A short set of articles on some of the confusing points of the D800 launch
Still Lots of Confusion I
April 2 updated (commentary)--When you get as much email as I do, you start to note some common questions/themes. One thing I haven't written about on the site before deals with this question: "If I never shoot at ISO 6400, why do I need a D4?"
D2x users know the answer to this question.
If all you ever do is shoot at base ISO, then it's true, you don't need to consider what the camera does at higher ISO values. The D2x may be the best 12mp crop sensor camera when considered in all respects at base ISO. That Decay image in the Teaching Point this week is a good example: there's almost nothing to complain about when you shoot the D2x at base ISO. Even shadow areas have decent recovery, and there's excellent micro contrast in the detail. My biggest issue with the D2x was at the highlight end, where you had to watch very carefully what happened above 2.5EV above middle gray lest you lose highlight detail.
But if you venture even a stop or two above base ISO on a D2x, things start to fall apart. Dynamic range disappears and noise intrudes very rapidly. The highlight limitation is pushing your exposure down to start with and now the shadows start to have real issues. ISO 800 on a D2x is almost unusable because of all the issues that creep in. From base ISO to ISO 800 there is a quick and linear drop in image quality, and that continues as you push higher.
And that's the answer to the question. If a camera still has good dynamic range and little noise at ISO 6400, the other ISOs between that and base are better, too. D3/D3s/D700 owners know that they can use base ISO to ISO 800 with near impunity. Having a camera with such a trait gives you at least three stops of shutter speed relief. If you're shooting wildlife at base ISO and the light drops so your shutter speed is now 1/125 (not good for a long lens in a vehicle that may be shaking and an animal that may be moving), you can boost that via ISO to 1/1000 without changing anything else. As long as you're not significantly losing other image quality factors as you boost ISO.
It's that last line that makes us want a camera that has really good high ISO capability. It's not necessarily that we need to shoot at ISO 6400, or 12800, or worse--though every now and again you run into a situation where that is useful--it's that the ISOs we do use don't steal away image quality.
Technically, pretty much every ISO boost, even on the best cameras, steals a small amount of image quality. But there's a tangible difference between a camera that takes three or four ISO boosts to lose a stop of dynamic range and one that loses nearly a stop of dynamic range per ISO boost.
[A couple of things to note about the previous paragraph. First, in the original I did not write that the D800 is worse in dynamic range than the D4 at higher ISO values. I wrote "it loses its dynamic range advantage." Some people are trying to put that paragraph's original wording together with another earlier paragraph and imply that I wrote that the D800's DR is significantly worse than the D4's at higher ISO values, but I didn't write that, because it's not true. I've edited the paragraph so that no one should incorrectly make that connection again.
But now we have some new work to do. At base ISO, the D800 is better than the D4. At about ISO 400 or so, the D800 loses that advantage (which was due to the D4's higher read noise in all likelihood). But both are significantly better than the D3/D3s until about IOS 1600. Above ISO 1600 all three cameras are very close in DR.
Note also, that I do not measure dynamic range the same way DxO or some others do. For over a decade I've been writing about usable dynamic range, which is a subjective thing, and the numbers I come up with are a bit different than other reports. I don't report these numbers (let alone to double decimal point precision as I've seen in some other cases, on a subjective measurement!) because they're not meaningful to the discussion. You either believe what I write about my subjective measurement observations or you don't; numbers won't help convince you of which to do.
I'll also point out that what's happening in the individual channels is important too, especially since Nikon seems to have turned down the color information lately. My D800 produces bigger RB shifts in white balance than my D4 does, which in turn produces more RB shift than my D3s does. I'm still evaluating this, but significantly red or significantly blue light is going to change how you evaluate low light performance, too.]
Sensors have gotten good enough--at least the FX ones--that we're now in a land where all the Nikon cameras (D700, D800, D3, D3s, D3x, D4) perform remarkably well in almost every common situation you'd shoot them in. What makes you pick one over the other tends to be the uncommon situation of your shooting. The D4 is the frame rate and AF champion for fast moving subjects. If I were shooting sports I'd want a D4 over a D3 or D3s just for those things. The D800 is the pixel champion for detailed subjects or large prints. If I were shooting landscapes for large gallery prints, I'd want a D800 over everything else.
