Canon 7D Zone Plate Shots

[Roberto Lequeux]

(how do I make those nice popups ??)

Below the reply box hit "Go Advanced", then click on "Manage Attachments"

This is a 100% crop of the still Morgan selected to show the bad side of the camera.

 

[Pawel Achtel]

1080p 24 - F4 85mm 1.2 Canon L Lens
http://www.redgrabs.com/up/1258857792.zip

720p 60 - F4 85mm 1.2 Canon L Lens
http://www.redgrabs.com/up/1258808878.zip

Is that chart still under warranty? I'd be asking for my money back. LOL

 

[Lauri Kettunen]

can you tell me what is the difference between cropping a sensor to 1/4th resolution or to skip line in the same amount of resolution?

Pat, sorry for not understanding your point. Line skipping implies the sampling is only partial as if somebody tried to tell the color of your clothes seeing only your sleeve. The debayer algorithms, however, assume implicitly that one has "full" sampling. Moreover, can't see how one could even build a reliable algorithm (that reproduced the desired image) based on partial sampling, for, in simple words, some data is missing and quite can't see how such data could be generated by some statistical methods.

Guess Canon looks at this from the other way around. Likely, from Canon's point of view the image obtained with partial sampling is much better than no image at all. Especially so as the image quality is satisfactory for many people.

 

[Rick Darge]

Don't understand why Red is shooting zone plates with a 7D. We love you Red. You don't always need to remind us that other cameras are inferior to yours. Yes Canon does some trickery, but then again, it's lightweight, good in low light and very very cheap. Almost 1/2 the cost of the Red EVF. I own both and love both for their respective needs. Just curious as to why Red is doing a test on other cameras..

 

[Graeme Nattress]

Because nobody else is??

Graeme

Don't understand why Red is shooting zone plates with a 7D. We love you Red. You don't always need to remind us that other cameras are inferior to yours. Yes Canon does some trickery, but then again, it's lightweight, good in low light and very very cheap. Almost 1/2 the cost of the Red EVF. I own both and love both for their respective needs. Just curious as to why Red is doing a test on other cameras..

 

[Rick Darge]

Yes, but why the 7D in particular?

 

[Sanjin Jukic]

Don't understand why Red is shooting zone plates with a 7D. We love you Red. You don't always need to remind us that other cameras are inferior to yours. Yes Canon does some trickery, but then again, it's lightweight, good in low light and very very cheap. Almost 1/2 the cost of the Red EVF. I own both and love both for their respective needs. Just curious as to why Red is doing a test on other cameras..

Richard,

just wait for my uncompressed and ProRes 4:2:2 HDMI 7D live feed recording output that I'm going to test at the end of the week.

Then we can talk different about this topic.

Also here is my teaser for that in this tread:

http://www.reduser.net/forum/showthread.php?t=38025

 

[Graeme Nattress]

We've done 5D2, and I've posted up GH1 too, I think. Some of the photo sites I read just say "look at the lovely HD" - and don't measure it at all and don't mention what we mention here. I did see one foreign photo site that mentioned the lower resolution of the HD though. But that's it.

Graeme

 

[morgan_moore]

IF you deliver 1080 then the 7d looks good on paper.. (and in some highly compressed web video)

S35 chip
24/25 FPS
High Iso
Flys on a $2000 steadicam
Choice of lenses that cost less than $1k

On paper the 7d is almost Red Beater if you deliver 1080 - you have to look a little harder to see that it is very very far from that - unsuprisingly Red are helping you to look

Personally delivering great 1080 from a small package is my current ambition, so the comparison is very relevant to me

Those looking to theater delivery - I cant see why they care about these little cams

S

 

[Graeme Nattress]

Yes, the spec on paper says 1080p. The measurements say otherwise. I like to measure all new cameras. I don't always get the chance, but when I do I always learn something.

Graeme

 

[Sanjin Jukic]

We've done 5D2, and I've posted up GH1 too, I think. Some of the photo sites I read just say "look at the lovely HD" - and don't measure it at all and don't mention what we mention here. I did see one foreign photo site that mentioned the lower resolution of the HD though. But that's it.

Graeme

Graeme,

dont't worry, my test could be only for something to use in a way you want TO MATCH 7D footage with R1 or later with Scarlet...

