Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Camera testing and comparisons

Status
Not open for further replies.
Peter Moretti said:
Why not converted from ProRes 4:4:4:4, since that's the highest quality the camera can currently record to?
Click to expand...


Because I work in Windows and those nice folks at Apple are still causing us the odd problem with 444, I decided to take the most commonly used format as far as I could gather.

The whole P2 workflow is a nightmare, I tried to install the P2 player and drivers but the drivers wouldn't install because I didn't have a camera attached.

Also bear in mind that these tests were done as a part of a workshop in Gothenburg and the students have final say as to what formats we shoot.

David is absolutely right, the goalposts keep moving, its why I keep testing, and getting different results, just compare the F900 in 2001 with it in 2100!!
 
geoff,

Maybe you can get someone w/ a Mac to do the Alexa conversion for you? I have to think some people here would be mroe than willing to help out. (I don't have a Mac either, or I'd gladly do it.)

But going from ProRes 4:4:4:4 Log-C to ProRes 4:2:2 to DPX sounds pretty unfair. Also there is the whole issue of employing a proper Log-C LUT, b/c w/o some kind of grade, the footage will look washed out.
 
OK,

I’ve finished the first page, the next one will be Canon 7D and SI2K and is a lot easier.

Alexa and RED MX are at…

http://www.cinematography.net/Gothenburg/Goth-Alexa-RED.html

This is not meant to be a shoot-out it’s for reference, I’ll add R3D files later for people to try and improve on the material themselves.

I’ll add Epic as soon as I can get one…
 
(...)The Alexa material was loaded into Iridas Framecycler and rendered out a single DPX frames and also to Rec709 using an Arri de-log file.

The RED material was loaded into Redcine-Xe and rendered out to REDFilmLog and REDGamma2 in each case using RedColor2.

These were the settings reccomended by RED and ARRI.

I stopped rendering material from any camera when I couldn't see where the exposure indicator at the bottom was.

I will also be uploading single frames in R3D format for you to render out tourselves, I'll be very interested in looking at any better results, I'll add them with a description of how they were done and by whom...

I'll also add Epic frames as soon as I can.(...)
Click to expand...

:thumbup1:

Now I'll just wait till the more experienced technical minds pre-masticate the results down little.
 
geoff,

Thanks for the very nice work :). I do have a ? about the Red images. It seems that you stopped printing them sooner than you did the Alex footage when testing highlight handling.

I know you say, "I stopped rendering material from any camera when I couldn't see where the exposure indicator at the bottom was."

But it seems that that has allowed Alexa images that look considerably more washed out/blown out to be included, while their Red counterparts were not.

Perhaps just add for the Alexa 709 test and all the Red tests images from minus seven to plus seven. This way we can decide what has useful info and what doesn't. Just an idea.

Thanks again!
 
Wow, that server is getting slammed. Very slow downloads. I can understand that a lot of people might be interested in the results. I am getting 0.1KB/s for one still which equates to about 14 hours. So, won't be able to do much with these until the bandwidth frees up.

Nonetheless, thank you for putting this out there Geoff!
 
I'll try and add the missing files up and down but it's really difficult when you can't read what the damn file is!

The server shouldn't be that slow, we pay a fortune for a fast system and I warned them in advance, it may be that they are reconfiguring to deal with demand.

The system we are on also hosts the Financial Times and a number of other huge institutions so generally speed isn't an issue :-(
 
Just tried downloading a file myself, my personal ISP isn't connected in any way to the CML one and you are absolutely right, it's saying 4 minutes for one full res frame.

How many people are on there!
 
OK, life is too short to try and deal with the files that I can't read the images of!!

I have uploaded a R3D of every shot so you can try for yourselves.

They're going up at the moment 4pm UK time and should be there pretty soon, you'll need to right click and save as to get them.
 
I would assume there are a LOT of people interested in your test Geoff. So they could be hitting it pretty hard. But perhaps it is just an initial adjustment period and it will speed up at some point.
 
geoffboyle said:
OK, life is too short to try and deal with the files that I can't read the images of!!

I have uploaded a R3D of every shot so you can try for yourselves.

They're going up at the moment 4pm UK time and should be there pretty soon, you'll need to right click and save as to get them.
Click to expand...

Much appreciated!
 
Apart from resolution, I'm impressed with Alexa's handling of noise & latitude.

I'll look after the R3D and see if I'm missing anything.

Patrick
 
My first short look at two files Alexa-norm-log0021.dpx and RED-norm-log.dpx showed up that RED MX was rated wrong at monitor out and EVF @ 400 ISO just because 800 ISO is sweet spot of MX sensor (that was repeated 100 times here at the forum).

Maybe that could be the main reason that testers couldn't go better with RED MX.

For my point of view the test felt short after that before even started...