But nothing has really diminished the capability of the D700, D3, D3s, or D3x. They're all still exceedingly good cameras. Every one of these cameras survives the ISO boost game better than the D2x did. For most people, that's all you really need: a camera that has little drop off in image quality in the ISO values you'd normally shoot in. You can add the D800 and D4 to that list if you want, but my point [before this revision] was that for many people they may already have a camera that's more than capable of great images.
What makes things a little more difficult is the tendency of a lot of users to put lesser lenses on their cameras. If you're using a consumer zoom that has a maximum aperture of f/5.6, you're giving up two stops right out of the box (compared to a constant f/2.8 aperture zoom). To put that in perspective, that's as if your base ISO is no longer 100, but rather 400 (or 800 if your base is 200). Now in that wildlife situation where you need another three stops of shutter speed, we're starting to reach fairly high in the ISO range (3200 for base 100, or 6400 for base 200). Even the best FX sensors start giving up a bit of image quality that can become visible as you reach that high in ISO.
Still Lots of Confusion II
April 3 (commentary)--With the mega megapixles of the 36mp sensor on the Nikon D800, the question of "small raw" has returned to the fore given the 75MB file size at the best capture quality. The 21mp Canon 5DII has three "raw" sizes (21, 10, and 5mp) after all. So why didn't Nikon do something similar?
The first problem we have is definitional: what is raw? In theory, raw is the actual data from the sensor, basically the digital version of the analog count of electrons at each photosite position. Anything that changes that is no longer "raw", it's something derived from raw data.
So just how would we create a "smaller raw"? We have quite a few possibilities:
- Bin the Bayer. Take each 2x2 block, average the greens, then create an RGB value for the block. This unfortunately only saves us 25% in file size, but we now have a 9mp output (3680x2456). You do get a SN ratio and dynamic range that improves by 3dB, maybe an extra stop worth. That still doesn't sound like the gain (25% size drop) outweighs the loss (75%)*.
- Site Skipping. Okay, how about if we just skip over photosites? The first row is RGRGRGRG and so on, so how about if we record RG...RG... etc., skipping every other column pair? Likewise, skip every other pair of rows. Once again we're down to 9mp (75%), but at least now we're down 50% in file size. One problem we have is that by leaving gaps in the coverage, we have the same aliasing/sampling issues we get with video, as we're subsampling the sensor.
- Interpolate Raw. Interpolation is coming up with "new (faux) data" from the data that exists. We could generate "virtual photosites" by interpolating existing photosites. By looking at the 18 million green photosites, we could interpolate data for 6, 9, 12, 15 green photosites, basically any number we want. Then we do the same for red and blue. We get size reduction, but the problem is that we move away from a direct connection to the data. Our interpolation routines need to be very good and not contributing new artifacts.
- Grow TIFF. Since TIFF supports the smaller image sizes, the only real issue with selecting TIFF over raw is the loss of data as we cut 14-bit to 8-bits. Of course, even 8-bit TIFF files start out as bigger than uncompressed raw files (108.2MB TIFF versus 74.4MB NEF), so expanding the TIFF to even 12 bits adds 50% to our file size and makes even the smallest 12-bit TIFF larger than the smallest raw compressed file (42MB versus 29MB). Oops. Even if we could get as good as compression as the lossless NEFs get (44%), we're still at 23MB, only a 6MB savings from the original raw file. Again, gain doesn't outweigh loss, especially since the camera is going to need to spend time doing a lot of processing in this scenario, which could impact frame rates. Plus we've lost post processing of camera settings.
Other possibilities exist, of course, but each has some form of downside. The real question is whether that downside is actually worse than the upside.
* How does Canon do it in the 5D/1D models? They bin. But they do much more than just a straight bin. It's highly likely that they've patented what they do, though I can't find the patent in the US files.
On a D800 we can create smaller image sizes for JPEG or TIFF, only they're 8-bit, not the 14-bit data coming off the sensor. Nikon basically says "yes, the other downsides are all worse, so just use our demosaic" I'm not sure they're entirely right about that, as they're missing the point of having post processing correction, but you do have an interpolated answer if you need smaller files.
If you're really that worried about raw file size with a D800 (why exactly did you buy it if you are?), the best choice you've been given is 12-bit Compressed, which nets you a file size of 40% the full raw size, in exchange for a loss of some highlight data that rarely ever shows up in images and 2 bits of data per photosite.
- If you can stand the bit and post processing loss, shoot JPEG or TIFF and pick whatever size you need.
