My thoughts on sensor size

Ian-T said:
Ok. I gotcha.

Question though (and this might sound stupid) but when the same camera shoots both 720 and 1080 do they use the same area on the sensor to make both pictures? Or is the 720p image obtained from a smaller part of the sensor?

My thought is if it uses a smaller portion of the sensor to make 720p then that's why the skew will not be as bad.
Click to expand...

Depending on the camera.

Example: GH-1 Uses the same area for 1080 and 720P. So you dont' get a crop when going to 720P. The change is scan rate, which is why the 1080 has a bit more jello (which is already handled quite damned well) versus the 720/60P. Same goes for the 720/30P as well.

However, with the RED, when you shoot @ 3K or 2K you DO Crop the sensor. This is the reason that just about everyone shoots 4K on the RED and not 3K or 2K unless using 16mm Primes, doing off speed acquisition, etc etc.

You don't get S35mm FOV if you're not in 4K.

This is actually the only reason I'll take an FF35 Scarlet after having thought about it. The OPTION to do as you wish will be awesome. The Resolution, don't care about it. I'd take a 1080 Scarlet so long as it's FOV was accurate, but the option to shoot Academy @ 4K or FF @ 5K (landscape detail would be SCREAMIN' I tell ya) is excellent.

So, the smaller portion fo the sensor doesn't apply to the GH-1. The reason I think that the MKii's cropped 720p (not downrezzed, but cropped and composed for Academy) would still have the same amount of Jello is because it's baked into the footage. If you're trying to get Academy FOV then you treat the camera as is and scoot back, etc etc. THe jello is going to be baked in as it is, though.

The only real way to know is to try it:

MKii @ FulL Frame w/50mm Prime. Compose, Pan at various speeds

MKii @ Academy Guides w/100mm Prime. Match Composition within the guides, Pan at various speeds.

Compare.
 
The only issue I've had so far with FF is that I'm finding I need a longer lens around 180-200mm. shallow dof isn't an issue because you stop down a couple of stops and the mark II has a couple of stops or more of extra low light sensitivity over the GH1. I'm going with a variation on this cropping theme and either cropping horizontally to get 2.39:1 and/or shooting anamorphic and scaling down horizontally. I'll do some tests when my adapters arrive. So far I've been letterboxing for 2.39:1 and I've been happy with the results. 720p is enough to go to film.
 
squig said:
The only issue I've had so far with FF is that I'm finding I need a longer lens around 180-200mm. shalow dof isn't an issue because you stop down a couple of stops and the mark II has a couple of stops or more of extra low light sensitivity over the GH1. I'm going with a variation on this cropping theme and either cropping horizontally to get 2.39:1 and/or shooting anamorphic and scaling down horizontally. I'll do some tests when my adapters arrive. So far I've been letterboxing for 2.39:1 and I've been happy with the results. 720p is enough to go to film.
Click to expand...


It's more like 1 stop, if that. Dynamic range is a different story. Sorry, that was just inaccurate information.
 
Ian-T said:
Question though (and this might sound stupid) but when the same camera shoots both 720 and 1080 do they use the same area on the sensor to make both pictures? Or is the 720p image obtained from a smaller part of the sensor?

My thought is if it uses a smaller portion of the sensor to make 720p then that's why the skew will not be as bad.
Click to expand...
Just speculating right now. We know that...

When the GH1 camera is set to 1080p it has to read 1080 lines.
When the camera is set to 720p it has to read 720 lines.

Using the same clock frequency you can read 720 lines much faster than 1080 lines. That's probably why the GH1 has reduced rolling shutter at 720p, and why it can shoot at a higher frame rate. It can read the frame almost twice faster and has less work to do.

Hypothetically, the GH1 should show more aliasing and artifacts under 720p since it is skipping more rows... Thats the problem in the first place. The more rows you skip, the more moire and aliasing. That's why its worse on the 5D, its skipping a LOT of rows with the 21 megapixel sensor.

I doubt the sensor frequency changes... Who knows!


Kholi,

Feel free to test, I encourage it... You will be surprised, it works. I have been using that method for years :)
 
Just going by what has been reported in the past and that side by side comparison that was posted on the web. The GH1 is a lot noisier though.

downscaling should reduce aliasing just as upscaling increases it. but I'm talking straight scaling not interpolated wizardry
 
I mean more in lines with the jello,. The method works, I know for sure. I've done it myself because of a messed up 35mm adapter. The Jello is the question.

I agree with you, though. It's a great workaround and 720P is great for it.

Just gots to figure out if it affects the jello. Besides, if the skew doesn't bother you in the first place then there's nothing to figure out.
 
squig said:
Just going by what has been reported in the past and that side by side comparison that was posted on the web. The GH1 is a lot noisier though.
Click to expand...

Links please. Only tests I've seen are with kit lenses. And the GH-1 Kit lens is an F4.

If you show me a test where both are at the same stop then you'll see that the DR is the separating factor. Dunno what you're talking about as far as noise though, didn't experience this either.

From FIRST HAND USE.. not conjecture.

Anyway, this isn't a GH-1 vs MKii thread. Daniel's advice is solid and I'd suggest it for anyone looking to battle the FF35 FOV. IN fact, it could be used well if you master @ 720.

Simply shoot your wide material @ 1080 (best detail) then for everything else use your homemade Academy Frame Guides. Beast.

Hell, you could go somewhere in between Academy and FF so long as it fits into actual display guides: 16:9, 2.35:1, etc.

Good idears.
 
