RAW noise in Sony verses other DSLRs Adam-T,David K you out there?

Hi Adam (and others) I have a RAW question:

I've seen Adam write several times that the Apha's RAW output isn't as good as other (10mp? high ISO) DSLRs. This has me thinking...I've seen so many reviews/articles state the RAW output in 10mp DSLRs is very close (not different enough to make any tangible difference). There are some reports that state Nikon hits it's RAW w/ very mild noise reduction.

I have only spent about 1.5 hours on a computer where images from my alpha and my buddies D80 images were on the same monitor. In fact, I was teaching him RAW conversion w/ CS2. From my short hands on experience, even a touch of noise reduction in Adobe RAW converter makes the images quality a draw at high ISO. Nikon does seem to have some noise reduction built in...I don't have any friends w/ a 10MP Canon DSLR so I only go from what I read. I know 6- 8MP RAW images should have lower detail but at high ISO be better w/ noise even in RAW.

Since others have more experience and several cameras at your disposal, I wonder if you could check out how the 10mp DSLR RAW images compare on a few standardized shots. And see just how much noise reduction Nikon does on its RAW shots w/ the same sensor the Sony camera has...It would be just as interesting to see how Pentax and Canon 10mp cameras do too.

My hypothesis is what ever RAW conversion there is with a subtle change in the color and luminosity noise reduction sliders there will be very little difference between any of the 10mp DSLR's output.

Can anyone test this?

The problem is that most users will not have access to both cameras in order to do this kind of comparison in RAW output. Different files have different levels of noise & therefore most comparisons taken from review sites are invalid.

Sony writes it's RAW files in a different way to other manufacturers & this affects the way in which they come out. All RAW converters have to translate the basic data in order to display the demosaiced image & this naturally affects any resulting image.

Other factors come into play, the inherent noise in sensors & the electronic circurts in cameras affect the signal versus electronic noise ratios. Although there are similarities in all the Sony 10MP sensors in different makers models it cannot be assumed that all these other factors are the same & therefore there are bound to be some differences.

RAW converters have to use the basic data in different ways & this can account for perceived noise levels , not all RAW converters are equal. The RAW output from Sony's own converer are different from , say, Adobes RAW converter. You will find that some files are better when using one converter wheras others will be handled better with another. Shadow, highlight areas, etc will shown noise better in some converters than others.

RAW conversion is not a simple one click operation & some changes in settings are required to get the best conversion coupled with further processing in other imaging programs. Lightroom for example does not at present handle sharpening very well & most users wil use another program to sharpen - this issue will be addressed in version two which is coming out shortly.

Keith-C

KEITH-C said:

Sony writes it's RAW files in a different way to other
manufacturers & this affects the way in which they come out. All
RAW converters have to translate the basic data in order to display
the demosaiced image & this naturally affects any resulting image.

Click to expand...

While it is true that Sony stores the RAW data differently than the other camera makers, it doesnt make any difference as to how this data is extracted by any RAW converter. Sony uses lossless compression in its ARW files. So, no matter what converter you use, it should decode the exact same data that was put in the file (as far as pixel data is concerned). If there is any difference, then there is a problem with the converter you are using.

I think the reason Nikon RAW files are percieved to have less noise is because they use lossy compression in their compressed NEFs. I don't have access to Nikon DSLRs, so I can't make a good comparison between compressed and uncompressed NEFs against RAW files from my A100.

--

IraqiGeek
http://www.iraqigeek.com

mark moe said:

Can anyone test this?

Click to expand...

Why not look at Phil's tests? His tests may not be definitive. But they are the best you are likely to get.
--
David Jacobson
http://www.pbase.com/dnjake

dnjake said:

mark moe said:

Can anyone test this?

Click to expand...

Why not look at Phil's tests? His tests may not be definitive.
But they are the best you are likely to get.
--
David Jacobson
http://www.pbase.com/dnjake

Click to expand...

I have looked at Phil's test but there isn't a lot to go from...Phil's test point to the conclusion that all 10mp DSLRs have basically the same RAW output using the same converter. He doesn't cover the changes that accompany ISO change.

mark moe said:

dnjake said:

mark moe said:

Can anyone test this?