- If want smaller NEF sizes, shoot 12-bit, or shoot Compressed, or both.
- If your destination is an 8x10" print, shoot 5:4 format in the first place, as you'll save some space.
Of course, the interesting thing is that you can combine all three of those strategies, which is what I'd suggest you do if file size plus post processing capability is a concern.
Personally, while I understand the desire for "smaller raw," in practice it doesn't make a heck of a lot of sense to me. Since I'm a control freak with my images, anyway, if I'm going to start throwing away raw data or change it to interpolated, I want to be in charge of that process and decide what gets thrown away or changed. I'll throw away bit depth (12-bit versus 14-bit) before I'll throw anything else out, then highlight data (Compressed NEF) before letting anything else start distorting my raw data via interpretation I don't control. That nets me files almost half the size. Why do I need more reduction than that and still need raw?
Full disclosure: I'm not a wedding photographer. It seems that a lot of the argument for Small Raw is coming from wedding shooters. Perhaps I don't see their argument clearly enough, but perhaps they also don't have enough computer in the first place ;~). The usual complaint seems to be "I shoot too many raws to post process them quickly." Dropping from 21mp to 10mp (sRaw1 on a Canon 5DII) drops you from 25.8MB to 14.8MB a file. I'll be kind and say this is "half the size," so it should result in half the post processing time, right? Is there another way to reduce the post processing time? Well, I asked one of the folk who wanted sRaw about his computer: two generations old, 2GB of RAM running Windows XP. At that point I was afraid to ask the rotational speed of his hard drives ;~). It's difficult for me to understand why you need a new 36mp camera if you are going to try to stick with a 10 year computer. That seems like flying in a supersonic jet across the ocean and then trying to bike the last 200 miles and being disturbed that you're late. Yeah, I know: harsh. But here's the thing: if you're going to pony up to use the best, most megapixeled DSLR ever made, is whining your best choice? (Yes, I know that's hypocritical, because I'll probably whine about something in the D800 in my review. ;~) Still, Nikon tried to put out a camera that will do incredible things, yet some people are complaining that it won't do less incredible things. Indeed, if you really want 12mp raw, why not just shoot with a D700? This notion that one camera has to "do everything" is never going to be met, IMHO.)
Still Lots of Confusion III
April 4 (commentary)--I'm almost afraid to post this one. Shields up. Warp speed.
The high demand and limited supply for the D800 has devolved into some strange and off-the-mark comments and accusations. The system is indeed broke, but it isn't broke the way most people think it is.
Let's deal with supply and demand for a moment. Nikon's ability to produce D800's works out to about 30,000 units a month. Whether they stockpiled a week, a month, or two month's worth of supply before shipping isn't important to our discussion at the moment: it's clear that demand is well beyond whatever was stockpiled.
We've got two things we need to deal with first:
- How do you know demand in advance? Until the camera was announced there was only vague indeterminate demand, and that demand was equally for a D700s and D700x, as I've pointed out many times. The D800's announcement went further than anyone thought a D700x would go, and by a wide margin (new AF, new metering, new video capabilities, new features, way more megapixels). That increased demand for the D700x version even more. Then the early results started to show up from a few pre-release shooters and that goosed demand to insane levels of desire because it looks like a D800 is both a D700s and D700x all in one. We now have demand levels we haven't seen before for an FX body at launch. Meanwhile, the factory is already producing at their 30k/mo max. So demand won't be met quickly because it was underestimated.
- Do you blow it all in one production run? The Sendai factory only makes two cameras, the D4 and D800. These FX type of cameras historically have been relatively low in volume (hundreds of thousands a year, not millions). If you were clairvoyant (see #1) you could, I suppose, build a factory to produce every unit you needed at launch. That would be a far bigger factory than Sendai is. It would require another shift of workers to train. It would take a huge up front commitment to parts and a place to put them. You'd probably have to ship the final products all via air, which is expensive. Meanwhile, both the D4 and D800 are two-to-four-year cameras in terms of life cycle. As good as they are, demand a year from now won't likely come close to matching today's demand. So what does Sendai do a year from now with that expanded factory and extra shift of workers you needed for launch?
