How about Dynamic range?

[bronxbombers]

the 40D did have pretty clean shadows in terms of nasty looking pattern banding noise, as you found, it has less than the 20D or 50D or even 5D2, which lets you mess around in the depths without bringing up ugly lines and bands more easily than with some others

although it does have more random spot noise than the 5D2 in the depths

the measured DR is also lower than the 5D2, although taking into account how ugly banding looks to the eye they effecitvely might be almost the same (of course the 5D2 will have better SNR and less random noise in the midtones by quite a bit)

i believe the 1D3 also does very well in the low ISO depths, i think it or perhaps the 1Ds3 have the highest usable ISO100 DR of any of the canon bodies

the 5D2 at ISO1600 has much higher DR than either the 1D3 or 40D though

Look I did a lot of shooting w different cameras in Raw only and processed in DPP and CS3 and AS SOON as you open the 40d file there's no need for peeping and tweaking..you can see it right away! my findings were that the 40d is better than the 50d (specially the shadows) WAY BETTER than the 30d,20d,XT (in highlights recovery also...40d can recover more highlight detail).

That said, 5d files were better than the 40d in DR (slightly)...and so were MkIII files but that is to be expected really given the sensors

Note: Before you jump on me I want to add that i'm a semi-professional photographer...and yes, I sell quite a lot of pics and I have a free-lance contract w a luxury magazine. These were MY FINDINGS shooting anything from events,jewellery and mostly wildlife.

--

Big lights will INSPIRE you, Mariano will RETIRE you, let's hear it for New York, New York, New York!!

 

[pixelpepper]

The calculated value was 1198.37, but since I didn't have good levels for white level, so assumed 1024 or black level, so assumed 0

I am lost. I have no idea, what you are referring to.

I don't know the actual ADU range used by the camera, they rarely use the full 4096 (or 16348 in 14 bit). Black is not always at zero, white is not at 4095. I don't know the actual full range. I don't know what the count for your '-2ev' actually is. Therefore, I had to make an assumption. The percentage error won't be large however, but it does make it pointless using a figure to six significant places, so I used one to two significant places.

And it was exactly 5 ADU. There's a co-incidence. How big was you patch of black?

I don't understand the question; I posted I measured it on a black frame . Therefor the black patch is 6080x4044 pixels large. I don't remember how large are I measured, but it is totally irrelevent .

When you look at measured figures, particularly something like a standard deviation, which is calculated from many individual observations, and integral value is pretty suspicious. So, my question is, does '5 ADU' mean '5.00 ADU' or is it a measurement to one significant place? If the latter, that imprecision in measurement will have a substantial effect as the scale of the read noise and shot noise becomes comparable. The reason for asking the size of the measured patch, is the number of pixels will directly affect the precision of your measurement of the standard deviation - it is far from irrelevent.

What was that in ADU? not sure whether you're referring to 0.5% of pixel value or 0.5% of full scale. Neither seems consistent with your other measurements.

1. It refers to the pixel value range ("full scale" in your terminology).

2. Forget about the ADU, it is useless.

But necessary if we are to complete the calculations. You could also use your '%' units, but you do have to know '%' of what.

If the raw data is lossless, then the pixel value range is from 128 to 4095,

Really? That's remarkably accurate adjustment of the capture chain. Generally there is some headroom for sample variation. Perhaps you were measuring the one camera which had the maximum permissible gain and just peaked at 4095.

but if it is lossy, then the range is from 256 to 8191; all measured values double. However, if you convert the lossy raw file with Adobe's DNG converter, it spreads the values up to 16383.

I wouldn't have though there's much point using lossy raw or a DNG processed file for these purposes.

Therefor, if one refers to pixel values, one has to say in which format they relate to.

Or, one has to refer to pixel values.

 

[Steen Bay]

The calculated value was 1198.37, but since I didn't have good levels for white level, so assumed 1024 or black level, so assumed 0

I am lost. I have no idea, what you are referring to.

I don't know the actual ADU range used by the camera, they rarely use the full 4096 (or 16348 in 14 bit). Black is not always at zero, white is not at 4095. I don't know the actual full range. I don't know what the count for your '-2ev' actually is. Therefore, I had to make an assumption. The percentage error won't be large however, but it does make it pointless using a figure to six significant places, so I used one to two significant places.

And it was exactly 5 ADU. There's a co-incidence. How big was you patch of black?

I don't understand the question; I posted I measured it on a black frame . Therefor the black patch is 6080x4044 pixels large. I don't remember how large are I measured, but it is totally irrelevent .

