Dynamic Range test (real one)

Status
Not open for further replies.
Ok great. Seems like a great tool with the automation.

But wouldn't you want to rate the camera as 5000k native then as that will boast the max dynamic range and color fidelity.

Also I can't really get why some wouldn't want to rate a cameras sensitivity. Most DSLR's seem to have 10 1/2 stop of RAW dynamic range and puts normal exposure about 4 1/2 stops below saturation, which translates to an over exposure latitude of about 2 1/6 of a stop, over the white patch. Wouldn't you want to do something alike with the RED.
 
Anders, we can choose what is native kelvin for the camera. Would you like 5000k?

We can rate a sensitivity or iso, but you can rate it yourself too, picking how many stops above and below you want. Looking at DPReview, DSLR's don't have 10.5 stops though - but again, hard to rate given that white clip is easy to spot, but black detail into noise floor is hard to determine.

Graeme
 
yes indeed.. great post Glen.

Filters are great for balancing things out, but as Graeme mentioned, putting on a color filter always results in a loss of light from the reduced transmission optically of the filter..

So you kinda defeat the advantage because by shifting down the highlights with a filter you also push down the lows past the black clip point.

Thats why DSP's use curves, and even better when they use curves in individual channels, to roll off highlights.

I think the Wedge test should be the new standard for digital DR tests.. its only a $40 chart (calibrated!!), is backlit which is usually easier to get more accurate results, and is a whole lot easier then the classic method with alot less margin for error.

As Graeme mentioned during the shoot today, ramping stops on lenses and using ND's really is testing the accuracy of the markings on the glass and the accuracy of the ND's.
 
That's right Jarred. I mean, exactly how accurate are those marks on the lens, and how accurate and repeatable can you set them. Given each camera tested on the CML graph used a different lens, I can see this as a big source of error. With this method, the accuracy of the lens' markings is eliminated, as is the user error of setting the lens to such marks. As you get down into the shadows, you're trying to see if there's still detail not being masked by sensor and quantisation noise. This is not easy. Practically any source of error will show up as distortion of the curve in the shadows. I'm not saying our test has no errors, but it was very repeatable and we did try to eliminate as many sources of error as possible, including full automation of the measuring of the chart.

Also, I think this chart is also going to show you the minimum dynamic range, rather than a more liberal range which the multiple shot method contributes to, as pointed out by Adam Wilt.

Graeme
 
Filters are great for balancing things out, but as Graeme mentioned, putting on a color filter always results in a loss of light from the reduced transmission optically of the filter..
That's ok since most of the time you will have enough light. Where you don't have enough light, it's not a big deal to ditch the filters to get the most low light possible.

2- What if you could change the color filters on the bayer sensor itself? (since each photoshop needs a color filter over it.)

In low light situations, perhaps you could optimize sensitivity by using filter choices that let in more light. i.e. the Sony F900 lets in a lot of infrared-ish light to increase sensitivity.

And for the color accuracy nuts, I'd like to see to a RGB+E filter option please. :D (Although Sony likely holds some sort of patent for that; it is arguably an obvious improvement though, if you're trying to reproduce the appropriate CIE CMFs).
On the other hand, I've never seen accurate color... so even if you do achieve it, it may be that no one cares.

FYI: Accurate color is really hard even if it's your eyes looking at the real world... look up metamerism.
 
The following is a CML post from Jim. You continue to impress and inspire me with your mad skills.


" s1st... my apologies (again) for stirring up a sh$tstorm.

We have struggled from the beginning here at CML. My introduction to Geoff was rocky. I'm betting we would do much better if we could start over.

I had hoped for constructive criticism when I 1st posted here. No one told me I needed a helmet. We really don't mean to be a disruptive force on this board, but it seems we are. Maybe it is our frustration showing. We recently posted, with Geoff's assistance, many test shots of green screen work done with the help of David Stump. I don't recall one post of feedback about the quality of the images. The next thing that happens is an exposure latitude chart that puts us below consumer cameras. You might imagine your reaction if this was your project and the results of that chart did not match up with the results you had in-house.

The testing of dynamic range is best done on a single exposure. The Stouffer T4110c is the best tool. If you take multiple exposures while adjusting the f-stop and using neutral density filters that are not calibrated... the results are influenced by several non-sensor factors. The DSLR world has already embraced this testing methodology for good reason. And like it or not, these are high speed DSLRs.

We set up our own lab for this test. It is easy and accurate. Graeme spent a whole day automating the results process. There is no speculation to the results. Our suggestion is that David Stump come to RED to supervise the process (we invited him today). We cannot tell anyone what to do with their testing process but we encourage all interested to investigate this methodology.

It sounds like Jessica would delete the "future cameras" from this list. I'm OK with that. We will be done soon. In some ways it might be better to wait until our project is finished before discussing it.

Geoff... I appreciate you hosting the RED images. I appreciate your early skepticism at NAB. It has motivated me to not only finish the project, but to deliver more than we originally promised. I do not want you to grovel, anymore than I want to. You have built a great home here. It is possible that there is just not a room for us. We are as passionate about this project as cinematographers are about their work. Everyone wants their work to be shown in the best light.

Jim "

Aloha
-A
 
I hope having to deal with this little "crisis" hasn't slowed REAL work on the camera, or tied up the RED Team. Testing is useful, but wouldn't it all be better if done on, at least, the next generation of RED test platform?

