DX versus FX (Again)

(commentary)

Someone asked me via email why they should buy a D600 instead of a D7100, which is less money. After all, they're both 24mp and have much the same feature set and build quality. 

Now technically, Nikon's Marketing Department ought to be the one answering this question, as Nikon is the one that should benefit from people understanding the difference. But what the heck, you all know that I'm not going to back off from doing their job when they fail at it ;~).

Unfortunately, things are a bit more confusing with digital these days than with film. With film the underlying grain structure was the same on 35mm and medium format (MF) film stocks. Thus, if your goal was to produce a 36" print, because you had to magnify the 35mm capture more than the MF capture, the grain also became more visible in the 35mm print, as it was magnified more. 

In digital--in particular with the D7100 versus D600 comparison--that low-level detail magnification problem is no longer quite true. Both DX and FX can capture the same number of discrete elements at the core (e.g., 24mp of data across the frame). In theory, if you use an equivalent lens and then magnify a base ISO DX (D7100) result to 36" you should get pretty much the same pixel-to-print relationship as when you magnify the base ISO FX (D600) result. Each feature you're magnifying in the image has the same number of sample spots (pixels) in it, after all. 

Some of you wonder why I've been harping on Nikon's DX lens absentees so much in the past year. Well, it's simple: we've now moved past the bar where the larger sensor is so clearly better than the smaller one that you'd always pick the larger one. If you don't believe me, go back and compare a Canon 1Ds image (full frame, or FX equivalent) with both a Nikon D100 and a Nikon D7100 image (both DX). Almost everyone would take the D7100 image over the 1Ds or D100. Thing is, our sensors today are very good at capturing and correctly converting every photon received into a digital bit count. You have to really stress the system (low number of photons due to low light) in order to see visible, meaningful differences between a DX and FX sensor with the same pixel count. For most of our picture taking, both the D7100 and D600 have more than enough dynamic range and noise avoidance. It's only in low light that any meaningful difference surfaces.

So the first of the marketing messages is this: all else equal, FX probably gives you another stop of high ISO capability over DX. By that I mean that--assuming everything has been managed perfectly equal, which isn't always the case--when we make a 24" print from a DX body at ISO 3200 we should probably get visibly indistinguishable results from an FX body at ISO 6400. This also ties into one of my sub-themes lately: we have fast FX lenses, but we don't have fast DX lenses, so the differences tend to be actually higher in practice because we have FX folk shooting ISO 6400 at f/1.4 and DX folk shooting ISO 3200 at f/2.8 or higher. Still, the marketing message regarding the body is the same: FX is a better choice for very low light work, while the DX system should hold its own against FX in almost any other amount of lighting.

Another aspect of the DX/FX difference generally doesn't get a lot of mention, though: pixel density. The DX body with 24mp packs smaller pixels than a FX body with the same count. That has other implications than just light collection. It has lens implications. Yes, I'm back on my repeated "where are the DX lenses" complaint. 

Modern lenses resolve quite well, so let me explain this a bit more theoretically than practically. Imagine a lens that can only resolve lines about 6 microns in size (that would be one photosite wide in a 24mp FX sensor). What happens when you put that lens on a camera that resolves 4 microns in size (the photosite of a 24mp DX sensor)? Right, the lens isn't up to the capabilities of the capture system behind it and you'll be getting a little bit of edge blur as discrete lines spill over into part of an adjacent pixel. Moreover, let's assume that you're handholding and you move the camera 1 micron while shooting your image. In FX we've got a 17% pixel blur, in DX we've got a 25% pixel blur. (Remember, this is all theoretically arbitrary to help you understand the point I'm trying to make; in practice things are much more complicated.) 

Technically, the FX camera is a more "relaxed" system than the DX camera when it comes to critical sharpness: FX can use lenses with a little less resolution than DX (all else equal), and you can handle the FX body a little more sloppily and get the same results you'd get in DX being more careful. While this may surprise you, this is not really different than the way it was with film. Medium format cameras could get by with lenses that were a little less resolution capable than 35mm cameras. That's not exactly the way people tended to use them, though: most medium format film users were trying to make bigger prints, so it was really the magnification issue that drove most of the 35mm/MF comparisons. As long as you didn't compromise the lens quality/handling in MF more than the extra magnification of 35mm, you were still better off, and if you didn't compromise those things at all, you were far better off.

Some of you were surprised at my D600 Lens Set comments recently. To some degree, the difference in needs at the photosite are part of that: the D600 is a bit more forgiving than the D800 in that respect. But we're talking DX versus FX here, so let's get back to that: the D600 is a bit more forgiving at the pixel level than the D7100 when you shoot equivalently from the same spot (e.g. D7100 with 200mm f/2 and D600 with 300mm f/2.8). 

