LensRentals.com

Say Whhaaat?

Posted by

Camera and lens manufacturers employ teams of professionals to write their copy, and others to edit it dozens of times. At least you would think they did. But what they actually publish is often a combination of market-speak, outright lies, and sometimes just oddly puzzling (and often funny) gibberish. Here's a few claims that I've run across in the last few days, all taken from the manufacturer's websites. (They'll edit them when they find out people are making fun of them so they might not be there long, though.)

Tamron

Tamron 10-24: "Wider and longer than others in its class."

Let me think. . . that’s the class that includes the Sigma 8-16, Nikon 10-24, Canon 10-22, etc. So exactly how is it wider and longer? Oh, wait. I get it. The lens is 3.9 inches long, which is longer than anything else in this class. Still not sure about the wider part, though.

Tamron SP 70-300 f/4.5-5.6 VC: "provides the ultimate in image quality and handling in the telephoto zoom class."

You said, that. Really? So it’s got better image quality than the Sigma 50-500, Canon 100-400 IS L or 70-300 IS L, Sigma 120-300, or Nikon 200-400? If you’re gonna say stuff like that, you better stop publishing your MTF charts - they make you look, at best, foolish.

Tamron 18-270: "The lens provides a 1:3.5 macro"

What, exactly, is a 1:3.5 macro? I've seen macro, and 1:3.5 magnification isn't  it.

Tokina

Tokina 12-24: "With the built-in motor, the lens can be used in AF mode with the Nikon D60 and D40 and other silent wave bodies."

I actually know what they are trying to say, but I sure can't find the term “silent wave body” anywhere. Sounds like some type of interpretive dance.

Sigma

Sigma 18-250 f/3.5-6.3 DC OS: "at the long end it can be used for sports actions."

Note to marketing department: never use “f/6.3” and “sports action” all in one sentence. That’s just really wrong.

Sigma 18-250 f/3.5-6.3 DC OS: Perfect for all shooting situations, at the wide end the Sigma’s 18-250mm is perfect for scenic, landscape or group photography; at the long end it can be used for sports actions or great shot from a day on safari. Of course, you can do great candid portraits with it as well.

Brilliant! If you’re gonna tell lies, might as well tell lots of lies.

Sigma 20 f/1.8: "provides great creative freedom to the photographer who wishes to exploit its capabilities for juxtaposing very near subject with the surrounding background."

Juxtapose vb: to place close together or side by side

OK, so am I supposed to put the “near subject” beside the background, or close to the background? I’m so confused. I guess I should read up on juxtaposition photography.

Sigma 120-400: "Rear focusing system corrects for fluctuation of aberration due to focusing."

There are lots of advantages (and a few disadvantages) to a rear focusing system, but correcting for “fluctuations of aberration” isn’t one that I’ve ever found. But if it is true, then why doesn’t the rear focusing system in all of your other telephoto zooms correct for ‘fluctuation of aberration’?

Sigma 300-800: "A perfect lens for surveillance use, its stealth black finish is unobtrusive and easily concealed."

OK, let me get this straight: the lens weighs 13 pounds,  is nearly 3 feet long with the hood mounted, and has a front element the size of a dinner plate. But by painting it black, it’s now “ unobtrusive and easily concealed”. If you mount it as a gun on a battleship, maybe.

Nikon

Nikon J1: "Set free from technicalities and complicated operation, you can capture the desired image faster and more unfailingly."

You mean I’ve spent thousands of hours over a decade learning how to do things like select aperture, choose ISO, frame shots, use strobes; but all this time the key was being set free from complicated technicalities. Oh, the waste of it all.

NIkon J1: "What's more, it will automatically select the most spectacular shot from the series of shots taken."

That’s good to know, because I can spend hours looking at a day’s shots in Lightroom and still not be able to do that.

Nikon J1: "The only buttons you will find on the top of the camera are the power switch, shutter-release button, and movie-record button so you will have an amazing range of functions at your fingertips."

Three apparently is the new “amazing range”.

Canon

Goverment Marketing Division Mission Statement: "The Canon Government Marketing Division represents Canon U.S.A., Inc. in the Government Market..."

Wow, am I glad they pointed that out. For a moment there I thought maybe Canon was marketing governments to small island nations or something.

The Mission of Canon Australia: To be the preferred supplier we will be easy to deal with and provide quality, value for the money, products and services.

My Australian friends tell me this rivals the Sigma 18-250 blurb for most lies in the fewest words. But I guess if you are the only supplier, you’re the preferred supplier.\

Canon G1x: "the new 1.5-inch High-Sensitivity CMOS sensor"

OK, your 1.5” sensor measures 22.3mm diagonally, which is actually 0.88 inches. So 0.88 is the new 1.5. Got it.

Canon G1x: "The new 14.3 Megapixel sensor, combined with a new lens and DIGIC 5 Image Processor, creates the HS SYSTEM that delivers image quality previously impossible in a compact camera."

Hmmm. Maybe so, but since the Sony NEX-5n and 7, Fuji X-100, Leica X-1, Panasonic GX-1, and Olympus E-P3 are all at least as compact, it’s quite a claim, don’t you think? Maybe you could show us a bit of data to back that up? No? OK.

