Are split neutral density filters needed with modern cameras?

[Ligament]

Lets assume one has access to a modern, higher end camera. I'm thinking D800e for example.

With the dynamic range sensors can capture now, are you still using split NDs vs. one capture or bracketed exposures?

I have a mountain / valley shoot coming up and am deciding whether I really need split NDs any longer.

[Hans Kruse]

I never use grad ND filters. The number of times I need to make and HDR file are not many. Even with the 5D mkIII! I have also a D800 which extends the possible shots without HDR. You can take a look at my pictures on http://www.hanskrusephotography.com/ and I think there is one photo in total taken with a grad ND filter. Use grad filters in Lightroom and the tone mapping controls and you are done. A 32 bit HDR file created by either Photoshop or the Photomatix 32 bit plugin for Lightroom can be edited (tonemapped) in Lightroom as a normal RAW file.

[francois]

Lets assume one has access to a modern, higher end camera. I'm thinking D800e for example.

With the dynamic range sensors can capture now, are you still using split NDs vs. one capture or bracketed exposures?

I have a mountain / valley shoot coming up and am deciding whether I really need split NDs any longer.

If you own NDs, then take them with you. If you don't then no need to spend money on a good set of filters. With a camera like the D800/D800e, you can certainly avoid using NDs. I wouldn't bother, personally.

[Paulo Bizarro]

Maybe they are not strictly needed (I use the Canon 6D), but I still use them. I like to get things right on the field as much as possible, rather than on the computer.

[PeterAit]

I agree with Hans. With today's sensors, software like Lightroom, and the availability of HDR, split filters are a relic of the past. Aside from being unneeded, they are limited in that the gradation between regions is always, at least in my experience, a straight line, which does not always suit the subject.

[Lonnie Utah]

There is one problem with the line of thinking that Graduated ND filters are no longer "needed" given the large dynamic range of modern sensors. While that statement accurate in some regards (Yes, you can process images with large variations for light to dark), it neglects to account for an inherent problem in pulling up the exposure in the darker and "shadow" areas of the image. When we start to increase the brightness in these areas we are also introducing noise into those areas of the image. It's simply physics and unavoidable. Now that noise might be well controlled, and it might be able to be reduced by NR software, but when we apply the NR we do it at the cost of sacrificing image details.  If we expose for those shadow areas, we risk blowing out the highlight (which, IMHO, is a bigger post processing challenge than dealing with shadows and noise).

While it might take a little more work, the best solution for the "best looking" (yes that's a subjective term), is using ND filters in landscapes and situations where there are big differences between light and dark.

Now there are some counter points to be made to these thesis, and I'm sure you'll all make them, but this is where I'll stand for now.

[Slobodan Blagojevic]

If one has the time to fiddle with NDs, one has the time to bracket just as well.

[Lonnie Utah]

If one has the time to fiddle with NDs, one has the time to bracket just as well.

Very true, but some of us prefer not to do exposure blending (Not that there's anything wrong with it, just as a matter of personal preference)

[stamper]

Spot metering off the brightest area in a sky - other than blown out areas - and raising EV by 2 stops lightens the dark areas which means that there is little need to worry about noise. My split filters are permanently in a drawer.

[hjulenissen]

There is one problem with the line of thinking that Graduated ND filters are no longer "needed" given the large dynamic range of modern sensors. While that statement accurate in some regards (Yes, you can process images with large variations for light to dark), it neglects to account for an inherent problem in pulling up the exposure in the darker and "shadow" areas of the image.

Either the DR of a camera is sufficient to capture the shadows with sufficient quality for a given scene, or it is not. If it is, then there is no visible problem to pull the shadows. The argument is kind of circular.

When we start to increase the brightness in these areas we are also introducing noise into those areas of the image. It's simply physics and unavoidable.

This is not physics. Increasing brightness does not _introduce_ noise. It will typically make the inherent noise of your capture more visible, though.

Now that noise might be well controlled, and it might be able to be reduced by NR software, but when we apply the NR we do it at the cost of sacrificing image details.  If we expose for those shadow areas, we risk blowing out the highlight (which, IMHO, is a bigger post processing challenge than dealing with shadows and noise).

