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.

Sony’s sensor dominance

iPhone 4S Camera Module (via Chipworks)
iPhone 4S Camera Module (via Chipworks)

Is Sony building a sensor monopoly?

It’s beginning to look like Sony is positioning itself as the Kodak of digital photography. Or perhaps the Intel.

Virtually all the exciting new cameras around are sporting Sony cameras — including Sony’s own RX-1, the Leica M, the Olympus OM-D EM-5, and the D-600. The new Leica is almost certainly using a Sony sensor. Fuji’s XF system seems to be based on the 16MP APS-C sensor with a custom color filter. Does the GH-3 use a Sony sensor too? And, if not, will it be competitive? I’d suggest that Panasonic relying on Sony sensors would be pretty disturbing given that for a long time Panasonic has been Sony’s only credible rival in video.

Now and for several years Sony’s sensors have dominated DxOMark’s rankings (perhaps Sony has cottoned onto DxOMark as the most cited sensor benchmark and is optimizing its sensors accordingly).

It’s also worth noting that Pentax’s K5 and K30 both use the same Sony sensor used in the Nikon D7000. And, for good measure, Sony has just invested a bunch of money in Olympus, giving it some kind of stake in Micro Four-Thirds. The camera module in the iPhone 4S is known to be from Sony, and it’s highly likely that in the iPhone 5 is also. (Who makes the camera modules in Nokia’s various “PureView” branded phones?) It seems like the only major holdouts are Samsung (who I assume are busily trying to clone Sony’s sensors) and Canon (and I for one am not going to buy a Canon DSLR to help prop them up until they give up on their ridiculous control layout).

The digital photography market is not an easy place to be right now. Smartphones with ridiculously good camera modules are eating out the ground beneath it, and at the high end Sony is cheerfully selling very nice sensor modules to everyone and letting them all kill each other. Sony (and Minolta) have never managed to dominate this market (aside from a brief period where their high-end point-and-shoots were all that enthusiasts could afford), so a chaotic melee where everyone ends up weakened and dependent on Sony suits them just fine. In this context, Sony’s injection of $400M into Olympus makes a lot of sense. In a world where Sony were trying to make NEX dominant it wouldn’t make sense to prop up NEX’s most credible competitor, but in a world where Sony just wants everyone else weak or dead it makes perfect sense.

Of course, Sony has been losing money for a couple of years now, so in order to capitalize on its success in the camera market (where it still makes money) it needs to get the rest of its house in order.

Sony’s Pellicle Cameras

Salient features of the new Sony pellicle cameras
Salient features of the new Sony pellicle cameras

So, as the rumor mills asserted, Sony has released four new interchangeable lens cameras with DX sensors, two of them DSLRs and the other two pellicle cameras designed to look like DSLRs. The new DSLRs are essentially very similar to Sony’s previous generation DSLRs (with the weird hybrid focusing trick for live view) while the pellicle cameras are actually very interesting. The new SLTA55 is already reviewed in depth at dpreview, so check that out. (The reviewers were unable to process the RAW images as of writing this so we won’t know just how good the low light/high ISO performance is with RAW, but just based on the JPEGs it looks like Sony has finally shaken off its IQ issues (i.e. the excellent performance of the NEX cameras was not a fluke).

Sony’s new pellicle cameras offer continuous live view (because the sensor is always getting 70% of the incoming light) and continuous phase-shift autofocus (i.e. DSLR-quality/speed autofocus) because the autofocus sensors are always getting 30% of the incoming light. It also means you don’t get an optical viewfinder, but instead get a 1.4MP EVF (which is better in some ways than an optical finder, and worse in others). If the new cameras have one significant weakness, it is that their continuous shooting seems somewhat compromised (essentially the live view — which is the only view — seizes up during 6fps or 10fps continuous shooting, and in 10fps mode you have no control over exposure).

And I should add that Sony has managed to seriously reduce the new cameras’ dimensions without significantly impacting usability. (Indeed, the Sony bodies are barely larger and heavier than the Panasonic G2, which is remarkable given the design constraints.)

So, to summarize, these cameras clean the Panasonic G/GH series clocks (especially since they have sensor-shift image stabilization), and give conventional DSLRs a run for their money as still cameras. If it weren’t for the new cameras’ continuous shooting compromises I would find them almost irresistible.

