DR Modes: The Truth and Nothing but the Truth - long post
There is often confusion among beginners, and even among some of those with more experience with Fuji cameras and photography in general, about what the DR expansion modes do and how they do it.
The most widespread belief and the typical answer for “what do the DR modes do?” is: “they change the exposure”
Even DPR’s Richard Butler writes: The DR100% mode is comparable to the way any other camera treats ISO. But when you switch to DR200%, the camera halves the exposure (which you'll see as a jump in the lowest available ISO setting) but leaves hardware amplification the same, brightening the image later while incorporating that extra stop of highlight information (essentially using a flatter JPEG tone curve). DR400% does this again: another 1EV drop in exposure without a change in amplification, to protect an extra stop of highlight information.
But is that really true?
To find out, let’s start with a very simple experiment:
Set your camera to ISO800, DR100, choose an aperture, and half-press the shutter button for the camera to determine the exposure time. Note the exposure time. Then, leaving everything else the same, switch to DR200 and half-press the shutter button. Note the exposure time. Finally, repeat for DR400.
You will see that the exposure time is always the same, and since aperture is constant as well, exposure is constant.
Invoking DR200 or DR400 does not change the exposure!
Now that that is debunked, what actually happens when the DR modes are engaged?
Let’s go one step back. The auto-exposure system on a camera is designed to balance exposure with the final, desired image brightness through amplification of the signal. Say, the target image brightness is that of middle gray; given a certain exposure, the system will adjust amplification through the ISO setting so that, ultimately, an image with the target brightness is generated. Most commonly, exposure is controlled by aperture and exposure time. Thus, we have that “triangle” of aperture, exposure time and ISO that must be balanced to give the target image brightness.
Note, the term originally coined by Bryan Peterson for this triangle is Photographic Triangle. Unfortunately, someone, somewhere, somehow made a mistake and used the term “exposure triangle”, and even more unfortunately, that term stuck, causing endless confusion and misunderstandings among photographers. But that is a discussion for some other time.
So, let’s do another simple experiment, one that you have probably already done countless times:
Set your camera to ISO200, DR100, choose an aperture, and half-press the shutter button for the camera to determine the exposure time. Note the exposure time. Then, leaving everything else the same, switch to ISO400 and half-press the shutter button. Note the exposure time. Finally, repeat for DR800.
You will see that the exposure times will decrease as the ISO increases, or in other words, the exposure will decrease with increasing ISO. That is the Photographic Triangle at work. When the exposure gets lower, the amount of amplification needs to be increased to maintain the target image brightness. Or the other way around: if one pre-sets increasing amounts of amplification upfront, the exposure will need to decrease to maintain the target image brightness.
So, we know that the ISO setting relates to the amplification of the captured signal. But we are dealing with digital cameras, and it turns out that there are two basic types of amplification: analog amplification and digital amplification. Analog amplification, in a nutshell, amplifies the signal (and noise) electronically before the conversion to digital numbers by the analog-to-digital converter (ADC). These digital numbers can typically be captured in form of “raw data” written to a “raw file”. The second type of amplification, digital amplification, acts on these numbers, simply by multiplication, and it happens anywhere along the image processing pipeline, from applying a constant multiplication factor to all numbers, applying different multiplication factors to the numbers by applying a tone curve, etc, even through adjusting the brightness of a monitor that is used to display the image.
It is important to point out that the ISO setting in the camera specifies the total amount of amplification applied to the captured signal, the sum of analog and digital amplification.
With that out of the way, we can take a closer look at what happened when we changed the DR setting in experiment 1 above. We know that ISO stayed constant throughout the experiment, thus the total amount of amplification stayed constant as well. But when we look at the raw data, we’ll see that the recorded intensities in fact decrease with increasing DR setting. If you want examine the raw data yourself, you’ll need a special tool, such as RawDigger (welcome, Iliah So, from all that we conclude that changing the DR setting changes the amount of analog amplification: DR200 reduces the amount of analog amplification by 1EV, DR400 reduces the amount of analog amplification by 2EV. However, since the total amount of amplification, specified by the ISO setting, hasn’t changed, the amount of digital amplification needs to increased by 1EV (for DR200) and by 2EV, respectively, for DR400. More on how that is done later.
Now we know what the DR modes do:
The DR settings, for a given ISO and a given exposure, change the relative amounts of analog and digital amplification while maintaining the total amount of amplification as specified by the ISO setting.
There is an interesting upshot to all of this: as mentioned already, exposure is typically adjusted through aperture and exposure time, but the third major parameter, scene luminance, is also in play. We can affect scene luminance, for example, with a flash or by mounting ND filters. And we know that the ISO setting refers to both analog and digital amplification. So, we now have five parameters that are under our control and that can be balanced to achieve a desired image brightness. The Photographic Triangle has become a Photographic Pentagon!
So then, where does the widespread notion come from that the DR settings change exposure? That notion is the result of an assumption that isn’t always valid, and it is a result of a misunderstanding of cause and effect.
