Hannover Hands On comparisons of RED F35 & D21

[Martin Stevens]

In regards to testing Red One footage.....I would have to say that seeing movies like "Knowing" and "Che" in the theater and on Blu-ray, as well as seeing Red demo footage at NAB via 4K projection have been the REAL test and the footage has been fantastic.

Yes we all want more dynamic range etc....and it's coming.

Oh, and perhaps it best to view mistakes made by Geoff as simply mistakes. We are only human after all.

 

[David Battistella]

Here below is an example how it should be done a proper test with RED ONE camera by any "professional".

The first you shoot with RED1 that has to be set up for a time-lapse shots with duration of one second

bacause to have a small R3D file to publish and later share with others who would like to process it by themselves.

Simply showing of JEPGs or even DPX Log don't work because R3D RAW files files were already have been processed.

You always have to give the original R3D file as your test reference file and no any other.

By the help of RED Team test instructions I did a pilot (test) for RED B20 and here we go:

Lens: Leica Summilux 50mm f/1.4 @ f/1.4 (wide open),

Tungsen light: (a big bulb E27 75W 220V) coming from a left side at approximately the same distance between object and camera.

Color light meter: Gossen Colormaster 3F

Color Charts: Color checker by GretaMacBeth, Gamma-3cP Chart by Gamma & Density Co, Grey Card by FotoWand (Neutral Grey DN=0.75 log D, 17.68 % Reflection).

The original R3D file is opened in REDAlert set to Camera RGB and RED Log still with no any other adjustment applied.

- camera's AUTO white balance = 2605K

- white balance tool in RED Alert = 2498K

- light color meter = 2450K

Then you can offer original R3D file for a quick download that community can have a look properly at the shot and see what happened there during the test.

Original file to download:

A003_C001_06053N_001.R3D >>>Right-click on the "Save As..."

The R3D file size is just about 934K.


----------------------


Hope this helps.

thanks sanjin!

david

 

[J. Eric Camp]

Alright,

First off I am aghast at some of the reactions and even more so to hear that Jim has already delete the ones that were more offensive. Some of us would do well to remember that reputation and perception are EVERYTHING in this industry. Being known as a brash, rude zealot will not serve anyone.

That being said there are in fact errors in the methodology of this test. Tests like this with omissions, liberties, and errors being proffered by respected professionals do in fact spread misinformation in an infectious manner through the industry.

It is unfortunate and does warrant public clarification in the same forums where is has been distributed. However it does not warrant the condemnation of a seasoned professional and a corresponding generation of peers.

 

[Todd M.]

Alright,

First off I am aghast at some of the reactions and even more so to hear that Jim has already delete the ones that were more offensive. Some of us would do well to remember that reputation and perception are EVERYTHING in this industry. Being known as a brash, rude zealot will not serve anyone.

That being said there are in fact errors in the methodology of this test. Tests like this with omissions, liberties, and errors being proffered by respected professionals do in fact spread misinformation in an infectious manner through the industry.

It is unfortunate and does warrant public clarification in the same forums where is has been distributed. However it does not warrant the condemnation of a seasoned professional and a corresponding generation of peers.

Well put.

 

[Duke M]

Of note is the fact that some of the loudest complaints have been posted by people that have never tried to conduct any tests and post them.

I have on other boards and can state that the best you can ever do is an approximately equal test due to the strengths and weaknesses of each camera, as well as simple human error in doing anything that requires repetition. This is especially true with multi camera tests because each camera has its own quirks to bring out the best of that camera.

If he has the original files and can correct the errors great. If not move on. There may be errors on the files of the other cameras too. If you think you can do better, we'd love to see your tests.

 

[Joel Kaye]

It is adjacent to comparing someone walking into a BestBuy that sells multiple brands of computers and saying "Apple sucks" vs. someone walking into the Apple store and saying "Apple sucks".

Hmmm. I never thought of it that way. That's a good point.

Anyway, it would be nice if there was an addendum to the test page noting the issues raised in this thread.

Seems like a link back to this thread would be fair considering this thread links there... just sayin'. Has there even been an acknowledgment of the channel swapping and 8 bit vs 10 bit issues? Clearly they are real.