And the Nikon D3100‘s reign as the only DSLR-class camera capable of shooting 1080p video while continuing to autofocus was not only very short-lived (i.e. about two days) but pretty much thoroughly outclassed (given how much better phase-shift autofocus tends to be than contrast-detect).


The a33 (pellicle) and a560 (relatively conventional DSLR) side-by-side size comparison
The a33 (pellicle) and a560 (relatively conventional DSLR) side-by-side size comparison

I took two photos from dpreview and scaled them so their lens mounts matched and here’s the result. The a33 is markedly smaller than the a560, despite sharing the same sensor and image-processing pipeline (and being announced on the same day) so this is the immediate size-saving Sony obtained by ditching the conventional SLR mirror. That’s pretty amazing.

Note that the new “conventional” DSLRs from Sony offer better continuous shooting (7fps with an optical viewfinder) but lack the focus-while-filming video (they still offer 1080p). If you recall, Sony’s “conventional” DSLRs are already pretty radical, having a bunch of wacky design tradeoffs (a smaller, dimmer optical viewfinder) in exchange for faster phase-change autofocus in live view. I’d have liked to see more differentiation between the new pellicle cameras (which I see as compromising still shooting in favor of video) and the “conventional” DSLRs, but perhaps we’ll be seeing a refresh of the a850 along those lines (and, after all, at $1950 the existing a850 is a serious bargain, assuming you’re willing to hunt for second-hand lenses or wait and hope for Sony to release more).

Even More

Interesting criticism of the a55 (et al) here, mostly along the lines that video capabilities (1080i, low bitrate, and no manual controls) in the new cameras is seriously (and intentionally) crippled. Steadishot makes the camera overheat after 9 minutes. The AF only works when the aperture is wide-open, video bitrates are no better than compacts. I have to say I find interlaced HD video to be actively offensive (it’s a holdover from analog and has no place in a modern pipeline).

And more technical criticisms here, along the lines that marketing futzed with the specs, the onboard computer hardware is laggy in general and specifically the way the camera seizes up after shooting (if image review is switched on) is a major bummer. These are finicky, but not minor criticisms, since one of the two main reasons for using DSLRs over compacts is handling.

There’s also some discussion as to “ghosting” caused by internal reflections in the mirror (which are unavoidable). This kind of thing could be post-processed out of the image (in-camera) since the effect is very consistent but it’s another drawback of the fundamental design. (And it’s quite noticeable in some of the examples in the thread.)

Digital Photography in Flux

It seems to me that this is a terrible time to buy a new digital camera, certainly a terrible time to buy anything other than a sub-$200 point-and-shoot.

Mid-range compacts ($200-300) are going to be made obsolete pretty soon as the high-end compacts ($300-500) are heavily discounted and then their features are moved into the more inexpensive models. Why? Because DSLRs are starting to sell for under $500 and the Micro Four-Thirds system is going to make small sensor compacts look like the garbage they are.

High-end compacts, such as the Canon G10, are expensive but fundamentally inferior to low-end DSLRs in image quality. The only reason you would buy one of these cameras is because it’s smaller than a DSLR, but they’re still not exactly compact and their image quality is only a small notch above cheaper cameras. New Micro Four-Thirds cameras will be as small or smaller, offer interchangeable lenses, and (very near) DSLR-level image quality.

Meanwhile low-end DSLRs are getting cheaper as they face competition from Micro Four-Thirds compacts, each other, and heavily discounted mid-range DSLRs (e.g. the street price of the Nikon D80 body is around $600 now — that makes a Nikon D60 or Canon XSi at a similar price just look silly).

As we get to the mid-range and high-end DSLRs we have a bunch of cameras that are already obsolete, except (arguably) for the Sony A900 (with its new 25MP full frame sensor, but it’s a Sony…) and the much hyped Canon 5D mkII (with its outmoded AF). Nikon’s D300, D700, and D3 all are beginning to look a little behind the times (they were introduced with fairly low pixel counts, and the market has just accelerated away from them). Canon’s new low-end range has the main virtue of being (relatively) cheap. The D90’s innovative video mode is already looking like a beta-quality product (Costco is taking pre-orders for D90 kits and the D90 isn’t even out yet). I love Nikon, but the D90 is looking like a loser. It has a video mode that doesn’t quite work, 12-bit image processing, and a price point sandwiched between the Canon 40D and 50D, both of which beat it on build quality, responsiveness, and still image quality. The sad thing is that while the 40D and 50D make the D90 look silly, they’re also both obsolete. Canon has already announced that the 50D’s successor will probably feature video capture.

