Dynamic Range of M8

[roblumba]

I heard people comment on the dynamic range of the M8, but is it really proven? I have experience looking at many of my 1D MarkII images and it seems to me that the M8 images are sometimes easily overexposed and blown out in the highlights. Perhaps this is also a function of the Leica glass.

This is really a plus for picking up subtle reflections and exagerating the effects of light on a darker scene and picking up light in the shadows, however, in daytime shots this can be hard to manage.

 

[blakley]

Have you looked at the histograms? I've not blown anything out with the M8; I'm metering manually most of the time - in fact mostly I'm just using manual settings I know to be correct and not even looking at the meter. If you're on A or using the internal meter, maybe you're overexposing.

 

[J. Borger]

Where most internal meters for digital cameras tend to underexpose the internal meter of the M8 tends to overexpose because it is heavy center-weighted .. i almost routinuously dial -1/3 or -1/2 EV in.
That said .. DR seems larger than anything i ever used .. i find it very difficult to blow highlights or block shadows with it.
Some modern Leica lenses are chalenging with the R-D1 in high contrast light .... the M8 handles the same contrast much easier ...

 

[Nick R.]

Are the highlights really blown? If you back off the exposure in whatever raw program you're using, does highlight detail start to appear. Not questioning your competence, just suggesting a method.

I heard people comment on the dynamic range of the M8, but is it really proven? I have experience looking at many of my 1D MarkII images and it seems to me that the M8 images are sometimes easily overexposed and blown out in the highlights. Perhaps this is also a function of the Leica glass.

This is really a plus for picking up subtle reflections and exagerating the effects of light on a darker scene and picking up light in the shadows, however, in daytime shots this can be hard to manage.

 

[MP Guy]

My experience with the M8 and DMR has shown me that weather you over or under expose, the detail is always there and PS brings it out.

 

[roblumba]

I didn't blow it

I didn't blow it

I'm just commenting on my observations of some of the photos I've seen in the gallery and some other comments on Leica glass. I haven't actually tried editing any photos from an M8 because I don't have one. From my experience you may be able to recover some detail in the highlights, but one or more color channels may have lost information. So the recovered details are sometimes off in color and posterized. This is especially noticable in portraits when the person's forehead is overexposed. Some amount of detail might still be there, but it's damaged and doesn't look natural anymore.

 

[MarkM6]

Question

Question

.. DR seems larger than anything i ever used ..

Please do tell what is that "anything" you have ever used? Thanks.
~M

 

[J. Borger]

Please do tell what is that "anything" you have ever used? Thanks.
~M

Canon D30, 1D, 1Ds and R-D1!

 

[jaapv]

I have a feeling dynamic range may be even larger with an IR cut filter, as an unnecessary part of the light is blocked.

 

[gogopix]

I have a feeling dynamic range may be even larger with an IR cut filter, as an unnecessary part of the light is blocked.

YES! Interesting. Those bits are not wasted on red IR but on visible.Another argument for the filters

 

[willie_901]

I have a feeling dynamic range may be even larger with an IR cut filter, as an unnecessary part of the light is blocked.

This certainly makes sense. No bits would be wasted on photons in the IR frequency range.

 

[kevin m]

I'd bet the DR of the M8 is still less than an M6 loaded with NPH.

How's that for a sentence full of jargon? 😀

 

[jaapv]

You think so? I would venture that the M8 covers about 11 stops, that is without resorting to post-processing tricks, as for instance most Canon sensors produce between 9 and 10. How much is the dynamic range of an ideal print in the zone system? 10 stops, of which 8 are actually used....😛

 

[MarkM6]

I want to believe!

I want to believe!

.. DR seems larger than anything i ever used ..

Please do tell what is that "anything" you have ever used? Thanks.
~M

Canon D30, 1D, 1Ds and R-D1!

I would care less about all this "problems" with the current M8 if someone can prove this claim by software such as Imatest that measures the Dynamic Range of the M8's sensor. Perhaps Sean will do it soon.

 

[x-ray]

You think so? I would venture that the M8 covers about 11 stops, that is without resorting to post-processing tricks, as for instance most Canon sensors produce between 9 and 10. How much is the dynamic range of an ideal print in the zone system? 10 stops, of which 8 are actually used....😛

Something you should know about raw processing and dynamic range. Every raw converter that I have ever used and i've used most since getting into digital seven years ago apply a contrast curve when converting. I shoot canon 1DsII cameras and process raw in photoshop CS2. I also use canon DPP from time to time and every once in a while Phase one pro. If you look carefully under the curves tab in the raw converter there will be a linear mode. The same is true for canon and phase one. Select this linear mode and you will see what the sensor is actually capturing. If you're comparing in the default mode you will see the hidtogram and image as the contrast curves are applied. This is defined by the software manufacturer and does not represent the true output of the camera. The curves will be different for each model of camera.

