Determining ISO...

[Dj Joofa]

And how are you measuring the noise threshold? How much noise is acceptable? A test like the Kodak test seems like the least scientific test possible. Is that mixed light? Same color temperature throughout the scene? How accurate is the spot meter?

The DSC Labs test is a great one. I should have said Stouffer-like which includes several calibrated charts available. But even with these tests, someone needs to determine the amount of acceptable noise which is very subjective.

Jim

Do you use ImaTest?

 

[Finner]

A test like the Kodak test seems like the least scientific test possible.

Jim

I agree with you there Jim but I find to often the "science" end of things looses the art end. I get paid to direct and shoot scenes not charts and though charts offer up some good information I find there is nothing better then shooting a high dynamic range shot and comparing multiple cameras that shot it. It gives tangeble real world results that I can come to rely on and trust. Learning from it how cameras will best be used.

To each person there own system. I have just found this system to be one of the best for me.

 

[Jannard]

Graeme uses ImaTest as one of his tools.

Jim

 

[Jarred Land]

I agree with you there Jim but I find to often the "science" end of things looses the art end. I get paid to direct and shoot scenes not charts and though charts offer up some good information I find there is nothing better then shooting a high dynamic range shot and comparing multiple cameras that shot it. It gives tangeble real world results that I can come to rely on and trust. Learning from it how cameras will best be used.

To each person there own system. I have just found this system to be one of the best for me.

I agree Finner.. the best blend of science and art in my opinion is to shoot a scene with both highs and lows in one frame and compare the results. Ive never been a fan of the traditional progressing luminance chart tests; a stouffer chart is one way that is more on the scientific end to shoot range in one shot, but still leaves a little of the subjective artistic aspect to the results as Jim said.

The days of using a light meter are numbered.. I guess that's why Minolta got out while they could.

 

[Brent J. Craig]

A few questions that kept coming up in our tests yesterday:

1. Can we believe the waveforms and histograms on the Red LCD? Are they generated from the RAW sensor data or from the Rec709 processed video?

2. Can we or will we be able to build a LUT and apply it to the video outputs?

3. Can we assign one of the programmable buttons to false-color?

 

[Graeme Nattress]

Test charts like the Stouffer are great, especially when calibrated as they're so much more repeatable than using a scene to measure dynamic range. Using a scene to compare two cameras is totally fraught with problems because it's very hard to visually trade highlight stops for shadow stops, especially when you're looking at non-linear data. With a Stouffer or similar, it doesn't matter at which box you clip on at the top, as you just count down until you can't see the join from one box to the next and call it. It doesn't matter if one camera sees the image brighter or darker than the other.

I tried Imatest and it's certainly fun and useful, but even though I was giving it data with a straight power law gamma curve, it would not correctly interpret it as such to get to the linear data for measurement. I also found that putting different curves, all of which were preserving full dynamic range, onto the data gave wildly different results. My own methods with the Stouffer couldn't care less what curve is on the data, and give the same results.

I've seen some images from users with terrific dynamic range. I have one lovely shot where the highlights are just below clipping, and the darkest shadows are 12 stops below and have detail (yes noisy, but detail) in them.

One of the nice things with the RED data is that I can access the linear light data, and knowing from measurement how that data responds to the input light, I can use the data as a light meter and look at which tones are recorded in which stop.

Graeme

 

[Dylan Macleod CSC]

And how are you measuring the noise threshold? How much noise is acceptable? A test like the Kodak test seems much less scientific a test than a calibrated chart. Is that mixed light? Same color temperature throughout the scene? Or almost the same... How accurate is the spot meter? Are between stop measurements available? Building an accurate plot from this test would be impossible.

The DSC Labs test is a great one. I should have said Stouffer-like which includes several calibrated charts available. But even with these tests, someone needs to determine the amount of acceptable noise which is very subjective.

Jim

Exactly...noise threshold is subjective. My "quick analysis" is again based on a "quick" look on a monitor that was not calibrated but we all agreed that we could see 10 separate levels of gray on the DSC chart.

We did the standard "over under" with a face and a macbeth and just judging what was acceptable by eye...it felt like 2 1/3 over...2 under.

Someone else might be more tolerant of "noise" and determine 3 over and 2 1/2 under as "acceptable".

Dylan Macleod
Cinematographer
Toronto, Canada
www.dylanmacleod.com

 

[Dylan Macleod CSC]

A few questions that kept coming up in our tests yesterday:

1. Can we believe the waveforms and histograms on the Red LCD? Are they generated from the RAW sensor data or from the Rec709 processed video?

2. Can we or will we be able to build a LUT and apply it to the video outputs?

3. Can we assign one of the programmable buttons to false-color?

This first question is how I started this thread. I know the RED TEAM are reading this thread, but no-one has commented????

The 2nd question is a great idea that Chris mentioned...also in this thread...

Any on the RED TEAM care to chime in?

Dylan Macleod
Cinematographer
Toronto, Canada
www.dylanmacleod.com

 

[David Mullen ASC]

Part of this is simply the difference between actual dynamic range as measured scientifically by an engineer... and practical, workable dynamic range as a cinematographer thinks about it.

For example, when shooting color negative film, I generally "think" of five stops over being near white and five stops under as being black, so essentially I'm working within a 10-stop practical range. But obviously Kodak engineers are not going to say that film negative only has a 10-stop range because they can measure more scientifically.

Also, the word "latitude" means something different than dynamic range. Latitude really describes the range of correction possible. Though material that is three-stops overexposed that is shot digitally may have some information throughout the image, that doesn't mean that it is correctable back to "normal" due to the amount of clipping. So you may say that three-stops over is the most over you generally want to go, but that doesn't necessarily mean that the dynamic range doesn't go beyond three-stops over.

