Luminous Landscape Forum
Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: Guillermo Luijk on February 15, 2009, 12:00:43 am
-
According to DPreview (http://www.dpreview.com/reviews/canoneos5Dmarkii/page25.asp), the Canon 5D MKII can capture the same DR from ISO100 to ISO1600. But that is not all folks, it can even capture more DR at ISO3200!
-
This is not the first camera; the Nikon D300 is said to have higher DR at ISO 400 than at ISO 200.
But that's not all.
1. the lowest ISO is said to have lower DR than the next one,
2. the fake ISOs are said to have some gain (in fact the DR is always a full stop lower with the fake ISO steps).
The Nikon D3 is an even more miraculous camera: from ISO 100 to 200 it gains 1.1 EV, it goes lower at ISO 400, then again higher at 800 and even higher at 1600, and it drops only 0.5 EV from 12800 to 25600.
-
BTW Gabor I wanted to check by myself the fake ISO6400 on this camera (which is contrary to what the camera user manual suggests), and found a combed histogram which makes me think of fake ISO as you said:
(http://www.guillermoluijk.com/article/isos5dmkii/histos.gif)
But the holes, strangely, are not 100% empty, there are pixels on them. In a negligible amount though.
How would you explain that? perhaps some rounding errors in building the ISO6400 RAW file from the ISO3200 capture?
BR
-
BTW Gabor I wanted to check by myself the fake ISO6400 on this camera (which is contrary to what the camera user manual suggests), and found a combed histogram which makes me think of fake ISO as you said:
Neither Canon, nor Nikon afford to declare the truth. Two fake ISOs at the top must be the maximum; basta. The vast majority of users don't have any idea of this issue anyway, and believe everything written in the papers.
I was confronted for a short while ago with a Nikon paper stating, that the D2H applies an analog gain depending on the white balance. Some Nikonists took it as hard cash. I have proven that this is not so, then the question was "why would someone from Nikon lie?"
But the holes, strangely, are not 100% empty, there are pixels on them. In a negligible amount though
That is sometimes a single pixel; sometimes a few. I have no idea how this happens.
-
According to DPreview (http://www.dpreview.com/reviews/canoneos5Dmarkii/page25.asp), the Canon 5D MKII can capture the same DR from ISO100 to ISO1600. But that is not all folks, it can even capture more DR at ISO3200!
Guillermo,
The figures to which you refer relate to default camera jpegs. Another way of looking at this would be to say that the default jpeg DR of the 5D2 is only 1 & 2/3rds stops down at ISO 3200, but almost 2 stops down at ISO 100 to ISO 1600. It might be the case that Canon have done a really good job of maintaining DR at ISO 3200, and that there is no further DR increase to be gained by shooting RAW at ISO 3200. Whereas, using ACR with 5D2 RAW images at ISO 100, allows for nearly 2 stops of additional DR to be extracted.
-
The figures to which you refer relate to default camera jpegs. Another way of looking at this would be to say that the default jpeg DR of the 5D2 is only 1 & 2/3rds stops down at ISO 3200, but almost 2 stops down at ISO 100 to ISO 1600. It might be the case that Canon have done a really good job of maintaining DR at ISO 3200, and that there is no further DR increase to be gained by shooting RAW at ISO 3200. Whereas, using ACR with 5D2 RAW images at ISO 100, allows for nearly 2 stops of additional DR to be extracted.
Ray IMO it's much simpler than that, and the reason has nothing to do with the camera's JPEGs: DPreview's DR figures are just crap since their measurement criteria is completely wrong from any scientifical point of view. I just opened the thread to make some fun and, perhaps, hope someone in the team reaches it and, perhaps, has the splendid idea to suggest a change in their methods.
If they were right, the conclusion would be that every JPEG photographer with a Canon 5D MKII should stick with Loctite their ISO adjustment to 3200, since it is clearly the optimum ISO value according to that table.
BR
-
If they were right, the conclusion would be that every JPEG photographer with a Canon 5D MKII should stick with Loctite their ISO adjustment to 3200, since it is clearly the optimum ISO value according to that table.
You might be right, but there is also the issue of noise. I get the impression from the DXOMark website that there is sometimes a significant difference between a camera's DR performance at a particular ISO and the SNR at the same ISO. In other words, camera A might have better DR than camera B, but worse SNR than camera B at the same ISO.
I'm suggesting that the 5D2 jpeg at ISO 3200 might have approximately the same DR as the default jpeg at ISO 100, but much more noise than the jpeg shot at ISO 100. You would therefore not wish to glue your ISO adjustment to ISO 3200. Dynamic Range is not everything.
-
It is strange, but it is correct that a camera A noisier in most of the range, can have higher DR than a less noisy in most of the range camera B, as long as camera A gets less noisy than camera B in the deep shadows, where the SNR threshold is set for the DR criteria chosen.
But we are talking of the same camera here, and the 5D MKII will by certain be noisier in the _whole_ range at ISO3200 than at ISO100 for any given RAW exposure. So, if we assume DR is deeply related to noise (in fact it's just the range in which SNR does not fall below a chosen SNR threshold), I cannot believe by any means that the in-camera JPEG generation is so badly designed that produces valid information along a larger range at ISO3200 than at ISO100. And of course, the ISO100 image will be less noisy than the ISO3200 in the _whole_ range if RAW exposure is kept on both ISO settings, as is expected (i.e. shutter/aperture vary accordingly to the ISO set).
-
I cannot believe by any means that in-camera JPEG generation is so crappy that produces valid information along a larger range at ISO3200 than at ISO100. And of course, the ISO100 image will be less noisy than the ISO3200 in the _whole_ range if RAW exposure is kept on both ISO settings as expected (i.e. shutter/aperture vary accordingly to the ISO set).
Whatever the DR of the 5D2 at ISO 3200, the noise appears to be so bad (according to dpreview) that its resolution advantage compared with the D700 is wiped out. DXOmark doesn't address resolution, but at ISO 1600 and higher, both noise and dynamic range for the 5D2 and D700 are the same at a normalised 8x12" size. It would seem that at ISO 3200 and above, the D700 has slightly greater resolution than the 5D2 at any print size, acording to dpreview.