The second thing you should know is about the sensitometry of film emulsions. In 1975 I studied the zone system as a student of Ansel Adams. I spent weeks with him in the field and in the darkroom. I learned a great deal that i use to this day. Look at film and you will see it is not a linear capture like digital capture is. Yes digital in the linear mode is a real linear capture. If you look at the sensitometry curve of film it is a true curve with a sholder and toe and a somewhat linear mid tone. On the shadow side the film can receive about two full stops of exposure above base density before it starts to register any serious detail. Thsi is zone 2. Now in the midtones it captures much like digital in a somewhat linear fashion. This is zones 3 to 7 or sometimes 8. Now at the upper highlight side the film starts to roll off and the next two stops no longer capture detail like the midtones. Linear digital captures highlight detail the same as midtones. This is zone 8 to 10. Ok, now you see that there are six zones or stops that actually capture information in a linear mode and the first and last 2 zones mainly increase or decrease in density with little detail captured. Digital in linear mode captures the same detail from zone 1 to zone 10. In reality every pro digital camera in linear mode captures about 10 stops or zones. Some digital backs for MF capture 14 zones.

Now what is this about? Look at the histogram on the back of the camera. If the histogram does not extend from the extreme left side (shadows) to the extreme right side (highlights) you are not using the entire dynamic range. Most subjects in nature do not use the entire dynamic range. If you do not shoot raw and process in linear you are not getting everything the sensor can produce. This again means that the subject must have a full tonality that equals the maximum capture potential of the sensor.

If you're not working in raw linear and 16 bit processing you may be throwing much of your detail and color information. Process a image in 8 bit and then edit with levels and curves in photoshop. Look at the histogram and you'll see serations and voids in the histogram compared to the original. This is lost data and the more you have the more you approach a posterized image. Now do the same by processing to 16 bit from the raw. Do the same curves and levels and now look at the histogram. It looks like the original. I ALWAYS do all raw processing from my 1DsII in 16 bit raw and whenever needed in linear with my own set of curves. If you don't do this you are throwing away important information. JPG files are the least quality of all.You immediately toss a great deal of information in the trash. You lose so much that I personally think you're wasting you money on a quality camera.

When I finish editing I conver at that point to 8 bit and release the images to my clients. At this point I have all the information that the 8 bit can carry and the histogram looks like the original in camera with no voids of detail and color information.

As to the canon dynamic range vs film. I feel from daily experience that it'sslightly better than transparency film such as RDP. The main advantage of digital is the ability to capture the dull range of detail unlike film which does not.

 

[roblumba]

Typical Curve

Typical Curve

The typical curve actually increases midtone contrast and decreases shadow and highlight contrast from what the sensor produces. And somehow that looks most natural to us. That's is an interesting thought. I wonder if that default curve is really optimized for the behavior of each sensor. I usually tweak the shadow, exposure, brightness, contrast on my own until the histogram looks evenly spread without highlight or shadow clipping. Of course, sometimes clipping is appropriate. But I wonder if each camera could have a better optimized curve profile that can be used as the default so that we make the most of the dynamic range and the specific characteristics of a sensor.

 

[pfogle]

Hi X-ray, thanks for your analysis.

Two minor points (questions, really, I'm not an expert)

first, you talk of 16 bit RAW, but the Canon and other 35mm format cameras use 12 bit RAW, so the 16 bit tif is already using some interpolation of the levels. I assume this interpolation is linear if you're in linear conversion mode.

second, the linear response of the chip doesn't really give equal detail to the highlight and shadows, as this depends on the distribution of bits - as per the 'shooting to the right end of the histogram' concept, the highlight end of the range contains the most levels (half in the top 1 stop), while the number of levels available at the shadow end is a lot less. In practice, however, it seems that 12 bit RAW provides enough levels (4096 in each channel) for this not to be a major limitation. As always, in both film and digital, the shadow detail is limited by noise.

 

[vicmortelmans]

Look at film and you will see it is not a linear capture like digital capture is. Yes digital in the linear mode is a real linear capture.

I don't want to say you're wrong here, but there's definitely confusion! Analog is linear in log-space, while digital is linear in linear space!