Also seems to me that Alexa file look sharper than RED MX that is natively 4K.

In that case Alexa probably was more accurate focused at the lens, could be a sort of test trick in favor to Alexa just because RED has always razor sharp image.

Then RED color rendering looks more natural than a bleaker Alexa image.

Waiting for R3Ds...

In the meantime have a look:

RED_MX_normal_CML1-detail.jpg

RED_MX_normal.dpx processed in Shake with GenArts Film Effect Kodak_5245/Kodak_2383 (the left image corner detail from 1080p file).

Alexa_normal_CML1-detail.jpg

Alexa_normal_CML1-detail.dpx processed in Shake with GenArts Film Effect Kodak_5245/Kodak_2383 (the left image corner detail from 1080p file).
 
Sanjin, in all fairness to Geoff, I wouldn't jump to that conclusion so fast. Alexa's sweet spot is also supposed to be around 800. There's more to it then just ISO rating.
 
Sanjin Jukic said:
My first short look at two files Alexa-norm-log0021.dpx and RED-norm-log.dpx showed up that RED MX was rated wrong @ monitor and EFF 400 ISO just because 800 ISO is sweet spot of MX sensor (that was repeated 100 times here at the forum).

Maybe that could be the main reason that testers couldn't go better with RED MX.
Click to expand...

At first I thought so too. But in that case the RED should have an advantage in the shadows since essentially you are giving it twice the amount of light that it should get.

However, if you look at the darkest files they match pretty much the Alexa, maybe the Alexa even does a tiny better with a bit less noise. So I don't think iso 800 is the problem here. If the RED was exposed for iso 800 the lightest files would have more detail but the darkest files would lose detail, there would not be more dynamic range.

I tried to download the R3D files but I guess they aren't uploaded yet. So then I went to RedCine-X, loaded an overexposed file that I had myself and see what happens when you output to Red filmlog: basically all the information will be in the file. Overexposed means information gone.

So I guess Alexa just looks good...
 
Steve Sherrick said:
Sanjin, in all fairness to Geoff, I wouldn't jump to that conclusion so fast. Alexa's sweet spot is also supposed to be around 800. There's more to it then just ISO rating.
Click to expand...

For me a fairness to Geoff means exactly the fact talk otherwise is not making sense...

According to a lot of user reports Alexa had ISO rating issues so then...all left in this direction of ISO 400...

That could be the main reason why Arri still can't deliver RAW almost a year after Alexa's launch (but they promised next week is finally coming up!!!???)...

Also the image rendering (resolution/color/latitude) between Alexa and RED MX is a difference something

like an aquarelle watercolor painting (Alexa) and Hyperrealism painting using oil (RED MX)

Both are difficult painting technologies but Hyperrealism painting when using oil can produce a higher realistic, sharper, more detailed image...

Btw, an aquarelle watercolor painting is more known for a smaller picture size like scathes (1080p/2K) than an oil painting that can be really in a huge wall size (4K, 4.5K, 5K+)...

All it's up to you and what would you like more...
 
Sanjin Jukic said:
For me a fairness to Geoff means exactly the fact talk otherwise is not making sense...

According to a lot of user reports Alexa had ISO rating issues so then...all left in this direction of ISO 400...

Also the image rendering (resolution/color/latitude) between Alexa and RED MX is something like an

aquarelle watercolor painting (Alexa) and Hyperrealism painting using oil (RED MX)

Both are difficult painting technologies but Hyperrealism painting using oil can produce a higher realistic, sharper, more detailed image...

All it's up to you and what would you like more...
Click to expand...
As Geoff mentions, this is not a shootout. These are results he came to given the set of parameters he used. YMMV.

The most important thing about viewing those Log-C files is that you are able to view them properly, i.e. with an accurate LUT. So just as we would want to have our R3Ds viewed correctly, same goes for Alexa Log-C. Hopefully you are able to do that to analyze the images properly.

What this test doesn't tell us is how ArriRAW stacks up, nor does it take into account Epic. Geoff didn't have access to them as far as I can tell.

I honestly believe that at this point, many DPs have tested M-X and they have tested Alexa. They will make choices based on which camera delivers what they need for that given project. The two variables in the Arri-RED debate are Alexa using ArriRAW which not many people have had the opportunity to test outside the select few, and Epic with its increased dynamic rage capabilities. Having seen all the cameras just mentioned in action, I have to say that I know what I would reach for in most situations. Everyone else must come to their own conclusions.
 
Thanks Geoff for your kind publication of the test.

I don't understand :

"I stopped rendering material from any camera when I couldn't see where the exposure indicator at the bottom was."

We still see a lot of details in the underexposed R3D, in the exposure indicator.

Pat
 
Status
Not open for further replies.
Back
Top