Sony

Sony SAL 30mm f/2.8: "While most Macro lenses are not image stabilized, this one is, because Sony builds SteadyShot® INSIDE image stabilization into every ‘A’ camera body."

And the camera body has what to do with the lens? Especially since you also push it as a small macro to shoot with an adapter on NEX bodies? And how come the Sony 100 Macro isn’t image stabilized? I read the blurb on it carefully and it doesn’t say a thing about it. I’d rather have IS on my 100.

And in a somewhat related topic, will the real “world’s fastest camera” please stand up?

Apparently "fastest"is going to take over from "most megapixels" in the marketing wars this year. After reading all these claims, I'm so confused. . .

Olympus E-P3Worlds Fastest Autofocus

Olympus OM-D “employs the FAST (Frequency Acceleration Sensor Technology) AF system, introduced in the PEN E-P3, but with even greater speed.”

Nikon V1: “First, we present the world's fastest autofocus camera”

Nikon J1: “World’s fastest continuous shooting, 10 fps”

Nikon D3s:11 fps in Dx mode

Nikon V1: “The world's fastest shooting time lag”

Sony NEX:The world's shortest shooting time lag

 

I'm sure there are dozens more entertaining marketing blurbs that I've missed, so please add any good ones you've run found in the comments. But reading these reminds me of why I insist on us writing our own "takes" on various items, rather than copy-pasting the manufacturer's blurbs like most sites do.

 

Roger Cicala
Lensrentals.com

February, 2012

35 Responses to “Say Whhaaat?”

Bob said:

To be fair to Sony, I think when they said that the NEX-7 "has an APS-C size HD image sensor the same as Sony builds into full-sized DSLRs", they were referring to the a77, which does indeed have the exact same sensor as the NEX-7.

Chris said:

The 1.5 inch sensor refrence is using the same one that originated with television tubes and is what would be the circumference of the tube surrounding the sensor area. Most video cameras use this (1/3, 2/3, 1/2 ect.) as well as the micro 4/3 (4/3 inch). Nikon 1. (1 inch) ect. I think pretty much every sensor size refrence smaller then aps-c goes by this measure,

Luke Fields said:

Shhhhhhhh!!!! Don't make these companies self conscious! We need them to continue to lie to the general public muggles so that said muggles will go out and purchase tons of mediocre glass. That way these companies have enough funds/profit to produce the truly wonderful, spectacular lenses and cameras that we can only dream of owning and instead must live vicariously through the lensrentals staff, or take on fate and rent the equipment ourselves! Don't question the system!

Simian Crease said:

Fun stuff, Roger. Reminds me of how marketing research data has apparently turned manufacturers on to using the letter "X" as much as possible. I look forward to the 2014 Canon "XXX Mark X", and the Nixon (Nikon's yet-to-be-made name-change mistake) FX-sensored "DX00" (pronounced "D X-Hundred")...

Jim Davenport said:

Marketing companies/departments are beginning to reflect the current state of education. I recently commented on Facebook about the Mercedes ad where they talk about their new coupe have "less windows" when it should have said "fewer windows." I posted this on the Mercedes FB page and they promptly deleted it.

A said:

Sigma 300-800

ROTFLYSST* @ Roger's comments

As for:
Sigma 18-250 f/3.5-6.3 DC OS: “at the long end it can be used for sports actions.”

In all fairness they didn't say it would be *good* for them. It could also be used as a hammer, or a wheel/roller, or... I'll also note that lawn bowling is theoretically classed a sport, so if you're covering slow moving and well lit sports it might be a candidate ;)

*Rolls On The Floor Yet Strangely Still Typing (nods to Bill Bailey)

Walter Freeman said:

"Sonic wave body" seems like a bizarre variant on Olympus' idiotic "Supersonic Wave Drive" name for ultrasonic motor focusing. By all accounts the SWD lenses do focus very fast like they claim, but could they come up with a less dumb name for them?

As for the "most macro lenses don't have image stabilization, but ours do because of sensor-shift IS" -- that's wrong on two levels, even! Nikon's 105 macro has VR, Canon and Sigma have new ones with it, and Olympus and Pentax have in-body IS.

But for real macro shooting IS won't help you all that much anyway, unless it's aware of translational shifts too. The only systems that do that are the one in the Canon 100L macro and the new Olympus EM-5.

Nicholas Condon said:

I don't think it's entirely unfair to categorize the Sigma superzoom as useful for sports action photography; I just think one wants to be careful as to the actual subject and use of such photos. You want a perfect shot of Jeremy Lin in the middle of a jump shot for the cover of "Sports Illustrated," then, as anyone actually trying for such a shot would tell you, the Sigma's not the way to go. You want a shot of your 10-year-old playing soccer for your family-only Flickr page, you could certainly do worse than a decent SLR with that Sigma mounted on it (e.g. a cameraphone or a cheap compact superzoom).

Bart said:

Olympus' EM-5 5-axis image stabilization is reached by moved the sensor along 3 different axis. Now, how is it again that they correct for pitch and yaw?

jim thomson said:

Sigma 300-800. This is shrewd marketing. It plays to many photographers predjudices.How many times have you seen comments about wanting black lenses because the big whites draw too much attention?