 

[Jannard]

I have received several private emails from redusers that posted here about Geoff's test and the intent to "stand up for me" because CML has been so disrespectful to me in the past. While I certainly appreciate the loyalty, it is not a good reason to be disrespectful to Geoff here. I have to say that I have been very frustrated about the way I have been treated at CML. But it is not good to resort to the same level of disrespect here. At some point, we are what we post. Reduser has to be above personal attacks and disrespectful posts... no matter what.

Information is power. I am all for accurate info, healthy debate and even disagreement based on perception. But name calling and disrespectful posts have no benefit to anyone. All they do is polarize.

I am happy for RED to stand on it's merits. We have plenty of examples that RED is successful in shooting feature films, episodics, commercials and documentaries without having to resort to inelegant trash talk.

Please honor our wishes to treat everyone here as "industry interested". Healthy debate is fine. Personal attacks are not.

Jim

 

[Ariana]

The dangerous thing about tests like this is that conclusions were made and stated confidently on CML by Geoff regarding where the cameras placed in the competion (ranking it turns the tests into a competition regardless of what the intention is).

What would happen if you replace Sony, ARRI, SI, and RED with Geoff, Claudio, Roberto, and Dave Stump in a competitive "test"?

Everyone loses in these tests. Mostly it's us users and owners.

 

[Jannard]

There are a 1000 ways to grade RED footage... from wrong to wonderful. If there is a test that goes beyond r3d's and includes grading, I'd certainly have a few people that I would pick to grade RED footage in any test.

Jim

 

[I Bloom]

Just not sure what all the excitement is about. Geoff just gave the Red a 2 bit handicap because it's 1/4 of the price. It's mathematically perfect I think

IBloom

 

[M Most]

There are a 1000 ways to grade RED footage... from wrong to wonderful. If there is a test that goes beyond r3d's and includes grading, I'd certainly have a few people that I would pick to grade RED footage in any test.

But therein lies part of the problem. You seem to be regarding processing ("developing") and grading as the same thing. They're not. Processing is a basic step to get from the original information in a form that isn't really an image yet to something that, well, IS an image. Grading is an interpretive step that is based on some technical principles, but is largely artistic and based on personal preference. In many ways, there is no "right" or "wrong" when it comes to taste. In the case of Geoff's material, he was not trying to grade anything. That was the whole point of the test. What he was trying to do is examine the capabilities of the camera - in fact, of all the cameras - in terms of the base image it yields, without the influence of either metadata or an interpretive grade. Clearly the methodology used for this was not correct, but that has nothing to do with "grading," it has to do with the basic parameters needed to yield a non-interpretive image that is the product of the camera system. The interpretive step would normally be done by a colorist, something Geoff is not (and he would be the first to tell you that). If grading was involved, all of the images he posted would look a hell of a lot better than they do, but displaying the base image without grading - in log form, something you would normally not look at in the first place - was, in this case, the whole idea.

I understand your sensitivity to having anything shown that represents an image from a Red camera in anything other than final, perfected form. I really do. But when certain tests are involved, that's not the intent. It's not meant to "look good," it's meant to demonstrate limitations, if any, in basic imaging capabilities. To see what's "really there." This is what cameramen need to know, and this is what a lot of these kinds of tests are designed to do. And yes, in that sense, this one did have some problems that I'm sure Geoff is already addressing, but even once addressed, the images won't be - and shouldn't be - "graded" because that's not part of the test.

 

[Gavin Greenwalt]

Can I make a suggestion. And I know I'm going to get shot down by a number of people here but.... Can we please use this experience as a lesson in why DPXs are a poor image format to choose for technical tests. How about EXRs? If you can output a linear EXR from all cameras in a test with little to no processing then everybody can process them, themselves. The data would be less molested. If you accidentally set the ISO wrong then then highlights would just be overbright. If you processed the shadows wrong the shadows would just be negative and easy to restore with a simple grade.

Seeing as storage size for a still is rather inconsequential and it doesn't need to be RT it seems like a more technical and forgiving file format would be more fitting in such circumstances. Although in this test that assumes that Speedgrade outputs EXRs with overbrights and negative values. Maybe it doesn't.

I also know first hand as do many others that getting all of the data out of an r3d file isn't always a foolproof system. I for one would really appreciate an increased level of idiot proofing in ensuring I wasn't clipping data when outputing from REDCine or REDAlert unnecessarily. Even like a checkbox "Compress Dynamic Range" which takes all of my color and curves etc but applies one last range compression to fit my output format.