Here are some of screen grabs from the "beauties" that I "shot" and focused them with manual @ handheld on 7D, ISO 1600, H264 compression today

and with my favorite lens that is 75 years old Carl Zeiss Olympia 180mm f/2.8 (Nikon F mount) on Canon 7D.

I'm just curious what could I get from 7D with uncompressed 4:2:2 in upcoming days?

Shot on Canon 7D with Carl Zeiss Olympia 180mm f/2.8 (Nikon F mount), no any filters, no any grade...

Shot on Canon 7D with Carl Zeiss Olympia 180mm f/2.8 (Nikon F mount), no any filters, no any grade...

Shot on Canon 7D with Carl Zeiss Olympia 180mm f/2.8 (Nikon F mount), no any filters, no any grade...

Shot on Canon 7D with Carl Zeiss Olympia 180mm f/2.8 (Nikon F mount), no any filters, no any grade...

 

[Daniel Browning]

Are you saying I have been suckered by false detail for all this time ?

Here's a simple test: which of the following images do you prefer?

If you prefer the first, then that means you like aliasing, and it explains why you prefer cameras with no anti-alias filter.

Remember all out put is digital anyway - printed blobs on a page or pixels on a computer monitor - all you need to do is outresolve the output medium

Usually, that's correct. Aliasing often gets better and better the more you outresolve the display. But sometimes it doesn't.

'Real World'

These images are flawed expecially the moireed one which I picked for fairness,

Most of them do not have much aliasing because there is very little in sharp focus.

Here is a 100% crop from the 4th image (1080p):



And here it is again, after downsizing to 720p, then up-rezzing back to 1080p:



On the 7D the difference is only in the aliasing, but if you do that with your EX1 (or any *real* 1080p camera), you'll see a huge difference.

...but what else could I have shot them on for less than $10k

Nothing.

• Large sensor
• Image quality
• Low price

Pick any two. The 7D gives you a low price and large sensor. That means low light performance and DOF control. But it also gives you poor image quality, which means very poor resolution, aliasing, skew, compression artifacts, etc., all of which are present in the frame captures.

 

[Eisen Feuer]

Graeme, first of all thank you so much for posting these so we can really see what going on inside of Canon's little black box. The effects of moiré and line skipping couldn't be made clearer.


Anyway,

Is it possible to know what the ideal look is for the zone plate? To be clearer, is it possible for you to render out an image of the zone plate under mathematically ideal conditions, to have a reference defining perfection?
With all the graphs you're pumping out, it seems there must be some perfect standard that these tests are being used to compare to.

For example the fuzzy concentric circles of the zone plate appear to be based on sin(radius²), which is something that can be represented perfectly (yay math!). I made my own in Shake just now, and at first glance it seems to indicate that mioré is formed simply when pattern density and resolution reach a certain threshold, meaning lower resolutions are more prone to moiré, and the comparison between 1080p and 4K frames is no comparison at all. (Not a bust on the test you did in this thread, more pondering the actual physical limits of the ideal camera)

Ultimately my questions would be:
A) Is there a perfect way to represent the zone plate?
B) Would it be at all practical to use a 'perfect' representation, as our eyes are the common denominator throughout?



Also I just remembered that you and I had spoken before on Apple's Discussion forums when I was having trouble repurposing the custom looks I had made in Shake into the more streamlined world of Color. Small world! Major kudos on the job at RED!

 

[Neil W. Smith]

Beautiful glass ...

Beautiful glass ...

and with my favorite lens that is 75 years old Carl Zeiss Olympia 180mm f/2.8 (Nikon F mount) on Canon 7D.

How the hell did you get hold of such a beautiful lens, Sanjin? .... was your father part of the Zeiss company? ... is that why you care so much about visual imagery?

You need to come to LA, dude ... bring your quirky dual RED/Canon rig with you and your most excellent glass .... Rodney thinks you're very cool so you could join our crazy team.

We need an Austrian with a sense of humour to offset German seriousness and OC style ... come and see us soon .... but bring your glass with you ... wrapped up carefully.

Neil

 

[Daniel Browning]

Is it possible to know what the ideal look is for the zone plate? To be clearer, is it possible for you to render out an image of the zone plate under mathematically ideal conditions, to have a reference defining perfection?