Hanspeter said:

"Simga [sic] 20 f/1.8"

Speaking of copy editing... :)

LensRentals Employee

Roger Cicala said:

Hanspeter wrote: “Simga [sic] 20 f/1.8″ Speaking of copy editing…

OUCH! Got me. :-)

Daniel Browning said:

You could fill a site -- a *lot* of sites -- with lies that Marketing says. And they have. Which is why I don't read them. (+1 Ratatouille.)

The ones you've listed here might fool a layman, but I think they wouldn't sneak past anyone who has an been into photography for a while. So the lies that annoy me the most are the ones that even veteran photographers swallow hook, line, and sinker. Like Canon's "14-bit for smoother gradations" balderdash -- photon shot noise already limits the gradations to 8 bits.

Carl said:

Some of those are pretty bad. Makes me wonder when someone will claim a camera is "turbocharged"...you know, that old chestnut...

What is "photon shot noise"?

Daniel Browning said:

Hello Carl, photon shot noise is described pretty well on wikipedia:

http://en.wikipedia.org/wiki/Shot_noise

Basically, it's part of the fundamental nature of light itself, so there is no technological way to work around it. The only way to truly improve the number of gradations (or "color depth") is to increase the total amount of light recorded, such as by increasing the effective FWC (full well capacity). But since that is so difficult, it's much easier to just increase bit depth and Marketing can spout bald-faced lies about how it improves things. In fact, in Canon's case, it makes no difference whatsoever.

The reason basically comes down to the difference between accuracy and precision. Imagine you're counting the number of water drops that fall into a set of buckets every minute. First you use hoses that drop water at extremely accurate intervals. With advanced electronic measuring equipment, you measure 1.0000 drops/sec for every bucket, every time. Then you try it with actual rain, and the numbers jump all around randomly for every bucket (e.g. 0.9751, 1.2453, 0.9853, etc.) in a Poisson distribution, even if the average rain is exactly 1.0000 drops/sec. The random variations mean that recording five significant digits of precision is no more accurate than just four or three, etc.

So in effect, Canon is increasing precision (12 bit to 14 bit) and claiming that it increases accuracy, when in fact they are just adding useless random numbers in the least significant digits.

Carl said:

Daniel, that is interesting. I could be mistaken, but your assertion also seems to hint at a point I made in another blog section, about the theory of light photon behavior, and evidence that the same photon can occupy two points in space simultaneously.

However, your overall point does seem a bit inconsistent with reality. For one thing, Canon's analog-to-digital conversion (for DSLR's) has been stated as 14 bits, since at least what, 2006? It's nothing recent, unless we are really only talking about the improvement the G1x may have, over the previous G12. Regarding DSLR's, it's only been Nikon's entry level "crop sensors" that have been 12 bit...and perhaps a few others I may be unaware of.

Are you suggesting that there is no use in having analog to digital conversion at more than 8 bits? Let me ask you this, do you shoot in RAW, or jpeg, and do you have a bias specifically against Canon? (i.e., are you a Nikon "fanboi"?).

I mean, 99% of the time, I shoot in RAW mode. Then I have to convert to a lossy, 8 bit jpeg, if I want to have a print done by most any photo printing lab. (Some labs have told me they can print from an 8 bit TIFF file, which would be better, since there is no lossy data compression.)

But, I (and perhaps most every serious photographer) can see a vast difference, just on a computer monitor (even a cheap one), in loss of color depth and gradation...going from the displayed RAW file (in Adobe Camera RAW, in my case), or from a 16 bit TIFF file...than going to an 8 bit jpeg, and viewing that on the screen.

What evidence do you offer, other than mathematical theory and imaginary raindrops in microscopic buckets, for your assertion? Has anyone performed tests on cameras or camera sensors, which would help prove the supposed "folly" of using 14 bit AD conversion over 12 or 8 bit conversion?

If you look at DXO Mark testing, you will see that "medium format" digital cameras, achieve the most color bit depth of any digital camera, in their "portrait test". Most or all of these cameras, actually use 16 bit analog to digital conversion, rather than "merely" 14 bit. They also obviously have much larger sensors and "sensels", which helps.

In the world of music and audio, there are people who will spout things like "nobody can hear the difference between a lossy MP3 file, and a 16 bit linear PCM file from a CD. Therefore the extra data on a CD is wasted". They especially don't think anyone can hear an improvement when listening to higher resolution digital formats, such as "SACD", or "24bit /96 kHz", or "24 bit /192 kHz", etc.

I think what these people really would like, is if nobody listened to any recorded music at all! I can promise you that there is an improvement which is easily heard, with audio...if you actually are interested in listening. You don't even need a system that is all that high in performance, or price. The key is, do you really care, or are you more interested in trying to equalize everyone's entertainment outcome, by stating that all measures of performance are useless, and thus there is never any improvement with higher resolution, or newer innovations in technology?

In looking at my RAW files in ACR, I can certainly see a vast difference between them, and an 8 bit jpeg. To me, they even look better than a converted 16 bit TIFF file, but I'm sure there's a reasonable explanation experts would have for that.