-
Whatever the DR of the 5D2 at ISO 3200, the noise appears to be so bad (according to dpreview) that its resolution advantage compared with the D700 is wiped out. DXOmark doesn't address resolution, but at ISO 1600 and higher, both noise and dynamic range for the 5D2 and D700 are the same at a normalised 8x12" size. It would seem that at ISO 3200 and above, the D700 has slightly greater resolution than the 5D2 at any print size, acording to dpreview.
Not sure what all the commotion is about. The Sony A900 blows all of them out of the water for DR (albeit at low ISO).
-
Not sure what all the commotion is about. The Sony A900 blows all of them out of the water for DR (albeit at low ISO).
Not quite. It's the Nikon D3X that blows all of them out of the water, as Bernard Languillier will be pleased to inform you .
-
But we are talking of the same camera here, and the 5D MKII will by certain be noisier in the _whole_ range at ISO3200 than at ISO100 for any given RAW exposure.
Guillermo,
It's interesting that DXOmark show approximately a 2 stop difference in DR between ISO 3200 and ISO 100. Dpreview claim slightly less than a 2 stop DR difference between the default jpeg at ISO 3200 and the auto-adjusted RAW ISO 100 image in ACR. Their methodology is different, but the results in some respects seem very close.
We know that in-camera processing can be very good, but it might not be the type of processing we prefer for any particular image.
It's conceibable to me, that Canon have made a decision with regard to in-camera processing of ISO 3200 images, to preserve the maximum dynamic range that the signal allows. At lower ISOs, Canon are probably applying different tone curves, and other adjustments, to make the image look as pleasing as possible, but at the sacrifice of some degree of DR.
Does that make sense?
-
Not quite. It's the Nikon D3X that blows all of them out of the water, as Bernard Languillier will be pleased to inform you .
Indeed. Well controlled highlights, noiseless shadows and a pleasant contrasty curve, without mentioning the fact that this was shot with a T/S lense.
(http://farm4.static.flickr.com/3469/3279020588_6ddf71aaab_o.jpg)
This doesn't take away anything from the A900's abilities though.
Cheers,
Bernard
-
Indeed, the dpreview DR measurings, but also the other sites' numbers for DR are quite confusing, they differ substantially (with DXO mark having the highest numbers) and in the end they don't tell much.
I have found Lloyd Chambers' blog ("DAP" on diglloyd.com) to give the most valuable information concerning IQ of the latest DSLRs. Lloyd is quite critical towards the 5dII, specially in terms of shadow noise (which tends to produce banding patterns) and colour. He shows picture samples to make his point, colour has been a critical point with Canon indeed, I have never really appreciated the slightly muted (or alternatively artificially boosted look of my 5d Mark I). Using high quality lenses à la Zeiss ZF betters it's output, but still doesn't make it as vivid and photogenic as the new top end nikon. Lloyd also shows that shadow noise compromises colour quality on low contrast pictures.
Unfortunately the D3x is too expensive for me, at least right now. It may take a D4 or a d700x to make me buy it.
Until then I use my Zeiss lenses on my Canon 5d via adapter and wait for a more affordable camera to come along.
But I have to admit that the D3x may be the first Nikon that really impresses me in terms of image quality. The 5d has always been an "almost there"- camera to me. Impressively sharp, but a bit weak on colour and vulnerable highlights.
regards
bernie
-
It's conceibable to me, that Canon have made a decision with regard to in-camera processing of ISO 3200 images, to preserve the maximum dynamic range that the signal allows. At lower ISOs, Canon are probably applying different tone curves, and other adjustments, to make the image look as pleasing as possible, but at the sacrifice of some degree of DR.
Does that make sense?
I wouldn't try to match conclusions about DR from different sites following so different criteria Ray. I think DPreview rely too much on the preadjusted modes (both JPEG camera modes or ACR RAW settings) when they could just go into the RAW data an measure SNR. But I agree that site is not so technically focused.
Indeed. Well controlled highlights, noiseless shadows and a pleasant contrasty curve, without mentioning the fact that this was shot with a T/S lense.
A D3X, with controlled highlights and noiseless shadows. 14-24 and TS lenses. Pano stuff. Moreover you enjoy living in Tokyo.
My conclusion Bernard is that your goal is not photography, but to ruin us all here
BTW do you have the 24mm PC-E? what could you tell about it in terms of geometric distortion, need to be corrected? and CA and sharpness?
Indeed, the dpreview DR measurings, but also the other sites' numbers for DR are quite confusing, they differ substantially (with DXO mark having the highest numbers) and in the end they don't tell much.
DxO Mark DR figures can be used to comare camera sensors. Moreover they provide the complete SNR curves (called 'Full SNR') that allow you to calculate the DR of any sensor applying your favourite criteria (they use the SNR>0dB criteria which is not very demanding, yielding such high DR figures).
-
I wouldn't try to match conclusions about DR from different sites following so different criteria Ray. I think DPreview rely too much on the preadjusted modes (both JPEG camera modes or ACR RAW settings) when they could just go into the RAW data an measure SNR. But I agree that site is not so technically focused.
I guess one has to appreciate the fact that Dpreview tries to cater to the needs of the majority of all those who buy cameras, including P&S cameras. I believe the majority of people who use cameras shoot in jpeg mode, as recommended by Ken Rockwell .
I think it's probably true also that more photographers use ACR than any other single RAW converter.
-
I guess one has to appreciate the fact that Dpreview tries to cater to the needs of the majority of all those who buy cameras, including P&S cameras. I believe the majority of people who use cameras shoot in jpeg mode, as recommended by Ken Rockwell .
I think it's probably true also that more photographers use ACR than any other single RAW converter.
(http://www.movie-blast.com/Groundhog_day.jpg)
-
A D3X, with controlled highlights and noiseless shadows. 14-24 and TS lenses. Pano stuff. Moreover you enjoy living in Tokyo.
