Ad
Ad
Ad
Pages: « 1 ... 6 7 [8]   Bottom of Page
Print
Author Topic: What about 36MP DSLRs?  (Read 31038 times)
fotometria gr
Sr. Member
****
Offline Offline

Posts: 568


WWW
« Reply #140 on: December 31, 2011, 04:04:05 AM »
ReplyReply

I'd have to agree with that.

Our Greek friend has still not replied to the following question: how do you explain that the data inside the raw files is linear relative to the illumination reaching the sensor?

His esoteric theories about amplification etc... could possibly be true. BUT, if they were true, they would translate into non linearity in the data stored in the raw file relative to incoming illumination.

In reality, these non linearities are simply not there.

Cheers,
Bernard

But I have Bernard, its not linear relative to the illumination that reaches the sensor, its linear to the illumination that the sensor records! Make a simple test, I think this may convince you: Choose a high contrast scene on a bright sunny day, a scene that includes the sun in your frame, use your D7000 at 100iso and fit a most resistant to flare lens, like the old 12-24DX or a zeiss, now with the camera set in manual make 21 different exposures, one at 1/250 - f8.0, ten overexposing that a stop at a time and another 10 underexposing a stop at a time (8stops until1/8000 f22 and for the last 2 use a -1 and -2 filter). Shoot that in Raw+jpeg, now go back and open the files with NX2.2 and do not apply anything to them, make sure you keep the exposure graph linear. Now starting with the +10stops blown image watch carefully each next image towards under exposure and compare it to the previous one. You may notice the following: 1. At the blown image nothing behaves linearly to the next stop. 2. When some of the image scene enters linearity as you move towards the under exposed images, the highlights still don't behave linearly. 3. When a part of the image exits linearity towards the shadows, it starts resisting to disappear and starts not to behave linearly. 4. When the actual planet of the sun starts forming its spherical shape from the glare around it (at around -2 to -3stops from f8.0, 1/250 depending on the atmosphere's level of humidity) it makes very little difference when you underexpose another stop, while if you still have correctly exposed parts of the scene, they do behave linearly. 5. If you apply -1 stop of digital underexposure from the Raw converter to the previous condition (no. 4) the image doesn't respond the same as when it was shot by another stop of underexposure, only the "mid" part of it does! 6. You can't make the sun disappear or behave linearly not even if you go to the mostly underexposed of your images, its still on the highlights! In fact it won't disappear even if it was -20 stops from 1/250, f8.0! Believe me I've done it! Regards, Theodoros. www.fotometria.gr
Logged
fotometria gr
Sr. Member
****
Offline Offline

Posts: 568


WWW
« Reply #141 on: December 31, 2011, 06:25:30 AM »
ReplyReply

But I have Bernard, its not linear relative to the illumination that reaches the sensor, its linear to the illumination that the sensor records! Make a simple test, I think this may convince you: Choose a high contrast scene on a bright sunny day, a scene that includes the sun in your frame, use your D7000 at 100iso and fit a most resistant to flare lens, like the old 12-24DX or a zeiss, now with the camera set in manual make 21 different exposures, one at 1/250 - f8.0, ten overexposing that a stop at a time and another 10 underexposing a stop at a time (8stops until1/8000 f22 and for the last 2 use a -1 and -2 filter). Shoot that in Raw+jpeg, now go back and open the files with NX2.2 and do not apply anything to them, make sure you keep the exposure graph linear. Now starting with the +10stops blown image watch carefully each next image towards under exposure and compare it to the previous one. You may notice the following: 1. At the blown image nothing behaves linearly to the next stop. 2. When some of the image scene enters linearity as you move towards the under exposed images, the highlights still don't behave linearly. 3. When a part of the image exits linearity towards the shadows, it starts resisting to disappear and starts not to behave linearly. 4. When the actual planet of the sun starts forming its spherical shape from the glare around it (at around -2 to -3stops from f8.0, 1/250 depending on the atmosphere's level of humidity) it makes very little difference when you underexpose another stop, while if you still have correctly expoe sed parts of the scene, they do behave linearly. 5. If you apply -1 stop of digital underexposure from the Raw converter to the previous condition (no. 4) the image doesn't respond the same as when it was shot by another stop of underexposure, only the "mid" part of it does! 6. You can't make the sun disappear or behave linearly not even if you go to the mostly underexposed of your images, its still on the highlights! In fact it won't disappear even if it was -20 stops from 1/250, f8.0! Believe me I've done it! Regards, Theodoros. www.fotometria.gr