Click to expand...

Why not look at Phil's tests? His tests may not be definitive.
But they are the best you are likely to get.
--
David Jacobson
http://www.pbase.com/dnjake

Click to expand...

I have looked at Phil's test but there isn't a lot to go
from...Phil's test point to the conclusion that all 10mp DSLRs have
basically the same RAW output using the same converter. He doesn't
cover the changes that accompany ISO change.

Click to expand...

He's not entirely right either. I have used all the 10 mpixel DSLRs and they all have very different output from RAW, different levels of noise and innate sharpness.

See this page:

http://www.pbase.com/davidkilpatrick/digital_photo_samples&page=2

You will find some RAW conversion comparisons of D80, 400D, A100 - the bookshelf shots.

David

(nt)

mark moe said:

tangible difference). There are some reports that state Nikon hits
it's RAW w/ very mild noise reduction.

Click to expand...

It's CCD RAW so it can't be hit with any Noise reduction - what Canon do on chip with its CMOS is Another matter entirely .

All I did was test the A100 at ISO400, 800 and 1600 and used the same raw converter and nothing I did with exposure could make the A100 as low noise as the D80, D200, D2X, ANY Canon whatsoever from the 3Mp D30 to the noisy 1DS Mk1 or even - Shock Horror - THE R1!! (which I originally thought was noisier but it underexposes at the wide end heavily and the shots I was looking at had been upped by a Stop - I redeveloped better samples I had) ..

This isn't trolling, I tried - was my friends camera faulty?, dunno, I doubt it - I'm not the first to think the A100 is noisy, if it wasn't, Barry Fitzgerald would probably own one.. As I said before, it's the cameras ony serious Flaw (and it takes away any Antishake advantages) , it has other niggles like all the rest but otherwise it's a great camera and I enjoyed my week with it and the old Minolta lenses..

--
Please ignore the Typos, I'm the world's worst Typist

IraqiGeek said:

KEITH-C said:

Sony writes it's RAW files in a different way to other
manufacturers & this affects the way in which they come out. All
RAW converters have to translate the basic data in order to display
the demosaiced image & this naturally affects any resulting image.

Click to expand...

While it is true that Sony stores the RAW data differently than the
other camera makers, it doesnt make any difference as to how this
data is extracted by any RAW converter. Sony uses lossless
compression in its ARW files. So, no matter what converter you use,
it should decode the exact same data that was put in the file (as
far as pixel data is concerned). If there is any difference, then
there is a problem with the converter you are using.

Click to expand...

Not true. Different RAW converters use different algorithms to come up with the final image based on the data from the sensor. Somewhere on the web (I don't recall what site) there is a comparison of the same RAW shot from the same camera but converted using different converters, and there are differences, sometimes significant. Some are better with some files than others, or better in certain areas and worse in others (e.g., color accuracy, dynamic range, noise, etc.). Thus, the RAW converter used can and does have a definite impact on how the image will look. But remember, because it is a RAW file that is being converted, you also can change many different parameters to the image to get a particular look. What I'm talking about, generally speaking, are the standard algorithms the different converters use.

--
Mark Van Bergh

He meant using the SAME raw converter with many diferent RAW files (like ACR for example), so there's no diference in data layout.

That the A100 has the highest noise level at high ISOs (400 and above) of all comparable camera, when working from raw files. Actually the JPEGs in camera are pretty good, and it is hard to match them by converting from raw, but they are also heavily processed.

However - the A100 also has better contrast without sacrificing too much dynamic range; better colour; and most of all, much better sharpness. It wipes the floor with any of the competitors at ISO 100 (though the R1, which Adam mentions, is a dark horse - much better than it has any right to be, and even at 3200 the files are usable, I know because I've done it).

I also have plenty of unusably noisy images from Canon, by the way. The 300D and 400D are nearly as bad as the A100 at 1600, and the best around is not Canon, but Nikon - the D80. Canon 400D suffers from random hot pixel colour speckles on JPEGs, and A100 doesn't get this at all.