Tom Cavanaugh said:

Here's a tutorial on how to shoot children's sports. It's actually fairly painful to watch, so be prepared. I have no doubt that the photographer is very good at some things, but it's clear he's never shot children's sports.

http://www.youtube.com/watch?v=WjwQieeSbeM

The Sony website has it embedded in their website here:

http://store.sony.com/webapp/wcs/stores/servlet/CategoryDisplay?catalogId=10551&storeId=10151&langId=-1&identifier=S_Sony_Howdini

Daniel Browning said:

> Canon’s analog-to-digital conversion (for DSLR’s) has been stated as
> 14 bits, since at least what, 2006?

Yes, and I'm glad you brought it up because that is one of the reasons that makes it so annoying. They've had six full years to either stop lying about it, or to actually bring the hardware up to the level of their lies (or at least some of the lies). But as it stands, even after six years, Canon's 14-bit files still have *no* amount of improvement over 12 bits whatsoever. (Crossing my fingers that the 1D-X finally changes that.)

> Are you suggesting that there is no use in having analog to digital
> conversion at more than 8 bits?

I am definitely not suggesting that. Gradations is only *one* of the raw file characteristics that is affected by bit depth. Dynamic range is another. So even though 8 bits would be enough for gradations, it would not be enough for dynamic range. But Canon didn't say that the 14-bits improved dynamic range (and if they did, that would have been false too, because their read noise limits it to less than even a 12-bit ADC can do).

> Let me ask you this, do you shoot in RAW, or jpeg?

Always raw.

> , and do you have a bias specifically against Canon? (i.e., are you
> a Nikon “fanboi”?).

First, I like to think that I am an equal opportunity hater. :D I can bash nikon, sony, olympus, pentax, and the rest just as well as canon for their nonsense claims and fantastically braindamaged cameras. Would you like a sample? The list of nikon flaws that annoy me include: star killer algorithm, hardcoded black clipping, white balance preconditioning, harmful analog gain, masked-off pixel removal, MTF charts that make up their own laws of physics, curtain *and* mirror cycling in liveview (finally fixed in the D800), "lossy" LUT that was never updated when they went from 12 to 14 bits, etc.

Second, if I do have a bit of a unrecognized bias specifically against canon, it may be because I use canon myself, and familiarity breeds contempt. So their particular set of braindamage may hold a special place in my heart. But at the same time, the reason I continue to use canon is because their cameras and lenses do offer the best compromise of benefits and flaws for my photographic interests. You could say it's a love/hate relationship. :)

> But, I (and perhaps most every serious photographer) can see a vast
> difference, just on a computer monitor (even a cheap one), in loss
> of color depth and gradation…going from the displayed RAW file (in
> Adobe Camera RAW, in my case), or from a 16 bit TIFF file…than
> going to an 8 bit jpeg, and viewing that on the screen.

There may be a few factors that you are not considering:

* The bit depth of intermediate and display formats have nothing to do with the bit depth of the ADC or raw file. I use 32-bit EXR files sometimes because it is superior to 16-bit tiff when I need the increased precision. But that doesn't mean that 14-bit is necessarily superior to 12-bit for the raw file. The answer to that depends on the camera and the performance metric being characterized.
* Degredation caused by JPEG compression (which may include cosine transforms, color space conversion, chroma subsampling, and more) has nothing to do with losses caused by bit depth conversion. There are significant losses just going from 8-bit tiff to 8-bit jpeg, but you obviously can't blame those on bit depth conversion, since no such conversion took place. If you had wanted to make a point about bit depth, you should have said that you saw significant degredation going from 16-bit tiff to 8-bit tiff.
* I imagine that you are not one of the 0.0000001% of photographers who have a fully working end-to-end 10-bit display setup (including a $2,000 monitor, 10-bit DisplayPort cable, etc.), so anytime you think you are viewing a 16-bit file, you are actually only viewing an 8-bit conversion of it on your display. The only folks I know that have truly 10-bit end-to-end setups right now are film colorists with $100,000 grading suites. The fact that you are seeing a difference between an 8-bit conversion of a 16-bit file and an 8-bit file means that something else is coming into play (such as JPEG compression).

> What evidence do you offer, other than mathematical theory and
> imaginary raindrops in microscopic buckets, for your assertion?

I'm happy to oblidge your requests for evidence as long as you try a little harder to keep the flowery compliments about my imaginary friends, er, raindrops, to a minimum. ;) Here is some evidence:

http://thebrownings.name/images/2009-03-bit-depth/John-Sheehy-14-bit-gaborstrapbits.jpg

Let me know if you would like the raw file to repeat the experiment yourself and I'll see if I can find it. It is a canon 40D raw file at ISO 1600, with a varied number of the least significant bits replaced with random values, then a standard raw conversion using dcraw with nothing more than offset subtraction (no noise reduction, no sharpening, etc.) at various bit depths. This shows a dynamic range that is far, far beyond what a normal photographer would ever use, but even then, a bit depth of 14 makes no difference. As you can see, 9 bits is sufficient for an ISO 1600 image even in the extreme shadows. If you repeat the experiement with an ISO 100 file, you'll find that just 11 bits is sufficient for any canon file, even in the extreme shadows at ISO 100.