EDIT:

Format | Read | Write

OpenEXR | Yes | No

Nevermind looks like that's a no-go in this test.

 

[geoffboyle]

I do certainly take exception to the fact that these tests don't seem to have been done with much care. 8 bit vs. 10 bit and transposed colors seems like a rather large error in processing. Because Geoff DOES have credibility and IS unbiased, this is a rather serious situation if people make a judgment of RED while thinking these tests make RED look extraordinarily bad relative to competitive cameras.
Jim

I will try processing again, I don't have Mac machines so Red Alert is not available to me.

I tried using RedCine but unfortunately it wouldn't render on my system, it used to but something has changed. I get messages saying to look at NVidea's help site, of course there is nothing there .

Let me check what I have done, what got processed how and try again.

I have no issue here.

My comments are all based on what we saw at Hannover, not what I uploaded, which I assumed was the same.

The material in Hannover was mostly processed in Digital Vision Film Master.

Oh and I know that I'll never do a test that will satisfy people, I've been publishing them for many years now and getting grief every time.

I have happily shot with the RED, as a lot of you know, it doesn't have the same latitude as the F35 but Hell! I don't expect it to for the price.

I was simply trying to understand what happens.

I'll happily mail R3D files to RED for processing, but let me re attempt the transfers.

Oh and I'm not a poor wimp who'll be broken by criticism, bring it on

I'm happy for people to be as offensive as they like, it tells us more about them than anything else.

Tell me where I went wrong and what I can do better and I will.

 

[Gunleik Groven]

Thanks Geoff!

A link to the RAW files would be a most appreciated addition to re-developed DPXs!



Gunleik

 

[I Bloom]

Tell me where I went wrong and what I can do better and I will.

Despite being in a default setting the software is not all created equal. There are many small details of math that make the different programs perform very differently. In fact you could say with some certainty that no two of the available programs work alike, so how the processed image is perceived really is dependent on what pipeline you are using, to a certain extent.

In the case of Red you really should use RedAlert (and publish the settings you used in the program) that will be the most baseline image you can create. Publishing raw files will also allow others to verify your test results.

One thing no one mentions is how much work these tests are. And boring and tedious work for guys used shooting movies. So despite the jokes it's good to have someone willing to do those tests and spend the time to publish the results. So thanks for that Geoff.

IBloom

 

[Sanjin Jukic]

But therein lies part of the problem. You seem to be regarding processing ("developing") and grading as the same thing. They're not. Processing is a basic step to get from the original information in a form that isn't really an image yet to something that, well, IS an image. Grading is an interpretive step that is based on some technical principles, but is largely artistic and based on personal preference. In many ways, there is no "right" or "wrong" when it comes to taste. In the case of Geoff's material, he was not trying to grade anything. That was the whole point of the test. What he was trying to do is examine the capabilities of the camera - in fact, of all the cameras - in terms of the base image it yields, without the influence of either metadata or an interpretive grade. Clearly the methodology used for this was not correct, but that has nothing to do with "grading," it has to do with the basic parameters needed to yield a non-interpretive image that is the product of the camera system. The interpretive step would normally be done by a colorist, something Geoff is not (and he would be the first to tell you that). If grading was involved, all of the images he posted would look a hell of a lot better than they do, but displaying the base image without grading - in log form, something you would normally not look at in the first place - was, in this case, the whole idea.

I understand your sensitivity to having anything shown that represents an image from a Red camera in anything other than final, perfected form. I really do. But when certain tests are involved, that's not the intent. It's not meant to "look good," it's meant to demonstrate limitations, if any, in basic imaging capabilities. To see what's "really there." This is what cameramen need to know, and this is what a lot of these kinds of tests are designed to do. And yes, in that sense, this one did have some problems that I'm sure Geoff is already addressing, but even once addressed, the images won't be - and shouldn't be - "graded" because that's not part of the test.

I do not know how Sony F35 or Arri D21 are working and processing image after that in a computer

but I do know pretty well how RED1 is working and also to make proper R1 image processing in a Mac.

It's a strange to see that R3D processing at "CML Hannover Hands of RED F35 and D21" test sees the lamp light like a blue instead of a beige/orange.

"Incorrect precessed" RED1 shot as a "blue" lamp is at the bottom of the image.