My conclusion Bernard is that your goal is not photography, but to ruin us all here
Tokyo is the place to be these days, but the fact is that I only need good gear to compensate for my poor shooting skills.
Speaking of panos, I just came up with a 510 megapixel monster with infinite DoF (172 images involved in this photographic orgy) that should print nicely in 2.5x1 m at 360DPI.
BTW do you have the 24mm PC-E? what could you tell about it in terms of geometric distortion, need to be corrected? and CA and sharpness?
It takes about a +3 correction in PS lens correction module to totally get rid of distorsion on the 24 PCE, +2 does most of the job when the lens is not shifted too far out.
Sharpness is excellent on my sample and I have not noticed any outrageous CA, but haven't really been looking too thoroughyl either.
I have been using the 24 T/S as my regular wide angle lens on many assignements, even when I don't necesseraly need the T/S capability.
Cheers,
Bernard
-
I'll be the last to defend dpreview. However, I always wish to be accurate when commenting on something and I also seek to find out why things appear to be they way they do. People who have started this thread should have noted and mentioned that dpreview 'measures' DR using the camera's default jpeg settings (FWIW). Any default High ISO noise reduction applied to the higher ISO images could have an effect of increasing apparent DR (resolution loss not withstanding) by lowering the noise floor. In fact, washing out all detail by applying an extreme NR would indeed raise DR the way Dpr is measuring it.
So one can bash the method by which dpr are measuring DR (and he would be very right to do so) but he should at least report the facts completely and accurately.
-
NikosR, did you spend the same time as here in writing to the DPreview team to let them know they are providing non advanced users with misleading figures because they consider noise reduction as a DR enhancer? (which BTW is just another evidence of how surrealistic their analysis are).
Because I think that is by far much more important than the completeness and accuracy of this thread.
-
NikosR, did you spend the same time as here in writing to the DPreview team to let them know they are providing non advanced users with misleading figures because they consider noise reduction as a DR enhancer? (which BTW is just another evidence of how surrealistic their analysis are).
Because I think that is by far much more important than the completeness and accuracy of this thread.
What else are the poor b******s supposed to do? NR can take many forms, viz the strange sensor in the Fuji cameras. You just take what comes out of the camera and rate it. What they could be criticized for is having a DR measurement method that can be fooled by NR. But we'll talk about that on the day somone can convince me that there is actually a meaningful way to *really* define DR.
Edmund
----------
Pay peanuts, get monkeys; be kind to strangers, get friends.
-
NikosR, did you spend the same time as here in writing to the DPreview team to let them know they are providing non advanced users with misleading figures because they consider noise reduction as a DR enhancer? (which BTW is just another evidence of how surrealistic their analysis are).
Because I think that is by far much more important than the completeness and accuracy of this thread.
I can't be bothered with dpreview, but I'm bothered with this site. You opened a thread 'as a joke' as you said presenting some ridiculous findings on the part of dpreview. You could have taken the time to try to find a reason behind those findings. This would not have made the findings less ridiculous but, since you care about readers' education, would have educated a few.
-
But we'll talk about that on the day somone can convince me that there is actually a meaningful way to *really* define DR.
There is one, and in fact is much easier to apply than DPreview's strange manoeuvres. Just calculate the RAW exposure with respect to saturation at which SNR falls below some threshold (I'd suggest 12dB). That would help a lot to obtain comparable figures among different cameras.
Of course there is still the issue of cameras applying noise reduction on RAW data, but we would be quite closer to a rigurous measurement than the DP team are now.
I can't be bothered with dpreview, but I'm bothered with this site. You opened a thread 'as a joke' as you said presenting some ridiculous findings on the part of dpreview. You could have taken the time to try to find a reason behind those findings. This would not have made the findings less ridiculous but, since you care about readers' education, would have educated a few.
I guess you didn't spend any time writing to the DPreview team.
-
There is one, and in fact is much easier to apply than DPreview's strange manoeuvres. Just calculate the RAW exposure with respect to saturation at which SNR falls below some threshold (I'd suggest 12dB). That would help a lot to obtain comparable figures among different cameras.
Of course there is still the issue of cameras applying noise reduction on RAW data, but we would be quite closer to a rigurous measurement than the DP team are now.
How d'you measure *this* SNR with respect to saturation ?
Edmund
-
Let's remove DPReview from the subject and concentrate on the underlying issues, namely the methodology of determining the dynamic range.
If a review is focusing on potentional "JPEG shooters" (as opposed to "raw shooters"), then it is logical to evaluate the camera's capability based on in-camera JPEGs. The reviewers decide the basic parameters: the acceptable noise level and the minimum intensity level. If the result of the evaluation is, that the camera throws away three stops of the DR @ ISO 100, then either
1. the methodology or the actual measurements are useless, or
2. the final verdict must be Not recommended.
However, if the review is aimed at users, who want to squeeze out the best of the camera, then the JPEG based evaluation is worthless.
Just calculate the RAW exposure with respect to saturation at which SNR falls below some threshold (I'd suggest 12dB)
Why do you think that you or anyone else should determine the level of acceptable noise? Why would any specific value be declared as a "universal threshold"?
Someone may find a certain level of noise acceptable @ ISO 100 in landscapes shot in bright daylight; however, someone else may be shooting night club scenes @ ISO 1600. Why on earth would one apply the same criteria for all situations?
-
Why do you think that you or anyone else should determine the level of acceptable noise? Why would any specific value be declared as a "universal threshold"?
Someone may find a certain level of noise acceptable @ ISO 100 in landscapes shot in bright daylight; however, someone else may be shooting night club scenes @ ISO 1600. Why on earth would one apply the same criteria for all situations?
Gabor,
I believe I understand and I quite agree with what you're saying here. Acceptable noise floor is both subjective and scene / photograph specific. Where does that leave us with respect to measuring DR, as a single reference noise floor must be established for any measurements to be of any comparative value? Maybe to Dxo Mark's definition of S/N=1?
-
Why do you think that you or anyone else should determine the level of acceptable noise? Why would any specific value be declared as a "universal threshold"?