> If you look at DXO Mark testing, you will see that “medium
> format” digital cameras, achieve the most color bit depth of any
> digital camera, in their “portrait test”. Most or all of these
> cameras, actually use 16 bit analog to digital conversion, rather
> than “merely” 14 bit. They also obviously have much larger sensors
> and “sensels”, which helps.

As I said before, the only factor in play here is max snr. An MFDB is optimized for max snr (by virtue of the fact that they all have to use old, off-the-shelf sensors from Kodak/Dalsa, and those in turn are optimized for it), but even then they are only slightly better than the average DLSR at any given spatial frequency (e.g. at the same pixel size). It's only when you compare different spatial frequencies (always a bad idea, IMHO), or you factor in the larger size of the MFDB (as you should in the general case), that they come in with a far greater max snr. The latter is precisely what DxoMark does. In fact is has nothing to do with bit depth.

> I can promise you that there is an improvement which is easily
> heard, with audio

I agree. If you are into production audio engineering at all, then we will be able to come to agreement very quickly. :) Canon's ADC is like all those cheapo 24-bit ADC on the market. They have so much self-noise that dithering from 24-bit to 16-bit loses no useful audio information -- the 24 bits are all just marketing nonsense. In fact, I have used an 16-bit ADC that is far superior than a 24-bit ADC. In the same way, canon is putting out 14-bit files that have so much noise, you can dither them down to 11 bits and still have the same result. And unlike audio, light has noise as a fundamental part of the signal itself, so bit depth does not affect the max snr, but it does affect the dynamic range.

Hope that helps.
--
Daniel

Carl said:

Daniel, interesting observations. Your image is hard to read; I had to scale it up to 400% to read the various bit rates you labelled.

I fail to see how your image proves your original point, I guess I am too dumb. It doesn't prove to me that your theory of "shot noise" somehow swamps a camera sensor and ruins its dynamic range. For one thing, the very nature of the sensor means it has the MOST dynamic range at or near the bottom of its ISO sensitivity...NOT AT I see tremendous noise even within the bar that is labeled "14 bit". The noise pattern appears worse within the bars that are labelled as the bit depth becomes less. It's sort of like airing a camera's dirty laundry (with "skid marks"), it's not nice to look at!

How are "gradations" different from "dynamic range"? I would think that the bit depth would have the biggest effect on gradations. Or maybe your idea of gradation is different from mine.

Whatever dynamic range and bit depth you think you have found with the Canon 40D, if you are using it at ISO 1600 very often, I can see why you would have a love-hate relationship with it. However, since you say Canon is your system of choice, and yet you are not happy with it, but you somehow think you would be less happy with other systems, this speaks of deeper issues of self-loathing you might have. I'm not a doctor, I just have learned this by watching sit-coms...

No, I don't have a world class monitor, and I agree that a single photograph probably doesn't need 14 bits of dynamic range, again...since most printers only use an 8 bit file, and I assume a photograph itself couldn't display 14 bits of dynamic range anyway. Even if perhaps it gets printed on aluminum, and lit for display by some sort of light source of infinite cost and snob appeal...shining only the most well-heeled photons who attended the best quantum mechanics schools.

I agree that pixels of similar size should be compared, when speaking of medium format and "DSLR", whether crop or full frame. (I think that is one of your points, but I'm not sure). I only mentioned them because they use 16 bit AD conversion, which is the highest of any digital camera system I have yet heard of. I assumed you would make the case that their bits are wasted too, but you held back on that.

I was aware that Kodak made the CCD sensors used in many medium format cameras. It makes me wonder, since they are supposedly going out of business for keeps now, who will take up the mantle for medium format? Since you have all the answers to any question a simple guy like me would have, I'm sure you've already typed the answer, you need only hit "enter". I'll type it for you: Canon.

That said, the Nikon D800 looks very good to me. It will be more interesting once more people test actual production units and compare it with previous bodies. The latest "rumors" are there will be more of a delay on this new camera than previously thought. Perhaps I'll just try a D700 next, since the prices are supposed to fall by $500 or more, on whatever new units will be left, perhaps through the end of the year. Or perhaps I will stick with Canon...I won't loose much sleep over it either way. I will however keep two of my Canon lenses, whatever I do next.

Carl said:

Sorry for the typo early on, my keyboard seems to be having problems! I was basically trying to say that ISO 1600 isn't the place to start talking about dynamic range and complain about noise, it seems to me.

Daniel Browning said:

> I fail to see how your image proves your original point, I guess I
> am too dumb.

Not at all -- that was my mistake, actually. I thought you had accepted that photon shot noise limited color depth and that we were on to discussing how canon's 14-bit doesn't have any other benefits, either. But I see now we're still on the original point. For evidence of the validity of that, here is the method that I offer:

* Take any 14-bit raw file with content in the top 8 bits (i.e. midtones and highlights).
* Converting one raw green channel to linear tiff (subtract offset, etc.)
* Truncate (or dither) to 8-bits, removing the least significant bits
* Convert both original 14-bit and truncated 8-bit using a common black clip (i.e. only the part of the dynamic range that both have in common, since we aren't arguing that 8-bit files have the same dynamic range).
* Compare the two images and see if the 14-bit has any sort of gradation/color depth advantage (or any difference at all).