Some time in the next year we’re going to see:

  • Micro Four-Thirds system cameras with in-body vibration reduction, HD-video capture, and a selection of superb lenses that don’t lock you to a single high-quality vendor (i.e. you’ll probably be able to buy Leica and Zuiko lenses for them at a minimum). At last we’ll be able to buy a pocketable camera that takes photos that are nearly as good as anything we could take with a “full-sized” camera.
  • Reasonably-priced Canon or Nikon bodies with HD-video capture that doesn’t suck, 15+ MPs (with no appreciable image quality reduction), usable ISO 6400 (the D90 is damn close), 14-bit image processing, and — just possibly — in body vibration reduction. I wouldn’t be surprised if full frame sensor cameras appear in the $1500-2000 price range (heck, Sony will probably be there with the A900).

Obviously, if you’re a pro, your camera pays for itself, and next year’s camera doesn’t matter. But for most of us, we don’t get to drop $1000 on a camera every year, and this seems like a truly awful year to buy a camera.

On top of all this, the economic downturn is going to force camera companies to cut into their hefty margins on higher end gear to keep volumes up. Once consumers become used to $300 entry-level DSLRs, they’ll never be able to bring the price back up (nor should they).

Sony Online Entertainment to Act as Honest Broker for Item Sales

Sony has decided to facilitate the sale of in-game items for real-world cash in EverQuest II. While lamentable, this is hardly surprising. They argue that 40% of their support issues have to do with such trade or activates related thereto.

While this is an interesting development, and there will be much discussion of it on the level of “this is a bad thing”, “this is a good thing”, or “this is regrettable but necessary”, as usual the root causes have to do with long held and unquestioned assumptions about what a “role-playing game” is, or aspires to be.

Leaving aside the fact that the only reason, in theory, you play an online game like EverQuest II is to feel some kind of empathy for your character, and that the concept of buying cool stuff using real world cash to make your character more uber so violates this concept that it makes you wonder why you’d play the game at all (or maybe for some people it’s more like accessorising Barbie). E.g. why not simply pay Sony to make a monster you can’t kill drop dead? Why not simply pay to make your character level 75 (or whatever the mostest highest level is)? This is entertainment, right? Would you watch a TV show where the hero escapes from scrapes by bribing the scriptwriter? Well, maybe if it were a comedy about the TV industry…

The root assumptions underlying all of this baloney are simple and can be explored by showing exactly how well the games industry (paper and computer) has utterly failed to provide the experience it has always aimed for in RPGs.

The quintessential source book for RPGs is The Lord of the Rings. The entire milieu of the D&D world and that of its imitators springs from The Lord of the Rings. This applies to assumptions about the way the world looks, people talks, who lives in it, and what they’re up to.

In Lord of the Rings, someone who is basically a country gentleman, of no special skill, and his gardener, a stout fellow, together with two well-meaning idiots (all halflings), go on an adventure which involves a long arduous journey. Along the way they pick up a professional military scout of Royal Lineage, the world’s second-most-powerful wizard (soon to be first), and three professional soldiers (human, elf, and dwarf). They are engaged in a number of battles, in which each contributes in some positive way making the best of both their training and natural abilities.

At the end, the gentleman is all but dead from his exertions, but all four of the halflings have gained confidence, and two have grown physically larger and stronger as a result of magical drafts. The world’s second most powerful wizard has become top wizard owing to the fall from grace of his boss; he may or may not have gotten more powerful. Aside from that, the characters have not much changed as a consequence of experience besides having new stories to tell, and (in the case of the military scout) having gotten married and gained high social position.

Given this inspiration, what did we get?

A game where no-one starts play with a character capable of doing anything much. If you want to play Aragorn, you have to start as a level 1 wannabe.

A game where the only thing anyone ever gets good at is killing stuff. It’s much too hard to make rules about herbalism, but we do have a LOT of different magic swords.

A game which can only represent one kind of hardship — being attacked by monsters. When was the last time anyone cared about going hungry or freezing to death in your RPG?

A game where the content of the game mainly comprises getting more powerful by accumulating experience, items, and money (the latter two by stealing from the dead).