Someone may find a certain level of noise acceptable @ ISO 100 in landscapes shot in bright daylight; however, someone else may be shooting night club scenes @ ISO 1600. Why on earth would one apply the same criteria for all situations?
It's simple Gabor, the goal of these analysis is not to determine the DR using a criteria according to any given situation (if so there should be thousands of criteria), but to get a figure that allows to compare DR on different cameras.
I chose 12dB since it seems to me a good trade-off to export the conclusions from that comparision, to real situations. But in order to stablish comparisions, any other criteria will be as good (DxO Mark use 0dB).
Personally I prefer much more to look at the complete SNR curve plots, which are the only ones that let you find situations like that reported by Ray: camera A with more DR than camera B for a given criteria, but camera B is less noisy than camera A in most of the range, which could happen to be the range of interest for my application.
BR
-
Personally I prefer much more to look at the complete SNR curve plots
It's like the difference between an MTF plot vs. just one MTF-50 number.
-
Where does that leave us with respect to measuring DR, as a single reference noise floor must be established for any measurements to be of any comparative value? Maybe to Dxo Mark's definition of S/N=1?
I don't see any reason to declare a single noise level as floor. GLuijk suggested SNR curve plots. I prefer a chart in Excel like format. The format is less interesting; the essence is, that I am producing data for many different noise levels and one can pick from the list any level and compare it to another camera of another ISO of the same camera.
I posted an example in the thread "1 Feb, 2008 - Quality vs. Value - When is Enough Enough?", post #85, and a crop of the results in #125. I make literally hundreds of measurements in an ISO serie of such shots:
EV -> noise
9.28 -> 32.5
9.32 -> 35.1
9.41 -> 34.9
9.51 -> 41.0
9.59 -> 40.0
9.69 -> 43.2
9.75 -> 43.9
9.78 -> 44.7
9.89 -> 46.4
9.95 -> 51.7
The first number shows the intensity of a selected patch from saturation downwards, the second number is the degree of noise, as the standard deviation in percentage of the average pixel values in the selected patch. 51.7% is roughly SNR=2 at -9.95 EV. This means, that if you accept SNR=2, then the DR is 10 EV (this is the 40D at ISO 200). If you accept only SNR=3, i.e. 33% noise, the DR is somewhere between 9.3 and 9.4 EV.
I have done this for several cameras, and I will put them in a single Excel chart for easy comparison. Unfortunately, I don't have suitable raw files from the 5D2, nor from the D3X.
-
I don't see any reason to declare a single noise level as floor. GLuijk suggested SNR curve plots. I prefer a chart in Excel like format. The format is less interesting; the essence is, that I am producing data for many different noise levels and one can pick from the list any level and compare it to another camera of another ISO of the same camera.
I posted an example in the thread "1 Feb, 2008 - Quality vs. Value - When is Enough Enough?", post #85, and a crop of the results in #125. I make literally hundreds of measurements in an ISO serie of such shots:
EV -> noise
9.28 -> 32.5
9.32 -> 35.1
9.41 -> 34.9
9.51 -> 41.0
9.59 -> 40.0
9.69 -> 43.2
9.75 -> 43.9
9.78 -> 44.7
9.89 -> 46.4
9.95 -> 51.7
The first number shows the intensity of a selected patch from saturation downwards, the second number is the degree of noise, as the standard deviation in percentage of the average pixel values in the selected patch. 51.7% is roughly SNR=2 at -9.95 EV. This means, that if you accept SNR=2, then the DR is 10 EV (this is the 40D at ISO 200). If you accept only SNR=3, i.e. 33% noise, the DR is somewhere between 9.3 and 9.4 EV.
I have done this for several cameras, and I will put them in a single Excel chart for easy comparison. Unfortunately, I don't have suitable raw files from the 5D2, nor from the D3X.
I'm back home so I can supply 5DII RAWs for you please detail how you want them shot
Marc
-
I'm back home so I can supply 5DII RAWs for you please detail how you want them shot
Marc
Hi Marc, I remember you wrote once that you were on a trip months long.
If you have a printer, then there is a good solution. The same issue came up with Eric (a fellow poster here). He has a color checker card, but that turned out not to be reliable (in terms of noisiness). Then he had an idea: printing a color checker card. The result is much better than the original (but it can not be used for color profiling).
You can download a color checker file from here: http://www.rags-int-inc.com/PhotoTechStuff...hTarget_Lab.tif (http://www.rags-int-inc.com/PhotoTechStuff/MacbethTarget/MacbethTarget_Lab.tif)
For the noise measurement the colors are irrelevant; the point is, that the different color squares offer many different intensity levels in all three channels.
Before printing it out, pls remove the layer with the calibration numbers.
Print it please on hard, not mat paper, which does not have any visible texture. Eric printed it on glossy paper, which he is using for proofing.
The illumination should be *very* even, and the shot underexposed by three stops at least. In order to avoid unnecessary uploading, pls make two or three shots @ ISO 100 with different underexposure and upload them; I verify if they are all right. I would need them for all full-stop ISOs up to 3200; ISO 50 too only for demonstration, that it is a plain overexposed ISO 100 (there are some, who don't want to believe that).
Pls use a lens at an aperture, which does not create much vignetting.
Thanks
-
2 questions Gabor:
- When you say -9.95EV, you mean 9.95EV from sensor's saturation point, or from the top end of the bitscale (2^n-1 being n the RAW bitdepth). IMO the first approach is more interesting, but you need to calculate that saturation point of course. Having the sat point informed for each camera and ISO is also very interesting.
- As Emil Martinec commented some day, being Nikon's RAWs pre-scaled in the R and B channels (pre-WB), the SNR and thus your measured stdev for a given EV will be worse in them than in the G channel. Are you taking this into account? In which channel are you measuring noise?
I would love to have that Excel datasheet, obtaining the plots can be very interesting. I was even thinking of some simple VBA app reading your Excel and allowing camera comparisions of any brand and/or ISO values.
BR
-
Hi Marc, I remember you wrote once that you were on a trip months long.
