontal resolution, and during panning it dropped to as low as 600. Can you point me to a link to your charts?
Great Barry - the charts are on page 2 of this thread - including ones at 1/60 shutter. I panned at a moderate speed, fast enough to make the peaking disappear.
I also explained earlier in the thread that I care deeply about rolling shutter artifacts - but that the topic was not relevant to this thread. The charts I shot were not properly set up to observe skew and many were trying to use them as such.
I use boujou and syntheyes routinely in my work, so I care.
I also explained earlier in the thread that I care deeply about rolling shutter artifacts - but that the topic was not relevant to this thread. The charts I shot were not properly set up to observe skew and many were trying to use them as such.
I use boujou and syntheyes routinely in my work, so I care.
Last edited:
TheMusician
Well-known member
I have not ever used a scope other than those in Vegas and would not know what the detail loss would look like, so I am truly speaking out of my arse here. Hopefully you guys who use them extensively can tell me if I am way off base. I was wondering if a rolling shutter can have a way of making an image LOOK softer on a scope during a pan, where visually it looks no different. In a still shot, all the detail in the horizontals and verticals are all perfectly lined up so we can confidently identify the resolution. On panning with a CMOS chip, the detail is still there but now, it is not horizontally in line with the row of pixels above and below it because of skew. I am wondering if the skew caused by panning causes a vertical line to lean, and thus look softer on a scope when in actuality - or visually - all the detail is still there it is just leaning? I'm wondering if the leaning causes all details to not be lined up on verticals and thus look softer on a scope, but is visually accurate. Give me your thoughts.
Last edited: