The Myth of the $500 FX Sensor

Bubble defects in a silicon wafer — SEM image
Bubble defects in a silicon wafer — SEM image

Disclaimer: I am not an electrical engineer and have no special knowledge about any of this.

Some time ago Thom Hogan estimated the cost of an FX camera sensor to be around $500 (I don’t have the reference, but I’m pretty sure this is true since he said as much recently in a comment thread). Similarly, E. J. Pelker, who is an electrical engineer, estimated an FX sensor to cost around $385 based on industry standard cost and defect rates in 2006. So it seems like there’s this general acceptance of the idea that an FX sensor costs more than 10x what a DX sensor costs (Pelker estimates $34 for a Canon APS sensor, which is slightly smaller than DX, and $385 for a 5D sensor).

My assumptions can be dramatically off but the result will be the same.

E.J. Pelker

I don’t mean to be mean to Pelker. It’s a great and very useful article — I just think it’s not that the assumptions he knows he’s making are off, it’s that he’s made tacit assumptions he doesn’t realize he’s made are completely and utterly wrong.

The assumption is that if you get an 80% yield making DX sensors then you’re get a 64% (80% squared) yield from FX sensors (let’s ignore the fact that you’ll get slightly fewer than half as many possible FX sensors from a wafer owing to fitting rectangles into circles).

Here are Peltzer’s “unknown unknowns”:

Sensors are fault-tolerant, CPUs aren’t

First, Peltzer assumes that a defect destroys a sensor. In fact if all the defect is doing is messing up a sensel then the camera company doesn’t care – it finds the bad sensel during QA, stores its location in firmware, and interpolates around it when capturing the image. How do we know? They tell us they do this. Whoa — you might say — I totally notice bad pixels on my HD monitors, I would totally notice bad pixels when I pixel peep my 36MP RAW files. Nope, you wouldn’t because the camera writes interpolated data into the RAW file and unless you shoot ridiculously detailed test charts and examine the images pixel by pixel or perform statistical analysis of large numbers of images you’ll never find the interpolated pixels. In any event (per the same linked article) camera sensors acquire more bad sensels as they age, and no-one seems to mind too much.

Sensor feature sizes are huge, so most “defects” won’t affect them

Next, Peltzer also assumes industry standard defect rates. But industry standard defect rates are for things like CPUs — which usually have very small features and cannot recover from even a single defect. The problem with this assumption is that the vast majority of a camera sensor comprises sensels and wires hooking them up. Each sensel in a 24MP FX sensor is roughly 4,000nm across, and the supporting wiring is maybe 500nm across, with 500nm spacing — which is over 17x the minimum feature size for 28nm process wafers. If you look at what a defect in a silicon wafer actually is, it’s a slight smearing of a circuit usually around the process size — if your feature size is 17x the process size, the defect rate will be vanishingly close to zero. So the only defects that affect a camera sensor will either be improbably huge or (more likely) in one of the areas with delicate supporting logic (i.e. a tiny proportion of any given camera sensor). If the supporting logic is similar in size to a CPU (which it isn’t) the yield rate will be more in line with CPUs (i.e. much higher).

This eliminates the whole diminishing yield argument (in fact, counter-intuitively, yield rates should be higher for larger sensors since their feature size is bigger and the proportion of the sensor given over to supporting logic is smaller).

(Note: there’s one issue here that I should mention. Defects are three dimensional, and the thickness of features is going to be constant. This may make yields of three dimensional wafers more problematic, e.g. BSI sensors. Thom Hogan recently suggested — I don’t know if he has inside information — that Sony’s new (i.e. BSI) FX sensors are turning out to have far lower yields — and thus far higher costs — than expected.)

Bottom Line

To sum up — an FX sensor would cost no more than slightly over double a DX sensor (defect rates are the same or lower, but you can fit slightly fewer than half as many sensors onto a die owing to geometry). So if a DX sensor costs $34, an FX sensor should cost no more than $70.

DX DSLR: IQ is tired; UX is wired.

Acronyms FTW!

Canon and Nikon have now both released their “high-end entry” DSLRs incorporating essentially everything they know about sensors, image processing, video, and what have you in an entry-level body. The big differentiators between $799 (for the D5100) and $1799 (for the 7D) are viewfinder size and accuracy, fairly abstruse things like micro-focus adjustment, “pro” features such as weather sealing, and marginalia such as continuous shooting speed and buffer size**, and number and type of autofocus sensors. Nikon has even included 14-bit RAW support in the D5100 (so its Dx0mark sensor score may equal or exceed the D7000’s, whereas the D5000 was slightly outscored by the D90).