If you have scenes that challenge the DR of your camera, it is going to be a trade-off, and perhaps single, unfiltered shots wont cut it.

While it might take a little more work, the best solution for the "best looking" (yes that's a subjective term), is using ND filters in landscapes and situations where there are big differences between light and dark.

Inserting glass in front of your lens does introduce image quality challenges of its own. My biggest gripe is that graduated ND filters have to be chosen from a small set of filters, positioned in the field. I prefer being able to flexibly chose my gradation in post.

-h

[hjulenissen]

If one has the time to fiddle with NDs, one has the time to bracket just as well.

GNDs have the benefit that the finished image will be captured at a single period of time. For some kinds of scene movement, this might be important?

-h

[Lonnie Utah]

Increasing brightness does not _introduce_ noise. It will typically make the inherent noise of your capture more visible, though.

Yes it does. As you increase the brightness of the image, your image processing software has to start interpolating color and luminance values for the pixels in the dark areas of the frame. As you increase the brightness or exposure value, the software cannot and will not interpret all of those pixels accurately. The ones that it calculates incorrectly translate as added noise in the image. The lower the signal to noise ratio in that part of the image, the greater the effect will be.

One can test this by intentionally underexposing an image by several steps and then increasing the brightness....

[sdwilsonsct]

GNDs have the benefit that the finished image will be captured at a single period of time. For some kinds of scene movement, this might be important?

-h

Very! It's almost always windy when there is anything in the frame that can be blown around.

[shadowblade]

They're only occasionally needed (as in, absolutely needed, not just 'nice to have'), but, when you need them, you really need them!

[Lonnie Utah]

I will amend my original statement to this,

Are they needed, No. ("Need" is a very big word, IMHO).

But one needs to understand the tradeoff's and limitations of each and all techniques used to increase and express the dynamic range in an image.

[Slobodan Blagojevic]

GNDs have the benefit that the finished image will be captured at a single period of time. For some kinds of scene movement, this might be important?

-h

Perhaps. But that type of situation becomes rarer and rarer, as today's cameras are capable of firing three (bracketed) shots within 0.5 seconds or faster.

As Lonnie mentioned, it is always a tradeoff (what in life isn't?). With GND, I used to fight (in film days) color casts, unnatural darkening in parts of the image that do not follow the straight line of the transition, too strong or too weak filter strengths (or carry a full assortment of strengths and transitions - hard vs. soft - which, when good quality, would cost more that a lens), ghosting when combining two, etc. etc.

[hjulenissen]

Yes it does. As you increase the brightness of the image, your image processing software has to start interpolating color and luminance values for the pixels in the dark areas of the frame. As you increase the brightness or exposure value, the software cannot and will not interpret all of those pixels accurately. The ones that it calculates incorrectly translate as added noise in the image. The lower the signal to noise ratio in that part of the image, the greater the effect will be.

I am sorry but this is wrong. You seem to mix up the terms "multiplication" and "interpolation"?

One can test this by intentionally underexposing an image by several steps and then increasing the brightness....

Yes I have, and the outcome of such a test is perfectly consistent with my claim that the noise exists in the raw file before touching it with the raw developer.

-k

[hjulenissen]

Perhaps.
...

I think we are in agreement. I was just offering suggestions why GND might still have some value, for some people, some of the time.

-h

[Lonnie Utah]

I am sorry but this is wrong. You seem to mix up the terms "multiplication" and "interpolation"?  Yes I have, and the outcome of such a test is perfectly consistent with my claim that the noise exists in the raw file before touching it with the raw developer.

The algorithms used by our software programs to increase the brightness in the dark areas of the frame are much more complex than simple "multiplication". 

[stamper]

The algorithms used by our software programs to increase the brightness in the dark areas of the frame are much more complex than simple "multiplication".  

What do you mean by " increase the brightness in the dark areas " ? At the risk of being accused of being pedantic there isn't any brightness in dark areas. That is why they are called dark areas. You mean moving dark areas towards mid tone values? I don't think interpolation takes place in a raw converter because there aren't any pixel to interpolate???