Generally when I have used another camera for a lot of shooting, I have been frustrated that I've lost the best pictures because I was not using the A100. I've got some great sets of shots on the D80 which would have been far better on the A100. I also have shots on the D80 which would be completely impossible on the A100, using Nikon Capture DX and their software D-Lighting. Trying to lift the shadows the same way on an A100 low light image would be disaster.

Example:



Unprocessed in camera JPEG, Nikon D80, ISO 100, time exposure on trodpipe



Nikon Capture NX processed from raw with rather extreme D-Lighting. Noise level apparently unchanged. Ridiculous degree of lightening to shadows before this point, where posterization just begins.

I don't actually like the colours much, that is another issue. Also, I would probably have shot with the Sony and no tripod

David

I found that when I tested an A100 that noise was unacceptably high at 1600 ISO compared to other cameras especially 6MP versions. This is a limitation which may or may not be a problem for most users - a lot depends on the type of photos which they shoot.

The good news is that the Alpha is hard to beat at ISO 100 where its sharpness & colour is accurate. It is to be hoped that the noise problem will be addressed in forthcoming Sony DSLRs - they have taken some flack over the noise issue of the A100 & need to re-gain some credibility.

Keith-C

Mark VB said:

IraqiGeek said:

KEITH-C said:

Sony writes it's RAW files in a different way to other
manufacturers & this affects the way in which they come out. All
RAW converters have to translate the basic data in order to display
the demosaiced image & this naturally affects any resulting image.

Click to expand...

While it is true that Sony stores the RAW data differently than the
other camera makers, it doesnt make any difference as to how this
data is extracted by any RAW converter. Sony uses lossless
compression in its ARW files. So, no matter what converter you use,
it should decode the exact same data that was put in the file (as
far as pixel data is concerned). If there is any difference, then
there is a problem with the converter you are using.

Click to expand...

Not true. Different RAW converters use different algorithms to
come up with the final image based on the data from the sensor.
Somewhere on the web (I don't recall what site) there is a
comparison of the same RAW shot from the same camera but converted
using different converters, and there are differences, sometimes
significant. Some are better with some files than others, or
better in certain areas and worse in others (e.g., color accuracy,
dynamic range, noise, etc.). Thus, the RAW converter used can and
does have a definite impact on how the image will look. But
remember, because it is a RAW file that is being converted, you
also can change many different parameters to the image to get a
particular look. What I'm talking about, generally speaking, are
the standard algorithms the different converters use.

Click to expand...

Yes, its true that different RAW converters use different interpolation algorithms. BUT, no matter what RAW converter you use, it should be able to extract exactly the same pixel data that the camera put in the file, without regard to which camera brand or model you are using, or how that data is written to the file.

Keith was saying that the way Sony writes ARW files affects how they are read by RAW converters, which is not correct. The fact that Sony writes the data differently in their ARW files, than say how Canon, Nikon (which BTW uses lossy compression in their compressed NEFs that give you 9.5 bits from the CCD's 12bit data), Olympus or Pentax writes it, has no impact on the accuracy of the extraction process of that data.

The pixel data that goes in the RAW file should come out the same, without a single bit of difference in values. Period.

--
IraqiGeek
http://www.iraqigeek.com

Hi David,

I wonder where the 30D or the D200 stand, compared to the other cameras you have tested (300D, 400D, D80, A100). Have you had a chance to test them?

--
Luis

Adam-T said:

Adam-T said:

tangible difference). There are some reports that state Nikon hits
it's RAW w/ very mild noise reduction.

Click to expand...

It's CCD RAW so it can't be hit with any Noise reduction - what
Canon do on chip with its CMOS is Another matter entirely .
All I did was test the A100 at ISO400, 800 and 1600 and used the
same raw converter and nothing I did with exposure could make the
A100 as low noise as the D80, D200, D2X, ANY Canon whatsoever from
the 3Mp D30 to the noisy 1DS Mk1 or even - Shock Horror - THE R1!!
(which I originally thought was noisier but it underexposes at the
wide end heavily and the shots I was looking at had been upped by a
Stop - I redeveloped better samples I had) ..
This isn't trolling, I tried - was my friends camera faulty?,
dunno, I doubt it - I'm not the first to think the A100 is noisy,
if it wasn't, Barry Fitzgerald would probably own one.. As I said
before, it's the cameras ony serious Flaw (and it takes away any
Antishake advantages) , it has other niggles like all the rest but
otherwise it's a great camera and I enjoyed my week with it and the
old Minolta lenses..

