The 5D2 has strong pattern noise at ISO 100

pwm2 wrote in post #8630115
The amount of detail you can get from the dark parts is a question of dynamic range. Each extra stop of dynamic range means one more stop of contrast i the dark region. That is why we want digital cameras to reach the dynamic range of film.

Yes, that's true as far as it goes, but with digital there's an additional factor that you simply don't have with film: how much tonal resolution there is.

Digital is a discrete format. The reason ETTR is useful is not simply to make maximum use of the available dynamic range (that's part of it), but because the tonal resolution increases as you move towards the lighter part of the tonal range. That is, the tonal resolution is not constant throughout the tonal range.

With digital, any given stop contains half the number of tonal values as the stop above it as you move towards the bright end of the range. So if the brightest stop consists of 4096 tonal values, the stop below it will contain 2048, the one below that will contain 1024, etc.

A digital camera with 14 stops of dynamic range and 14 bits per color channel of tonal resolution will have 8192 different values per color channel in its brightest stop -- and 1 value per color channel in its dimmest stop. If that same digital camera had only 9 stops of dynamic range with the same 14 bits per color channel then the brightest stop would still have the same number of discrete color values, but the lowest would now have 32 values.


And all of this is because the sensor is a linear recording device, and by talking about how it performs in stops, we are shoehorning it into a logarithmic model.

No it won't. See above.

It might be able to capture some of the same floor values in the tonal range (that is, successfully "see" the same light intensities), but that doesn't mean it'll be able to capture those values with the same tonal resolution.

Film is an analog medium. There are an infinite number of tonal values being recorded at every stop*. This is why film has a reputation for smooth tonal gradations.

* Bounded by quantum limits, of course. But I don't think the film we're typically talking about is capable of being pushed to the degree that you'd see the difference between one photon and two -- that difference would surely get lost in the noise.

Yes, and that comes at the expense of capturing detail in the shadows. Note here that I'm talking about tonal details, not spatial details.

But it's very different because of how it does it. As you approach the dark end of the dynamic range of film (more precisely, as you continue to push the film and examine the darkest areas), it gets noisier: the average difference in tonal values from one miniscule area to the other increases. But the gradations remain continuous, if noisy. In digital, the gradations are not continuous.

In both cases, what would otherwise be smooth tonal gradations (in the case of film) or very visible discrete tonal jumps (in the case of digital) are hidden by noise, but the fundamental characteristics are nevertheless very different between the two.

You could argue that, at the end of the day, when there's that much noise to contend with, there really isn't much difference between the two, and you'd be basically correct about that. In both cases, pushing the medium too hard will get you poor results.

In a way, you're saying the same thing here as I am. What I'm explaining above is why you capture more detail when exposing to the right with digital.

This fundamental difference (logarithmic versus linear, continuous versus discrete) between film and digital is precisely why you're better off exposing to the right with digital, while the same thing is not (from what I've seen and understand) true of negative film. In film, my (rather limited) experience is that you're best off nailing your exposure (or if you want to record the most dynamic range, exposing for the shadows).

Bottom line: you expose film (for shadows or highlights, depending on whether the film is slide or negative) to the left or to the right to maximize the range of light that can be recorded, but that is all. You expose digital to the right to maximize the range of light that can be recorded and to increase the tonal resolution of the shadows. Even if you didn't care about additional dynamic range, you would still want to expose digital to the right to get better tonal resolution in the shadows. With film, if the dynamic range is sufficient for you then you needn't change your exposure at all.

See the difference?

Think it through. The tonal range of the camera is merely the range of values reported by the camera's sensor. The dynamic range is the range of incoming light intensities that the sensor can "see" and which is mapped by the camera to the values stored in the image file (RAW files are a 1:1 mapping between values reported by the sensor and values stored in the image file).

The lower the dynamic range of the camera, the smaller the difference in original intensity between the brightest light recorded and the dimmest light recorded. Decrease that and you by definition increase the contrast.

This is readily apparent by looking, with your own eyes, at a scene and examining a shot taken of the same scene. The camera records as black areas of the scene which your eyes see as a dim shade, and it records as white areas of the scene which your eyes see as a brighter shade, at the same time. That is contrast multiplication.

The more dynamic range you have in the camera, the greater the difference in light intensities you can record. That says nothing about how well you can record them. You can have a camera with a huge dynamic range and a very low color resolution, and you can have a camera with a small dynamic range and a very high color resolution.

