Reading mode: Light Dark

S3 Important discoveries !!

Go to last post
Z

Zarathustra

Senior Member
Messages
1,797
Reaction score
0
Location
Pangaea, US
I don't now if this has been discussed before but I think I came across a very interesting finding.

Apparantly the SR sensor can deliver more resolution, less noise and better dynamic range using certain technique.

Now, don't get too excited yet, since this discovery was for my F700 which happen to have a SR sensor. And the extra performance has a little drawback.

Does any one here have any clue what might possible this be. ?

If not I can explain how to unleash this 'hidden' S3 power, but we need some volunteers to make some RAW tests or send me some RAW files.

If the test turn possitive, I think I'm going to run and buy one of these S3 wonders.

Any clues ?

--
 
Is simple.

Studying what the SR sensor does with some tests I did last night, I realized that my F700 was not utilizing its full sensor capacity, the camera was using half its magapixel count (Only S sensors).
So the R sensors were negleted, for the final in-camera jpeg.

The consecuence was that the camera was delivering about half its 6MP resolution.

To 'Activate the other half of pixels (R sensors) you have to dramaticaly overexpose.

The trick is finding the right overexposure level to balance the S and R pixels.
IOW. You need to expose so both S and R pixels have an optimum exposure.
That is watching for clipping highlights and shadows in both.

This requires 3 things.
1. To shoot in RAW.
2. Longer exposures. (And this is the only drawback).
3. S7RAW or Adobe CS2. (Both take advantage of the R sensor)

-The extra resolution is gained by getting more bright pixels to contribute to the final image. I won't be surprise if you can realize the full 12MP potencial with this trick.

-The extra dynamic range, comes from getting also a more even distributed tonal range from the S and R images.

-The less noise comes from the R sensor being less noisy because is exposed 'to the right'.

Bottom line:
You have to overexpose by about +2.5 and shoot RAW.

The picture will come awful at your camera screen but when the image is process by S7RAW, you'll see the glory.

If you want to test it yourself.
Get a tripod, and fire several shots from exposure compensation 0 to +3.
Find a subject that is resolution challenging with some text on it.
Not necessarily has to be a high dynamic range scene.
Extarct all the S and R data as independent photos using S7RAW.
Then compare the samples.
The one that have the best pair is the perfect exposure for the S3Pro.
Process that shot and copare it to the 0EV compensated shot.

Let me know what you find.
And don't hesitate to ask questions.

I might post some pictures if I find some time later on.
The results were amazing !

--
 
1.In S7RAW hit the reset buttom to make sure all the settings are to default values.

2. Select the Preview buttom in the top tool bar. (This program is slow, but preview makes things faster by working with a low resolution proxy.)

3. Go to the Tab called "Adjust2"

There is where the SR Mixing console is. (No other RAW software have this AFAIK.)
Make sure to UNcheck the "Easy Mode". (This will open the advanced options)

To extract the S picture:
Select from the dropdown list "S only".
Disregard the checks on S, R and the sliders.

Now try to make the best of that picture using the EV slider found in the "Adjust1" tab. I recommend you turn the Highlight/Shadow warning to avoid clipping.

Once you are happy, go to save the image, but make sure you save it in Full resolution mode by selecting the Full green square at the top tool bar.
Make sure you use "Save As" to give meaninful names to the files.

To extract the S picture:
Select from the dropdown list "R only".
Disregard the checks on S, R and the sliders.
And do the same as before.

Once you do this with 3 pictures at different levels of exposures, compare the pairs of S&R images to the other pairs.
The file that has the best pair, is the one that will give you the best quality.

Then you can go back to S7RAW or ACR and render your full photo using the RAW file that showed the best pair.

The idea of separating the S&R channels is only to find out what shot will give you the best exposure.
So next time you'll have an idea on how much overexposure is needed.

From my test with the F700, it is about +2.5 EV.

Obviously, the huge drawback to this is that you need plenty of light to achive practical shutter speeds.
For Landscape and Architecture, this seem to be ideal solution.

--
 
You said that camera RAW does take all the advantage of S+R pixels,were that right?

Im just checking and it is really a far bit more complex a S3 RAW than a D70s RAW i can tell....

greatz,
Minim3
 
minim3 said:
You said that camera RAW does take all the advantage of S+R
pixels,were that right?
Click to expand...
So it seems, based on my own, comment from reviewers and some S3 owners.

I'm willing to bet that the reason why ACR renders are noisier than HU is because it is using the generally underexposed (noisy) R pixels.
minim3 said:
Im just checking and it is really a far bit more complex a S3 RAW
than a D70s RAW i can tell....
Click to expand...
Yes, but is not that much complex.

Think of the S3 as a camera that takes 2 exposures. One for the highlights and one for the shadows. Is that simple.
Or like having a RrGgBb sensor array.