If you have a printer, then there is a good solution. The same issue came up with Eric (a fellow poster here). He has a color checker card, but that turned out not to be reliable (in terms of noisiness). Then he had an idea: printing a color checker card. The result is much better than the original (but it can not be used for color profiling).
You can download a color checker file from here: http://www.rags-int-inc.com/PhotoTechStuff...hTarget_Lab.tif (http://www.rags-int-inc.com/PhotoTechStuff/MacbethTarget/MacbethTarget_Lab.tif)
For the noise measurement the colors are irrelevant; the point is, that the different color squares offer many different intensity levels in all three channels.
Before printing it out, pls remove the layer with the calibration numbers.
Print it please on hard, not mat paper, which does not have any visible texture. Eric printed it on glossy paper, which he is using for proofing.
The illumination should be *very* even, and the shot underexposed by three stops at least. In order to avoid unnecessary uploading, pls make two or three shots @ ISO 100 with different underexposure and upload them; I verify if they are all right. I would need them for all full-stop ISOs up to 3200; ISO 50 too only for demonstration, that it is a plain overexposed ISO 100 (there are some, who don't want to believe that).
Pls use a lens at an aperture, which does not create much vignetting.
Thanks
Give me a day or two
Marc
-
- When you say -9.95EV, you mean 9.95EV from sensor's saturation point, or from the top end of the bitscale (2^n-1 being n the RAW bitdepth)
Of course I calculate with the saturation level. This questions is offensive :-)
you need to calculate that saturation point of course
Actually I do not calculate it; I read it from an overexposed shot. Although there are some variations between copies of a camera model, but that is negligable small.
being Nikon's RAWs pre-scaled in the R and B channels (pre-WB)
I don't know about this myth.
I would love to have that Excel datasheet, obtaining the plots can be very interesting
At the moment I have the results in "raw" format. They have to be cleaned up; sometimes outlandish values occur. I uploaded now the 50D values for you, measured on a Stouffer transmission wedge: tab separated form (http://www.panopeeper.com/Noise/Canon50D_Noise.tab)
Sometimes I create a graph from them, but I find the numbers more useful. Attached the visualized charts of the 50D.
-
Those data look great.
I will let Emil know about in case he wants to comment something on the Nikon pre-WB. Anyway if you have some Nikon test, it would be clarifying to recalculate separate noise datasets for each RAW channel. If all behave the same it can be concluded there was no scaling.
-
Anyway if you have some Nikon test, it would be clarifying to recalculate separate noise datasets for each RAW channel. If all behave the same it can be concluded there was no scaling.
There ARE differences, but not due to WB. Nikon published something like Canon's "white papers" are, re the D2X, and declared that the analog stage applies WB. However, this is not so. (I think I mentioned the D2H in some other thread, but that's incorrect.)
-
Those data look great.
I will let Emil know about in case he wants to comment something on the Nikon pre-WB. Anyway if you have some Nikon test, it would be clarifying to recalculate separate noise datasets for each RAW channel. If all behave the same it can be concluded there was no scaling.
Nikon does a fixed scaling of the R and B channels after quantization whose purpose remains a mystery to me, since it introduces additional sources of roundoff error and reduces DR in those channels. It is easy to see from the RAW data, a histogram or list of populated RAW levels shows a regular pattern of gaps. From the pattern of gaps one can infer the multiplier by taking the list of numbers (1 2 3 4 5 6 7 ...) and doing integer multiplication and division, rounding off at each step. If the correct multiplier is chosen, then the output will be the list of populated levels. It is easy to get in the right ballpark via the fraction of levels that are unpopulated.
The multiplier is also quite evident in the fact that the gain (e-/RAW level) differs for the three color channels, being smaller for R and B than it is for G.
-
Give me a day or two
Marc
In focus or slightly out of focus? and so that I get this correct let the camera meter the correct exposure then under expose by 3 stops.
-
In focus or slightly out of focus?
This should not matter if your printer delivers a smooth, uniform color. Please do not touch the surface with bare fingers if you want to avoid being fingerprinted.
so that I get this correct let the camera meter the correct exposure then under expose by 3 stops
Honestly I have no idea, how auto metering would look like; that's the reason I suggested to make first at least two shows with different underexposure. The darkest patch has to be in the 12th stop; this requires huuuge underexposure. Give it a try with -3 and -4 EV and let's see.
You can pretest it in ACR: pick WB on the white patch, reset everything to 0 (blacks too!), curves linear, then the black patch should show R,G,B in the range 2 to 4.
ADDED
I am an airhead. If you make only one shot (let's say with -4 EV), then I can determine, how much more or less is necessary.
-
Hi,
Just a comment on the CC card, I was using a mini color checker that accentuates the problem. Originally I just introduced it as a reference, being aware of a potential problem with granularity. Exchanging some e-mails with Gabor i felt I needed a smooth copy of the CC card so we came up with the idea of printing it. The first copy I had got dusty and it seems that it's also sensitive for fingerprints, so be careful with those prints.
For illumination I had a couple of halogen lights standing on a table at about 45 degrees angle on each side. The lamps were slightly below so illumination was somewhat uneven, but seemingly good enough. Using strobes would be probably better, but I have no strobes and exposure may have been harder to control.
Best regards
Erik
Give me a day or two
Marc
-
Hi,
On the last series I shot for Panopeeper I used manual exposure, starting from EV 0 (meter reading) and going down to EV -5. I have some considerations regarding going below say 1/20 s, because illumination varies with alternating current frequency. This flicker may not be noticable with floodlights but was very obvious with my simple flourescent daylight tube arrangement. Gabor suggested that I get an electronic ballast, that would solve the problem.
Best regards
Erik
This should not matter if your printer delivers a smooth, uniform color. Please do not touch the surface with bare fingers if you want to avoid being fingerprinted.
Honestly I have no idea, how auto metering would look like; that's the reason I suggested to make first at least two shows with different underexposure. The darkest patch has to be in the 12th stop; this requires huuuge underexposure. Give it a try with -3 and -4 EV and let's see.
