Pages: 1 2 [3] 4 5 ... 8   Go Down

Author Topic: What about 36MP DSLRs?  (Read 45646 times)

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12536
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: What about 36MP DSLRs?
« Reply #40 on: December 24, 2011, 09:28:36 am »

Highlight DR does simply not exist with linear digital sensors.

Cheers,
Bernard

fotometria gr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • www.fotometria.gr
Re: What about 36MP DSLRs?
« Reply #41 on: December 24, 2011, 10:00:30 am »

Highlight DR does simply not exist with linear digital sensors.

Cheers,
Bernard

I won't try to convince you for the opposite, you are simply wrong! Merry Christmas, Theodoros. www.fotometria.gr
 P.S. The sensor may be linear but the output signal is not, nor is the translation (electrically) to the sensor of the input signal!
« Last Edit: December 24, 2011, 10:07:53 am by fotometria gr »
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12536
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: What about 36MP DSLRs?
« Reply #42 on: December 24, 2011, 10:48:52 am »

I won't try to convince you for the opposite, you are simply wrong! Merry Christmas, Theodoros. www.fotometria.gr
 P.S. The sensor may be linear but the output signal is not, nor is the translation (electrically) to the sensor of the input signal!

The result of the adc from the commercially available CMOS/CCD sensor is linear relative to the incoming illumination. Up to a well defined point where the sensels reach saturation. This what linear means.

As a result highlight DR does not exist.

What does exist is:
- under-rated ISO/exposure calibration (real ISO 50 is called ISO 100 to generate on purpose under exposure),
- smart raw conversion algos giving the illusion of highlight DR.

The Fuji S5 belongs to the first category, it's array of smaller secondary sensels has a lower real ISO than th main sensels for which it's ISO is calibrated. It does therefore underexpose.

The only choice is btwn a camera providing you real information about the data it captured or one giving you the illusion of highlight DR. I prefer the former.

Cheers,
Bernard
« Last Edit: December 24, 2011, 10:51:54 am by BernardLanguillier »
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3350
Re: What about 36MP DSLRs?
« Reply #43 on: December 24, 2011, 11:32:29 am »

I won't try to convince you for the opposite, you are simply wrong! Merry Christmas, Theodoros. www.fotometria.gr
 P.S. The sensor may be linear but the output signal is not, nor is the translation (electrically) to the sensor of the input signal!

I have to agree with Bernard. The concept of highlight and shadow DR harkens back to film, which has a shoulder and knee in the characteristic curve. This concept has been carried over by DPReview in their reviews. They use a Stouffer wedge and look at JPEGS to which an S curve has been applied. However, the raw file output is linear if there is no clipping as shown below for the Nikon D3. A Stouffer wedge was photographed and the file was rendered linearly with IRIS.

I also wish merry Christmas to all.

Regards,

Bill
Logged

fotometria gr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • www.fotometria.gr
Re: What about 36MP DSLRs?
« Reply #44 on: December 24, 2011, 11:33:07 am »

The result of the adc from the commercially available CMOS/CCD sensor is linear relative to the incoming illumination. Up to a well defined point where the sensels reach saturation. This what linear means.

As a result highlight DR does not exist.

What does exist is:
- under-rated ISO/exposure calibration (real ISO 50 is called ISO 100 to generate on purpose under exposure),
- smart raw conversion algos giving the illusion of highlight DR.

The Fuji S5 belongs to the first category, it's array of smaller secondary sensels has a lower real ISO than th main sensels for which it's ISO is calibrated. It does therefore underexpose.

The only choice is btwn a camera providing you real information about the data it captured or one giving you the illusion of highlight DR. I prefer the former.

Cheers,
Bernard

It's not what I said and doesn't have anything to do with what I said. The sensor may be linear but the S-slope is not and nor is the sensors outcome, the input is not imported linearly to the sensor either. In fact to produce the S-slope the sensels are controlled at different Iso value depending on the light intensity they receive. This means that depending on the shot, the sensor directs some pixels to higher, some others to even higher, some others to different, some more to lower (etc)... sensitivity to achieve the S-slope. Thats not very linear behavior is it? That is exactly why you see noise at the deep shadows, even if you set sensitivity to the minimum. In fact the logic (control circuit) of the sensor is trying to fool the sensor and make it behave like film, its just that they haven't succeed yet, but its getting better all the time..., lets hope this time they will succeed! Merry Christmas, Theodoros. www.fotometria.gr
Logged

fotometria gr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • www.fotometria.gr
Re: What about 36MP DSLRs?
« Reply #45 on: December 24, 2011, 11:42:29 am »

I have to agree with Bernard. The concept of highlight and shadow DR harkens back to film, which has a shoulder and knee in the characteristic curve. This concept has been carried over by DPReview in their reviews. They use a Stouffer wedge and look at JPEGS to which an S curve has been applied. However, the raw file output is linear if there is no clipping as shown below for the Nikon D3. A Stouffer wedge was photographed and the file was rendered linearly with IRIS.