Now, I won't let Nikon completely off the hook for either #1 or #2. It was clear from my yearly surveys that demand for a D700 replacement of any type was high and growing. The new camera would need to be made in larger quantities than the D700 was made. The release at the last minute in Nikon's fiscal year also seems a little contrived to me. Things would have been smoother with another month's production build-up. And the reliance on trade shows for the launches put a fixed date for launch in place, too. In other words, Nikon committed to February launch and March ship in moderate quantities some time ago. Both definitely influenced the supply/demand balance.
Likewise, Sendai seems a little inflexible. The D4/D800 production numbers aren't that far off the old D3/D700 numbers. Nikon doesn't seem to be planning for the amount of FX growth that is already present, and will likely grow.
What's that? FX volume will grow? Yes, I think it's an inevitable response, both by customers and by camera makers. The APS/DX demand is slackening as the number of new users coming into those formats (traditional crop DSLR) have dropped off. We're in a mostly upgrade market now, so APS/DX has to offer something compelling for existing users to ditch their current, perfectly fine camera, and move up. That's getting more difficult, as there's nothing really wrong with the existing APS/DX bodies. The natural progression for a D7000 user isn't likely to be another DX body, but more likely an FX body. Why? Because the going from 16mp to 24mp doesn't produce all that much gain, and the 24mp users all want higher performance lenses than we've got (especially at the wide end). That the D800 is almost a perfect D7000 as well as an FX body makes it a natural upgrade point for D7000 users. Adding another FX body below the D800 makes as much, if not more, sense than adding another DX body above the D7000. In fact, moving the DX user to FX gives yet another upgrade path to push them on.
Still, even if Nikon had done better at #1 and #2, I'd bet we still wouldn't have supply meet demand. After all, the FX bodies aren't a mass market product like an iPad or iPhone, which sell in the tens of millions of units, but high end niche products that will barely break into the millions for unit volume this year and next. There's also a difference of price and sophistication. An iPad is maximized for production in volume, as it has minimal parts and labor costs. A D4 and D800 is maximized for quality in low volumes, and has way more parts and assembly issues.
It'll take a few months before supply is more in line with demand for the D800. That's especially true because the D800 is turning out better than people's expectations. It's quite a good camera. More camera than most people need, but one that has a huge potential for photographer growth, too. It's one of those buy-and-use-for-a-long-time products. Given that a lot of its buyers are older grew-up-with-film-DSLR folk, there's a chance the D800 will be a lot of users' last camera.
That said, the D800 will stay in high demand for awhile, is my best guess. It'll be a few months before D800's actually start to sit on a dealer's shelf for more than a week or two.
The other part of the supply/demand equation is the manner in which NikonUSA distributed what they brought into the country. The culprit here is the NPS (Nikon Professional Services) Priority Purchase program (I'll call it NPS PP from here on out). (Full disclosure: I used NPS PP to order my D4, D800, and D800E.)
It's arguable whether NPS PP is the right thing to do or not. On the one hand, these are professional cameras, and it behooves Nikon to get them into professionals' hands quickly. First, they'll learn more about any unforeseen issues faster, as these NPS PP cameras will definitely get used big time. Second, professionals are the center of word-of-mouth branding for Nikon. Without the pros on board, word-of-mouth can go from good to bad very quickly and that has ramifications down the line. Third, the pro cycles are centered around both known pro upgrade cycles and a big event: the Olympics. No pro wants to go to the Olympics with a last minute "finally got it" body--they want some time and experience with it before they go to the Big Show. Likewise, Nikon wants time to be able to iron out any unforeseen wrinkles before hundreds of those cameras show up in the photo boxes on the sidelines of the Olympic events. The Olympics are the big Red Carpet moment for cameras and have been for decades. That's one reason why the usual cycle is tailored to give them a full year between launch and the Olympics.
On the other hand, there is the argument as to why a pro's money is any different than an amateur's. Some of Nikon's most loyal customers are the serious enthusiast base, who've been buying and upgrading Nikon gear most of their life, and who really drive much of the volume of the higher end cameras, even the highest end (yes, even the "why is it so expensive" D3x).
Frankly, even if there hadn't been an NPS PP on the D800, we'd probably still be in mostly the same place: lots of people waiting for one. Moreover, the amateurs still waiting would be asking "so how good is it really?" and not getting nearly the answer they are today because a lot of pros they want to hear from would still be waiting, too. It's not like we pros can sit around in the middle of the day waiting for B&H to open up their pre-order queue (more on pre-orders in a moment), after all.