After 6 months with a D90 jello really isn't an issue anymore and I don't shoot anything without the steadytracker. There aren't any circumstances where I really need to shoot handheld if there where I'd certainly look at the GH1
 
Kholi said:
Links please. Only tests I've seen are with kit lenses. And the GH-1 Kit lens is an F4.

If you show me a test where both are at the same stop then you'll see that the DR is the separating factor. Dunno what you're talking about as far as noise though, didn't experience this either.

From FIRST HAND USE.. not conjecture.
Click to expand...

was the stuff shot over the water with the power pole, we talking the same stuff?

In comparing noise I'm judging by looking at a noise test I saw yesterday of the D5000 vs the GH1 then comparing how noisy the D90 (same sensor/internals as D5000) vs the MKII. The MKII is way cleaner than the D90 and the D5000 had the edge by a good margin over the GH1 in the tests. I know I haven't run a side by side test with the GH1 but even if the noise level is comparable to the D90 it doesn't come close to the MKII.
 
To be clear I'm not trying to highlight any specific jello problem. I agree with you Squig, jello on the 5D is not something I worry about... However, the idea is that you can do something about it if necessary.

Yes, you will lose your 1080p. But in the worse case scenario, a jello free 720p is better than a very distorted 1080p.

You can also remove jello in post with software. This has been shown to work very well.. but that is for another thread...
 
A 1.5x crop @ 1080p is probably closest to the ideal size for filmout but I don't see any major issues with either full frame or 2x crop. lens selection at either end is all that really changes. The important thing is to know exactly what your delivery format is gonna be and how best to get there.
 
It occurs to me that when shooting a 5d cropped a software could be developed that takes a parrallelogram from the frame that counters the tilt in the (panned) image

S
 
Last edited:
cropping does reduce jello

cropping does reduce jello

The 5DII sensor read speed has been measured by Pango of Cinema5D, and others at Dvinfo, at 43200 lines per second.

43200/1080=40, thus it takes 1/40 seconds to read a 1080p frame.

43200/720=60, thus it takes 1/60 seconds to read a 720p crop of the same frame.

Therefore, cropping to 720p yields a 50% reduction in jello over 1080p.

Another option is to crop to 2.40:1 Cinemascope. Then you're talking 1920x800. 43200/800=54, so you're at 1/54 seconds per frame.

But I think Trammell Hudson's crop lines are for 2.35:1 in Magic Lantern, so we're talking:

(16/9)/(2.35)=0.757

(0.757)(1080)=817

43200/817=53

(1/40)/(1/53)=1.325

So, shooting 1920x817 with Trammell's Cinemascope crop lines, you'll see a 32.5% reduction in jello over straight 1080p.
 
Thank you Paul, you explained it a lot better than I could.

One of the things I would like to see in Hudson's firmware is 720p crop lines.
 
PaulM said:
So, shooting 1920x817 with Trammell's Cinemascope crop lines, you'll see a 32.5% reduction in jello over straight 1080p.
Click to expand...
That's what I thought. You might loose some resolution but in that area of the picture you get much less skew. This is why I said what I said earlier about the GH-1 when it comes to it filming 24p/1080 vs 60p/720.
 
Kholi said:
Isn't that in camera though? Not in post?
Click to expand...

You can't record anything but 1080p or 480p to the CF card on the 5DII, even with Magic Lantern.

Magic Lantern will give you 4:3 and Cinemascope crop overlays, but the camera still always records 1080p or 480p.

Cropping is always done in post.

Any vertical cropping, done in post, obviously, will reduce skew.

In fact, the 480p image exhibits more skew than the 1080p image in the 5DII. This is because the 5DII sensor only has one read speed, 43200 lines per second, and it always reads out the full 3:2 sensor. This is why the 16:9 overlays are just shaded, not black. There's still an image under them.
 
Right. I agree. But were talking in camera crops or hardware adjustments. The idea proposed here isn't a hardware adjustment it's a software adjust mention after the camera has already recorded what sort of skew it exhibits.

Cropping in post dosnt change anything.

So you use longer lenses to adjust for academy which means jellos exaggerated right? Then you crop it.

But I think I may be thinkIng about it from the wrong angle.

I know that shooing 2k or 3k on red is a different story than shooting 4k. And Gh1 1080 vs 720. But cropping. I'm not sure about a post crop though.

This is a actually a very intriguing test to me. I'm interested in experimening with different crops and daniel does make a grea point. There's a lot of roo
for new effects and such with full frame.

Hmmmm. You may have sold me on keeping the mkii just based on the post effects you could pull off with the extra scan area.

720 is abosultely enough to use for just about anything to me. And there are amazing tools to uprez to 1080. Mastering at 720 makes for a very resolute downrez from 1080 to 720 as well.

Hmmmmmm!!!!! Must run some After effects tracking and crop tests. This could be fun.
 
If you want to utilize a 720P crop of the 5D2's 1080P frame to take advantage of reduced skew, fine.. But be prepared to lose a LOT of resolution..

Keep this in mind. You can't equate a 720P crop from the 5D2 to a 720P native cam like the GH1. The 720P from the 5D will have WAY less resolved resolution than the native 720P of the GH1. If you want to frame the exact same FOV scene with the GH1 and 5D2(with the 5D2 cropped to 720P) The 5D2's clip will be much lower resolution than the GH1's. Why? The GH1's 720P is down converted from the full width of the GH1's sensor, utilizing all of the pixels (minus any skipping algorithm it employs). While the 5D2's 720P is simply a crop of the downsized 1080P, not utilizing the full width of the sensor.
 
You must log in or register to reply here.
Back
Top