This essentially marks the point at which Nikon and Canon have given up trying to use basic image quality as a differentiator between APS-C models. (Ken Rockwell has been arguing for years that Nikon’s best IQ — modulo sensor size — is always found in its most recent camera, regardless of price, unless it’s the D3000). But Nikon and Canon now aren’t even using megapixels or dynamic range*  as a differentiator  for imbeciles and enthusiasts — respectively — now. You can compare them here — this should link to dpreview’s D5100 High ISO page — to my eye, the D5100 is as good as the D7000 and K5, and better than every competing APS-C or Micro-four-thirds camera.

* Note that while Canon has offered 14-bit RAW support for years as standard feature throughout its range, the measured dynamic range has lagged behind Nikon’s 12-bit pipeline in actual testing.

** According to Nikon’s product information (unless it’s changed since I linked it) the D5100 has a 16 frame RAW shooting buffer. This would make its shooting buffer larger than the D7000’s (and buffer size is/was perhaps the single biggest criticism of the D7000).

Not coincidentally, the low-end cameras in both ranges are now, unambiguously, the best DSLRs offered by either vendor for shooting video. The D5100 even supports 1080p30 , which begs the question as to whether the D7000 will get a firmware upgrade to match it.

So, that’s the line in the sand. From now on DX cameras will compete, as did their analog forebears, on user experience, something neither Canon nor Nikon are really much good at.

Aside: Pentax seems to be more thoughtful about DSLR design than either Nikon or Canon. (It also does a better job of delivering enthusiast lenses.) For example, Pentax introduced an exposure mode where the photographer picks Aperture and Shutter Speed and the camera picks ISO (which is so obvious) which the others haven’t picked up on. Similarly I believe there’s a program mode where one dial selects ISO and the other works like the dial in Program (or “Professional” as Rockwell puts it) mode — adjusting Aperture and Shutter in opposite directions. Both modes make more sense than any exposure mode on a Canon or Nikon.

Olympus E-P2 Announced


Olympus has announced a newer (and more expensive) successor to its first micro four-thirds camera, the E-P1. From my point of view, the key features of the new camera are electronic viewfinder support (hotshoe-mounted — very similar to the GF-1), continuous-tracking autofocus (will this make it suitable for photographing toddlers? according to dcresource the E-P1 and E-P2’s focusing have both improved since initial release — thanks to firmware upgrades — but both remain lackluster compared to the GF-1), and the option to shoot video with full manual control. Aside from these two new features, the new camera is black — sad, because I think the new model is markedly less attractive — and has a few irrelevant gimmicks.

Along with the new camera, Olympus will be offering a 9-18mm (18-36mm) wideangle zoom, and a 14-150mm (28-300mm equivalent) superzoom, making Micro Four-Thirds a pretty complete ecology (and quite likely offering superior optics at a given price-point than Canon or Nikon). This will leave Nikon and Canon sandwiched between Panasonic and Olympus (offering superior compact cameras) and Leica (offering superior high-end cameras). If Panasonic and Olympus can only get some camera bodies out at a reasonable price, they could do some serious damage.

While these may seem pretty minor changes, assuming the continuous-tracking autofocus doesn’t suck, it significantly changes things relative to the GF-1, since the E-P2 has almost everything (except a built-in flash, which I don’t care about) that the GF-1 offers along with image stabilization and better controls. In the end, however, this is really still a first generation product. I imagine the true second generation micro four-thirds cameras will be truly compelling.

The E-P2’s SRP of $1100 (which includes the new EVF) is steep, which remains a deterrent.



It’s interesting to see Canon backing off its relentless pursuit of the megapixel, with the 1D Mkiv and G-10 both signaling that maybe it’s time to concentrate on pixel quality. (What’s with the zany crop factor of the Mkiv though? I’d have thought we’re past that kind of odd compromise.) It’s also interesting to see Nikon persisting with 720p-only cameras — how hard would it be to offer 1080p @24fps?

It seems to me that Canon right now has two very compelling cameras (the 7D and 5D mkii) while Nikon’s range is starting to look dated. (The 7D’s hefty introductory price tag does manage to make the 300s seem reasonably priced though.) Nikon’s cameras are all great still cameras, but their half-assed video support seems more like an attempt to tick a checkbox than offer serious functionality.

Right now, I want Panasonic, Olympus, and Leica to give Nikon and Canon more serious competition, because it seems that they’re all just milking their customers — especially early adopters. (It’s amazing that Leica’s cameras seem reasonably priced compared to Nikon’s. Indeed, the Leica X1 looks pretty competitive with the E-P1/E-P2/GF-1 coupled with the (~$500) 20mm f1.7. If you don’t need autofocus, full frame Nikon prices look ridiculous.)

Look on my works, ye mighty… Part 2

There’s a very thorough review of the new (and as yet unobtainable) Panasonic GH-1 just posted on Edit: and Cameralabs has just posted their review as well (link is to the verdict — it’s worth noting that after carefully managing to give all the current generation of contenders — D5000, XTi, E620 — identical scores of 88%, the GH-1 scores 89%).