JPEG defines the color resolution of the typical storage medium. Its limited color resolution is why so many (such as myself) prefer to work with RAW. You don't lose dynamic range of the light that was recorded when you're working with JPEG versus raw, because the light does remain recorded, but you do lose tonal resolution.

CyberPet wrote in post #8632969
So to conclude:

This is how the 5D Mk II should be like:

1. Have 99 focus points, all implanted in your brain, so it'll focus on what you want, just in case you forgot to change the focus point.

2. A 99 stop range with absolutely no noise even at ISO 24,000... just in case, and the files in ISO 100 should be so clean you can cut your retinas if you were to pixel peep.

3. The ability to shoot in Infrared by a switch of a button.

4. Being able to do X-rays, in case your buddy falls and breaks a bone and you're no where near a phone.

5. Wait, strike that last part... it should have a built-in phone (3G turbo of course) so you can upload your images while on the shoot.

6. Video in any frame-rate you want and with a built in steady-cam feature with a push of a button.

7. Total weather sealing, so you can dive with it... or if you're just clumsy and fall off a dock while backing up to shoot a duck.

8. Battery life that last about 100,000,000 frames so you can go onto a world trip with just one battery.

9. Super compressed RAW files that are smaller than a full res jpeg, and yet you can fit 99x more images on a 8 GB card than you can now.

10. Finally.... it'll only cost $99 and will only be sold to those who knows how to use the M mode.

Yes? No? Maybe?


No I'm not drunk... just having fun

Wow.

I dunno what kind of stripped down crap you're pandering; you left off WiFi, GPS, and CDMA (for the USA).

really detailed post!

cdifoto wrote in post #8633043
Wow.

I dunno what kind of stripped down crap you're pandering; you left off WiFi, GPS, and CDMA (for the USA).

I can't believe you left out the cup holders.

D'oh! How could I leave out a basic necessity?!?!?

Oh and they gotta be able to hold a Big Gulp.

Lowner wrote in post #8630519
I've come to realise that digital is not the magic bullet people expect it to be. With film we expect to see grain and with the current digital sensors we get pattern noise in deep shadows.

yeah, except the fact everyone loves to forget is the patterns are a canon issue on "low end" bodies. The 1 series doesn't have it, nikon doesn't have it, sony doesn't have it. etc etc etc.

Sure, you can make some of those have it under very extreme situations, but the rebels, x0D and 5 series all have it with minimal work to show it.

CyberPet wrote in post #8630590
And it's a known fact that even at ISO 100, if there's underexposed shadow areas you'll get noise and patterns. Nothing new at all.

with canon's current sensors in bodies that don't start with one, sure.

yogestee wrote in post #8630764
Pixel peeping is a disease that must be stamped out,,it is counter productive and opens up a whole new can of worms..

pixel peeping is only when you're looking at 100% crops. Look at that church shot once it's pushed enough to get some details in the shadows. You telling me you wouldn't see that in a print?

I'd bet it'd be easily seen in a 4x6 by someone with zero clue what they're looking for...

c00lpix wrote in post #8632213
It's obvious there are two camps about this:

A) It doesn't matter, the exposure was wrong
B) Even if the exposure is wrong the camera shouldn't do this by design + other cameras in this class do not do this (D700, D3).

what about my shots? My exposures aren't wrong? Neither is the shot of the STi someone else posted (the gray car)??? Yes the church is a dramatic example, but that's not the only example floating around here people!

c00lpix wrote in post #8632560
I'm curious what the result would be if the OP used ISO 160 instead of 100:

http://www.flickr.com …4/sets/72157617​903991680/

yeah, that is very interesting. I wonder if I'd been using 160 all along if I would have ever even seen my issues? It sucks because I learned back in the 10D day that everything but whole stopes were "faked" so they were noisier. For that reason I've always stuck to even/whole ISOs (100/200/400/800/1600/​3200 and 6400). I even go so far as to turn off the others as options in the menu..... I'll have to test 160 against the D700's 200 and see if it helps/fixes my shadows issue....

kcbrown wrote in post #8632979
Yes, that's true as far as it goes, but with digital there's an additional factor that you simply don't have with film: how much tonal resolution there is.

Different technologies often have different terminology. But that doesn't mean that you often can compare technologies and find similar phenomena.