Let's define some words and do some math:

- intensity (light) is a value linearly related to the amount of light (photons, if you want). E.g. if you open a lens 1 stop, the amount of light hitting the film is doubled, so the intensity is times two. Note: definition is not ok, since I also take an increase of exposure time in account, so really look at this as 'amount of light'

- density (film) is the logarithm (base 10) of the intensity of back-light divided by the intensity of transmitted ligth

- raw pixel value (digital) is the pixel value as you find it in a (16 bit?) raw output from your digital camera

Now, let's have a look at the film behaviour. The characteristic curve is described as such:
- horizontal axis = logarithm of intensity (amount of light hitting the film)
- vertical axis = density
- behaviour = as x-ray describes; linear in the middle, and a toe and shoulder at the extreme low/high intensity area's; the slope is [\], meaning you have a negative image

Then let's look at digital film behaviour. The workings of a detector are indeed linear, and could be put in a graph as such:
- horizontal axis = intensity (no logarithm!!)
- vertical axis = pixel value
- behaviour: linear through the whole range, meaning that at some point (highest pixel value reached), the pixel is 'burnt' by even higher exposures; the slope is [/], so you have a positive image

So already here we can state that both systems behave linearly, at least in some part of the 'dynamic range', but the variables that behave linearly are quite different, so actually, there's nothing to compare!

This difference points out the main handicap of the digital system, if you know that the eye's sensitivity is also log-based (we perceive an duplication of light intensity equally along the intensity range, while duplication a low intensity adds much less 'amount of light' than duplication of a high intensity). The anolog (film) system matches this nicely, but if you look at the digital system, the difference between pixel values 60000 and 60010 in your image is invisible (a fraction of a stop in photographic terms), while the difference between pixel values 10 and 20 is much more visible (1 stop)... So on the one end of the scale, values are wasted to store invisible intensity differences, while on the other end there are not enough values available to store even the visible intensity differences.

There's much more to say about this, but it's an ever recurring problem that people are talking about math and forget to clearly identify the variables and their proper definitions, leading to wishful or wrong conclusions.

Groeten,
Vic

 

[jaapv]

I don't want to say you're wrong here, but there's definitely confusion! Analog is linear in log-space, while digital is linear in linear space!

Let's define some words and do some math:

- intensity (light) is a value linearly related to the amount of light (photons, if you want). E.g. if you open a lens 1 stop, the amount of light hitting the film is doubled, so the intensity is times two. Note: definition is not ok, since I also take an increase of exposure time in account, so really look at this as 'amount of light'

- density (film) is the logarithm (base 10) of the intensity of back-light divided by the intensity of transmitted ligth

- raw pixel value (digital) is the pixel value as you find it in a (16 bit?) raw output from your digital camera

Now, let's have a look at the film behaviour. The characteristic curve is described as such:
- horizontal axis = logarithm of intensity (amount of light hitting the film)
- vertical axis = density
- behaviour = as x-ray describes; linear in the middle, and a toe and shoulder at the extreme low/high intensity area's; the slope is [\], meaning you have a negative image

Then let's look at digital film behaviour. The workings of a detector are indeed linear, and could be put in a graph as such:
- horizontal axis = intensity (no logarithm!!)
- vertical axis = pixel value
- behaviour: linear through the whole range, meaning that at some point (highest pixel value reached), the pixel is 'burnt' by even higher exposures; the slope is [/], so you have a positive image

So already here we can state that both systems behave linearly, at least in some part of the 'dynamic range', but the variables that behave linearly are quite different, so actually, there's nothing to compare!

This difference points out the main handicap of the digital system, if you know that the eye's sensitivity is also log-based (we perceive an duplication of light intensity equally along the intensity range, while duplication a low intensity adds much less 'amount of light' than duplication of a high intensity). The anolog (film) system matches this nicely, but if you look at the digital system, the difference between pixel values 60000 and 60010 in your image is invisible (a fraction of a stop in photographic terms), while the difference between pixel values 10 and 20 is much more visible (1 stop)... So on the one end of the scale, values are wasted to store invisible intensity differences, while on the other end there are not enough values available to store even the visible intensity differences.

There's much more to say about this, but it's an ever recurring problem that people are talking about math and forget to clearly identify the variables and their proper definitions, leading to wishful or wrong conclusions.

Groeten,
Vic

Which is a very interesting discussion (I mean it -not being sarcastic) but in practice the dynamic range of even the older Canon 10D exceeds that of Kodachrome. Not that that is the best film in that respect. I am sure TriX performs better than most sensors, but then it performs better than most films as well.

 

[x-ray]

My current experience is with canons so this is what I reference too and E-6 film since I've primarily shot transparencies over my professional career.

The short answer to dynamic range is that digital may not have quite the dynamic range of color neg but it equals transparency film and normally processed B&W. B&W can offer the greatest range if processed in pyro or other compensating developer. The advantage of digital is that it has more usable detail in the shadows and highlights compared to film. Digital allows you as the shooter, provided you use raw format, to design your own look. With film you are bound to the look that the engineers and chemists have designed into the film. With digital you can alter each channel any way you want and create your own curves. This is the big advantage of digital. You are in controll of your image with digital.