--
Please ignore the Typos, I'm the world's worst Typist

Click to expand...

Well a number of reasons I dont have an A-100, mostly due to only recently starting paid work, I have to purchase "essential items" ist. And I have been caught between getting a 5/7D instead of an A-100.

As far as my needs, will likely not be the same as others. If you are in a church shooting a wedding, 98% of the time you cannot use flash, and some places are super dim, even that fast prime needs you to bump into higher ISO often.

At ISO 1600 I get good results from the 5D, and with a bit of latitude too, which you do need. I can also print fairly big if needed. Even ISO 3200 isnt bad, fine for smaller print work.

A-100 cannot match that, you wouldnt worry about being in ISO 800/1600 on the 5D.

As for the R1, yes at the time it was thought of as a bit noisy, have to say that it beats the A-100. What gives? You tell me!

I would like sony to upgrade the sensor.....that would do me fine. I dont expect 10mp to be as good as 6mp for high ISO, but I would like something not a million miles away.

A-100 is a great camera for good light work, in some ways I could use it plenty, landscapes, people work, no problem. Inside with low light and no flash, I would reach for the 5D every time.

Sadly it will likely be next year before they upgrade the A-100.

--



Clint is on holiday! Soon to return! ;-)

IraqiGeek said:

Mark VB said:

IraqiGeek said:

KEITH-C said:

Sony writes it's RAW files in a different way to other
manufacturers & this affects the way in which they come out. All
RAW converters have to translate the basic data in order to display
the demosaiced image & this naturally affects any resulting image.

Click to expand...

While it is true that Sony stores the RAW data differently than the
other camera makers, it doesnt make any difference as to how this
data is extracted by any RAW converter. Sony uses lossless
compression in its ARW files. So, no matter what converter you use,
it should decode the exact same data that was put in the file (as
far as pixel data is concerned). If there is any difference, then
there is a problem with the converter you are using.

Click to expand...

Not true. Different RAW converters use different algorithms to
come up with the final image based on the data from the sensor.
Somewhere on the web (I don't recall what site) there is a
comparison of the same RAW shot from the same camera but converted
using different converters, and there are differences, sometimes
significant. Some are better with some files than others, or
better in certain areas and worse in others (e.g., color accuracy,
dynamic range, noise, etc.). Thus, the RAW converter used can and
does have a definite impact on how the image will look. But
remember, because it is a RAW file that is being converted, you
also can change many different parameters to the image to get a
particular look. What I'm talking about, generally speaking, are
the standard algorithms the different converters use.

Click to expand...

Yes, its true that different RAW converters use different
interpolation algorithms. BUT, no matter what RAW converter you
use, it should be able to extract exactly the same pixel data that
the camera put in the file, without regard to which camera brand or
model you are using, or how that data is written to the file.

Click to expand...

Agreed. Any RAW converter should be able to use the pixel data from the camera created file, regardless of the camera brand or model (unless of course it's a brand new model and the converter has not been updated to be able to read files from that new model).

IraqiGeek said:

Keith was saying that the way Sony writes ARW files affects how
they are read by RAW converters, which is not correct. The fact
that Sony writes the data differently in their ARW files, than say
how Canon, Nikon (which BTW uses lossy compression in their
compressed NEFs that give you 9.5 bits from the CCD's 12bit data),
Olympus or Pentax writes it, has no impact on the accuracy of the
extraction process of that data.

The pixel data that goes in the RAW file should come out the same,
without a single bit of difference in values. Period.

Click to expand...

I think we probably agree here as well, but it's just that we were talking about two different things. My point was not how accurately the RAW converter extracts the data from the RAW file, but how it interprets that data to "create" (or perhaps better said, display) the image. It is the conversion process, from extracted RAW data to "final" image, where the RAW converters differ (i.e., they use different algorithms to interpret the data and display the final image). This is what leads to different results for different RAW converters, using the same extracted RAW data.