For example my R1 sees a light exactly as it is.

Here are two recent examples:

Example 1>>>

LED light (500W) example, shot handheld on RED1 with Canon 50mm f/1.2 FD lens @ f/22, no filters, processed in REDAlert...

See the footage>>>


Example 2>>>

Ecological low bulb light (11W) at the floor lamp that has a couple of gel rolls inside to make custom color change, shot on RED1 with Canon 85mm f/1.2 FD lens @ f/5.6, no filters,...
The footage and picture were directly output from QT proxy file, no grade, no other app processing except that is exactly how QT proxy file shows the shot...

See the footage>>>

Hope this helps to finally put RED1 testing in the right direction.

 

[geoffboyle]

I screwed up

I screwed up

OK, I've looked at the files and also the workflow.

I screwed up, and not only the RED material but some D-21 that hasn't been uploaded yet.

I rendered everything at full res for my own use, I don't have a fast enough connection to upload a lot of 4K files.

Instead of rendering everything again in SG to 1080, which I should have done and have now done, I used Xnview to generate reduced res files.

I had no idea that this would result in 8 bit files, I am totally used to all DPX files being 10 bit.

I have corrected this and re-rendered everything from scratch.

I have also added the files that I originally omitted.

OK, let the criticism fly!

Of course the point of doing things like this is to learn.

And I use DPX and not any other format because that's the industry standard.

Oh and if anyone wants to donate a Mac and software I'll happily try that and again if anyone wants to pay for a leased line to my home then I'll upload all the original R3D files.

I do have to deal with the realities of life!!

 

[Sanjin Jukic]

OK, I've looked at the files and also the workflow.

I screwed up, and not only the RED material but some D-21 that hasn't been uploaded yet.

I rendered everything at full res for my own use, I don't have a fast enough connection to upload a lot of 4K files.

Instead of rendering everything again in SG to 1080, which I should have done and have now done, I used Xnview to generate reduced res files.

I had no idea that this would result in 8 bit files, I am totally used to all DPX files being 10 bit.

I have corrected this and re-rendered everything from scratch.

I have also added the files that I originally omitted.

OK, let the criticism fly!

Of course the point of doing things like this is to learn.

And I use DPX and not any other format because that's the industry standard.

Oh and if anyone wants to donate a Mac and software I'll happily try that and again if anyone wants to pay for a leased line to my home then I'll upload all the original R3D files.

I do have to deal with the realities of life!!

Thanks and it's good to know that even a "professional" has to learn.

Hope that here at the forum the other "professionals" would follow your (learning) steps...

Also btw, one of the best R3D app from the beginning on is not on Mac based.

It's about Scratch and works on PC only (unfortunately).

Try to get Scratch free 15 days evaluation and then do all properly to conform at DPX as you've said it's the industry standard.

Hope this helps.

 

[geoffboyle]

It's about Scratch and works on PC only (unfortunately).

Try to get Scratch free 15 days evaluation and then do all properly to conform at DPX as you've said it's the industry standard.

Hope this helps.

I've used Scratch in the past, Lucas kindly loaned me a system installed on a MF of a laptop!

It worked well isn't RedCine based on this?

I use SpeedGrade, which works very well, mainly because I use SG OnSet on every shoot and can't imagine going anywhere without it!

Anyway, the software wasn't at fault here, I was.

 

[geoffboyle]

I have no stake in any of this. But, let me say something reasonable about it.

Geoff's error may have already caused some real-world damage to Red's reputation across the world with people who saw the first results and will not see any corrections.

Given that, I think Geoff should do several things to make amends:

1) Reprocess the files according to the best possible settings, possibly using some other software (Redalert)???

2) Offer the original R3D files to the Red team so that they can also reprocess the files, and compare against the future results.

3) Offer a correction to the original article, that most who read the original are likely to see.

The name-calling is ugly.

I think I've done all that now.

I really don't think that a couple of days of incorrect images will do a hell of a lot to Red's reputation

Oh and Iridas are also blameless in this.

It is all my fault, trying to save time.

I wasn't paying any attention to the images as I uploaded them, I've been looking at them for weeks before this!!

Only the ones that I was looking at were correctly processed :-(

I hope that the newly uploaded ones resolve everything, I'm off to take a good look at them myself