Comparison shots from 4 bodies

The arrival of my 50D prompted me to have a shoot off between the four bodies I have in my bag....1DmkII, 5D, 30D and the new 50D. I chose my next door neighbour's TV antenna as the subject due to the detail contained and ease of cropping to an identical frame. The 50D's impressive pixel count lead me to believe that the 5D would be left for dead in terms of resolution....certainly far away from my normal 5D subject matter.
All shots were taken using the 600/4 IS, AF centre point, bodies in AV mode and picture style set to standard. I err'd slightly in that the 1D2 was using centre metering whilst the others were set to evaluative, hence the over exposure, and I apolgise for this, I decided not to make corrections in PP and this allows the test to retain some validity.
PP was limited to cropping all images to the same view and a post crop sharpening with the same parameters. Focus was on the brown rectangle in the centre of the structure.

The 5D sensor clearly exhibits the need for a clean but, as it is seldom used above f/4, it hadn't shown itself on my other images.
I hope it gives some insight into pixel density benefits.
Bob

1DmkII f/8


1DmkII f/16


30D f/8


30D f/16


please see next post due to image limit

continuing.....
5D f/8


5D f/16


50D f/8


50D f/16


If the posts in this thread impinge on forum rules then can the moderator please cut the links. I believe that test images require a certain size and resolution to remain valid.

This is a real help!

Why is the blue sky broken up in the 5D. Similarly the 50D. That doesn't seem like dirt. How was this developed from RAW?

Asher Kelman said:

This is a real help!

Why is the blue sky broken up in the 5D. Similarly the 50D. That doesn't seem like dirt. How was this developed from RAW?

Click to expand...

Asher,

I'll confess to not understanding the terminology..."broken up"?

Canon's DPP 3.5 was used to simply crop the images. They were then resized in Irfanview followed by 100% sharpen with a 0.2 pixel threshhold.

Bob

Look at the 5D at 58. The blue of the sky at the top of the image is not smooth but has a texture.

Asher

Asher Kelman said:

Look at the 5D at 58. The blue of the sky at the top of the image is not smooth but has a texture.

Asher

Click to expand...

I believe the texture you're referring to is jpeg "over compression"....the result of knocking a 12Mb RAW down to a 140k jpeg.
I've checked the original RAW's and large JPEG's from the camera and the textured sky is absent on the output from each body. It may be that Irfanview makes a poor job of this compared to CS3.
The images could be processed in a better way to enhance the strengths of each body but the exercise was primarily aimed at looking at the output from the body and concessions for web publication neccesitated a certain amount of processing.

Bob

Bob Latham said:

The images could be processed in a better way to enhance the strengths of each body but the exercise was primarily aimed at looking at the output from the body and concessions for web publication neccesitated a certain amount of processing.

Click to expand...

Hi Bob,

Did you select 'Lanczos' as resampling method in Irfanview? That is a requirement to avoid downsampling artifacts.

Bart

Bart,

I did have lanczos selected (by default). I'll do some experimentation and see at what compression level the deterioration starts to occur. My initial analysis was that the crops were reasonable as it's still possible to see the individual strands off the spider's web hanging from the structure (the white vertical lines glinting in the sun and having the appearance of scrathes).
There's also the possibility that DPP 3.5 RAW conversion is the culprit...it was delivered with the 50D and this would have been my first usage of it.

Bob

Hi Bob,

I think that this is a great thread, thanks for starting it.

Just to make sure that we understand you correctly, I have a couple of questions:
1) What are we trying to achieve here ? Comparison of IQ, resolution or some other factors? Knowing this will help us validate the test method and evaluate the results.
2) I assume that the FF images are cropped to provide the same image composition since you indicated that you worked with a fixed length 600mm lens?

Cheers,

Yes Cem,

This is very good. We do need a standard way of presenting the images to see whether or not the new cameras with the smaller pixels render materials well. So the sky and also steel or aluminum are among a list of characteristics I'd like to see. We can get more detail in textures but can we retain the smoothness in non-textured areas too?