--
 
Not sure how to interpret the results, except that both the R and S pixel pictures (I did 0, +1.5, +2.5) were best (showing most detail, best color rendition) at +1.5, though they did need some more pushing up of the exposure in S7 to get it looking nice. The +2.5 were defintely too washed out, then moved to too dark when in S7 I tried to adjust the exposure.

Now I will convert all three from RAW with both R&S pixels active and see what I get.

Does any of this make sense?
Crystal
--
http://treehuggergirl.zenfolio.com/
 
That means that the S3 is shooting to get S and R to good exposures.
Or
That you are doing something wrong.

The most apparent benefit I'm getting is less noise and better resolution.

Which of your shots had the best pair?

Are you sure you are making some adjustment to the single channels S&R when extracting them so they look the best it can?

If the best pair was the 0EV compensated shot, then the S3 is doing its best.

But I doubt it, since that means that by default the S3 has to be 2EV less sensittive than the other 'normal' cameras.

I think reason why, the F700 is the way it is.

Oh well, thanks for trying.
I was just curious.
Maybe someone else can confirm yours or my findings.

--
 
was defintely NOT the O EV. When I convert them for comparison, do I need to keep them uncompressed as TIFFs or can I save them as JPEGs? That way I can post what I did.

The S pixel shots were definitely washed out when I added + EV, the detail in the highlights was lost and not recoverable by adjusting the EV or color channels in S7. But from what I gather that is to be expected, as the R pixels are there to save the blown hightlights, right?

Now the R pixel image at on my 0 EV shot was very, very dark, but the color in the highlights was good. If I adjust its exposure in S7, then I lose the hightlight detail and color for the sake of an overall lighter picture.

I think the S3 DOES make the right exposure for both R and S pixels when the histogram shows a good bell curve. There was no comparison in my fully converted files. The O EV looked the best.

I too hope someone who knows more than me can chime in here.
:)
Crystal
--
http://treehuggergirl.zenfolio.com/
 
Detail in shadows, because you will use more saturation of R pixels and will lose a small partition of S saturation for register in mid lights in +1EV.

If you overexpose +1EV and recover it in PS after shot, you will gain resolution in shadows, maintain resolution in mid lights and lost resolution in highlights, in relation w/ 0EV. ;)
acrystalball said:
detail, not shadow? Or are they for both?
Scratching my head ~
Crystal
--
http://treehuggergirl.zenfolio.com/
Click to expand...
--

 
Leo Terra said:
Detail in shadows, because you will use more saturation of R pixels
and will lose a small partition of S saturation for register in mid
lights in +1EV.
If you overexpose +1EV and recover it in PS after shot, you will
gain resolution in shadows, maintain resolution in mid lights and
lost resolution in highlights, in relation w/ 0EV. ;)
Click to expand...
Well, I think that by overexposing to +1EV or larger you will:
-Gain S Shadow resolution. (Now S shadow detail is visible)
-Gain S&R Midtones resolution. (S & R meet and add pixels to Midtones)
-Maintain Highlight resolution. (S has clipped but R is still pumping tones)

This technique is sort of 'expoxing to the right' but for the R photosites.
We might call it "Expose to the Ultra Right" ;)

--
 
Hi Zara,

I used another image, one with not as much contrast or DR, and HOLY SMOKES the detail resolution in the +2.5 EV shot, after PP was amazing. The R pixel +2.5 image was the best of the Rs. I used text and woodgrain in the image for examining detail. I'm going to run the test again, making sure I use the same parameters, and then post the results. It is kind of unbelievable.

The only thing was that the S pixel colors were washed out, and I don't know how to recover it. The S pixel 0 EV was way better colorwise.

So could I theoretically blend the S pixel image from the O EV and the R pixel image from the +2.5 EV in PS, and get a vastly superior image? Or is there some way to extract properly satuarated colors in overexposed images?

This has been fun, thanks for posting your theory.
:)
Crystal
--
http://treehuggergirl.zenfolio.com/
 
Sorry the images are so big but I wanted y'all to easily see the difference.

First text, both are 100% cropped, no PS except resizing:
0 EV is on top, +2.5 EV on bottom





And woodgrain, again 100% crops, no PS except resizing:
0 EV is top, +2.5 EV is bottom





So Zarathrustra, are these results consistent with what you found?
---
http://treehuggergirl.zenfolio.com/
 
Here I did this to analize the S & R content independently.
Crops from the test scenes.

Things to check for: Noise in flat areas, detail of text and specular highlights.

And another area of the scene.

One curiousphenomenom is to see noise for the S +3EV image.

I can only guess that as the R pixels gain signal, they 'contaminate' of noise the S pixels !?

But remember this is a different camera than the S3 so your results might be different.

In the next post, the bottom line, the Final image.

--
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top