So what happened when NikonUSA released the initial D800's they received from Sendai into the sales channel? Well, the high quantity of NPS PP orders took their toll. Indeed, even the staff at B&H is a little flummoxed about what happened, as only a few D800's showed up in New York at the B&H offices. Instead, all of B&H's NPS PP orders were drop-shipped by NikonUSA from the Louisville warehouse (as opposed to shipping them to B&H and then B&H shipping them to their NPS PP customer). That took a bit of time off the transit, but it also meant all the B&H employees only saw a few boxes, and that appears to be shaping what they're saying to customers ("we didn't get many").
Yes, many local dealers got a small shipment of non NPS PP D800's or an additional non NPS PP unit or two. That has to do with the US regulations and law: you can't treat one dealer differently than another under the same contract. If you ship 5 boxes to Dealer A, you need to ship 5 boxes to Dealer B, C, D...n unless there is a clear bit of the contract that allows something different, and the way the contract allows that can't violate US fairness laws. From what I saw, I didn't notice any disparities between dealers of the same class: NikonUSA seems to be pretty good about trying to distribute fairly amongst any class of dealers (note though that NikonUSA has several levels of "dealers").
Which brings us to "pre-orders." Here's the problem with pre-orders: the company taking them (originally Amazon, but now several of the large dealers, including Adorama and B&H) has no idea how many cameras they're going to get and when they're going to get them. When there's huge demand for a product that's in short supply, it's likely that they won't be able to fill all the pre-orders they do get before they run afoul of the Federal Trade Commission's regulations on timely shipping. Worse still, because there isn't any penalty for pre-ordering, quite a few folk think that they can "up their odds of getting one" by ordering everywhere, waiting for someone to ship to them, then cancelling all the other pre-orders they made.
Both these things make the pre-order queue completely unreliable and unpredictable. In talking with one company that is taking pre-orders, the justification I got back is "we had to do it because our competitors are." Personally, I think everyone is going to get tarnished by the pre-order fiascoes that are brewing. Indeed, the cancellation of the affiliate programs here in Pennsylvania because of our stupid government's unconstitutional move coincided with the D800 pre-order queues opening (yes, the whole affiliate nexus thing was just ruled by yet another US District Court--this time in Colorado--as being unconstitutional, for the very same reasons that the Supreme Court originally decided the issue in Quill v. North Dakota. It's about time someone held these politician's that think they know better what the law actually says up for judgment). Thus, I was able to (at least partially) avoid stepping into the pre-order fiasco with both feet myself.
Indeed, I'm going to take a moral stand here. I still have one affiliate program who's sticking it out in PA and the others will be back once Pennsylvania's not-interested-in-small-business Governor gets spanked hard enough. Starting today, I'm not going to post any more "pre-order" links. Not without assurance that the company in question knows how many units they'll get and stops their pre-order queue with that number. This whole pre-order thing has turned into "we'll take your order even though we have no idea if and when we can ship it, but we're afraid that if you pre-order from someone else they may actually get lucky enough to deliver it and thus we won't profit from you." That's pure, unadulterated balderdash. It doesn't respect the customer at all.
I ask all of you to ask those companies asking for your pre-orders a few questions: (1) Do you know how many units you'll get? (2) Do you know when you'll get them? (3) How many pre-orders do you currently have? (4) Why shouldn't I just pre-order with everyone and see who delivers first?
And here's the big kicker, because they'll have to answer "no" (otherwise they're lying to you): (5) Can you guarantee you'll deliver my order in the delivery period the FTC requires?
Note that the dealers aren't the only ones guilty in the pre-order game. The camera companies should have cut this practice off at the pass a long time ago. Of course, to do that, the camera makers would have to be more forthcoming to their dealers in how many units they'll get and when, and then deliver on that. After the quake, tsunami, and flood, the camera makers are all leery of any promise of delivery. But you know what? Open, timely, and clear information flow solves all wounds. Not a single soul would get upset if a week before a quake they got a message "10 units coming in 14 days" and after the quake they got another message saying "units expected will not arrive, more details on shipment when we know them."
Bottom line: we have lots of lemmings causing chaos. The lemmings of "gotta have a D800" folk who don't really need it but want it because it's the hot product of the day. The lemmings of dealers taking pre-orders because the other lemmings are doing it. The lemmings of the camera companies who all use the same announce-and-release tactics. Don't be a lemming.