$2500 budget - Which would you pick?

[canoli]

Thanks for the link to the RAW file. I'm curious how you got the .jpg version. Did you process it as an HDR and then convert to .jpg?


That shot really illustrates how it's underexposure that produces noise and not high ISOs. You helped me understand that awhile back, but this shot really shows that concept in action. I still can't believe how much misinformation is out there regarding noise and high ISOs...


thanks again. If you feel like sharing your conversion process I'd be interested to know.

[Daniel Browning]

I'm curious how you got the .jpg version. Did you process it as an HDR and then convert to .jpg?

That's pretty much it. I made two separate raw conversions in Raw Therapee then combined them in Photoshop with layers.

[canoli]

Ah thanks. I'm sure there's a few ways but that one seemed obvious.


btw, do you work in ("think" in) stops when you're converting RAW files? You mentioned Raw Therapee, which I've never used. I ask because none of the converters I use deal in stops, just the usual "exposure," "blacks," "brightness" etc. ACR, DPP, LR - they may use different words for the DR in an image but none of them use "stops."


You said that .jpg was processed for 11.5 stops and that if you process for 8 stops you won't see that banding. How do you know how many stops you're processing for? Is it just a subjective thing or are there converters out there that actually work in "stops of DR"?


thanks again!


[edit: checked Raw Therapee - thanks for mentioning it. One question, it says SSE is desirable but it doesn't specify the subsets. My Penryn goes up to 4.1 but not 4.2...wonder if it'll work well enough...any idea?]

[Jon Ruyle]

Oh no, is that sample really at ISO100? Looks like a bright sunny day...I assume it was processed to augment the banding in the shadows? - please tell me that's the case...

I believe there is way more dynamic in this pic than the jpeg makes it seem.

No, but it was processed for about 11.5
stops of dynamic range, and the banding appears around 9 stops or so.
If you process the image for only 8 stops of dynamic range, then the
foreground building will be completely black, so you wont see any
banding.

What this means is that if you viewed the raw, the face of the building would look black. Correct me if I am wrong, Daniel, but I guess you could produce a similar demo with the 1DsIII (or any other camera, I suppose), you just might require more dynamic range in the original (a blacker building face) to make it happen.


In other words, the real information about 5DII pattern noise came when Daniel told us it was processed for 11.5 stops of dynamic range, not when he showed us the jpeg.


One thing more: the 5DII much has less read noise at ISO 1600 than at ISO 100.

Had the reviews unanimously praised it I would've considered buying a new 7D. It really does have everything I could want in a camera body...except a 35mm-sized sensor.

I wouldn't let 7D detractors keep me from buying it. It is without a doubt the best 1.6 fovcf camera canon has made. On the other hand, I would most definitely let the 1.6 crop factor keep me from buying it if I wanted full frame. []

[canoli]

thanks Joh - I "hear ya" when it comes to the 7D reviews, I know you have to factor in the source always. But Bryan's review where he shows us the softness issues have been documented on other sites as well, and that is a little disturbing. But like I said, it doesn't matter much anyway (to me) because I really don't want another 1.6x body.


So the "stops of DR," is that a quantitative thing or is it subjective? How did Daniel know that shot was processed for 11.5 stops? Is there a converter that deals in stops, or is it all in the eye of the beholder so to speak?


thanks again for your reply!

[Daniel Browning]

btw, do you work in ("think" in) stops when you're converting RAW files?

No. I probably would if the raw converters made it easier.

I ask because none of the converters I use deal in stops, just the usual "exposure," "blacks," "brightness" etc. ACR, DPP, LR - they may use different words for the DR in an image but none of them use "stops."

Right. DPP's exposure slider is exactly 1 stop per number, while Adobe's exposure slider is slightly off from that. Some of the sliders (e.g. "fill light") do correlate with stops, but you have to convert it in your head ("10" is one stop, "20" is two, or something like that).

You said that .jpg was processed for 11.5 stops and that if you process for 8 stops you won't see that banding. How do you know how many stops you're processing for?

I measured it with Rawnalyze.