When you look at measured figures, particularly something like a standard deviation, which is calculated from many individual observations, and integral value is pretty suspicious. So, my question is, does '5 ADU' mean '5.00 ADU' or is it a measurement to one significant place? If the latter, that imprecision in measurement will have a substantial effect as the scale of the read noise and shot noise becomes comparable. The reason for asking the size of the measured patch, is the number of pixels will directly affect the precision of your measurement of the standard deviation - it is far from irrelevent.

What was that in ADU? not sure whether you're referring to 0.5% of pixel value or 0.5% of full scale. Neither seems consistent with your other measurements.

1. It refers to the pixel value range ("full scale" in your terminology).

2. Forget about the ADU, it is useless.

But necessary if we are to complete the calculations. You could also use your '%' units, but you do have to know '%' of what.

If the raw data is lossless, then the pixel value range is from 128 to 4095,

Really? That's remarkably accurate adjustment of the capture chain. Generally there is some headroom for sample variation. Perhaps you were measuring the one camera which had the maximum permissible gain and just peaked at 4095.

but if it is lossy, then the range is from 256 to 8191; all measured values double. However, if you convert the lossy raw file with Adobe's DNG converter, it spreads the values up to 16383.

I wouldn't have though there's much point using lossy raw or a DNG processed file for these purposes.

Therefor, if one refers to pixel values, one has to say in which format they relate to.

Or, one has to refer to pixel values.

This is all very interesting, but what does a simple 'real world' comparison tell us? Haven't someone in the Nikon forum tried to compare for exemple a correctly exposed ISO 3200, D3x image, to a 3 stops under-exposed (and 3 stops pushed) ISO 400 image? Does such a comparison show us that the pushed ISO 400 image is just (or pretty much) as good as the ISO 3200 image?

 

[pixelpepper]

The calculated value was 1198.37, but since I didn't have good levels for white level, so assumed 1024 or black level, so assumed 0

I am lost. I have no idea, what you are referring to.

I don't know the actual ADU range used by the camera, they rarely use the full 4096 (or 16348 in 14 bit). Black is not always at zero, white is not at 4095. I don't know the actual full range. I don't know what the count for your '-2ev' actually is. Therefore, I had to make an assumption. The percentage error won't be large however, but it does make it pointless using a figure to six significant places, so I used one to two significant places.

And it was exactly 5 ADU. There's a co-incidence. How big was you patch of black?

I don't understand the question; I posted I measured it on a black frame . Therefor the black patch is 6080x4044 pixels large. I don't remember how large are I measured, but it is totally irrelevent .

When you look at measured figures, particularly something like a standard deviation, which is calculated from many individual observations, and integral value is pretty suspicious. So, my question is, does '5 ADU' mean '5.00 ADU' or is it a measurement to one significant place? If the latter, that imprecision in measurement will have a substantial effect as the scale of the read noise and shot noise becomes comparable. The reason for asking the size of the measured patch, is the number of pixels will directly affect the precision of your measurement of the standard deviation - it is far from irrelevent.

What was that in ADU? not sure whether you're referring to 0.5% of pixel value or 0.5% of full scale. Neither seems consistent with your other measurements.

1. It refers to the pixel value range ("full scale" in your terminology).

2. Forget about the ADU, it is useless.

But necessary if we are to complete the calculations. You could also use your '%' units, but you do have to know '%' of what.

If the raw data is lossless, then the pixel value range is from 128 to 4095,

Really? That's remarkably accurate adjustment of the capture chain. Generally there is some headroom for sample variation. Perhaps you were measuring the one camera which had the maximum permissible gain and just peaked at 4095.

but if it is lossy, then the range is from 256 to 8191; all measured values double. However, if you convert the lossy raw file with Adobe's DNG converter, it spreads the values up to 16383.

I wouldn't have though there's much point using lossy raw or a DNG processed file for these purposes.

Therefor, if one refers to pixel values, one has to say in which format they relate to.

Or, one has to refer to pixel values.

This is all very interesting, but what does a simple 'real world' comparison tell us? Haven't someone in the Nikon forum tried to compare for exemple a correctly exposed ISO 3200, D3x image, to a 3 stops under-exposed (and 3 stops pushed) ISO 400 image? Does such a comparison show us that the pushed ISO 400 image is just (or pretty much) as good as the ISO 3200 image?

Steen, sweetheart, in the real world people don't bother about the n'th degree of performance. They just buy the camera they like best, set it to the ISO that seems to suit the conditions and snap away happily. Only a few of us get our kicks peering into raw files, speculating about camera design and debating how many electrons can dance on a sensel. And they dare to call us sad...

