Choosing an Exposure Setting: Dynamic Range vs. ISO Speed

I understand sensor dynamic range reduces with increase in DSLR ISO speed.

I am interested in knowing more on this subject when shooting in RAW format, and how noise and dynamic range can be weighed against each other. ( under short exposure times )

For example, I read an article claiming it better to set a low ISO speed and to underexpose then recover shadows in raw post processing, as an alternative to increasing ISO speed to obtain a normal exposure level and faster shutter speed. The former minimising noise and max'ing DR, whilst the later causing noise and limiting dynamic range.

Can this work? Are there any drawbacks? Is there any rule of thumb like drop -0.5 EV for each doubling of ISO?

There's quite a technical back and forth about this topic within the thread entitled Is DR higher, the higher the ISO at a given aperture/shutter?.

In short, you should use a higher ISO if you need it (instead of using EC afterwards). Using exposure compensation recovers shadows using digital processing *after* the photo has been captured at the sensor, whereas using a higher in-camera ISO recovers shadows using analog processing *before* the photo has been recorded as the final file*. The latter method is virtually always better.

As you point out, DR does generally decreases for higher ISO speeds, although this is usually negligible between about ISO 100 and 400.

Thanks for unambiguous information. I also read the other (linked) thread with full understanding.

I totally agree proper exposure with low ISO is the correct way to go, when shooting circumstances permit. I would just like to extend the discussion to the higher ISO range.
I use D300 and from ISO 200 to 800/1000 noise & DR is not an issue for me. However in the ISO 3200 to 6400 range (for which I don't have a lot of experience) the captured results can be quite poor. I presume the noise & DR performance of the D300 sensor is therefore logarithmically deteriorating. So much so that NeatImage is unable to clear the noise components from the RAW converted to 16 bit TIF.

Just to close this question, would you also offer the same advice at the higher ISO band 6400 to 3200, from the point of view of getting a clean image with post noise reduction.

I always hesitate to give blanket advice because there are usually exceptions. However, even for very high ISO speeds, I still stand by the assertion that you will achieve better quality images by using a higher ISO, as opposed to achieving this sensitivity through exposure compensation in RAW post-processing.

Keeping in mind also (assuming a 12 bit A/D converter) that of the 4096 levels, a full 2048 of them are devoted to resolving the brightest stop of light. Of the remaining 2048 levels, 1024 of these are used resolving the next brightest stop of light.

So ...

... If you under expose by 2 stops with the intention of pulling it up in post-processing then (a) you've lost 2 stops of dynamic range, (b) you've of the 4096 levels available to you, you've thrown away a full three quarters of them , and (c) attempting to increase in post processing with raise the noise floor and probably give you posterisation in your shadows.

With more modern 14 bit converters you've got a bit more leeway, but not as much as you'd think (one of those theory -v- practice things) - hence the saying "with Digital you expose for the highlights, and post-process for the shadows".

Hope this helps.

Cheers,

Colin - pbase.com/cjsouthern

It puzzled me to read in the first post that there was advice to underexpose and 'recover' the shadows in PP, which seemed contrary to everything I had heard.

Glad to be reassured by Colin; I'd agree with the concept of 'expose to the right'.

Personally, I think the "expose to the right" rule is a good starting point - but it really depends on how contrasty the scene is - and also whether or not you can afford to have certain areas of a scene blown - which is why I like to use the hightlight warning feature in conjunction with the histogram.

Case in point - if you shoot a bride and groom in a proverbial "57 chevvy" with chrome bumpers on a bright day you'll almost certainly get specular reflections from the chrome - that'll spike the histogram at it's right-hand end - and if you expose for that then the B&G will endup literally being "shadows of their former selves" (pun intended!) - so if you take a shot - find you have "the blinkies" - it often pays to see just where they are to decide if they're important before you adjust the exposure. That's the joy of incidence light meters - they centre the range of reflectances from white and black right on the 18% grey point - and if you have specular reflections (or strong back lighting - eg windows) then they blow, but the most important bits still expose correctly.

On the other hand if you're in a high-contrast situation (eg sunset with foreground detail) then you'll literally want to push the histogram as far as you can without too much clipping to have any hope of recovering any shadow detail (assuming not using GND filters).

However - although "expose to the right" is the mantra, if you have a relatively flat scene then usually if you expose to the right you end up with an "over-exposed looking" scene that in theory you can pull back by dramaticaly raising the black clipping point, in practice I find that it skews the gamma to a point where it's difficult to get the tones looking properly balanced - so i usually expose more towards the centre to avoid "making a rod for my own back".

The "dilema" is that there are only about 4 stops difference between light reflected off a white object -v- a black one - but most cameras can handle (with the likes of Canon's Highlght Tone Priority) about +4 stops and -3 Stops (without digging too deep into the shadows) - and a medium (18% grey) tends to be quite a bit lower than what is ideal in terms of what's ideal as far as shooting to the right goes. When shooting sunrises etc I often cheat by spot metering the brightest highlights with my hand-held meter (the meter assumes I'm metering a medium grey tone), then I shift the reading around 2 1/2 to 3 stops down (ie longer exposure) so the "mid-tones" that the meter thinks I'm spot metering get moved up into the highlight areas where they should be, but without blowing. (Note" although in theory you can shift them as far as 4 stops if highlight tone priority is on, in practice you need to leave a bit of leeway because during a multi-minute exposure the sun will rise further and of course become brighter).