[edit: checked Raw Therapee - thanks for mentioning it. One question, it says SSE is desirable but it doesn't specify the subsets. My Penryn goes up to 4.1 but not 4.2...wonder if it'll work well enough...any idea?]

Yes, I'm certain yours will work great.

What this means is that if you viewed the raw, the face of the building would look black.

What you mean here by "viewed the raw" is actually "viewed a JPEG that was converted from the raw file with the settings that are typical for today's raw converters."

I think the difference is important, because the number of stops is really just an arbitrary choice. 8 years ago the default settings used 4-5 stops, today the defaults are set for 6-8 stops, and 8 years from now the defaults might be 9-14 stops.

When you actually view the raw (before the default gamma, tone curve, black clip, white point, etc.), you see only 1 stop of dynamic range. Almost every image would be entirely black.

In other words, the real information about 5DII pattern noise came when Daniel told us it was processed for 11.5 stops of dynamic range, not when he showed us the jpeg.

Agreed.

So the "stops of DR," is that a quantitative thing or is it subjective?

It's quantitative. I used Rawnalyze to measure it.

[Jon Ruyle]

What you mean here by "viewed the raw" is actually "viewed a JPEG that was converted from the raw file with the settings that are typical for today's raw converters."

I actually meant "click on it and have your default .cr2 viewer display it" []

I think the difference is important, because the number of stops is really just an arbitrary choice. 8 years ago the default settings used 4-5 stops, today the defaults are set for 6-8 stops,

I agree that the way a viewer or jpeg converter makes the thing look on screen is an arbitrary choice, and that this arbitrary choice is a moving target.


My point is just that when you say "here is a jpeg showing iso 100 with the 5DII", people who are used to jpegs created with today's standard converters look at it and say "hey, why is there so much noise in broad daylight at iso 100? The 5DII must totally suck."


I don't think you're trying to mislead anyone- quite the opposite. I just want to emphasize what you already said, which is that the noise is there because the picture represents a scene with 11.5 stops of dynamic range. This is not what you expect to see when you set the camera to iso 100 and jpeg only, then go out into the sunshine and take some snapshots.

and 8 years from now the defaults might be 9-14 stops.

AFAIK, digital l seems to have the potential for almost limitless dynamic range, or at least far more than film. We're stuck with photon noise, but I guess (and I'm basing this guess on pretty much no knowledge[] ) read noise will continue to decrease for a long time. One day we may look back on the days when we did multiple exposures to capture HDR and laugh with snobby, self-satisfied giddiness.


Oooh... I get goosebumps even thinking about it...

[Daniel Browning]

I just want to emphasize what you already said, which is that the noise is there because the picture represents a scene with 11.5 stops of dynamic range. This is not what you expect to see when you set the camera to iso 100 and jpeg only, then go out into the sunshine and take some snapshots.

Agreed.

I get goosebumps even thinking about it

Me too. []

[Colin]

You guys make me laugh, because I can identify [8-|]

[Colin500]

Interesting discussion, I was just wondering:

I agree that the way a viewer or jpeg converter makes the thing look on screen is an arbitrary choice, and that this arbitrary choice is a moving target.


My point is just that when you say "here is a jpeg showing iso 100 with the 5DII", people who are used to jpegs created with today's standard converters look at it and say "hey, why is there so much noise in broad daylight at iso 100? The 5DII must totally suck."


I don't think you're trying to mislead anyone- quite the opposite. I just want to emphasize what you already said, which is that the noise is there because the picture represents a scene with 11.5 stops of dynamic range. This is not what you expect to see when you set the camera to iso 100 and jpeg only, then go out into the sunshine and take some snapshots.

And it's "unfair" to compress 11.5 stops of dynamic range into an 8bit jpeg, and/or render it to an 8bit display, because it exaggerates the noise?


(If I convert to 8bit by cutting of the low bits, much of the noise disappears; if I convert by cutting of some high bits that would be even worse, in the sense of amplifying the noise even more, but that's hardly something that is done in practice(?).)


Regards, Colin