Navigating Through Chaff


Beware. You ARE what you read...

The whole run-up to the D3/D300 announcement and the follow-on reactions have produced the usual set of overreactions, paranoiac responses, ill-informed posts, and outright silly misinformation. Let me guide you through some of it.

Nikon has abandoned small sensor users
Nonsense. As they said at the press conference, DX (small) and FX (larger) sensor, camera, and lens development will now continue in parallel. The D3 and D300 announcement is just one example of that (I hesitate to start more camera rumors--see right column--but what I know about future Nikon products says we'll see more of this type of dual progress).

Another variant on this are posts like "How long ago was it that Nikon said they saw no need for a full frame sensor?" Actually no one at Nikon Japan has ever said that as far as I can tell. When asked specifically about the 36x24mm sensor size (now dubbed FX by Nikon), they said they were working on it, but added that they wouldn't introduce one until it became financially tenable for users. Hmm, same price as the D1, D1x, and D2x seems to be the point they deemed "financially tenable," and we've now reached it. In my opinion, Nikon did exactly what they said they would do.

From both a financial and quality standpoint, true consumer DSLRs won't support larger sensors for a long, long time, if ever. That's because the quality is good enough already at the smaller size and the smaller size provides an 8x or higher improvement on cost for the most expensive part in the camera. The vast majority of the market wants sub-US$500 DSLRs, and that requires DX sensors.

Just to be clear, Nikon executives have said two things that should be interpreted as backing up their "develop in parallel" quote: (1) They expect to have to sell a sub-US$300 DSLR in the not-too-distant future; and (2) They expect to produce higher end FX DSLRs in the future (not specified as to time frame). Interpret that as you wish.

Another faux point that keeps coming up in Internet discussions is that the FX nature of all five new lenses that were introduced means no more DX lenses. Also wrong. Future DX lenses were said to be in development by Nikon. Moreover, this introduction was about professional gear and a Nikon professional may now be carrying both FX and DX cameras in his bag, and thus needs lenses that cover both. All five lenses that were introduced work well in that respect. Put another way: expect the next DX lenses to be announced when Nikon next introduces a DX consumer body.

Nikon did a better job keeping things quiet before the announcement
Actually, only partially. Things were about the same as previous professional announcements from my viewpoint (note that I correctly predicted the introduction would be pre-Osaka Games and hinted very clearly at the actual products prior to the announcement--those were the result of multiple leaks).

The important thing to note is that most of the public Internet leaks come from one of two places: (1) leaks of marketing and documentation materials for products in production; and (2) early, inadvertent postings on Web sites.

But when announcements are made in Japan on products that are just entering into the hardware production cycle (and don't yet have manuals or other materials deep into translation and distribution), it's a lot easier to keep the lid on specifics. When you have to turn on a factory in a foreign country (the D3 production is at the Sendai plant in Japan and already started, the D300 production will be in Thailand, but has just started), order huge quantities of parts from third-party suppliers, and get distribution channels like Best Buy informed up front, you have a lot more leak points. Thus, the D40/D80 type of camera and the 18-135/55-200mm type of lens (e.g. consumer-oriented) tend to get leaked early and often.

The one thing Nikon did better this round was get more of the key Internet sites under NDA (not me, by the way--no one from Nikon has said boo to me for over a year; well, okay, Nikon NPS rejected my order for a D3 and D300 after the announcement saying it was "premature") and restrict the distribution of the press release and other materials until just before the announcement. All those invited to the launch announcement were basically on good behavior, and that included most (but not all) of the influential photo press. Now that Nikon owns the distribution in more countries (e.g. Australia), the "official" Web site leakage is also less likely to happen now (the former Australian distributor, Maxwell's, was a notorious point of leakage).

Still, you might want to go back and read some of what I've written over the past year or so. I've known quite a bit about Nikon's product plans for some time now. Very little of what was announced was new to me. So, yes, there were plenty of leaks. Go back and read what I wrote the week before the announcement and you'll see that I was hinting at exactly what Nikon was up to (I purposely used "about a half dozen lenses" to mask the number five, just in case it was a plant to see where a leak was coming from).

More bits mean more dynamic range
This one has been floating around for awhile, but Canon's and Nikon's move to 14-bit ADCs and 16-bit processing have fanned the flames.

Bits don't equate directly into dynamic range. Sensors have a fixed well size and a minimum noise threshold, which determines the dynamic range. You can put an accurate 8-bit or 128-bit ADC on a sensor and the dynamic range possible doesn't change. Most APS sensors max out with less than 60k electrons in a well, thus, in theory, an accurate 8-bit ADC should get you the right values.