Here's an unretouched 100% crop from an ISO 3200 sample shot
Here's an unretouched 100% crop from an ISO 3200 sample shot. (If you're not into the minutiae of digital photography and for some reason you're still reading this blog entry, a digital photo taken at this sensitivity until the current generation of cameras looked like absolute garbage.)

You can find more (still) samples here.

Here’s the bad news:

  • It’s still listed at $1499 (which is not unreasonable given it comes bundled with a very good 14-140mm lens, but still)
  • Its electronic viewfinder is apparently a bit unpleasant to use in low light conditions (not sure how low — need to play with one)
  • Its continuous shooting rate is 3.3 fps (roughly the same as a Canon 500D/XTi, but not as good at the Nikon 5000D). Edit: Camerlabs notes that the EVF stops updating between shots during continuous shooting, which is actually a pretty crippling deficiency.
  • The built-in Flash is kind of lame (oh right, I don’t use flash)

Aside from that, the news is all good. Shockingly good. Notably:

  • Better detail (JPEG or RAW) than the Canon XTi/500D. (Brief pause while you adjust your lower jaw.)
  • Equal or even slightly better low light performance (JPEG or RAW) than the Nikon D5000. (Damn, it’s just not going to heal, is it?) Edit: Cameralabs low light tests (which are both more consistent and realistic than dpreviews) favor the Nikon D90 (which performs identically to the D5000) slightly over the GH-1.
  • Much better at avoiding “jello” effect than any other DSLR (apparently its circuitry is optimized for rapidly grabbing frames from the sensor)
  • 1080p at 24fps, 720p at 60fps (AVCHD) or 30fps (MJPEG). 720p footage described in the review as “broadcast quality” and favorably compared to footage from semipro camcorders.
  • Viewfinder size is roughly similar to Canon 1DS Mk III (i.e. full-frame pro DSLR)
  • Generally excellent UI, handling, menus, customization, physical dials
  • No “video mode” — dedicated video button is always available

So, we’re here. Panasonic appears to simply be better at doing image processing than Nikon or Canon (how else to explain equal or better image quality from a smaller sensor?), and they’ve produced a camera that is simply a better still camera than Nikon or Canon can (currently) make (bear in mind that the image quality on the D5000 and 500D/XTi is, in essence, as good as anything Canon or Nikon can offer, modulo sensor size) and is actually a credible video camera (versus a half-assed video camera with sub-par frame rate, jello effect, and manual focus while shooting). All in a camera that’s smaller than any DSLR on the market and has the most flexible lens mount there is*.

So, now all we need to do is wait for the GH-1 to actually appear in stores and the price to drop. (Then there’s the Pentax K-7 just around the corner, which is quite small and weather sealed.)

Note: * micro four-thirds cameras can currently use any Panasonic, Olympus, or Leica four-thirds or micro four-thirds lens, and Olympus OM lenses; there’s no reason not to expect adapters for Nikon, Canon, and Pentax mounts in the near future thanks to the micro four-thirds design. Of course you will probably need to focus manually with everything except the micro four-thirds lenses.

Olympus E-P1

At last, a small camera that takes DSLR quality images.
At last, a small camera that takes DSLR quality images.

Olympus has finally announced its first micro 4/3 camera. Unlike the Panasonic GH-1 it “only” does 720p video (but the samples I’ve seen are gorgeous), has no viewfinder or built-in flash, has a low resolution (but large) fixed LCD, is actually quite small, and is priced competitively (MSRP $699 body only, $799 with a 14-42mm Zuiko lens).

One really intriguing feature of the camera is that in addition to being able to share lenses with all other micro 4/3 cameras (e.g. Panasonic’s 14-140mm and 7-14mm lenses, and various Leica lenses), it can use normal 4/3 lenses via an adapter, and it could potentially use pretty much any 35mm or APS-C (or even Leica M-series) lens via adapters (Olympus is providing an OM-system adapter for starters). This makes it almost the perfect “backup” camera for serious photographers since it could easily share lenses with any other camera one happens to use.

This camera is almost pocketable, offers similar low-light performance to the Canon T1i (i.e. slightly inferior to the D5000/D90, slightly better than the GH-1), has built-in image stabilization, dual control dials, full manual control, all-metal construction — indeed the same feature set as Olympus’s E-620 and E-30 DSLRs. And it looks lovely.

If it had a viewfinder, or even a higher resolution LCD, I think it would be ridiculously compelling. If, as many expect, its street price quickly drops below $400 it will be ridiculously compelling anyway. I, for one, don’t much care about the lack of flash, since I avoid flash (especially built-in flash) as much as possible.

Oh, and it’s perhaps the most physically beautiful digital camera I’ve seen.