Dynamic Range and Bit Depth - 20D / 1Ds Mark II

[jonathan_bray]

Dynamic range and bit depth are critical in the specification of film

scanners. I am about to make the change from film to digital, but am

struggling to get my mind around dynamic range and bit depth in

relation to digital cameras.<b<

 

My understanding of the terms is that dynamic range indicates the

brightness range that a scanner or camera can capture without

clipping at either end, and a higher bit depth enables smaller

increments to be recorded within this range.<p>

 

Although specifications and reviews of good film scanners almost

always quote the dynamic range and bit depth, they are rarely

mentioned in camera specifications and reviews.

 

I want to capture images with a high resolution (to produce sharp

images up to 20? x 16?), have colors faithfully rendered, and have

sufficient data to enable images to be manipulated in image

processing software.

 

The Canon 20D seems to be the first digital camera at a reasonable

price that can achieve results comparable to (or even better than)

conventional film. But before I commit to it I?d like to be sure

that the 20D is likely to satisfy my objectives, and that I should

not be taking the plunge and instead buying a 1Ds Mark II.

 

NB. I do not need most of the features that distinguish the 1Ds Mark

II from the 20D; the higher pixel count would be nice, but the main

reason for my opting for the 1Ds Mark II would be if the quality of

the information captured in each pixel were markedly better.

 

So, I would welcome:

 

1. Any general comments on whether I am right about the importance

of dynamic range and bit depth in relation to digital cameras (and,

if so, views as to why they are not widely quoted).

 

2. Any specific comments on the likely differences in the image

quality between the 1Ds Mark II and the 20D (putting aside the fact

that the intrinsic resolution would be higher).

[coconutdaydream]

i think you're worrying too much, or not educated well enough about digital. many people will try to refute this...but i use a d60 and i can print GREAT 20x30's, without a problem.

[mike_p2]

Jonathan,

 

I noticed that you mentioned film. Unfortunately, the dynamic range of the modern sensors is yet too far from the dynamic range of the film (even slide film). And it clearly shows despite all the numbers people may give you.

[phil vaughan - yorkshire u]

Mike, remember Jonathan has asked how the CMOS sensor compares with a quality scanner, NOT the film.

 

Jonathan

 

There the answer is much more complicated, no-one does quote figures for dynamic range, however it's generally reported that the 20d and 1d mkII sensors are just behind slide film, therefore they're just in front of all but the most expensive scanners.

 

There should be enough information in a 20d image to produce wonderful 20x16's I have got only a few large prints from digital, and they beat anything I ever had printed large from 35mm film. Further this can be achieved at much higher ISO's than I ever dared use film at.

 

You pay's your money... the upshot is, if you give specific examples of where you fear digital may not be good enough, then you will get honest answers, but if you ask questions about 'how good is it', then you'll just get lots of rhetoric, as above (including mine).

[jim_larson1]

<i> The Canon 20D seems to be the first digital camera at a reasonable price that can achieve results comparable to (or even better than) conventional film</i>

<p>

Just curious: What is the basis of this statement? And in what way does the 10D or Nikon D70 fall short?

<p>

I have seen a bunch of vendor (Canon) claims regarding dynamic range of the 20D and 1D-II, but I have not really seen *any* of these claims backed up quantitatively by an unbiased (unpaid) source. Heck, I am not sure I have seen any SUBJECTIVE claims . . .

<p>

Now, on the topics of "image noise" and "resolution", information abounds. But the question here is "range" and "color rendition". => unfortunately most reviewers (and most readers) don't really know what these terms mean.

[brad_w]

Hi Jonathon, I'll try to answer your 1st qu as directly as possible:

 

About Dynamic Range: In my crude implementation of the Zone System, I counted on

four stops of useful range from Velvia and Provia 100F (high-contrast chromes). My

20D histogram covers five stops and in practice this feels about the same as Velvia

for most images. I could dig pretty far into the shadows of scans with my little

ScanDual III, but in truth when I felt I needed to do that I was dealing with a poorly

exposed frame and usually got a so-so print. Even though the shadow information is

recorded by the film, I don't feel is particularly useful - properly exposing the film

always gave better results.