I also wish merry Christmas to all.

Regards,

Bill
This is measuring the sensor Bill, not the whole process of capturing/storing data. The sensor IS linear. Merry Christmas to you and your family. Regards, Theodoros. www.fotometria.gr
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11249
    • Echophoto
Re: What about 36MP DSLRs?
« Reply #46 on: December 24, 2011, 03:13:13 pm »

Hi,

Merry Christmas to all and peace on Earth!

What Bill shows is that the raw data is linear. Now, we are never looking at raw data directly. The raw data is converted to a visible image. That conversion normally involves the use of an S-curve, but we are really free to choose any curve or straight line. Chemical rendering has the S-curve built in. It has a shoulder and a toe. This is typical of chemical processes.

Digital images have problems in extreme highlights. Film saturates but digital clips. Little to do about it. If the highlights are not to extreme we can expose for highlights, but it is often not possible.

Best regards
Erik


This is measuring the sensor Bill, not the whole process of capturing/storing data. The sensor IS linear. Merry Christmas to you and your family. Regards, Theodoros. www.fotometria.gr
Logged
Erik Kaffehr
 

kers

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2256
    • Pieter Kers
Re: What about 36MP DSLRs?
« Reply #47 on: December 24, 2011, 03:46:38 pm »

Using a D3x the last three years and making architectural work, what i would like to see is not so much more pixels, but better color at higher iso values, - as we see on the D3s- and a gentle way to handle dust.
I spend far too much time cleaning my images- on the computer, -with a blower in the camera and 0 have to get it cleaned at Nikon every two months...
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12536
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: What about 36MP DSLRs?
« Reply #48 on: December 24, 2011, 06:50:14 pm »

It's not what I said and doesn't have anything to do with what I said. The sensor may be linear but the S-slope is not and nor is the sensors outcome, the input is not imported linearly to the sensor either. In fact to produce the S-slope the sensels are controlled at different Iso value depending on the light intensity they receive. This means that depending on the shot, the sensor directs some pixels to higher, some others to even higher, some others to different, some more to lower (etc)... sensitivity to achieve the S-slope. Thats not very linear behavior is it? That is exactly why you see noise at the deep shadows, even if you set sensitivity to the minimum. In fact the logic (control circuit) of the sensor is trying to fool the sensor and make it behave like film, its just that they haven't succeed yet, but its getting better all the time..., lets hope this time they will succeed! Merry Christmas, Theodoros. www.fotometria.gr

The outcome of the sensor is a raw file that, at least at base ISO, does not contain any S curve. The S curve is applied when converting the raw file.

I guess that the point you are trying to make may be:
- Some sensors apply some form of amplification of the electric signal at higher ISOs, and this amplification may not be applied linearly. I would be interested in factual data you might have about this?
- Not all sensors have the same quality of analog to digital conversion near saturation, which may result in harsher transitions from non blown to blown areas. In essence, this would mean that some sensors behave non linearly near saturation. This may be true accross a very small range of illuminations, but would still basically not explain the highlight DR you claim exists.

Sorry, I still don't see anything in your argument that would explain highlight DR. Until proven otherwise, I'll stick tot the well accepted proposal that DR with linear sensors is ONLY depednant on shadow noise. The rest is just a pleasant illusion.

Merry x-Mas to you as well!

Cheers,
Bernard
« Last Edit: December 24, 2011, 08:00:25 pm by BernardLanguillier »
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12536
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: What about 36MP DSLRs?
« Reply #49 on: December 24, 2011, 06:51:21 pm »

This is measuring the sensor Bill, not the whole process of capturing/storing data. The sensor IS linear. Merry Christmas to you and your family. Regards, Theodoros. www.fotometria.gr

No, this does measure the content of the raw file. Which is all we have.