So Bart, could you suggest how this might be best done?

Asher

Cem,

My intention was to see whether the 50D's pixel count was as beneficial as predicted for what people were terming "increased reach" (ie, the ability to crop further whilst maintaining IQ and resolution). I added in the 1D2 and 5D for interest but expected that the 5D would be pixelly challenged.....it appears that it stood up quite well considering.

All 8 images were taken from the same position and cropped to provide the same FOV.

I also had some idea that a 15Mp sensor might well be beyond the resolution of some lenses but the 600/4 was not the best choice to prove or disprove this theory.

Bob

Bob Latham said:

...My intention was to see whether the 50D's pixel count was as beneficial as predicted for what people were terming "increased reach" (ie, the ability to crop further whilst maintaining IQ and resolution)...

Click to expand...

Hi Bob,

Thanks for this clarification. Allow me to think this through with you please.

In order to see the real effect of the additional pixel count so that we can crop to get the same results, I would suggest to test as follows:

Firstly, we have to start with camera's having the same sensor dimensions such as 30D & 50D or 5D & 1DsMk II. 1DMkII has a different sensor size so it does not match with any of the rest.

We then have to shoot a frame on the body with the lower pixel count and then shoot the same subject using a wider lens on the body with the higher pixel count. Then, you can crop the second picture to the same frame as the first one. So you then end up with virtually the same picture, one not cropped and one cropped. It is also good to use the same lens for the test so it will have to be a zoom lens.

In practice, I'd take the 30D with a zoom lens such as the 70-200mm. Mount the camera on a tripod and don't move it afterwards. Set the lens to 200mm and take a frame of any stationary subject. Next, mount the lens on the 50D and mount the camera on the tripod. Set the lens length to about 140mm (divide 200 by square root of 2). Take another picture. Now if you crop this second one by reducing the canvas to 70% size, you'll end up having to almost identical pictures, with the same frame and approximately the same number of pixels (30D has around the half of the pixels as the 50D, so this is a bit of an approximation really).

Then we can compare those two shots. What we will see is the difference introduced by the pixel density due to the higher number of pixels in the same area and of course also the IQ due to the DIGIC IV engine and micro lenses in the 50D. But it is still worth doing this test.

Looking forward to hear about your opinion on this.

Cheers,

Cem,

I'll think about the implications and whether it's a diversion from what I intended with the original shoot. I'm away until Thursday but didn't want you to think I was ignoring you or had just switched the subject off.

Bob

One difficulty with this methodology of course is that it is necessary to discard information from the 5D and 1D2 bodies to get equal crops. The alternative of using different lenses to get the same field of view present a new variable in the lens choice! Having said that, it would be interesting to see some 100% crops of the in and out of focus areas of the antenna.

Cheers, interesting thread,

Mike

Thanks so much Bob for this wonderfully helpful peep at the 50-D.

Having more pixels in a feature is no issue. We can show the feature of interest native at 100%, after all that's what we get when we take the picture. As long as the same lens is used one has enough brain power to glean benefit from a 100% crop, all other things being unchanged.

At f5.6, these highly packed sensors are still safe from the aperture-induced diffraction issues. At f8, such effects are usually not major. So to me, I just would like to see what simple uniform surfaces look like and how detailed feature are written with each sensor.

So we just need a standard RAW to jpg work flow so that a 100% cut out will be not degraded by the conversion.

So for this, for example, what should the jpg compression quality be? In these examples, I'd prefer then to be at level 10.

Asher

I've just been further investigating the textured sky syndrome seen in the original post. The JPG's uploaded from my PC to the hosting site (Photobucket) do not contain the texture but viewing the ones on Photobucket do have it.
Checking the file size has shown me that Photobucket apply further compression to uploaded images and hence degrade the IQ.

I'll have to figure a way around this before testing further.

Bob