 

There is a big difference in highlight rendition between dit and film: the non-linear

response of the emulsion "blends" into over-exposure more nicely that the after-the-

fact profiling of dit. My old 10D clipped highlights pretty hard and that seemed to

draw my eye toward the big flat white spots on a print. I really had to protect the

highlights during exposure. The 20D seems much better -I'm not sure if I've just

gotten used to working around it or if the profiles really are better.

 

About bit depth: TIFF's and JPG's are 8 bit CRW is 12. Everything goes to 8 bits for

printing, even LightJet prints. From where I sit, then, the extra bits are only in there

for "wiggle-room" if you mess-up the WB or exposure. Like I said, dit exposure's a

no-brainer. In my experience, 8 bits still gives enough latitiude to correct the color

vagaries of mercury vapor lamps. Also, I tend to "tighten-up" the levels just for

printing anyway (white value <255, black value >1), so I've still got a few bits to work

with at either end.

 

In case you haven't guessed yet, I shoot jpg's and use the WB presets (chosen

manually). Minor levels or White point adjustment and a little USM is about all I need

or want to do in PS.

 

-brad

[PuppyDigs]

True, digital sensors are a long way from the dynamic range of negative film and even less

forgiving of slight overexposure than chromes. Nevertheless, you adjust and find

workarounds and all is well. Although I still love shooting film, I find the near grainless

appearance of digital a big plus. I can't stand to use film faster than ISO 200 after shooting

digital for 18 months (10D). I find large prints of landscapes more detailed when taken

with slow film, but my 10D sports a lowly 6.3MP sensor. Large portrait prints fare much

better and I prefer the smoother texture of digital over film.

 

To extend the dynamic range of digital I use double exposures (one for highlights &

another for shadows) and blend the layers in PS. For grab shots I blend RAW images

processed with different amounts of digital exposure compensation.

[doug_landrum]

Hi Jonathan, I acquired a 20D about a week ago.

 

I shot and intend to continue to shoot film that I scan and print. I use Leica M and R cameras and lenses, a Nikon Coolscan IV and Fuji Provia 100. My resulting image files from film are tweaked in Photoshop CS - almost exclusively sharpened with unsharp mask after sizing to print size with Genuine Fractals 3.0. Prints to 12X18 on an Epson 2200 using Epson papers are sharp, saturated, contrasty and better than Cibachrome prints made in the past by professional printers and myself.

 

So far as I have progressed on the learning curve about digital capture, I believe the 20D will perform as well or better than film but differently. My best prints so far have been limited to a print area of 9-3/4 X 6 7/8 - admittedly a strange size.

 

Here is my procedure: Images are captured in RAW. RAW files from the 20D are called into Photoshop CS using the Adobe RAW plug in v. 2.4 beta. Files are tweaked in the RAW file window to adjust color temperature and a PPI of 340 is selected. (A PPI of 584 will yield an exact 4 X 6 sized print image). Myriad adjustments are available in the RAW file window. Then I complete importing into Photoshop, sharpen with unsharp mask and print.

 

Files from the 20D will take a lot more unsharp mask without showing artifacts than files from the Coolscan. From this I am hopeful that bigger prints from the 20D will be smooth and sharp.

 

Colors are pretty punchy and saturated.

 

I think the lens selection will make a difference. I opted for the 24-70 f/2.8L. It's a real monster in size for me but it's not heavy. I am used to Leica R primes and the weight of an R7 with an R Drive and 100mm APO-Macro-Elmarit-R is heavier.

 

In sum, the 20D should meet exacting standards of a professional or picky amateur like me.

 

Here is an image from the 20D captured yesterday using the Program mode - it is worked over quite a bit in Photoshop with Shaddow Highlight and Transform to straighten the buildings, then sharpened.

[awindsor]

</b>

[roger krueger]

Digital still sucks for scenes with an extreme range of brightness--things that'd get N-

development in the Zone system. I've got a 1ds II, even at the top the problem's there. The

good news is, unlike a scanner, it's usually not a problem.

 

I do still shoot film sometimes when I need extended dynamic range--at night, and for

situations where I need to capture detail in both deep shade and bright sunlight (And, of

course, situations where a big, noisy SLR is just the wrong tool).