Cheers,
Bernard

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12536
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: What about 36MP DSLRs?
« Reply #50 on: December 24, 2011, 06:52:59 pm »

Using a D3x the last three years and making architectural work, what i would like to see is not so much more pixels, but better color at higher iso values, - as we see on the D3s-

Just genuinely curious, may I ask why you are using high ISOs for architecture work?

Is it related to the amount of lighting you are willing to use?

If that is the case, why not stitch with a D3s? Or wait a few weeks until the D4 is announced/made available.

cheers,
Bernard
« Last Edit: December 24, 2011, 08:42:02 pm by BernardLanguillier »
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11249
    • Echophoto
Re: What about 36MP DSLRs?
« Reply #51 on: December 25, 2011, 01:51:38 am »

Hi,

I use an Arctic Butterfly on my Sony Alpha 900, it works well, and I never needed to do wet cleaning on the camera or send it in for cleaning.

Best regards
Erik


Using a D3x the last three years and making architectural work, what i would like to see is not so much more pixels, but better color at higher iso values, - as we see on the D3s- and a gentle way to handle dust.
I spend far too much time cleaning my images- on the computer, -with a blower in the camera and 0 have to get it cleaned at Nikon every two months...

Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11249
    • Echophoto
Re: What about 36MP DSLRs?
« Reply #52 on: December 25, 2011, 02:03:09 am »

Hi,

In my view we have three issues here:

1) Film essentially compresses extreme highlights. Adding more light increases density very slightly (on negative film).

2) To capture highlights we need to expose for highlights. Handling specular highlights may be not possible, see #1.

3) If all is perfectly done we have wide DR and a boring image.



The image on the left is the "original" image processed using Lightroom controls. The image on the right is from the same exposure, but using some intensive manipulation is PS.

Here is how that image has been processed: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/61-hdr-tone-mapping-on-ordinary-image

Please note that even if HDR mapping is used the image is based on a single ETTR exposure.

Best regards
Erik

The outcome of the sensor is a raw file that, at least at base ISO, does not contain any S curve. The S curve is applied when converting the raw file.

I guess that the point you are trying to make may be:
- Some sensors apply some form of amplification of the electric signal at higher ISOs, and this amplification may not be applied linearly. I would be interested in factual data you might have about this?
- Not all sensors have the same quality of analog to digital conversion near saturation, which may result in harsher transitions from non blown to blown areas. In essence, this would mean that some sensors behave non linearly near saturation. This may be true accross a very small range of illuminations, but would still basically not explain the highlight DR you claim exists.

Sorry, I still don't see anything in your argument that would explain highlight DR. Until proven otherwise, I'll stick tot the well accepted proposal that DR with linear sensors is ONLY depednant on shadow noise. The rest is just a pleasant illusion.

Merry x-Mas to you as well!

Cheers,
Bernard

Logged
Erik Kaffehr
 

fotometria gr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • www.fotometria.gr
Re: What about 36MP DSLRs?
« Reply #53 on: December 25, 2011, 05:23:27 am »

The outcome of the sensor is a raw file that, at least at base ISO, does not contain any S curve. The S curve is applied when converting the raw file.

I guess that the point you are trying to make may be:
- Some sensors apply some form of amplification of the electric signal at higher ISOs, and this amplification may not be applied linearly. I would be interested in factual data you might have about this?
- Not all sensors have the same quality of analog to digital conversion near saturation, which may result in harsher transitions from non blown to blown areas. In essence, this would mean that some sensors behave non linearly near saturation. This may be true accross a very small range of illuminations, but would still basically not explain the highlight DR you claim exists.

Sorry, I still don't see anything in your argument that would explain highlight DR. Until proven otherwise, I'll stick tot the well accepted proposal that DR with linear sensors is ONLY depednant on shadow noise. The rest is just a pleasant illusion.

Merry x-Mas to you as well!