The problem is more of how those bits get assigned to portions of the dynamic range due to the linear nature of sensors, especially if you're going to do any post processing of the data (which, by definition, includes all JPEG images done in the camera). Simply put, highlights get more refined data because they use all the bits, while shadows don't retain data well because they use only a few of the bits. It may take more bits to see what the camera is really capable of doing with its sensor. Indeed, we saw a perfect example of that with the D2h to D2hs transition, where Nikon changed from internal 8-bit processing to 12-bit processing for JPEGs, and noise suddenly got lower and shadow detail better.

Will a D300 or D3 get better dynamic range than a D200 or D2x? Yes. But most of that will come from the lower noise floor, I think. The rest will come from accuracy in dealing with data that has more bits. We'll get a pretty good idea of what's going on bit wise when we use the 12-bit versus 14-bit NEF feature. If there's really an efficiency to be gained, we'll see it when we convert our images. But I suspect that the primary advantage will be better gradated tonal ramps and less rounding errors on post processing.

The LCD isn't 920k pixels
This is actually true, but the same was true of 230k LCDs: they weren't 230k pixels (I've changed my Current DSLR chart to better reflect this). The specification for most camera LCDs is in dots, and you need an R, G, and B dot to form color images. Thus, you tend to have to divide by 3 to get the true number of "pixels" being displayed on LCDs. Bottom line: Nikon improved the resolution of the LCD by a factor of four. That's all you needed to know. Still isn't HD TV, if that's what you're expecting ;~).

We'll need better lenses with higher megapixels
Not really. Technically, even Canon's 1DsIII and its 21mp are a long way from blowing past the resolution limits of current lenses. On the other hand, better sensors do better reveal flaws in existing lenses, just not resolution flaws. Still, the new Nikons try to deal with some of those flaws by simply correcting them in software (e.g. chromatic aberration).

Most people don't understand MTF (contrast measurements of lenses) and antialiasing, two things that you need to consider to figure out what "resolves" in a digital image. Even without an AA filter over the sensor, any analog-to-digital conversion is going to have a brick wall limit to "resolving." I'm going to oversimplify here, but consider that the smallest white to black transition you can get "right" is essentially two photosites wide (one photosite has to register the white, the next the black). Beyond that, you get gray aliased values. So, does anyone believe that our current lenses can't resolve more than 2500 white lines on a black surface? Certainly at the center of the frame, virtually all current lenses have plenty of ability to resolve more than any current digital camera can record.

See, the answer is much more complex: digital cameras resolve pretty much perfectly up to the Nyquist frequency (or slightly under depending upon the AA filter), then all things become muddy. There's still some "data" in that muddiness, but it's not accurate. Worse still, diffraction issues come into play as you decrease the photosite size. As with film, you have to know what your camera+lens can and can't do. And as with film, more recording resolution (megapixels) has both it's pluses (more "resolution") and minuses (more exposure of a variety of lens issues). I should also note in passing that a lot of sites reporting MTF figures for lenses are using a digital camera to do the testing and software like Norman Koren's excellent Imatest to do the measuring. The problem is that the digital camera's antialiasing becomes a bar beyond which data can't be obtained. If you were to test the same lens on a D70s and a D2xs, for instance, you'd get two different "maximum resolution" numbers for the lens because the D2xs has a higher Nyquist frequency than the D70s. Likewise, most of those lens tests are done using default settings and in-camera JPEG processing, which further masks the actual number. A true MTF number for a lens needs very expensive equipment to obtain, and very precise testing procedures. The lens tests that are floating around the Web these days tend to be done in a way that measures what the combination of equipment can do, not what the lens can do.

 
 


Did I get it right?

My roadmap says a D2h replacement comes in December 2007. Nikon says the D3 will ship in November 2007. Coincidence? Okay, I missed on the full frame aspect (remember, I wrote those words over a year ago), but the megapixels and frame rates and improved AF system bits were about right.

So what am I predicting now? My roadmap still expects a higher resolution pro body in mid-2008 (sounds like pre-Photokina to me).

I'll also note that at least one Nikon executive said to one reporter in passing that he thought that the consumer model cycles would shorten (they currently run about 18 months). Make of that what you will, but I wouldn't be surprised to see D40x and D80 "refreshings" in the 14-18 month range after their introduction. And we still don't have that US$399 DSLR Nikon said was certain to come.

There, I've fanned the flames afresh ;~). Satisfied?



bythom.com | Nikon | Gadgets | Writing | imho | Travel | Privacy statement | contact Thom at thom_hogan@msn.com


All material on www.bythom.com is Copyright 2007 Thom Hogan. All rights reserved.
Unauthorized use of writing or photos published on this site is illegal, not to mention a bit of an ethical lapse. Please respect my rights.