In the absence of a hand-held light meter a "quick and dirty trick" for Canon cameras is to put the camera in manual mode with spot metering, then point at tones of interest to see on the meter where they're going to expose. You can literally take all the guess work out of exposing the shot. It's actually a bit of a show-off party trick - you can meter the shot - then tell anyone handy that eg "the fence will be white - the road will be a neutral grey - the shadow detail in the tyres will just be visable, but the clouds will blow" - take the shot - and hand over the camera without even looking at it :) - and you'll be "spot on" (another pun!). Not sure if you can do it the same way with Nikon cameras - I would have assumed so, but when I had a go on one I couldn't figure out what the metering was trying to tell me. That's one of the joys of the 1Ds3 - you can spot meter up to 8 different parts of a scene and display them all at once on the exposure meter :)

Cheers,

Colin - pbase.com/cjsouthern

Quote:

---

Originally Posted by Colin Southern

a "quick and dirty trick" for Canon cameras is to put the camera in manual mode with spot metering, then point at tones of interest to see on the meter where they're going to expose. You can literally take all the guess work out of exposing the shot. It's actually a bit of a show-off party trick -..... Not sure if you can do it the same way with Nikon cameras - I would have assumed so, but when I had a go on one I couldn't figure out what the metering was trying to tell me.

Cheers,

Colin - photo.net/photos/colinsouthern

---

Oh, I thought this was the normal way of shooting!! :) On Nikon D80 I always have it metering (& focussing) at the centre position of the grid of 12....except when I have tried to adjust something else using the 4 way toggle in the wrong mode and it has fooled me by moving it to the edge position, but it soon gives itself away by refusing to focus somewhere there is nothing to focus on:rolleyes:.

Chris

PS on the subject of the thread, I have found it possible to get RAW histo overshooting at both ends whether ISO 200 (normal default UK inland) or ISO 1000 (highest I go unless desperate). That is what RAW is for, giving you a choice of exposure in conditions where there is a full light range?? More usually though, with hideous grey sky, the histo is a pathetic little stalagmite somewhere centre-left.

Quote:

---

Oh, I thought this was the normal way of shooting!! :) On Nikon D80 I always have it metering (& focussing) at the centre position of the grid of 12....except when I have tried to adjust something else using the 4 way toggle in the wrong mode and it has fooled me by moving it to the edge position, but it soon gives itself away by refusing to focus somewhere there is nothing to focus on:rolleyes:.

---

Hi Chris,

I must admit to being a Canon man - but a few months ago I was giving a little one-on-one help to a colleague who had a Nikon D300. On Canon cameras we have a dot on a horizontal or vertical scale to indicate what the currently spot-metered point is going to expose as (eg shadow / highlight / mid-tone / blown etc) - but when he put the D300 in manual metering mode we ended up with what I could best describe as a "horizontal bar graph"; that appeared to eminate from the centre of the scale (mid-tone I assume).

I assume that the peak of the graph is probably equivalent to the position of the dot in a Canon system, but I really didn't understand why they made it a graph, which implies to me a range of values (that I couldn't relate to what I would have thought was a single value since we were spot metering).

Additionally, on Canon 1D series cameras you can spot meter up to 8 values and have them display as seperate dots on the scale so you can see how everything relates to everything else, and shift the whole shooting match up or down to suit - I would have assumed that Nikon would do something similar on the likes of their D3 / D3x type cameras, but don't know how they could do that with a bar graph.

Hopefully some Nikon shooters can shed some light on this for me?

Cheers,

Colin - pbase.com/cjsouthern

Quote:

---

Originally Posted by Colin Southern

..........Hopefully some Nikon shooters can shed some light on this for me?

Cheers,

Cp;om - photo.net/photos/colinsouthern

---

Hi Colin,

I am new here and was directed to your post by Chris.

I have a Nikon D2X, so not the latest, but in 'spot metering' mode the meter follows the A/F point and follows your description of the D300.
Bar graph with mid point indicating correct exposure and then + or - either side of that. I do not believe it can read multiple values as your Canon but as I mainly use manual exposure with my old manual focus AIs lenses I have not found that to be a problem. I still do a lot of calculations in my head and use the D2X like a digital version of the old F2 which I used with hand held Gossen Profisix as a light meter.

Photography has been a hobby since the 1960's but I am still relativly new ( 3 years ) to the world of digital imaging, so still lots to learn. ;)

Don

Hi Don,

Welcome :)

Thanks for your reply - makes sense. It was quite a few months ago when we did the little comparison (and I've had a few lunches since then, so I've probably forgotten things!). I figured that's probably what it was trying to tell me, but I remember at the time that something just didn't seem to add up (I think it was the actual EV variations for a number of objects in the office with our cameras set to same shutterspeed / aperture / ISO). I'll have to have another crack at it next time he comes around.

Is the bar graph type indicator standard issue on most/all Nikon gear?

Cheers,

Colin - pbase.com/cjsouthern

Hi guys, you seem to have just about figured it out ok for the Nikon, so there is only a bit more to add.

Spot metering:
Camera will meter a 4mm diameter circle centred on the current focus point, (so can utilise an off centre point, a bit like I believe you suggested with the Canon)

Centre Weighted:
Camera meters the whole frame, but assigns greatest weight in the centre of the frame (12mm dia circle in the centre of viewfinder)

3D Colour Matrix:
Camera meters a wide area of the whole frame.

These are the case with full frame, dimensions will vary with DX Nikon cameras, (3mm and 8mm respectively) and the above metering options will vary also with some of the older non CPU lenses.