I should add to the above that No:5 is the most important, in fact I may suggest a more accurate test: Shoot a backlit grayscale card (like the one they use in DPreview), one with at least 15 stops on it, with + and - 1 stop of exposure,  then apply the above test with + and - 1stop of digital exposure compensation on the mid of the 3 raws that will result, when you open it in your raw converter and compare the images that have resulted from digital exposure compensation, to the ones that have been produced be real exposure compensation, it will prove to you that the RAWs that have been recorded during capture were recorded linearly but in reality are linearized, they contain more info than they should in highlights and lowlights respectively (or they lack more info that they should from the other side of the tonal range respectively). Happy new year to everybody, Theodoros. www.fotometria.gr
« Last Edit: December 31, 2011, 07:06:13 AM by fotometria gr » Logged
michael
Administrator
Sr. Member
*****
Offline Offline

Posts: 4894



« Reply #142 on: December 31, 2011, 08:24:58 AM »
ReplyReply

This is an interesting and worthwhile debate, but really kids! Time to show some manners.

Either play nicely together or you'll be sent to your rooms (and this thread will be closed).

Michael
Logged
ErikKaffehr
Sr. Member
****
Online Online

Posts: 7648


WWW
« Reply #143 on: December 31, 2011, 09:12:51 AM »
ReplyReply

Hi,

I may try your suggestions, when I have time.

I have done some tests with the Stouffer wedge and found that the response was linear. BUT, the default settings in LR will not give a linear image, and the image will be hard to interpret. I used Imatest for my evaluation and it is presented here:

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/56-an-lr-view-of-the-stouffer-wedge?start=2


Anyway, this is a discussion on 36-MP DSLRs and not sensor linearity, so I would strongly suggest that you start a new topic named "Are sensors linear?" or so.


Best regards
Erik


But I have Bernard, its not linear relative to the illumination that reaches the sensor, its linear to the illumination that the sensor records! Make a simple test, I think this may convince you: Choose a high contrast scene on a bright sunny day, a scene that includes the sun in your frame, use your D7000 at 100iso and fit a most resistant to flare lens, like the old 12-24DX or a zeiss, now with the camera set in manual make 21 different exposures, one at 1/250 - f8.0, ten overexposing that a stop at a time and another 10 underexposing a stop at a time (8stops until1/8000 f22 and for the last 2 use a -1 and -2 filter). Shoot that in Raw+jpeg, now go back and open the files with NX2.2 and do not apply anything to them, make sure you keep the exposure graph linear. Now starting with the +10stops blown image watch carefully each next image towards under exposure and compare it to the previous one. You may notice the following: 1. At the blown image nothing behaves linearly to the next stop. 2. When some of the image scene enters linearity as you move towards the under exposed images, the highlights still don't behave linearly. 3. When a part of the image exits linearity towards the shadows, it starts resisting to disappear and starts not to behave linearly. 4. When the actual planet of the sun starts forming its spherical shape from the glare around it (at around -2 to -3stops from f8.0, 1/250 depending on the atmosphere's level of humidity) it makes very little difference when you underexpose another stop, while if you still have correctly exposed parts of the scene, they do behave linearly. 5. If you apply -1 stop of digital underexposure from the Raw converter to the previous condition (no. 4) the image doesn't respond the same as when it was shot by another stop of underexposure, only the "mid" part of it does! 6. You can't make the sun disappear or behave linearly not even if you go to the mostly underexposed of your images, its still on the highlights! In fact it won't disappear even if it was -20 stops from 1/250, f8.0! Believe me I've done it! Regards, Theodoros. www.fotometria.gr

« Last Edit: December 31, 2011, 09:15:18 AM by ErikKaffehr » Logged

PierreVandevenne
Sr. Member
****
Offline Offline

Posts: 510


WWW
« Reply #144 on: December 31, 2011, 10:25:54 AM »
ReplyReply

In reality, these non linearities are simply not there.

It's really hard to be categorical on those issues. Really. There is a huge number of things going on in the background at the sensor level and Sony isn't exactly forthcoming with details about the sensors it doesn't sell to the general public/small scale hardware developer. It could vey well be that they combine several exposures in a single one - they do it in some sensors of the Exmor line (possibly all recent sensors). Very roughly, the idea is to take short exposures on the pixels that receive a lot of photons while doing the main exposure and then combining the result. The mecanism is described here in general terms

http://pro.sony.com/bbsccms/assets/files/cat/camsec/brochures/prodbroch_viewdr_ve.pdf

Nothing prevents them from doing it multiple times in their DSLR sensors, virtually increasing the well capacity of over exposed pixels, and then massage the result into some linear looking file. It could be possible to test for that by shooting a wide DR scene at 1/1000th of a second and 1/30th of a second: if the second scene has a wider DR towards the highlights, it would show that in-sensor exposure summing is at play. (there may be more to it to create a good test, haven't fully thought about it.)