 

[ejmartin]

This is all very interesting, but what does a simple 'real world' comparison tell us? Haven't someone in the Nikon forum tried to compare for exemple a correctly exposed ISO 3200, D3x image, to a 3 stops under-exposed (and 3 stops pushed) ISO 400 image? Does such a comparison show us that the pushed ISO 400 image is just (or pretty much) as good as the ISO 3200 image?

Steen, sweetheart, in the real world people don't bother about the n'th degree of performance. They just buy the camera they like best, set it to the ISO that seems to suit the conditions and snap away happily. Only a few of us get our kicks peering into raw files, speculating about camera design and debating how many electrons can dance on a sensel. And they dare to call us sad...

Actually, Iliah Borg does both. IIRC he recommends not setting the D3x above ISO 800, underexposing if needed, at ISO 800.

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/

 

[pixelpepper]

This is all very interesting, but what does a simple 'real world' comparison tell us? Haven't someone in the Nikon forum tried to compare for exemple a correctly exposed ISO 3200, D3x image, to a 3 stops under-exposed (and 3 stops pushed) ISO 400 image? Does such a comparison show us that the pushed ISO 400 image is just (or pretty much) as good as the ISO 3200 image?

Steen, sweetheart, in the real world people don't bother about the n'th degree of performance. They just buy the camera they like best, set it to the ISO that seems to suit the conditions and snap away happily. Only a few of us get our kicks peering into raw files, speculating about camera design and debating how many electrons can dance on a sensel. And they dare to call us sad...

Actually, Iliah Borg does both. IIRC he recommends not setting the D3x above ISO 800, underexposing if needed, at ISO 800.

Yes, I have heard a rumour that Iliah actually takes photographs, unlike the rest of us. Sounds unlikely to me, and I've yet to see any substantiating evidence.

 

[John Sheehy]

This is all very interesting, but what does a simple 'real world' comparison tell us? Haven't someone in the Nikon forum tried to compare for exemple a correctly exposed ISO 3200, D3x image, to a 3 stops under-exposed (and 3 stops pushed) ISO 400 image? Does such a comparison show us that the pushed ISO 400 image is just (or pretty much) as good as the ISO 3200 image?

Based on DxO's chart, it seems the noise is proportional to ISO setting above 400. However, they just base their chart on std dev, and do no analysis of noise character (such as how much banding is present; banding makes a very small impact on std dev). However, noise looks consistent in the camera, so I would suspect that under-exposing ISO 400 would be sufficient, and leave a lot more headroom.

I often use my 5D2 at night in the city set to 800, with manual exposure, and just let the actual exposure index float where it may, as shooting at ISO 3200 is only significantly better at vertical banding, but vertical banding is not particularly visible at ISO 800 and above in the camera.

--
John

 

[DSPographer]

Since Canon tends to clip their RAW data low, and Nikon does not, there's probably an additional 1/4 stop benefit to the D3X.

What do you mean by this? The Nikon clips everything below the black point and the Canon doesn't.

 

[John Sheehy]

Since Canon tends to clip their RAW data low, and Nikon does not, there's probably an additional 1/4 stop benefit to the D3X.

What do you mean by this? The Nikon clips everything below the black point and the Canon doesn't.

Yes, that's part of what I was saying, but there is also additional clipping at the highlight end in the recent Canons, except at some ISOs. Most Canons only have about 11000 to 14000 14-bit levels.

--
John

 

[DSPographer]

Thanks for the explanation. When I read clips low I was reading "at the low end" instead of a low highlight clip. I don't think it matters much that Canon doesn't use the full 14 bit count range since they don't achieve anywhere near 14 bit DR anyway. Maybe Canon will really improve the dynamic range in their future cameras so wasting counts would then be significant.

 

[John Sheehy]

Thanks for the explanation. When I read clips low I was reading "at the low end" instead of a low highlight clip. I don't think it matters much that Canon doesn't use the full 14 bit count range since they don't achieve anywhere near 14 bit DR anyway. Maybe Canon will really improve the dynamic range in their future cameras so wasting counts would then be significant.

It is significant now, when considering/calculating DR based on std dev, whch needs a context.

The D3X is the only camera I know of, worthy of 14 linear bits.

--
John

 

[John Sheehy]

For John Sheehy that asked the blackframes also of a A900:

Without knowing where the RAWs would clip at the highlights, it is hard to determine DR. The RAWs are peculiar as they open in IRIS; there are no odd RAW values except a tiny number, probably from interpolating bad pixels from their neighbors. There is also a dearth of "258" pixels (black is 256), implying some kind of expansion beyond the doubling.

In the 100/200/400 progression, the read noise does not quite double. It is about 2.05, 2.68, 5.04 ADU respectively.