--
Mark Van Bergh

Adam-T said:

Adam-T said:

tangible difference). There are some reports that state Nikon hits
it's RAW w/ very mild noise reduction.

Click to expand...

It's CCD RAW so it can't be hit with any Noise reduction - what
Canon do on chip with its CMOS is Another matter entirely .
All I did was test the A100 at ISO400, 800 and 1600 and used the
same raw converter and nothing I did with exposure could make the
A100 as low noise as the D80, D200, D2X, ANY Canon whatsoever from
the 3Mp D30 to the noisy 1DS Mk1 or even - Shock Horror - THE R1!!
(which I originally thought was noisier but it underexposes at the
wide end heavily and the shots I was looking at had been upped by a
Stop - I redeveloped better samples I had) ..
This isn't trolling, I tried - was my friends camera faulty?,
dunno, I doubt it - I'm not the first to think the A100 is noisy,
if it wasn't, Barry Fitzgerald would probably own one.. As I said
before, it's the cameras ony serious Flaw (and it takes away any
Antishake advantages) , it has other niggles like all the rest but
otherwise it's a great camera and I enjoyed my week with it and the
old Minolta lenses..

--
Please ignore the Typos, I'm the world's worst Typist

Click to expand...

A couple of points: 1. SSS isn't obviated by high ISO performance. e.g. I shoot a lot of 250mm daylight shots that sss hand a hand in steady movement free images. I agree that great ISO performance plus antishake is the way to go.

Secondly, I still think that my hypothesis of RAW performance on the same sensor a-100, d80, k10, d40x... (and perhaps other 10MP sensors too) can be equalized via a using the same RAW converter w/ slightly different setting for color and luminance noise control.

I believe that circuitry difference that results in different RAW output is not too significant when one considers the RAW conversion part of the equation. (I may be tainted here since I only have used CS2 for RAW conversion.

Luis Villablanca said:

Hi David,

I wonder where the 30D or the D200 stand, compared to the other
cameras you have tested (300D, 400D, D80, A100). Have you had a
chance to test them?

Click to expand...

The D200 is the next noisiest after the A100 - much worse than the D80. When I tested the D200, I was also using a Kodak P880. When editing two images for double page magazine spreads, I picked the lowest noise, sharpest one and thought it was the Nikon. Wrong... I nearly used the Kodak consumer cam image (from raw) in place of the Nikon (from raw) because it was actually better. I had to do a last minute edit and rewrite two articles.

The 30D/350D/20D sensor I have not tested properly, oddly enough, but everything I have seen from other users convinces me that the 8 megapixel sensor is lower in noise than the 300D or the 400D. I've only taken a few 350D images (trade show camera use) and I've received good 20D/30D rock concert images for publication which were just in-camera JPEGs and showed very little noise at 800 and 1600.

David

mark moe said:

I believe that circuitry difference that results in different RAW
output is not too significant when one considers the RAW conversion
part of the equation. (I may be tainted here since I only have used
CS2 for RAW conversion.

Click to expand...

I dont feel you will be able to match the other sensors with the A-100 at high ISO, even doing your own processing. IF that matters is another issue.

--



Clint is on holiday! Soon to return! ;-)

David, allow me to hijack this older thread.

I probably buy an A100 to accompany my 5D so I currently reread older posts about the A100.

David Kilpatrick said:

That the A100 has the highest noise level at high ISOs (400 and
above) of all comparable camera, when working from raw files.
Actually the JPEGs in camera are pretty good, and it is hard to
match them by converting from raw, but they are also heavily
processed.

Click to expand...

So do you tend to use your A100 in JPEG mode for that reason?

I exclusively shoot RAW with my 5D and that saved me quite a lot of shots that otherwise would have suffered from unrecoverable blown highlights and occasionally wrong WB.

If in camera processing in the A100 is hard to match by converting from raw and with the additional help of DRO+ (which is only available in JPEG mode) I'm wondering, whether JPEG is preferable over RAW on the A100? A least giving probably comparable results with less PP. Maybe the JPEG compression is an issue for bigger prints.

Like to hear your opinion.

Cecco