You can pretest it in ACR: pick WB on the white patch, reset everything to 0 (blacks too!), curves linear, then the black patch should show R,G,B in the range 2 to 4.
ADDED
I am an airhead. If you make only one shot (let's say with -4 EV), then I can determine, how much more or less is necessary.
-
I printed the chart A3+ (gloss paper) and was going to shoot it during the day using diffuse sunlight if that's OK? or does it have to be repeatable day to day?
Are 1 stop increments OK (-3 &-4) or are finer increments necessary?
Marc
-
I printed the chart A3+ (gloss paper) and was going to shoot it during the day using diffuse sunlight if that's OK? or does it have to be repeatable day to day?
1. Repeatability is not important.
2. Even the shots of the serie you are shooting could be of different setup, light, etc., I would not care. However, for demo purposes it is good if the setup is controlled and the shots are nicely with 1 EV difference as the ISO increases.
3. Sunshine is ok, but pls watch out, that you or the camera do not cause any shadow, not even indirect. I found out, that changing my body posture behind the tripod was enough to change the illumination, even though I was not between the light source and the scenery. Light reflected from the surrounding may change. I hope there won't be moving clouds.
Are 1 stop increments OK (-3 &-4) or are finer increments necessary?
One stop increment should be fine.
Thanks; now I am off to bed.
-
Gabor I was planning today to shoot a professional IT8 card I borrowed from a friend to calibrate my camera. Would you be interested in having the RAW files of it for an old fashioned little 350D? I gues I have to shoot all ISOs and at different exposure levels to ensure you have both blown and very deep shadows data.
BR
-
Would you be interested in having the RAW files of it for an old fashioned little 350D?
I don't know if anyone is interested today for the noise characteristics of the 350D, but perhaps for "historical reference" yes (as a documentation of the development).
I gues I have to shoot all ISOs and at different exposure levels to ensure you have both blown and very deep shadows data
I don't need any highlights, I know the saturation levels, among others from your shots of a computer monitor. The 350D behaves nicely, it always clips at 4095. I don't know if there are intermediate ISO steps and if yes how they behave, but I don't think that is important for anything.
-
Not sure if I am happy, proud or offended to have an historical camera. Find here (http://www.guillermoluijk.com/misc/canon350dit8.cr2) the RAW file at ISO100.
However I have to shoot it again because I had a lot of problems with reflections for being this IT8 so glossy. If you prefer to wait till I get a more uniform copy and several ISOs I intend to do that tomorrow.
This RAW file provides content for a quite wide range of EV:
(http://img8.imageshack.us/img8/2121/mg5539hiskr7.gif)
-
However I have to shoot it again because I had a lot of problems with reflections for being this IT8 so glossy
There is a problem with the card: it's surface appears pearly. Otherwise the gray wedge strip at the bottom would be very suitable, but the texture gets counted as noise. See the attachment.
-
-3, -4, -5
https://www.yousendit.com/transfer.php?acti...e9d73cd5317b701 (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654701736&email=dc1625501f8018ac8e9d73cd5317b701)
https://www.yousendit.com/transfer.php?acti...6dd61f8f18954b4 (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654701208&email=2c51656fb3755420c6dd61f8f18954b4)
https://www.yousendit.com/transfer.php?acti...b28da103c21da0a (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654699362&email=d23243de7a2ecbf20b28da103c21da0a)
Marc
-
There is a problem with the card: it's surface appears pearly. Otherwise the gray wedge strip at the bottom would be very suitable, but the texture gets counted as noise. See the attachment.
Yes it's because of the reflections. Tomorrow I'll try some darkening strategies (no idea which but I will; indoor rear lighting, defocusing a bit,...).
-
Marc and GLuijk,
the exposure of the middle one, #1201 is ok; shutter 1/80. However, the illumination is not even, particularly in the bottom row, and there was some reflection on the last two patches. I see a pearly structure here too; GLuijk thinks that it comes from the reflection, but I am not sure if that is so.
Can you print it semi-glossy/satin, and on thick paper? One of the reasons of the pearliness is, that the paper is thin and its structure becomes visible. Hold the paper (not printed on) towards the light; is the structure visible?
-
Satin paper!
https://www.yousendit.com/transfer.php?acti...6177490fcea515b (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654808480&email=ea42b200381e079886177490fcea515b)
Marc
-
Satin paper!
Excellent! The illumination is very even, the surface is clean (i.e. free of visible texture) and the exposure goes into the 11th stop.
Pls make the ISO set with this setup.
-
Excellent! The illumination is very even, the surface is clean (i.e. free of visible texture) and the exposure goes into the 11th stop.
Pls make the ISO set with this setup.
OK ISO 50, 100, 200, 400, 800, 1600 & 3200 same aperture same exposure just changing shutter speed. right?
Marc
-
OK ISO 50, 100, 200, 400, 800, 1600 & 3200 same aperture same exposure just changing shutter speed. right?
Right!
-
There is a problem with the card: it's surface appears pearly. Otherwise the gray wedge strip at the bottom would be very suitable, but the texture gets counted as noise. See the attachment.
Shoot pairs of images at each setting and take the difference image (and don't forget to divide the resulting std dev in the difference image by Sqrt[2]). Surface texture will then be largely irrelevant unless reflectance varies substantially across a patch -- such effects are second order in N/S. If one doesn't use difference images, one is opening the experimental technique to substantial systematic errors -- surface texture, illumination gradients, etc.
-
Shoot pairs of images at each setting and take the difference image
This requires a high degree of repeatability, which I do not assume. I take a different route: I carefully check the patches for even illumination, clean surface and close noise levels on parts of the patch. As the quasy same intensity level occurs in different patches and different channels, I can cross-check the results.