Film may be analog, and our Canon sensors digital. But you get into the same problem if you run out of bits to capture the tonals or if your signal drops below the noise level. Canon could of course have made the sensor logarithmic. This would not have changed the issue - just added IQ problems from the logarithmic multiplier not being perfectly logarithmic - just as the film isn't perfectly logarithmic.

Correct. And what I did say in my post.

Correct. And what I did say in my post.

Still just more ways of saying the same thing.

Exactly. Which is why it wastes space on the memory card for an excessive amount of levels in the bright parts of the photo.

Here comes the important part.

First off is that you shouldn't stress the "infinite" part too much about analog film. "Infinite" would only have been applicable if film was noise-free. As it is, you have to limit your view of "infinite" at a suitable point below the noise level. Too much below the noise level, and you will no longer see a detail modulated by noise.

A linear sensor will have more tonal values in the brigtest stops. Way more than we can see. Film will have a constant tonal range in each stop, since it is a semi-logarithmic media. But that is only relevant for the brightest stops. When you get to the dark sections, both a linear and a logarithmic sensor will fail, even if for different reasons. The ADC of the digital sensor will have noise and it will have a limited number of available bits. So you get few and discrete steps. And the selected steps may be wrong because of noise. But the film will also get into troubles. It does not have a limited number of steps in the same way, but instead each grain in the film will get the value based on a very large dice. If you do capture a dark and evenly lit surface with a small intensity gradient with a digital sensor and with film, both will show the gradient with a semi-infinite tonal range until you zoom in. The physical function may be different, but both alternatives will contain noise averaged around the expected value. Switching to a patterned surface on the other hand, the spatial resolution will make a difference. When a detail is small, the noise will no longer be able to average around the expected value. In the end, you will either lose the detail, or you will get the wrong tonal value.

And the important thing: That happens with both film and with a digital sensor.

In the end, it is possible to create a sensor with a logarithmic or linear capture. But which one you select doesn't matter for the tonal range, as long as the linear sensor has enough bits compared to the dynamic range. If we keep just the top 8 stops, then 14 bits isn't a problem. With a 10-stop image aligned to the right, you still have enough bits to capture details with enough tonality. If you clamp away the high stops on the other hand, then 14 bits will not be enough. That is why we must do ETTR so that the dark stops are dark on our prints too, making the discrete tonal steps too small to see even if the percentage difference between the steps is quite high.

So is - hopefully - everyone in this thread.

How something works isn't always important. It is normally irrelevant what tonality range the film has, when the tonality is completely masked by noise. How does the piccolo flute sound, if masked by a jackhammer that is many dB louder than the flute? Does it then matter if the flute player is good or bad, or if the instrument is tuned?

Don't think there is any contention about that.

Same thing - not contended.

Yes, same as I did, but in different wordings.

But once more - linear or logarithmic doesn't really matter as long as the linear alternative has enough bits. When a linear sensor has enough bits, the only difference is that it will fill the memory card faster, by capturing too much tonality for the bright parts of the picture. A logarithmic sensor is evenly good/bad all through the range. A linear sensor is excessively good at the high end. There are logarithmic AD converters available, but they are normally way worse than linear. Just as your logarithmic film has tonality problems thruogh a number of stops.

Not really true. RAW is almost an 1:1 mapping of what was digitized from the sensor. How much almost will depend on manufacturer and model. The thing you should think through, is just that a logarithmic film is constantly changing scale, while a digital sensor with linear digitalisation has one big, continuous tonality range all through. For scientific work, this allows the linear sensor to capture much more details in light regions. For our eyes, it doesn't matter poop as long as you have enough bits for the dark regions.

Contrast isn't normally a term to use for a sensor. Contrast is normally something to use when talking about the final print. And the contrast will then depend on how you used curves to compress the dynamic range of the sensor into a narrower dynamic range of a print or a monitor.

No. That is contrast clipping. But besides this clipping, that happens in the sensor, there is a compression of the dynamic range when the raw data gets converted to a jpeg image. The main focus of this thread is to what degree you should be allowed to perform that dynamic compression before the noise gets visible. In the case of the shadwed church, the dynamic compression was too large, so it allowed us to see the patterned noise. The OP started this thread because he considers that the dynamic compression should not have been too large, and that Canon doesn't deliver as much as they say that they deliver.