 

But as long as you haven't blown your highlights (or missed your exposure), there is no

other downside to the limited dynamic range.

 

The difference between a single-digit Canon (1d, 1ds, etc.) and a two-digit Canon (10d,

20d) is all about build quality, metering options, and AF quality. If you want to shoot a fast

lens wide open, or need predictive AF of moving subjects that actually works, a single-

digit body is what you want. But for dynamic range there's little (if any) difference. The

sensor technology is substantially the same, just different size.

[jeff medkeff anchorage, a]

<p>Let's just see if we can </b> close that bold tag....

[shootme]

I Hope it helps a little, I have used both the 20d and the 1ds mark II and the later gives better results, you can make bigger prints. However they are no match to film and a good drum scan.

[jonathan_bray]

Many thanks for your responses - they've been extremely hepful. It was particularly good to get some clearly unbiased views from both 1Ds users and 20D users. I've also taken on board various points such as the need to avoid blowing out the highlights, and that 20D images will take more sharpening than Coolscan scans. Whilst the build quality and predictive AF of the 1Ds would be nice, since image quality (other than resolution) is not markedly different, I could not justify the extra cost for my own needs at this stage, and so I think that I'll be opting for the 20D.

[sean_mclennan]

It's important to note the difference between film and CMOS sensors.

 

Light is the main element required for image capture on either device, but film requires a fair bit of light and the longitude of a set ISO film isn't that long. Bottom line though, more light is preferred over less light. Darkroom magic can bring back a badly over exposed image whereas not enough light means the information simply isn't there to be recovered.

 

Digital sensors are the opposite. They do caprture information with very little light...so if you underexpose images, it is fairly easy to pull the image back and save it. However, in order to provide greater longitude than film (for a single ISO rated film compared to dslr's complete ISO coverage) the sensors have mechanical/physical limits to light saturation. Once reached, that's it, you've gone beyond the level the sensor can take and that pixel gets clipped. Once clipped, you cannot save it with ANY amount of pixel pushing or post-op in Shop.

 

These digital camera's are fantastic, but let's try to keep in mind their limitations. The benefit of being able to switch ISO with a button rather than changing films is a huge plus, but the camera still has limitations. With film, you needed to learn/try/test hundreds of films at various speeds to find the right film for *your* photography and while you can push/pull and play a little with film, you had to pretty much find ONE film that was right. With digital, the camera's has to cover ALL our various preconceptions and experiences with certain films, which is no small task. A specialized product will always perform better under it's target conditions than a general purpose one. Convenience of digital, compared to the specialization of film. Compared head-to-head in specific applications, film still wins, no doubt. However, the flexibility of digital is such a huge change for photography, it promotes much more freedom of exploration. (in my opinion)

 

If you shoot alot in very bright conditions, than you really should stock some ND filters. With these, you would be hard pressed to still have blow outs. you can shoot for your highlights and recover your shadows post-op. To me, this is still much better flexibility than film ever offered.

 

sean

[ron c sunshine coast,qld,a]

As you can see from most of the comments Jonathan,bit depth is not really a problem with digital but dynamic range could be improved for sure.

<BR>This really does tend to be a mixed up opinionated argument because of the false info and missunderstanding on the subject.Being something that is so tricky to measure doesn't help at all!

<BR>One often sees tests done online by someone but there is the problem of what to measure,how much noise can you put up with in the shaddows and how much tallent one has with photoslop

 

<BR>Overall canon's digic (D60,300D,10D ) has about 5&1/2 stops of dynamic range.Anything outside of that really depends on too many if's and buts.

<br>I did a thorough test once and got maybe 6&1/2 stops from my 300D

<P>I think canon claimed about one stop improvement with digic II (20D 1DmkII,etc).Some online reviews showed some improvement ,some didn't.If i remember correctly michael richeman at luminouslandscape showed a good half stop of noticable/improved DR.

<br>In any case it was light years away from the quote i read from chuck westfall of canon usa.He said the 1D mkII could give 8&12 stops!There's no way in hell that's gonna happen.They must be counting one bright pixel in a hundred or something....