-
https://www.yousendit.com/transfer.php?acti...10f5cbc4d6c137b (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654867592&email=2bb05ea50be85e2e210f5cbc4d6c137b)
https://www.yousendit.com/transfer.php?acti...26bb4fd579c7678 (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654867056&email=5567895a7b8a9f3ee26bb4fd579c7678)
https://www.yousendit.com/transfer.php?acti...3db6f3a7e56d311 (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654866674&email=eda28c5152e3771273db6f3a7e56d311)
https://www.yousendit.com/transfer.php?acti...7e1e3f947ae5341 (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654866206&email=f438267af1bd2f8647e1e3f947ae5341)
https://www.yousendit.com/transfer.php?acti...be951787aaa1907 (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654864918&email=48df067a731aa9a84be951787aaa1907)
https://www.yousendit.com/transfer.php?acti...87d0db86e4670a4 (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654864564&email=31a601aee3d5bcd1a87d0db86e4670a4)
https://www.yousendit.com/transfer.php?acti...8dfc473be0c4203 (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654863994&email=acc255b59553f3b9b8dfc473be0c4203)
Marc
-
Marc,
thanks, I got them. It takes some time to evaluate them, I will email you.
-
This requires a high degree of repeatability, which I do not assume. I take a different route: I carefully check the patches for even illumination, clean surface and close noise levels on parts of the patch. As the quasy same intensity level occurs in different patches and different channels, I can cross-check the results.
On the contrary, taking difference images does not require a high degree of repeatability between shots, within reason (ie assuming one is using standard good technique -- tripod, manual exposure, manual focus, MLU, etc). As I stated, variations between shots lead to effects that are at most (N/S)^2 in the difference image, while variation within a patch for a single shot is of order N/S, where "N" here denotes the variation in surface reflectivity of the target. Have the person with the "pearly" color chart take a pair of images and see for yourself. As long as you're having people take images to order, it's worth doing it right.
-
Hi Marc,
Which paper did you use? I used a glossy paper and Gabor still seems to see some structure in my images. I could try to use the same paper as you did.
Best regards
Erik
Satin paper!
https://www.yousendit.com/transfer.php?acti...6177490fcea515b (https://www.yousendit.com/transfer.php?action=batch_download&send_id=654808480&email=ea42b200381e079886177490fcea515b)
Marc
-
Hi Marc,
Which paper did you use? I used a glossy paper and Gabor still seems to see some structure in my images. I could try to use the same paper as you did.
Best regards
Erik
Why don't you try to shoot a bit out of focus?
-
Hi Marc,
Which paper did you use? I used a glossy paper and Gabor still seems to see some structure in my images. I could try to use the same paper as you did.
Best regards
Erik
Erik
Ink Jet Art Micro Ceramic Luster
I would have used their Duo Bright Matt (less texture) but I was out
The luster had a few reflections so I carefully positioned 2 daylight fluorescents one on each side
Marc
-
I'm suggesting that the 5D2 jpeg at ISO 3200 might have approximately the same DR as the default jpeg at ISO 100, but much more noise than the jpeg shot at ISO 100.
Assuming that the mapping of incident photon count and value recorded is linear and constant, this isn't possible; the noise is what determines dynamic range.
-
[attachment=11601:Marc_LL_...lections.GIF]The shots of the color checker provided by Marc are excellent for measurement. The attached graphs confirm, that the result is reliable; the values from the three raw channels are picked from all 24 patches of the color checker without contradiction (no manual adjustment took place).
What one can see from the graphs directly is:
1. ISO 50 and 100 are virtually identical, as they are supposed to be,
2. ISO 3200 is exactly 1 EV from ISO 1600. This is important, for ISO 3200 is not a fake ISO, nevertheless it should not be used when shooting raw, for it reduces the DR by a full stop,
3. the increasing spacing of the graphs with increasing ISO demonstrates, that the higher the ISO step, the less the real gain from the previous step (i.e. the more loss of the DR).
I am not a big fan of these graphs, I find them pretty useless. I will put the numbers in Excel charts with other cameras' data. I have now reliable measurements from the Canon 10D, 20D, 40D, 50D, 1DMkII, 1DMkII and Nikon D3, and Erik will provide the shots with a Sony A900.
UPDATE
Somehow I messed up the attachments; the formerly displayed graphs were from the Canon 50D. Sorry for that.
-
I am not a big fan of these graphs, I find them pretty useless.
If one plots (noise)^2 vs signal, the inverse of the slope is the gain in electrons/raw level, and the intercept at zero signal is the read noise squared. The other thing that is useful to plot is log(S/N) vs log(S), which provides a more complete picture of image quality over the dynamic range of a capture.
-
Not quite. It's the Nikon D3X that blows all of them out of the water, as Bernard Languillier will be pleased to inform you
Indeed, as DPReview confirms at http://www.dpreview.com/reviews/nikond3x/page21.asp (http://www.dpreview.com/reviews/nikond3x/page21.asp) !
From ISO 1600 to ISO 3200, the D3X gains a stunning 0.8 stops in the shadows, so total range varies like this:
8.4-8.6 at ISO 100 to 1600
9.3 at ISO 3200!
All I can say is that this is a measure of the combined effects of sensor performance, pre-amplification, tone curves, noise reduction etc.
Once a camera has options for dealing with contrasty scenes (ones of high subject brightness range) it is rather pointless to attach any value to high DR measurements from default JPEG settings. The default settings of all cameras are providing completely adequate DR for normal scenes, and are not usually intended to be used with highly contrasty scenes. Indeed, JPEG's that score well for DR will probably give a flatter look to typical scenes, due to using a flatter tone curve.
The options for dealing with high SBR scenes do not need to be just "shoot RAW and fix it in post"; they can be things like Nikon's D-lighting, or as simple as lower contrast settings.
-
According to DPreview (http://www.dpreview.com/reviews/canoneos5Dmarkii/page25.asp), the Canon 5D MKII can capture the same DR from ISO100 to ISO1600. But that is not all folks, it can even capture more DR at ISO3200!
Yes, too bad the 5DII DR is at best equal to, but in most cases less than, the DR of the D3X over the same range. This is according to DPReview, so the same measuring methodology was used in both cases. In fact, at ISO 3200, the D3X DR is 9.3 stops.
