3,5-camera look at resolution and noise: Dragon (+ADD), Alexa, F55 – pt 8 – Final thoughts

Geoff did a hell of a job with the Bristol files, and what little I have done here is just meant as supplementary analysis.

Initially I was curios about the Alexa and the Dragon and how resolution could affect the image.
Which I looked into here.

And amongst other things, found this:

G001_C001_0206JZ

 

After looking into the resolution of the F55 and getting these results, that triggered a bit of curiosity.

 

02b_1000px_crop_both

 

Essentially, according to the charts, the Alexa resolves less than 1080 (or at least a less precise 1080) than any of the two others, despite having “sufficient oversampling for proper 1080″, while from the charts it would appear that the F55 was as sharp as the Dragon.

The first part of this conclusion does not come as a surprise, and is essentially the reason for the Alexas scalability. The camera isn’t really sharp and doesn’t really resolve 1080, so you wouldn’t see a huge difference in a 2k or 4k delivery from the Alexa.
STILL it does its softness with enough pixels to give smooth scaling of the unsharpness. Thus the “Does resolution matter” part of that comparison.
Alexa is a very cleverly thought out camera resolution-wise, even without going into color and DR.
It’s cool to have been able to quantify some of the differences between Alexa RAW and Alexa RGB/prores, and that difference makes me curious if ARRI has improved the in-camera de-bayering before releasing the Amira. I certainly hope that the de-bayering in-camera for the Amira will be closer to the software/post de-bayring routines available for Alexa/ARRIraw..

Now, the second part of the “chart-based” conclusion was a bit of a surprise, and I could not really make that fit with my other observations. Dragon and Alexa both comes out as very pleasant cameras (though very different) image-wise, while the F55 appears to me a bit more crude.
But the charts were not capable of disclosing any meaningful difference in how the resolution and noise in each cam affect a practical image, and that is why I wanted to have a look at the three-way comparison with practical images.
I could not intuitively recognise similarity in resolution between the F55 and the Dragon, yet the chart told me they should be “pretty close”.
Becomes a pride sort of thing… Was I approaching a level of “Blind Fanboyism”  on behalf of the Dragon? (I admittedly do like that camera a lot…)

But after looking a bit more into this (the last 7 pages, really), images like these, do not really support the virtues of the F55 over Dragon/Alexa
04_int_TIFF_comparison

And frankly, this is not a “bad” thing for the F55. It is more like expected. It’s not supposed to deliver an image on-par with Alexa/Dragon. And it does not.
I am however even more curious about the F65 after seeing this. The F55 is after all supposed to be based on the F65.

What surprises me is that the F55 can read so well on charts, and still fare so badly in practical images on the exact same parameters (resolution/detail) as the charts were meant to reveal information about exactly that.

Which probably gets me back to my little pet-subject of how to conduct tests.
Because you will only get practical answers in the area you test, and it is hard to postulate much outside exactly what you tested for.

Thus: If you are shooting resolution charts, the F55 is probably a good choice.
For other subjects, the viable options are probably a few more.

Charts and purely technical images mostly have their value in disclosing “what might be wrong” with an image, if you notice something wrong, but they are not very good at disclosing “what is good”.

I have planned to make a 1080 film with 1080 scales and a 1080 crop from UHD from this material as well, but frankly I have learned what I needed to learn on this particular subject for these particular cameras.

I should probably mention that I planned to put the Canon C500 in here as well, but as I did not have any exterior images from that and the results from the C500 was so erratic from super-good, to really questionable, I rather hope I get the chance to spend some more time with the C500 before writing about it.

Finally:

The RAW files from SONY was developed into 4k DPXs with Sony’s RAWviewer.
As the SONY results have been questioned, here is a screengrab of the settings:

Screen Shot 2014-03-25 at 9.45.38 pm

 

For his development of the CML/UWE files, Geoff went through Resolve/ACES workflow.

For the Alexa I used ARRIs ARRIRAW converter to unscaled TIFF and for the RED files I used RedCine-X to TIFF.

 

I have to mention that this triggered this discussion on the Sony forums and I can clearly see why.
Here is the discussion.

The reason I do not just declare “defeat and I am a stupid ass”, is because I tried to use what seemed like good settings for the F55, and I recognize I have to look into that.

For what I did, the results are good. And ifthere are better ways, I will look into those.

Now: I have to move on to some more practical image purposes… :)

 

Cheers
If you like me to publish tests and have use of what I do, I do accept and appreciate contributions to the work through Paypal




______________________________________________________________________________________________
All imagery on gunleik.com is © Gunleik Groven, unless otherwise stated. For TFP shoots, the images belong to me and the model.
It is possible to buy most images for editorial, commercial or stock usage. I’ll even sell you large prints of them….
Contact: gunleikgroven@gmail.com

No comments yet.

Leave a Comment

You must be logged in to post a comment.