The noise has a fine, smooth character with no visible patterns.

--
John

 

[Hendrik123]

simply you are covered if you have the 40D and 5d II..not sure why would you upgrade to the 7D unless you are a collector or addicted to upgrading..

dynamic range well does it really matter that much i mean the 40D makes outstanding photos..and for better dynamic range you got the 5D II

instead of upgrading..i would buy a good lense

Hi everyone,

After reading the review of the 7d I still have a important question which is not yet clear to me.

Currently I have a 40d for backup purposes only and I am thinking of upgrading it to the 7d. I've got a 5d-mark2 too which I like very much.

I know there's a 8 megapixel difference which in practice will not bring me a lot of extra quality. And 4 - 5 fps is more than enough for me since I hardly do sports.

But how about dynamic range? In fact: wasn't the 40d one of the best 1.6 crop camera's on DR? Where does it stand compared to a 50d (from which I heard it was worse than the 40d?) and the 7d?
http://www.dpreview.com/reviews/canoneos7d/page17.asp

I do a lot of landscape photography, so to me, DR could be more important than 18mp's or a better lcd and viewfinder. And if my 40d is in fact miles ahead in DR, I'll probably invest in lenses again next year! ;-)

Does anyone have a thought about this? I'm esspecialy interested in opnions regarding switchers from 40d to 7d or maybe someone who can compare DR of the 7d to the 5d-mark2?

Thanks and greetings from Holland. (in autumn we don't have tulps)

 

[Steen Bay]

This is all very interesting, but what does a simple 'real world' comparison tell us? Haven't someone in the Nikon forum tried to compare for exemple a correctly exposed ISO 3200, D3x image, to a 3 stops under-exposed (and 3 stops pushed) ISO 400 image? Does such a comparison show us that the pushed ISO 400 image is just (or pretty much) as good as the ISO 3200 image?

Steen, sweetheart, in the real world people don't bother about the n'th degree of performance. They just buy the camera they like best, set it to the ISO that seems to suit the conditions and snap away happily. Only a few of us get our kicks peering into raw files, speculating about camera design and debating how many electrons can dance on a sensel. And they dare to call us sad...

Sorry, my mistake! ;-) And btw, maybe "Dammit Wangkin's" real world 7D test wasn't totally hopeless after all :

http://forums.dpreview.com/forums/read.asp?forum=1019&message=33710576

 

[jpr2]

simply you are covered if you have the 40D and 5d II..not sure why would you upgrade to the 7D unless you are a collector or addicted to upgrading..

dynamic range well does it really matter that much i mean the 40D makes outstanding photos..and for better dynamic range you got the 5D II

many reasons, but the most important would be the outstanding AF
speed - neither 40d nor (especially) 5d2 can even approach 7d in this
respect - however, DR is then an important concern;

supposedly, MF with auto ISO should deliver the best possible DR in a
given light, but this means a highest possible ISO as well (giving sort of
automatic "shoot to the right" response), but this gives rise to noise
concerns - that is unless I'm very much mistaken in my reasoning?!

jpr2
--
~
street candids (non-interactive):
http://www.flickriver.com/photos/qmusaget/sets/72157609618638319/
music and dance:
http://www.flickriver.com/photos/qmusaget/sets/72157600341265280/
wildlife, macro, B&W, and 'interactive' street:
http://www.flickriver.com/photos/qmusaget/sets/72157600341377106/
Comments and critique are always welcome!
~

 

[pixelpepper]

This is all very interesting, but what does a simple 'real world' comparison tell us? Haven't someone in the Nikon forum tried to compare for exemple a correctly exposed ISO 3200, D3x image, to a 3 stops under-exposed (and 3 stops pushed) ISO 400 image? Does such a comparison show us that the pushed ISO 400 image is just (or pretty much) as good as the ISO 3200 image?

Steen, sweetheart, in the real world people don't bother about the n'th degree of performance. They just buy the camera they like best, set it to the ISO that seems to suit the conditions and snap away happily. Only a few of us get our kicks peering into raw files, speculating about camera design and debating how many electrons can dance on a sensel. And they dare to call us sad...

Sorry, my mistake! ;-) And btw, maybe "Dammit Wangkin's" real world 7D test wasn't totally hopeless after all :

http://forums.dpreview.com/forums/read.asp?forum=1019&message=33710576

Yup, Dangly Winkle's discovered that, setting for setting, DPP applies less sharpening to 7D images than other Canon's. Probably because, for most purposes, 7D image need less sharpening, because the AA filter's well clear of critical detail, and DPP tends to oversharpen anyway.