Note that Sony mentions the superiority of its technique over competitors who use partial charge transfers to extend the dynamic range of their sensors.

Partial charge transfers  are well known for introducing non linearities, see this article for example

www.mdpi.com/1424-8220/9/12/9452/pdf

That's why I restate my "no simple categorical answer exists". :-)
Logged
fotometria gr
Sr. Member
****
Offline Offline

Posts: 568


WWW
« Reply #145 on: December 31, 2011, 10:27:25 AM »
ReplyReply

Hi,

I may try your suggestions, when I have time.

I have done some tests with the Stouffer wedge and found that the response was linear. BUT, the default settings in LR will not give a linear image, and the image will be hard to interpret. I used Imatest for my evaluation and it is presented here:

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/56-an-lr-view-of-the-stouffer-wedge?start=2


Anyway, this is a discussion on 36-MP DSLRs and not sensor linearity, so I would strongly suggest that you start a new topic named "Are sensors linear?" or so.


Best regards
Erik


I remind you that It was never my intention to change the subject Erik, I was about the importance of highlight DR and the negative impact that resolution increase would have to it and it was Bernard that quoted that HDR doesn't exist because dada is linear, which you supported. Anyway, since finally (I suppose) that we all agree that HLDR does exist and it benefits from larger pixels as is LLDR and the whole DR for that matter, wouldn't you agree that its the best for our photography to deal with these matters (DR and NR) first  and let resolution advance in the future, when inevitably pixels may be smaller but will be able to retain what has to be gained in performance? Or shall we retain performance to the low (IMO) standards that it has now and advance resolution, which will only increase the already huge possible prints that can be done? Happy new year to everybody, Theodoros. www.fotometria.gr
Logged
Hans van Driest
Newbie
*
Offline Offline

Posts: 20


« Reply #146 on: December 31, 2011, 11:55:03 AM »
ReplyReply

It is indeed hard to know what is done, by the manufatcturer, on the sensor. On the other hand, if one beliefs the results of Dx0, and look at the Nikon D3s, D3x and d7000 (in screen mode, the results per pixels), the dynamic range increases as the pixel size decreases.
The method with multiple exposures per pixel, if the light intensity is high, seems less likely on the type of sensor used in a DSLR. All of them , as far as I know,  need a shutter, at least to terminate the exposure.
Logged
PierreVandevenne
Sr. Member
****
Offline Offline

Posts: 510


WWW
« Reply #147 on: December 31, 2011, 12:27:39 PM »
ReplyReply

as far as I know,  need a shutter, at least to terminate the exposure.

Not really - how would they expose in video mode/ do live view if they needed a physical shutter, rotating wheel, whatever... In the Exmor line used in surveillance cameras, it is very clear from the data sheets and litterature that they are essentially widening DR by doingmultiple exposures at the pixel level. You "simply" need a not too noisy, characterizable method to read and drain the charge in a particular pixel if it fills fast enough. "simply" is a a way of speaking of course as it involves a lot a work at a very small scale...

Logged
fotometria gr
Sr. Member
****
Offline Offline

Posts: 568


WWW
« Reply #148 on: December 31, 2011, 01:06:16 PM »
ReplyReply

It is indeed hard to know what is done, by the manufatcturer, on the sensor. On the other hand, if one beliefs the results of Dx0, and look at the Nikon D3s, D3x and d7000 (in screen mode, the results per pixels), the dynamic range increases as the pixel size decreases.
The method with multiple exposures per pixel, if the light intensity is high, seems less likely on the type of sensor used in a DSLR. All of them , as far as I know,  need a shutter, at least to terminate the exposure.

Hi, DR as measured by DXO, is different to HLDR, its a matter of noise acceptance of the test in LLDR that decides the total DR, without being wrong in their measurment, DXO finds MFDBs to be worst than most DSLRs which is not the case for any photographer that has used a MFDB (far from it), in photography HLDR is far more important (to most photographers) than LLDR, in MFDBs the noise level at deep shadows is far less than DSLRs and to some tests this appears as less LLDR extension, which of course affects the total result of the test. I personally find my MFDB (Imacon 528c) to be almost as good as Fuji S5Pro in HLDR and both of them to be more than a stop better than either D3X or D700. OTH, I find them, both Fuji and my MFDB to be about a stop worst than film on HLDR. Happy new year to everybody, Theodoros. www.fotometria.gr
Logged
Pages: « 1 ... 6 7 [8]   Top of Page
Print
Jump to:  

Ad
Ad
Ad