Do you agree that the above is a good experiment for proving the hypothesis about photon shot noise limiting color gradations? If not, then can you suggest an alternate experiment?

> It doesn’t prove to me that your theory of “shot noise” somehow
> swamps a camera sensor and ruins its dynamic range.

That is not a very good characterization of my position. What I said is that photon shot noise limits the gradations/color depth, not bit depth, and that increasing the snr (typically max snr) is the only way to improve gradations.

> How are “gradations” different from “dynamic range”?

DxOMark calls the gradations "tonal range":

http://www.dxomark.com/index.php/About/In-depth-measurements/Measurements/Noise

They measure it for every camrea. Most DSLR only have 8 bits or so of tonal range (and that's only at the max SNR).

> I would think that the bit depth would have the biggest effect
> on gradations.

In raw files, it does not. SNR is the one and only effect.

> if you are using it at ISO 1600 very often, I can see why you
> would have a love-hate relationship with it.

Actually, I've always been very happy with the high ISO performance on my Canon cameras, that's one of the reasons I've stayed with Canon. Sure, they're a little behind the competition these days (I have a 5D2 now), but it's still close enough for me. It's a wide variety of other problems that I dislike about it.

In any case, I don't see why you're still so obsessed with my personal camera choices, but oh well.

> However, since you say Canon is your system of choice, and
> yet you are not happy with it, but you somehow think you
> would be less happy with other systems, this speaks of
> deeper issues of self-loathing you might have. I’m not a
> doctor, I just have learned this by watching sit-coms.

Thank you for the kind words, but unlike some photographers, I do not feel compelled to ignore the flaws in the equipment I own, even if it is the best possible equipment for my needs.

> shining only the most well-heeled photons who attended
> the best quantum mechanics schools.

Good one, heh heh.

> I agree that pixels of similar size should be compared,
> when speaking of medium format and “DSLR”, whether crop
> or full frame. (I think that is one of your points, but
> I’m not sure).

Yes, that's one way to put it.

> I only mentioned them because they use 16 bit AD
> conversion, which is the highest of any digital camera
> system I have yet heard of. I assumed you would make
> the case that their bits are wasted too, but you held
> back on that.

I was trying not to go off on too many tangents, but yes, I do maintain that the MFDB 16-bit raw files are a collosal waste, and have been for years, for the same reason that Canon's 14-bit files are. They were 16-bit back when Canon was only 12-bit, and even then the MFDB 16-bit contents were no better than 12 bits.

> who will take up the mantle for medium format?

I don't know.

Carl said:

Daniel, your explanation is more clear now, thank you. However, I confess I don't know how to perform your test. I will try to learn. It's not the type of processing I do.

That said...I still think it's silly to be as disgusted as you seem to be, with your camera system of choice. Maybe it's more of an act?

Has anyone other than you decided that 14 bit AD conversion in cameras is wasted data? If so, do they agree about the cause? This is really the first I have ever heard it.

In any case, the RAW file size doesn't seem to limit the processing speed, unless you are wanting to shoot at video frame rates or something. At that point you would really need electronic shutter and of course no reflex mirror...

My only strong criticism of Canon, is that their top tier supertelephoto lenses have risen 30% in price, with perhaps 8% increase in performance. Makes no sense. It's like they suddenly think of themselves as a college or something. I'm glad I attended back when it cost what it was worth...hahah.

Unlike you, I'm not very happy with the noise performance. But given my Canon is now nearly 4 years old, and works like new, there is still a feeling of high quality it imparts with use. I can't say the same for their Rebel line, which seems like it's made by a different manufacturer...and sort of is I guess!

Daniel Browning said:

> That said…I still think it’s silly to be as disgusted as you seem to be,
> with your camera system of choice.

It's because of wasted potential.

Imagine if you bought a $10,000 car brand new, that leaked oil, broken cruise control, no heater, no A/C, manual windows/locks, belches black smoke, and a host of other problems. A 2000-pound logo is welded to the car for marketing, and mileage is only 10 MPG. Now imagine that all the problems are very easy to fix -- it would only take a skilled mechanic several hours to make the car ten times better. But the hood is welded shut and no one can work on it, except the manufacturer, who hasn't bothered to fix any of the problems in years. Normally I'd just buy a different brand -- but they are only worse. Despite all its problems, it actually has the least annoying of them.

Of course, the core basics still work great. It gets me from point a to point b. The areas where the manufacturers compete with eachother -- acceleration, handling, aesthetics -- are all very good. Other drivers *love* the car. They have no idea that 10 MPG is actually very poor for this class of vehicle, and that it could be getting 20 MPG very easily. They've never even imagined features like a heater, air conditioner, power windows, or cruise control, because they've never seen a car that had them. In fact, when a manufacturer finally does announce a certain new feature, most drivers make fun of it for a year or two until they realize just how critical and useful it is, and yet they do the same thing the next time around. (This happened with auto ISO, liveview, movie mode, mirrorless, and more.)

I spent ten grand on my cameras and lenses, so I have 10 grand worth of expectations. All the problems that bother me the most could be easily fixed by anyone who has access to the camera software source code. But only the manufacturer has it, like a hood that is welded shut.

> Has anyone other than you decided that 14 bit AD conversion in cameras
> is wasted data? If so, do they agree about the cause?