 

[GaborSch]

supposedly, MF with auto ISO should deliver the best possible DR in a
given light, but this means a highest possible ISO as well (giving sort of
automatic "shoot to the right" response), but this gives rise to noise
concerns

In all of today's DSLR the dynamic range is determined by the noise. Increasing the ISO reduces the noise (up to a certain ISO level) from the same amount of captured light , but the reduction in amount of captured light due to the higher ISO more than makes up for that.

The effect is, that the dynamic range is getting smaller and smaller with increasing ISO. Take a look at the following graphs, showing the 5D2's noise levels; pick a level of noise on the y axis you are accepting and you can see the dynamic range with the different ISOs on the x axis; note, that the dynamic range is lower with the higher ISOs.



However, increasing the ISO does increase the dynamic range in such settings, where the exposure can not be increased, but the shot is not close to clipping yet. Unfortunately, that's not how Auto ISO works, at least not on the 40D.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

 

[kyklops]

Hi,

I did a small experiment with my 5d2, 50 f1.4 with shutter 1/30, f5.6 three shots, one shot ISO 1600 (-1 EV), one ISO 3200 (0 EV), and the last at ISO 6400 (+1 EV).

Then I only adjusted exposure (iPhoto) +1 for the ISO 1600 and -1 for the ISO 6400 shots. Then when I try to judge noise in the three images the 6400 looks best to my eyes, then followed by 3200, and worst the underexposed ISO 1600 shot.

Is this in line with the graph you presented? If so how should I interpret the graph to reach this conclusion?

If the 6400 shot looks best from a noise perspective how come it has so much lower DR (if I extrapolate your graph for 6400)? Less noise resulting in less DR?

Sorry, but I am getting confused.

k

 

[GaborSch]

Instead of speculation, why not taking a close look at the shots? Can you upload the raw files somewhere? If you don't have web space, pls use yousendit.com.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

 

[GaborSch]

(I received the raw files from kyklops)

This test shows, that ISO 3200 can be a bit better than ISO 1600, although the difference is very small.

This is not the first case I have seen this, though some other tests do not exhibit any advantage of 3200 over 1600. I will try to get more such tests; maybe it depends on the camera copy?

The point in this test is, that three shots were made with the same illumination and same exposure, i.e. capturing the same amount of light, but with ISO 1600, 3200 and 6400. This exposure was metered for ISO 3200; as the exposure was fixed, the ISO 1600 shot was 1 EV "underexposed" and the ISO 6400 shot 1 EV "overexposed".

Let's see a very dark, really noisy spot in all three shots. The following captures show the raw data, not demosaiced, only the intensity adjusted so much, that they appear with the same brightness. The red, green and blue marked numbers in the top line show the average intensity in the selected patch (the selection itself is marked by the orangey rectangle); the numbers in the second line show the noise, expressed as the standard deviation of the pixel values in the selection, in percentage of the average pixel values. This is the reciprocate of SNR; for example 55.7% is about SNR=1.8 and 105% is SNR=1.05.

The intensity is expressed relatively to the clipping level; for example -7.43 means, that the average pixel intensity is 7.43 stops under clipping, i.e. it is in the middle of the 8th stop of the dynamic range.

The intensities are roughly 1 EV apart in the three shots (the difference is max 0.1 EV). I don't know the reason of their not being exactly 1 EV apart; it may be caused by a tiny change in the illumination, or even by changing the posture of the photographer behind the tripod (I made that error already). Such small difference may be caused even by the hardware (shutter inaccuracy, my 20D has been doing that regularly). Anyway, they are close enough for this purpose.

Anyway, the noise with 1600 is a bit less than with 3200, and 6400 is the same as 3200:







The following captures show only the "red" raw channel; here it is even better apparent, that the nosie is practically the same.







The reason for ISO 6400 being practically identical to ISO 3200 is, that the 5D2 does not have ISO 6400 handware support. If 6400 is selected, thewn the shot is made with 3200 and the resulting pixel values get adjusted by the DIGIC.

The following captures show the raw channel histograms "up close". Every column or red, green and blue represents one specific pixel level, starting with 1024 respectively 1026 (see the number in the "Black point" field).

First ISO 1600; nothing special:



Then ISO 3200; it is like expected:



However, ISO 6400 is somewhat strange: there are gaps between the pixel values, showing, that these pixel values are numerical derivatives of the values from the ISO 3200 shot: the processor has doubled the values in order to "increase the intensity" (a totally brainless solution from Canon). The small "dirt" in same places of the gaps show, that the processor has applied some more adjustment, i.e. the raw data is less raw than one would beleive it.



--
Gabor

http://www.panopeeper.com/panorama/pano.htm