Here you go:

Test patterns hosted by reduser's John Beale

(It has both the sine wave and square wave versions. Graeme was going easy on the 7D by using the sine; in real life the transitions can be much more abrupt.)

I made my own in Shake just now, and at first glance it seems to indicate that mioré is formed simply when pattern density and resolution reach a certain threshold, meaning lower resolutions are more prone to moiré.

No, but perhaps you mixed in a down-sizing somewhere? That introduces moiré. The zone plate has to be generated with the exact final dimensions without any downsampling or interpolation.

 

[Tom Lowe]

Don't understand why Red is shooting zone plates with a 7D.

It's good to know all the facts. And the fact is, these Canon DSLRs are not 1080p by any stretch of the imagination. I use them, I like them for some types of shooting, but it's never a bad thing to have more information about cameras and their abilities.

 

[Rick Darge]

I agree with you Tom. It's good to know the facts and all. I just find it a little strange that this test is coming from inside Red HQ. If it was coming from a Reduser, like Sanjin, then it wouldn't matter, but the fact that Red is doing this test to blow the doors of Canon is what's weird to me. Red doesn't need to prove anything, they already have, 100-fold. The Canon is less than $2,000. The Red 8X Scarlet will be over $3,000 with a 2/3" sensor and fixed lens. The Red One is over $30k. I don't see the point. I don't see Canon shooting zone plates of the the R1 and claiming that it isn't true 4k so why is Red trying to prove how 'sucky' the 7D is?

 

[jbeale]

In the spirit of scientific inquiry, I think there should be no limits about what experiments anyone can do. As long as you document exactly what you did, so others can try and see if they can duplicate it, then it's a valid experiment. If you do the same experiment and get different results, then that's interesting and worth discussing. My $0.02

 

[Eisen Feuer]

No, but perhaps you mixed in a down-sizing somewhere? That introduces moiré. The zone plate has to be generated with the exact final dimensions without any downsampling or interpolation.

No downsizing, I included the .shk file in there if you have Shake, just increase the resolution of the Color1 node to see what I'm talking about. Pure math, and it eventually moirés when the pattern gets dense enough. If you look at the wedge pattern you linked me to, at 100% it moirés in the center: http://www.bealecorner.org/red/test-patterns/star-chart-bars-full-600dpi.png

At first glance of course this is a ridiculous nitpick, but it's a proof of concept that shoehorning any pattern onto a pixel grid will eventually moiré, and when framing the same object with different cameras, lower resolutions are mathematically more prone to moiré.

Anyway ultimately I was wondering whether there was a perfect reference by which a perfect comparison could be made, and it seems those charts you linked me to are just that. There was even a warning that most of them are made for computers to analyze, heh.

 

[morgan_moore]

.....

Daniel once again thanks for your considered thoughts

The first images are too small to really comment, neither are that appealing

but I can see the difference

-----

I can see that a filter ideally should be used - I think in some cameras it is too strong and over blurs the image which has been the case on some DSLR models - I dont think these VDSLRs are inhibited by the filter but by skipping

----

Im using the phrase HD 1080 in quotes with the EX1 because I dont think that is 1080 either (just a guess)

I would define a proper 1080 image as one that looks as good as a 1080 image dowscaled from a Raw still - the EX1 does not do that

(not to mention being able to fix colour in post which raw provides or touching on DR, or pleasant transitions to blown highlights)

My mind is clear on these cameras now

For now I will keep the 7d and use it for the sort of shots we see - simple tight ones

When the 5d framerate comes out I will use that again and sell the 7d

I think the 5d is better in every day use for issues of lens coverage and diffraction rather than being zone plate better than the 5

I will keep my EX1 and use it on complex/wide shots

I will not buy the 1ds4 becasue I dont want to spend that much money on a flawed device (both sound and resolution)

Hopefully by mid next year someone will offer an EX1 beating device for less than $10k that accepts still lenses and has a S35 or larger chip

I will then trade them all for that whoever makes it

I think my concept of having a device that can equal my D3 for stills capture and offerering EX1 beating S35 1080 motion is not likely to occur for some time

One thing I think red have got a little wrong is..

I would prefer a FF or S35 RAW 1080 camera to a 1/2 3k one - I think I am not unique in this because those larger resolutions outperform most commercially used display devices and create PP issues ie 4k is not needed by your average corporate or broadcast client


S