Now you moved into yet another off-topic area.

The jpeg file is both limited in the number of discrete steps it can record. But it also uses separate compression levels for luminance and chrominanse, since our eyes have better resolution for intensity than color. But this is yet another side track.

cdifoto wrote in post #8633043
Wow.

I dunno what kind of stripped down crap you're pandering; you left off WiFi, GPS, and CDMA (for the USA).

Well, at least WiFi and GPS are supported, even if not included.

Ok..

Let's imagine the year is 2050 and the new Canon camera is Canon i50D MkII.

The DR is true 16 bit range, that is, 2^16 = 65536 tones if my memory is correct.

And by saying, true dynamic range, I mean, the quality of darkest pixels is the same as the quality of brightest one.

What could photgrapher with such a machine do which is totally impossible today?

Yes, he could easily invert the photo, let the brightest bits be the darkest and via versa. All the same, the quality of the image, would be 100% the same, 100% inverted.

Today we have a race between makers and they are cheating. The true DR isn't close to 10 and above, not at 100 ISO, and lesser on higher ISO.

Still, very strange to see all this defensive posts here. Why? Why aren't you guys who hate such a discussions, read ones, respond and act like a troll?

cdifoto wrote in post #8633043
Wow.

I dunno what kind of stripped down crap you're pandering; you left off WiFi, GPS, and CDMA (for the USA).

Sorry mate!

jacobsen1 wrote in post #8633234
with canon's current sensors in bodies that don't start with one, sure.

Sorry mate.... you must be lost, the Nikon forum is that way ---->

Those who argue pattern noise and limited DR isn't a problem probably rarely shoot landscapes.

There's a reason almost every second or third landscape I shoot is HDR. And it's not just mid-afternoon sun, either - it's the same thing with sunrises and sunsets, when the horizon/treeline/citys​cape is too jagged to use a GND filter, or even when shooting a building or monument at dawn or dusk, when one side is brightly lit while the other side (or any alcoves or concavities in the building) are in darkness..

Probably none of this is a problem if you mostly shoot people or events, or if you can control the lighting with flash and modifiers (i.e. not when shooting a mountain, or shooting a cloudscape from a plane).

But, in any case, is increased DR ever a bad thing?

Shadowblade wrote in post #8633658
But, in any case, is increased DR ever a bad thing?

Yes, it is! It will require us to fork up a lot of money since increased DR will arrive first on the most expensive models. But a lot of people would be very willing to do that.

pwm2 wrote in post #8633722
Yes, it is! It will require us to fork up a lot of money since increased DR will arrive first on the most expensive models. But a lot of people would be very willing to do that.

Canon rarely debuts any feature on a top-of-the-line model, apart from the ones which cost a lot more to manufacture (e.g. full-frame vs crop). Pattern noise is largely a firmware fix.

When it does, you can bet it will trickle down to a lower price point within a few months.

In any case, the release of a new model in no way decreases the performance of an older one...

Ben,

jacobsen1 wrote in post #8633234
what about my shots? My exposures aren't wrong? Neither is the shot of the STi someone else posted (the gray car)??? Yes the church is a dramatic example, but that's not the only example floating around here people!

Actually what I was trying to say (perhaps poorly) is that some people would like to sweep the issue under the rug claiming the exposure was wrong therefore any after effects (PP) are rightfully deserved. Or that one should have shot "to the right". I actually don't agree with that sentiment, I think even if the exposure was wrong (and I'm not saying it was!) what we see in front of us shouldn't be so bad.

We know the 5D II's sensor was derived from the 1Ds-3, but (and I'm going on a limb by myself here) it seems like the addition of video probably required some sensor design changes. That or the microlens "improvements" weren't actually improvements. People with the 1Ds3 don't seem to have the same results.

jacobsen1 wrote in post #8633234
yeah, that is very interesting. I wonder if I'd been using 160 all along if I would have ever even seen my issues? It sucks because I learned back in the 10D day that everything but whole stopes were "faked" so they were noisier. For that reason I've always stuck to even/whole ISOs (100/200/400/800/1600/​3200 and 6400). I even go so far as to turn off the others as options in the menu..... I'll have to test 160 against the D700's 200 and see if it helps/fixes my shadows issue....

I guess if someone is willing to do the experiment, all of us would benefit =).

-c00lpix