This is why we're going to constantly hear different dynamic range figures tossed around. Someone shooting digitally is going to think he has about three over and five under to work within, which is only 8 stops in practical terms -- but the range may be 11 stops when measured scientifically.

 

[Chris Nuzzaco]

What type of file format will Red One take for user designed LUTs?

I have a small program called XLUT that lets me make LUTs for all sorts of programs like Combustion, and even my Andromeda camera. I can set the input and output bit depths to pretty much anything I want.

 

[David Mullen ASC]

Obviously anyone can take the 8-bit 720P 4:2:2 REC 709 output for monitoring and stick whatever LUT they want onto it, but obviously it already is corrected for monitoring purposes.

The question is whether a LOG output would be any more informative for waveforms, and then sticking a LUT post that point on the chain for monitoring.

Considering the RAW recording is not LOG, maybe a LOG signal would not be more informative, but it all depends on whether the REC 709 linear monitor output contains or can contain all the luminence information. We know it doesn't when you use different ASA ratings, but is there an ASA rating setting where the REC 709 signal would contain all the luminence information on the RAW? If not, can there be the option of one?

 

[Chris Nuzzaco]

If I can monitor a linear signal via a linear 12 bit to 10 bit LUT with my Andromeda camera... I would think Red could do the same. I do realize its not the sensors true linearity due to the bit depth conversion, but I'm just saying... its possible.

Last I heard, there was a 12 bit linear Andromeda prototype working before Reel Stream closed its doors.

Correction: My monitor LUTs are 10 bit to 8 bit linear LUTs.

 

[David Mullen ASC]

Well, again, the Andromeda rig is monitoring the RGB digital video signal coming off the A-D processor, right? Whereas the RED generates a RAW signal, which cannot be monitored using traditional video equipment like a waveform until it is converted to RGB (I assume).

Maybe the issue is just where in the chain one monitors the 720P signal, or maybe the process of live conversion makes it hard to have access to all the luminence values for monitoring on waveforms and other scopes.

 

[Chris Nuzzaco]

The Andromeda camera is simply pulling in RAW sensor data like the Red One. It needs to be processed before it can be used.

But you have a good point about the RGB monitoring. I'm not sure, maybe thats what all of the installed hardware does, but that also opens the question as to what bit depth that RGB signal is.

Also, a CORRECTION:

I just noticed, my MONITOR LUTs are 10 bit to 8 bit. But again, they are Linear.

 

[Dylan Macleod CSC]

Could this be reason to bring back the "RAW PORT"?

This isn't really practical, I guess, because you don't want to give up the convenience of CF.

But there was a plan at one time to offer a RAW signal...which then could allow for different monitoring options...720 RGB with a custom LUTS? Some sort of conversion to a format that would offer a useful waveform signal measurement...etc.

Might start to get messy.

I am pretty new to RAW data capture, so I am talking a little beyond my means...

Dylan Macleod
Cinematographer
Toronto, Canada
www.dylanmacleod.com

 

[Harry Clark]

I have to say that I agree with the earlier poster who like the false color mode on the monitor. It's very quick and intuitive.
I wish there was a way that you could hit a button on the monitor to switch between normal and false color. Maybe one of the assignable user buttons on the camera?
I will also say that I am slightly amused by the debate of EXACTLY how much latitude the camera has. I'm more holistic in my approach, like David. If the camera has a rough range that you can get used to, like you favorite film stocks, then it becomes a bit academic whether or not there is useful information above and below that range. It's very subjective way out at the ends, and while it's true that charts are an absolute way to measure latitude, the real world is not.
Part of beauty the fast feedback loop of any electronic imaging system is the ability to trust that feedback system, like we trust our meters. Which by the way Jarred, are still excellent tools for scouting and prelighting.
Cheers,
Harry

 

[David Mullen ASC]

The Andromeda camera is simply pulling in RAW sensor data like the Red One. It needs to be processed before it can be used.

Chris, the Andromeda is a modification of a 3-CCD camera. "RAW" refers to the what you get from a single sensor with a Bayer pattern before it gets processed into RGB.

A 3-CCD camera naturally produces RGB. What you get with Andromeda is just an uncompressed 4:4:4 RGB signal. It just isn't in a format that can be monitored without conversion to a standard broadcast signal format, but it is still RGB.

 

[Chris Nuzzaco]

Chris, the Andromeda is a modification of a 3-CCD camera. "RAW" refers to the what you get from a single sensor with a Bayer pattern before it gets processed into RGB.

A 3-CCD camera naturally produces RGB. What you get with Andromeda is just an uncompressed 4:4:4 RGB signal. It just isn't in a format that can be monitored without conversion to a standard broadcast signal format, but it is still RGB.

The Andromeda actually treats all three chips like a Bayer filtered sensor due to the pixel shift of the sensors, so it works in the same manner, even though it does have three sensors. Like I said earlier, it needs to be processed in a RAW conversion program, complete with sharpening settings, output format, etc...

 

[David Mullen ASC]

That may be how it reconstructs the image later, but my point is that with 3 separate sensors capturing red, green, and blue separately, it's a lot easier to design a way of measuring their values being sent to the data recorder than with a RAW Bayer file pre-conversion to RGB.

According to their website, three channels are recorded separately:
http://www.reel-stream.com/andromeda_techspecs.php

 

[Chris Nuzzaco]

I see what you mean. Yeah, from a hardware point of view, it probably is easier to have three dedicated sensors.