<P>Fuji have a very interesting duel pixel sensor for increasing DR on the S3 Pro.Sadly it doesn't seem to be MUCH better than the far cheaper 20D

<P>There are various techniques that can be used to increase dynamic range but they mostly depend on having a static subject one can take mulitple shots of

 

<BR>With fast moving subjects such as sport you are stuck with RAW capture and 'duel chanel mix editing'.That's good for a maximum of about 7 stops total dynamic range

[brian ball]

Dynamic range is a genuine concept, but there is no standard for measuring it. Scanner manufacturers make up their own standards, camera makers have shown more reluctance. Nonetheless, one may make distinctions between film and digital and amongst digital devices. I've been wading in the digital waters with a Canon S45 and am eager to get a DSLR. For whatever it's worth, I tested the dynamic range of the S45 by taking shots of an IT8 profile. I encountered a difference of 12 stops from the most overexposed image in which I could distinguish grayscale bars 11 and 12 to the most underexposed image in which I could make such a distinction. These were shot in RAW mode, and I used Photoshop Camera RAW for exposure compensation, albeit much more extreme compensation than I would ever use normally, and of course these extremes look hideous. So what does one make of this? Well, with film, exposure levels beneath a certain threshold become eliminated in processing whereas with digital, any small amount of light will get recorded. Of course, there is a major caveat: sensors

also produce noise, and with less exposure S/N ration is poor and, at some point, renders dark shadows as essentially devoid of meaningful detail. On the other side, film handles overexposure rather gracefully, while digital just hits a wall. Shooting RAW can help recover overexposed areas in RAW capture, but the highlights may appear undersaturated as certain color channels might not contain any meaningful information. A higher bit depth can render a greater number of gradations and can keep counting beyond 256 levels. Nearly all DSLRs have 12/bit A/D converters. Scanners often have more, but dynamic range and bit depth for scanning have slightly different implications. Slide film irrespective of how well it captures highlights or shadows produces a transparency with a Dmax of around 4.

A hihgher actual Dmax than the thickest chrome will not improve the scan. Negative film has a lower density and requires less scanner dynamic range even though negative film can record more levels of highlights and shadows.

 

What does any of this mean in practical terms? Think of it this way:

whether the intention is for output to paper print, transparency print, or video, an output device can only disply a certain amount of

contrast, and this amount is more often than not lower than the range of contrast one sees through the camera lens. If an outputted picture contains a great detail of shadow and highlight detail, midtone contrast will be sacrificed, and the opposite is true as well. Thus while one can be greatful for the abilty to capture highlights and shadows without any particular ugliness, actually displaying a high dynamic range does in fact have its costs. If punchy midtones are desireable, one can look at the dynamic range limitations of cameras and films as useful guidelines.

[ron c sunshine coast,qld,a]

Brian you test is interesting.

<BR>It shows how hard it is to measure such a thing though- how many stops difference was there between the dark bars and the light bars of the test target.This difference would have to be deducted from each end of your result

 

<P>In my test i took pics of an allmost plain wall.It had small imperfections that i used as something to recognise in very under or over exposed images but i still wonder how much those marks (being of higher contrast than the surrounding areas) contributed to an optomistic result

<P>In his S3 pro review,michael reichemann equalised the white points in photoshop then judged the images based on how much detail showed in shaddows.A good practical way to do it but i still would be interested if there are any exact test method

[brian ball]

I'm guessing that the difference in the bars I was measuring would amount to not more than 1/4 or 1/3 f-stops, these being the 12th and 13th bars in a 24 level white to black grayscale, so it would be more or less insignificant. I figured that I would use this method so that I would only attach genuine significance to a result in which relatively subtle differences were evident. If I were to test using a series of thick, solid, black and white lines, I could probably achieve a greater range of possible exposures. Perhaps one could say that an underexposed image (in this case, a 4 megapixel, 8 bit image) becomes a lower resolution, lower bit image. I was doing some more test shots last night, and I was amazed at how well I could recover detail from terribly underexposed images using Photoshop Camera Raw. The images were indeed quite ugly, but I could make out detail, and they could have some use in certain circumstances if truly necessary.

I suppose one could see this as the digital equivalent of extreme push-processing.