Yes and yes. For example, I highly recommend Emil Martinec's essay, "Noise, Dynamic Range and Bit Depth in Digital SLRs":

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/

He covers bit depth in section 3. Two quotes:

"Both Canon and Nikon have introduced a finer level quantization of the sensor signal in digitizing and recording the raw data, passing from 12-bit tonal gradation in older models to 14-bit tonal depth in newer models. A priori, one might expect this transition to bring an improvement in image quality -- after all, doesn't 14-bit data have over four times the levels (16384) compared to 12-bit data (4096)? It would seem obvious that 14-bit tonal depth would allow for smoother tonal transitions, and perhaps less possibility of posterization. Well, those expectations are unmet, and the culprit is noise."

"[...] many of the levels used to record higher exposure zones are unneeded and indeed wasteful -- photon shot noise is much larger than the level spacing in midtones and highlights on a typical DSLR".

Emil is not even a professional photographer or photograhpy book author -- he's just a talented amateur. But he is very smart; he's one of the guys that developed string theory.

> This is really the first I have ever heard it.

I'm not surprised. The bit depth thing is just one of many areas where 99.999% of photographers are being misled by myth and misunderstanding. Of course, this is not really the place to go into them.

Frank said:

@Carl & Daniel Browning:

I'm another one of those who say 14bit resolution (or more) is wasteful. It's all about the SNR and the noise level of the ADC. Like in every other electronic application where very low voltages are measured, noise is always the enemy. Every part of the circuit adds noise to the signal until it finally exits the ADC.
So high resolution of ADCs is mostly useless, especially in low voltage applications like a digital camera; 14 bit has 16384 steps, mapped to the output voltage range of a digital image sensor (i don't know exactly but i would assume the maximum output voltage is surely below 1V) you get in the µV (microvolt!) range where even the surrounding electromagnetic fields from electric equipment, wires, etc. induce 'tonal transitions' into any part of the analogue system not heavily shielded, turning the least bits of a 14bit reading into random numbers. (In software development for microcontrollers readings from an unconnected ADC input (pin left open, acts as an 'antenna' for noise) are often used to create 'real' random numbers for internal functions which is extremely useful in case you have an unused ADC input left.)
Sensor manufacturers must produce sensors with a higher SNR, giving higher output voltages with less light to be able to put actual values into the LSB of 12 or 14bit data, not just mostly noise.
Noise can't be eliminated by software, it's analogue.

A bit off-topic:
Eizo states that my monitor uses 8bit on it's DVI input (which i'm using of course) and a maximum of 10bit through the Displayport. They also state that the internal display controller has a 12bit(!) LUT. The idea is to stretch any 8 or 10 bit data to the length of a 12bit string and fill the 'gaps' with the data from the LUT. Now that's ambitious since it's just 'guessing'. After all the panel itself probably works with 8bit..

Frank

Carl said:

The least annoying, and yet you are obviously very annoyed. You seem to describe aspects of the car I own now! Except it has a 4 ton searchlight welded to the hood, which I use with a 6 foot diameter steel drainpipe, as a snoot...to add hair highlights to mountains in the background of my evening landscape shots.

If you spent 10 grand on your camera and lenses, then go out and use them, and be happy. I've spent a lot less, and I get lost in another world whenever I take pictures. It's not usually an out of body experience, but it's always been fun.

You are more than welcome to lend me whatever super telephoto lenses you have, for a few months. This will solve all of your problems! Even better would be to just sell it all, and loan the money to me so I can day trade facebook the day it goes public! If I happen to lose a few thousand of it, so be it.

Thank you for quoting it so I didn't have to read the whole thing, haha...

String theory is just a theory, and has evolved into membrane theory. But honestly, do you think humans will ever know any of that for a fact? Are we all just made up of unimaginably complex vibrating pretzels that have a shape that occupies all space and no space, and curve in every direction and in no direction? It does fascinate me, but in the end, I don't think humans will ever know "everything". It's nice that we try. I don't see us ever knowing what's at the other end of a black hole, or if there is a multiverse, or what really happened before the big bang...or why there was expansion which seemed to occur at a rate faster than light...Or for that matter how some scientists have recently measured particles ACTUALLY travelling faster than light...or why women love to play farmville!

I haven't noticed any posterization in my pictures; I only see noise when ISO is increased. I assume you're saying that the noise could be improved if less bit depth was written in the first place? But with less bit depth, won't there be more posterization, and not less?

Carl said:

The quote seems to say the bit depth is wasted because the "photon shot noise" overwhelms the recorded signal in the midtones and highlights...but I thought you had said it overwhelms the signal in the darker end. Certainly there is less bit depth in the darker end.

The only real criticism I have read that has interested me, regarding the difference between what film could do, and what digital sensors can do...is that digital sensors supposedly lose color saturation in the brighter third or so of the picture, where film supposedly didn't. And yet this is where most of the bit depth is. It really should have a broader range of color tonality as brightness increases, it seems to me. But I'm sure there's a theoretical, mathematical explanation which will include the word "cosine" and a lot of formulae...with some guy like tv's "sheldon cooper" telling me if I was only a bit smarter, I would "get it", and my life wouldn't be such a waste!