The assumption is that the reference for what the DR modes do is ISO200. At ISO200, with a certain exposure set, the amount of analog amplification is already minimum. It cannot be lowered further. So, if we want to use a DR mode, we need to first go up in ISO by at least 1EV (for DR200) or at least 2EV (for DR400). As we have seen, the ISO setting specifies the total amount of amplification; raising the ISO increases the amplification, and in order to keep the Photographic Pentagon balanced, we need to reduce exposure. Therefore, it is easy to conclude that changing DR changes the exposure. But it is in fact the change in ISO that changes the exposure. As we’ve seen above, DR has no effect on exposure when ISO does not change.
Why is it important to point out that it is not DR that changes the exposure, but that it is a change in ISO when such a change is required as a consequence of a change in DR? It’s not semantics. Understanding this point helps understanding why the DR modes are so interesting.
The practical purpose of the DR modes is to reduce the chances of clipping.
Again, one step back: As a good approximation, clipping can occur at three stages:
- the sensor receives more light than it can accurately and precisely respond to
- analog amplification exceeds the range of the ADC
- digital amplification exceeds what a given data format can specify
- This type of clipping is sensor clipping: the only way to reduce sensor clipping is by reducing exposure.
- This type of clipping is the result of analog amplification. The sensor capacity may not have been exceeded, but subsequent analog amplification of the readout exceeds the capacity of the ADC: this is where the DR modes come in. For the JPEG shooter, they provide a unique way of addressing this type of clipping while maintaining the overall image brightness. It cannot be achieved in any other way.
- This type of clipping is a result of digital amplification: applying less brightening, using different tone curves, and even decreasing the monitor brightness will address this type of clipping.
Now, let’s look at two typical scenarios for using the DR modes, and we are looking at them from the point of view of the JPEG shooter, i.e., someone who wants to walk away with an SOOC file that encodes an image with proper brightness:
- We sit at ISO200, and we have dialed in the exposure, but we are afraid that important highlights might get clipped. At ISO200, the amount of analog amplification is already minimum, so we know that any clipping here is likely caused by either over-exposure of the sensor or by too much digital amplification. If we assume that the JPEG engine won’t blow out highlights on its own (not a really good assumption), then we can conclude that clipping here is a result from too much light falling onto the sensor. We can simply reduce the exposure (by means of EC compensation) and not get any clipping, without ever invoking the DR modes. Or we can invoke the DR modes, which will then require to raise the ISO, which in turn will reduce the exposure to maintain image brightness. Same effect. The drawback of this approach, though, is that the DR modes work in 1EV increments. Often, much smaller adjustments are sufficient to make sure the sensor isn’t over-exposed, thus small EC adjustments are probably preferable, rather than using the DR modes.
- We sit at a higher ISO, say ISO800, which typically happens when there is rather little light, e.g., in indoor scenarios. The AE system has determined some exposure setting, but we are afraid that important highlights are going to be clipped. Very likely, the sensor isn’t saturated in this scenario, so we know the culprit isn’t sensor clipping. It is likely clipping caused by the analog amplification stage and/or clipping caused by the digital amplification stage. Using the DR modes, we can now reduce the amount of clipping due to analog amplification and then push in post (automatically, if your raw converter of choice recognizes the DR flag) to compensate. That, in my opinion, is a much more appropriate scenario than scenario 1. But mind you, scenario 2 emerges practically automatically from scenario 1 as soon as we change the ISO.
But wait! you might say. If I reduce analog amplification by 1EV and then boost it again afterwards digitally by 1EV, won’t I get clipping again? Yes, you probably would. But here is the rub: the DR modes don’t make up for reduced analog amplification by “mindlessly” adding 1EV to every intensity. Instead, a tone curve is applied that is different from the normal tone curve that is used with DR100. The modified tone curve applies more amplification to the shadow regions than it does to the highlight regions. Overall, the total amount of amplification, on average, is still preserved, according to the ISO setting, but how the amplification is applied to specific intensities is different, compared to DR100. It’s similar to pseudo-HDR.
IMO, the approach that the DR modes represent is very smart. There is no other way for the JPEG shooter to affect something (analog amplification) that they would not normally think of adjusting. And it is the only way to address this point of potential clipping in the image-generation pipeline.
To conclude, what are the DR Expansion Modes called this way?
As we have seen, the DR modes reduce the amount of analog amplification. The amount of analog amplification also determines the DR of the sensor. The DR is largest at the lowest amount of analog amplification (typically that specified by the base-ISO setting) and it then decreases with increasing amounts of analog amplification. At a given ISO setting, reducing the amount of analog amplification through use of the DR modes increases the DR compared to DR100. But keep in mind that the DR can never exceed that of the base-ISO setting.
|Post (hide subjects)||Posted by||When|
|6 months ago||17|
|6 months ago|
|6 months ago|
|6 months ago|
|6 months ago|
|6 months ago|
|6 months ago|
|6 months ago||1|
|6 months ago|
|6 months ago||2|
|6 months ago|
|6 months ago||3|
|2 months ago||1|
|1 month ago||1|
|1 month ago|
|1 month ago|
|1 month ago|
|Christine by JP Zanotti|
from Car wreck
|Fangorn Forest by cand1d|
|Yosemite Falls with Moonbow by Jonathan Shapiro|
from Best Landscape of the Week 4