Won't critics simply suggest that whatever results are obtained today not be reflective of the 2nd and 3rd generation of RED prototypes? I remember something similar just last year with the HVX200 and results of first beta units compared to shipping ones...

I'll say it again, you guys should just go information DARK until those handful of field ready prototypes are pulling in footage. All this drama from the "full disclosure" of process seems ultimately distracting from the real job at hand...
 
Jim... this is a valid point of view. We do have our hands full. And the drama is distracting. But we feel obligated to keep our reservation holders informed. The "open development" process can't just be when it is convenient to us. We are committed to it, for better or worse.

Jim
 
Graeme_Nattress said:
. I mean, exactly how accurate are those marks on the lens, and how accurate and repeatable can you set them.
Graeme
Click to expand...

Graeme,

Cooke & Zeiss lenses are indivulally calibrated for T stops & Focus marks, one of the reasons they are expensive.

Stephen
 
Stephen... are you sure that you can set f4 EXACTLY on a Cooke lens? We can't. And do you use calibrated NDs when performing a multiple shot dynamic range test with NDs? A single shot Stouffer T4110c chart takes the guess work and imprecision out of the process. Case in point... the CML test showed the RED camera at 8 stops. Our test (same methodology as dpreview) shows 11 1/3 stops. Out test is repeatable. The CML test has no published methodology. Our 12 bit A-D suggests a 12 stop possibility. The sensor specs call for >66db. Repeat testing using the Stouffer chart and careful testing validate the specs. Which do you think is the accurate testing method?

Jim
 
Last edited:
Jannard said:
Stephen... are you sure that you can set f4 EXACTLY on a Cooke lens? We can't. And do you use calibrated NDs when performing a multiple shot dynamic range test with NDs? A single shot Stouffer T4110c chart takes the guess work and imprecision out of the process. Case in point... the CML test showed the RED camera at 8 stops. Our test (same methodology as dpreview) shows 11 1/3 stops. Out test is repeatable. The CML test has no published methodology. Our 12 bit A-D suggests a 12 stop possibility. The sensor specs call for >66db. Repeat testing using the Stouffer chart and careful testing validate the specs. Which do you think is the accurate testing method?

Jim
Click to expand...

Hi Jim,

I never stated that I could set a Cooke S4 lens exactly. I am sure I would be within 1/10th stop, as the wedges are 1/2 stop intervals, I don't see that as a major problem myself.

I don't use ND filters with wedge tests (neither does Geoff Boyle). I think you need to ask David Stump that question.

The CLM methodology I understand as follows:- Light the chart for T2 and stop down, relight the chart for T22 & open up. The beauty is that anybody can repeat the tests themselves at home.

I understand that the sensor has been tweaked to produce more than 2 extra stops of dynamic range, that would be confirmed by the difference of the 2 results.

I don't have a problem with David Stumps methodology of testing or Dprevew's. I would be interested for other cameras to be tested with a Stouffer chart.

In conclusion, I don't think there is anything wrong with your sensor, I am waiting patiently in line for may turn to test the camera!

Best wishes

Stephen
 
[shrugs] Sometime next year somebody will get all of the market's 4K cameras together and do a proper test using one lens swapped between bodies. Then we'll know.

Until then, I'd love to see the test frame if possible. I understand if it's not one of those things that the team wants posted.
 
Brook... I am not sure if you know what you are asking for. The frame is dark (RAW). Do you have the tools to interpret it? The images at dpreview are re-manufactured from data off the RAW image.

1166070872.jpg


Jim
 
Oh, don't worry, I understand that it's dark. I have the tools handy to process and analyze the RAW file... anything from Photoshop to shake or SCRATCH, depending on what format we're talking about here. I'm interested because I'm planning on shooting a similar test with 5217 and 5218 in the next few weeks [ingesting@2K via ARRISCAN] and a discussion regarding dynamic range and grain/gain structure could be quite interesting.
 
Brook Willard said:
[shrugs] Sometime next year somebody will get all of the market's 4K cameras together and do a proper test using one lens swapped between bodies. Then we'll know.
.
Click to expand...

Hi Brook,

There will only be 2 cameras AFAIK, better to include 2K & HD cameras IMHO.

Stephen
 
Stephen Williams said:
Hi Brook,

There will only be 2 cameras AFAIK, better to include 2K & HD cameras IMHO.

Stephen
Click to expand...
Agreed, the more the merrier. I'd just love to see the test performed in the same environment with the same glass... that was my point.
 
Brook Willard said:
Agreed, the more the merrier. I'd just love to see the test performed in the same environment with the same glass... that was my point.
Click to expand...

Hi Brook,

Unfortunately the Genesis, Viper & Sony cameras do not support PL mount. I guess Genesis will use PV glass, Viper & Sony Zeiss digi primes, the rest Zeiss or Cooke glass.

Stephen
 
Jim,

I was actually referring to the kinds of images posted in their dpreview.com D200 review (not synthesized). Like this one:

dpreview.jpg


All the best with the camera! I can't wait to use it (admittedly I'll probably be renting...).

Bruce
 
Bugger, shows the extent of my experience with said cameras. I'm so locked into working with Arris where I go to school and where I work that I often forget to look beyond.
 
Status
Not open for further replies.
Back
Top