Cheers,
Bernard

The only thing that I am saying is that the linear raws that result from the analog to digital conversion, have compressed highlights and shadows, ie they are a linearized S-slope, that tries to behave like film was, its obvious when you open the files on your raw converter, just observe a high contrast scene carefully. This is different to the latitude that the data may have which allows us to produce a new S-slope during PP (D/A conversion). Regards, Theodoros. www.fotometria.gr
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12536
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: What about 36MP DSLRs?
« Reply #54 on: December 25, 2011, 06:35:03 am »

The only thing that I am saying is that the linear raws that result from the analog to digital conversion, have compressed highlights and shadows, ie they are a linearized S-slope, that tries to behave like film was, its obvious when you open the files on your raw converter, just observe a high contrast scene carefully. This is different to the latitude that the data may have which allows us to produce a new S-slope during PP (D/A conversion). Regards, Theodoros. www.fotometria.gr

Well, the raw file is not trying to do anything, it just contains the linear information recorded by the sensor. It has no idead what film was, it just records numerical values that are simply proportional to the amount of illuminations that reached the sensor, up to a point where the sensor saturates and beyond which you only record 255,255,255 whatever the amount of additional illumination.
 
Back to our initial point, do you now agree that highlight DR does not exist?

Cheers,
Bernard

fotometria gr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • www.fotometria.gr
Re: What about 36MP DSLRs?
« Reply #55 on: December 25, 2011, 07:34:02 am »

Well, the raw file is not trying to do anything, it just contains the linear information recorded by the sensor. It has no idead what film was, it just records numerical values that are simply proportional to the amount of illuminations that reached the sensor, up to a point where the sensor saturates and beyond which you only record 255,255,255 whatever the amount of additional illumination.
 
Back to our initial point, do you now agree that highlight DR does not exist?

Cheers,
Bernard

No it doesn't, but of course you are entitled to believe "its the linear info recorded by the sensor", its linear alright but not as it would have been recorded by the sensor if light/noise/sensitivity wasn't "tweaked" by the various "expeed, digitech, ...etc" engines! ...No it does exist, but you are entitled to believe it doesn't! Regards, Theodoros. www.fotometria.gr
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 12536
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: What about 36MP DSLRs?
« Reply #56 on: December 25, 2011, 07:42:38 am »

No it doesn't, but of course you are entitled to believe "its the linear info recorded by the sensor", its linear alright but not as it would have been recorded by the sensor if light/noise/sensitivity wasn't "tweaked" by the various "expeed, digitech, ...etc" engines! ...No it does exist, but you are entitled to believe it doesn't! Regards, Theodoros.

Hum... I guess we must be using different forms of logic. I'd say that mine is pretty standard, yours must be an improved version. :)

Cheers,
Bernard


« Last Edit: December 25, 2011, 07:44:11 am by BernardLanguillier »
Logged

fotometria gr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • www.fotometria.gr
Re: What about 36MP DSLRs?
« Reply #57 on: December 25, 2011, 08:09:12 am »

Hum... I guess we must be using different forms of logic. I'd say that mine is pretty standard, yours must be an improved version. :)

Cheers,
Bernard



+1  8) Cheers, Theodoros. www.fotometria.gr
Logged

hjulenissen

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2051
Re: What about 36MP DSLRs?
« Reply #58 on: December 25, 2011, 06:32:39 pm »

No it doesn't, but of course you are entitled to believe "its the linear info recorded by the sensor", its linear alright but not as it would have been recorded by the sensor if light/noise/sensitivity wasn't "tweaked" by the various "expeed, digitech, ...etc" engines! ...No it does exist, but you are entitled to believe it doesn't! Regards, Theodoros. www.fotometria.gr
You are entitled to an opinion, but I think that yours is wrong. I don't know if it is the translation, but I really don't see what you are trying to say with these posts. Do you have something concrete, of value, to teach us? Or are you just popping up once in a while with far-out claims?

-h
« Last Edit: December 25, 2011, 06:41:33 pm by hjulenissen »
Logged

fotometria gr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 568
    • www.fotometria.gr
Re: What about 36MP DSLRs?
« Reply #59 on: December 25, 2011, 06:47:24 pm »

Either it is linear, or it is not. Linear means no shoulder, no tweak, no digitech, no compressed highlights and shadows... but a linear response to the number of photons (within the known model limitations of noise and saturation). Since it seems that you have a different opinion from most of us (including the first sensible hits on google) on this issue, I would say that the burden of evidence is yours.

So do you have an image of a reference scene showing this? A reference to a paper? Some theory explaining what you belive to be happening?

-h
Please read my other quotes, you are out of subject which is (was) if highlight DR exists or not, this doesn't have to do with Raw dada being linear but whether the linearization comes from an A/D conversion of an S-slope, which it is! Regards, Theodoros. www.fotometria.gr
Logged
Pages: 1 2 [3] 4 5 ... 8   Go Up