Here's the second marketing message: if you're buying solely for pixel count, you need better lenses and shot discipline with DX to deliver the same optimal pixel data that you would get with FX. Not by a lot, but every little bit can hurt you if you're not careful. Plus, what you save on camera body may be lost in lens, support, and other costs as you try to drive your quality up as high as the camera itself is truly capable of. Of course, this is why Nikon's marketing department isn't saying anything here: it's a negative message, and one made worse by the fact that Nikon doesn't have many better DX lenses to sell you, doesn't sell tripods, and isn't in the photography instruction business. Thus, they'll mention the higher number ("now with 24mp!") but stop at telling you what the consequences of that might mean if you're quality driven. 

Of course, not everyone is printing large and trying to extract every last iota of quality out of their gear. What if you never really output to anything bigger than an HD TV? That's 1920x1080 pixels, which on a 24mp camera suggests that you'll downsize each pixel on the horizontal axis by about 3 (i.e. take 6000 pixels and make them 1920). Does it now matter how good the lens is, how much you moved the camera during the exposure, or even how noisy the high ISO results are? Not really. All those things tend to dive under the surface to invisibility when you downsize like that. 

So the third marketing message is this: For email, social site use, Web display, and even slide shows (we don't use slide projectors any more, we use HD video outputs), you're not going to see a difference between DX and FX, so buy what you can afford. Hmm. That would raise DX sales and lower FX sales if Nikon made that marketing message, and the profit margins on FX are higher, so any guesses as to why they don't trumpet this message? ;~)

Most people asking the D7100 or D600 question are thinking mostly along the lines of image quality without really paying attention to what their needs are. For me, the question gets back to pixel density: do I need to put more pixels on a distant subject? The DX body allows me to do that at the sensor (smaller area with same number of pixels), the FX body requires me to buy longer lenses (by 1.5x the focal length). This is why a lot of sports and wildlife photographers on a budget all loved the D300. While the D3 was clearly better in the low light realm, it was more expensive and pushed people 1.5x upwards in lens focal length from the D300 (and until the 800mm arrived, there wasn't always a 1.5x longer lens available). Both cameras had 12mp, so it wasn't about pixel count. It was a simple tradeoff: low light performance versus pixel density. If you could tolerate the D300's low light performance, the cost equation favored it. If you couldn't, you had to bite the bullet and pay for the D3 and longer lenses.

The D7100 and D600 have the same relationship, only not quite as separated in cost as the D3 and D300 were, nor are they quite as separated in low light performance, either. Still, costs add up if you're not careful. But again, this brings us back to lens choice. At the consumer zoom level, the D7100 has plenty of decent choice: 10-24mm, 16-85mm or 18-105mm, 70-300mm. This matches up reasonably well against the D600's 18-35mm, 24-85mm, and again the 70-300mm (though the crop of the D7100 sensor makes that lens have more "reach" on that camera for the pixel density crowd). 

But here's the strange part: look at the 10-24mm versus the 18-35mm, for example. The FX lens is a little cheaper than the DX lens. Ditto the 16-85mm versus 24-85mm (and over Christmas, the 24-85mm was essentially free if you bought a D600, a price that DX couldn't match even with the 18-105mm kit lens). The fully-burdened system cost equation between DX and FX has never been quite so close. Still, US$800 is US$800 (and over time I expect the body cost difference to probably average closer to US$1000). That's another lens you could buy, or a very good tripod and head, or a lot of photo instruction.

Which brings me to the last marketing point: the D7100 is the "value" product in this comparison. Maybe not as big of a difference in value as in the DX versus FX past, but still, it's essentially a slightly improved D600 body with a DX sensor at a lower price. 

Realistically, though, what's happened in this generation of cameras is that we've gotten four products that are not hugely apart in terms of their features and performance. The 24mp D3200 is the "stripped down" product, the 24mp D5200 is the "modestly equipped" version, the 24mp D7100 is the "fully loaded but with the smaller engine" product, and the D600 is the same thing with the bigger engine. We're exactly where we are with many automobile product lines: enough choice to confuse you and put you at the mercy of the clever salesperson, who will almost certainly make some attempt to up sell you. 

So here are the key differences between a D7100 and D600: US$800, one stop difference in high ISO shooting, 1.5x difference in pixel density on a distant subject with the same lens. One of those, or some combination of those, is what should determine which camera you buy. Be realistic, though. Are you really always shooting at ISO 3200 and 6400? Do you often go out and try to do distant wildlife shooting? Could you use US$800 in your bank account for something else? Answer those questions before anything else.

Yes, there are some other differences I've been ignoring: the breadth of the focus sensors across the frame, the perceived brightness of the viewfinder, buffer size differences, and a few others. But I think it unlikely that those things are really driving your primary decision. These might tip you the rest of the way, but it's those three things I mentioned in the previous paragraph that are they key to your decision. 

text and images © 2017 Thom Hogan
portions Copyright 1999-2016 Thom Hogan-- All Rights Reserved
Follow us on Twitter: @bythom, hashtags #bythom, #dslrbodies