[Leica] Re: the dynamic range of digital

[davison_m at msn.com (MARK DAVISON)]

I have experience with three Nikon DSLRs:  the D100, D2h, and D200.

On each camera I have performed George's suggested experiment of shooting a 
textured card to find out when the camera will blow the highlights.

The results are nearly uniform, with some minor variation depending on 
whether you shoot .jpeg or .nef (Nikon's RAW format).

The sensor clips hard at around 2 stops "overexposure", or more precisely, 2 
stops exposure greater than the metered value.

There is usable data down to about 5 stops "underexposure", or more 
precisely, 5 stops exposure less than the metered value. There is visible 
data with more stops underexposure, but it is filled with ugly noise.  (This 
result depends to a great degree on what ISO you have set, and to what 
extent you are willing to peform noise reduction in software outside the 
camera.)

Therefore the meter is calibrated to give a response range which is not 
symmetric around the metered exposure.

If you are shooting in flat light, where there is no more than 4 stops of 
scene brightness range (+ and - 2 stops from the metered value) then you do 
not need exposure compensation.

In contrasty light, I typically find I have at least 6 stops of scene 
brightness range (+ and - 3 stops from the metered value) and need to 
introduce an exposure compensation of -1 stop (when using the Nikon matrix 
metering.)

When I dial in -1 exposure compensation when shooting in contrasty light, I 
am not "underexposing"--I am just biasing my exposure to correct for the 
meter's calibration, in order to place the histogram just to the right. 
Another way to think about it is that I am fitting the scene brightness 
range into the recording range of the camera's sensor.

When shooting static subjects (like landscapes) you can take the time to 
spot meter the brightest highlight where you want to record detail, and then 
expose about 1 and 2/3 stops more.  In zone system parlance you are 
"placing" the highlight at Zone 6 and 2/3. I then look at the histogram(s) 
and make further small adjustments.

In shooting sports on bright contrasty days, I sometimes take test shots and 
adjust my exposure compensation to get the R,G,B histogram right.  This 
allows me to tailor the exposure compensation to prevailing conditions.

One weakness of the D100 and D2h is the lack of separate R, G, and B 
histograms--these cameras appear to simply use the G channel to represent 
luminosity.  This can lead to blowing the R channel when shooting flowers, 
sunsets, and red soccer jerseys and you get strange color shifts.

When I get time I'll check the sensor response of my R-D1 and report on 
that.

I hope my Nikon experience helps others.

Has anyone done these experiments with any of the Canon DSLRs?  Do they 
center the recordable brightness range around the metered value, or do they 
clip at the same 2 stops over the metered value?  How about the Olympus 4/3 
cameras?

Please note that the poor camera engineers have a hard choice to make in 
setting the meter calibration.  Setting the meter to read 2 stops below 
highlight saturation makes full use of the sensor range when shooting in 
flat light (giving the most accurate tonal detail), but it ensures blown 
highlights when shooting in contrasty light.

Also note that careful metering with an incident meter does not change the 
problem.  If there are objects in the scene which spot meter more than 2 
stops over the incident meter reading, then they will become blown 
highlights if the you set the camera to the exposure from the incident 
meter.

Mark Davison