I'm more like foghorn leghorn..."But how did you...?...I know boy, but don't bother...I wouldn't understand it anyway".

Carl said:

So, when shooting a vivid flower in bright sunlight, unless I set exposure compensation back quite a bit, the color seems more pastel-looking. To me it also certainly becomes one-dimensional looking, and has the effect home theater critics call "over-saturated primary colors". In short, it looks less vivid, less real. So in editing, I might dial back the exposure, add a bit of brightness, and take down the saturation slider, and perhaps tweak the individual color sliders. Then the flower suddenly has surface detail AND natural color...but then, so does everyone else's who entered the contest! Can the world support 6 billion photographers? No...so at some point I guess all photography will grind to a halt!

To conclude, I suggest you build your own camera from the ground up, and make it better than everybody else’s. Then after a couple of years of us moronic photographers making fun of all your innovations, we’ll suddenly stop playing angry birds, and buy your design hand over fist…making you rich…so you can go off and buy more gadgets that annoy you
.
I guess it’s good this isn’t twitter. Obviously I can’t express a thought in under 18k characters…but then, does anyone actually express anything on there, other than the verbal equivalent of passing gas?

Daniel Browning said:

Frank, Great post. Thank you for contributing in a positive way to the discussion.

Micke said:

Sigma 18-250 f/3.5-6.3 DC OS: “at the long end it can be used for sports actions.” Mountaineering and chess are two sports that come to mind!

Carl said:

You can use an f/6.3 lens for sports action fairly easily, as long as you don't need autofocus (many SLR cameras won't AF with a lens darker than f/5.6).

About three years ago, I shot pics at near sunset, of friends' kids riding their dirtbikes...with a cheapie manual focus, Pro Optic F/6.3 500mm "mirror lens" (mounted on a monopod). The pics turned out fine. Sure there was noise, but PS took care of most of it. The shots might not look sharp at 14x20 inches, but nobody wanted prints that big. They looked fine as 8x10's.

To get the shots I simply focused on the location where I was going to snap the shots, then as the kids rode through the creek and splashed water (using 6 frames per second), I got some decent shots...with the splashed muddy water stopped mid air in the pictures. I think I used ISO 1600 or 2000, and the shutter speed was around 1/800. The ambient light was about dark enough that you would begin to consider turning headlights on, to drive a car (not so much to see, as be seen).

I will admit that selling this lens wasn't so easy; I wound up letting it go for under $90. The image quality really wasn't terrible, it just didn't compare to a ~400 to 500mm lens costing over $1000. But if it sells for $160 new, and weighs only a fraction of those, why should it?

Anyway, it's really just the loss of AF that would hinder an f/6.3 lens for sports action, and not necessarily the lack of a fast shutter speed at low ISO. Certainly I got shots of 1/2000 and even 1/4000 second at ISO 200 and less, during very bright sunlight for mid afternoon shots.

Of course "pro bodies" will usually AF somewhat down to F/8, but since much of this discussion centers around consumer zooms, it's not relevant...other than if using a TC on a supertelephoto.

I rented the "often overlooked" Canon 400 f/5.6L "prime", and plan on trying my 2x TC on it, for an 800mm f/11. It also has no IS. I probably won't attempt to shoot sports...or hummingbirds in the moonlight...or rather, bats...with this combo, though!

Philip Tyre said:

Wow, Daniel. That's a lot of new information for me, and you really did a great job presenting it in an understandable way. It's not every day that you feel smarter after reading blog comments. :)

Thanks for taking the time to share what you've learned.

Daniel Browning said:

Thanks, Philip! You're very welcome.

MikeG said:

This has been one of the most informative, entertaining and generally good natured set of comments I have read in many a long year. Thank you both Daniel and Carl. And Daniel, I have learned a great deal from your essays .. far more than I should have and far less than I now want to :)

Carl said:

MikeG, you're most welcome. I've enjoyed it too. I'm guessing if the manufacturers ever start to think more like Daniel, it would surely help them increase the speed of new models...since abandoning 14 bit file depth in favor of less, would drastically reduce file sizes. I don't know if I am completely convinced, but there does seem to be valid evidence.

I guess I am just still hung up on sites like DXOmark, because their testing and comparisons are so useful. I look forward to their findings on all the new cameras, especially the D800 and 5D3. In using the 400mm f/5.6 lens I rented this week, I am finding that shooting in low light, really means a need for ISO 6400, 12,800, and beyond...so that would cause me to lean more toward the 5D3 than the D800. The D800 would absolutely rule when using wide angle lenses though, it seems to me (detail-wise anyway). And the D800 would benefit most from being able to store files at less than 14 bits, since there are so many pixels.

You would think the entire camera industry would pursue it, if Daniel's opinions are even leaning toward correct. Yet instead, they keep finding ways to increase file size, rather than reduce it. Most would probably just tell you to shoot jpegs, if you're worried about file size...which isn't really a good solution!

rtbox said:

I'd just like to point out in defense of the Tokina statement, Silent Wave Body kind of makes sense. The S in AF-S means silent wave, and they are the only lenses those bodies can focus. Dubbing them Silent Wave Bodies isn't exactly a leap as most people will want them to AF.

Leave a Reply