-
From ISO 1600 to ISO 3200, the D3X gains a stunning 0.8 stops in the shadows, so total range varies like this:
8.4-8.6 at ISO 100 to 1600
9.3 at ISO 3200!
Well, gaining 0.8 EV in the shadows from ISO 1600 to 3200 is stunning (if it is really so). The 5D2 gains nothing from 1600 to 3200, and only about 0.4 EV or less from ISO 800 to 1600.
However, if the gain in the shadows is 0.8 EV and the DR is 8.6 EV @ ISO 1600, then the DR @ 3200 is 8.4 EV. The 9.3 is a myth.
-
Gabor
Did you update your spreadsheet yet?
Marc
-
Gabor
Did you update your spreadsheet yet?
Marc,
I filled the spreadsheet with reliable data (i.e. measured on Stouffer transmission wedge or on the even better printed color chart) from Nikon D3, Canon 20D, 40D, 50D and 5DMkII. The Canon 1DMkII and 1DMkIII are coming.
It is downloadable as an Excel chart (http://www.panopeeper.com/Demo/NoiseAnalysis.xls); however, I am not happy with that. It does contain all the measurement data, but it is not easy to use. It can be sorted, of course, but that's not enough for me.
The data beside the camera model and ISO is the relative intensity of the measured points, in EV from clipping backwards, and the correlated noise in percentage of the average pixel intensity.
Example: the Canon 5DMkII with ISO 100 has 21.6% noise when the intensity is -8.9 EV (i.e. close to the end of the 9th stop of the dynamic range).
Now, if you want to compare this with the D3, then look for a row among the ISO 100 rows, containing the closest value to 8.9 (this is at the moment 8.92); the correlated noise is 17%. In turn, if you want to know the difference in EV, then look for a row with close to 21.6%; I interpolate this to -9.25 EV, i.e. the D3 has 0.35 EV advantage at ISO 100 (the accuracy of the measurements does not justify to talk about hundredths of an EV).
I would like to make a better presentation, but I don't want to invest too much work in that; anyway, at the moment I am still collecting data (measuring other cameras' results).
-
Assuming that the mapping of incident photon count and value recorded is linear and constant, this isn't possible; the noise is what determines dynamic range.
Sorry, I missed that comment. This surprisingly similar DR at different ISOs results from in-camera processing of the raw data. Noise is only one factor in determining DR. You can reduce the DR of any image by simply applying a tone curve that brightens the highlights and darkens the shadows. This is usually what tends to happen with in-camera jpegs. The images are processed to look 'punchy' and there's a consequent loss of detail in the brightest highlights and the darkest shadows..
There are a couple of comments from Dpreview in the comparison between the 5D2 and D700, which give an indication as to what might be going on here. There's apparently a quite significant loss of resolution in 5D2 images at ISO 3200 which brings resolution down to the level of the D700 at ISO 3200. One might conclude that this is a result of excessive in-camera noise reduction being applied to the 5D2 image, more than is being applied to the D700. However, Dpreview also note that D700 RAW images are also a whisker more detailed than 5D2 RAW images at ISO 3200.
This tends to suggest that the reason for the high DR of 5D2 in-camera jpegs at ISO 3200 is not excessive noise reduction, but a flatter or less contrasty (or simply different) tone curve which preserves all the DR that is present in the RAW data. Such a view tends also to be confirmed by Dpreview's comparison of RAW DR with jpeg DR at ISO 3200. Jpeg DR at ISO 3200 is almost 2 stops down from the best ACR conversion at base ISO. DXOmark similarly claim that the 5D2's sensor at ISO 3200 has 2 stops less DR than at base ISO.
One might therefore draw the conclusion that Cannon have simply done a first rate job with in-camera processing of ISO 3200 RAW data, preserving all the DR that exists. However, at base ISO, and other ISOs in between, they have made a decision to sacrifice a certain amount of DR in the interests of a better looking result straight out of the box.
Those who already own a 5D2 should be able to confirm this hypothesis by comparing RAW conversions of ISO 3200 images with in-camera jpegs of the same scene taken at ISO 3200. Can you extract any more DR than the in-camera processing does, from the RAW image at ISO 3200?
-
It is downloadable as an Excel chart (http://www.panopeeper.com/Demo/NoiseAnalysis.xls); however, I am not happy with that. It does contain all the measurement data, but it is not easy to use. It can be sorted, of course, but that's not enough for me.
That's excellent stuff Gabor, pity you didn't produce data for the upper f-stops. I think they can also be interesting in terms of noise, for example if the user plans to systematically do exposure bracketing to work only in the cleanest ranges of the sensor (like I do).
I took the freedom to plot your 5D MKII data in the form of SNR plots in dB:
(http://img520.imageshack.us/img520/8425/snr5dmkii.gif)
Conclusions:
Per-pixel DR of this camera with SNR>12dB criteria would be around:
9.1EV at ISO100
9.0EV at ISO200
8.7EV at ISO400
8.2EV at ISO800
7.5EV at ISO1600
6.5EV at ISO3200
It's easy to see on them that ISO50 is a true ISO100 as you found out.
I also plotted the SNR improvement curve obtained by rising ISO, and as Emil predicted according to his measures and I have analysed in some RAW files from this camera, there is no improvement in reaching ISO3200 even if it's an electronic ISO value (DR at ISO3200 is exacly 1EV less than at ISO1600).
Just for checking I compared the value for ISO1600 at -6EV, which provides SNR=17dB -> SNR=2,8EV which agrees with Emil's plot here (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise1ds3.gif) for the similar sensor 1Ds MKIII.
Find the Excel file here (http://www.guillermoluijk.com/misc/noiseanalysis_panopeeper.zip). I plotted the graphs one by one and layered them in PS.
The same graph from the 50D data:
(http://img211.imageshack.us/img211/1065/snr50d.gif)
Per-pixel DR of this camera with SNR>12dB criteria would be around:
8.7EV at ISO100
8.7EV at ISO200
8.25EV at ISO400
7.5EV at ISO800
6.7EV at ISO1600
BR
-
Now if only we could get Bernard to take some shots with his Dx3!
Marc