Economics of Digital Photo Sensors

[jeffrey moore]

Given that we are always kicking around the "full-frame" debate around here,

you might find the linked article interesting. The author explains the

economics of digital sensor manufacture in a clear and concise manner that I

have not read before. And the best thing about the article is its brevity. :-)

 

http://www.naturescapes.net/092006/ej0906.htm

[david_h._hartman]

Jeffrey,<br>

<br>

Seams like a very good article to me. I would have thought all of

the sensors from the 1.6x up to the 1x size to have cost more.

This is good news. R&D is a big part of the cost of lower

volume, high tech products. Against the cost of a Hasselblad 503CW

system I dont and never did think the EOS 1Ds Mark II was

outrageously priced.<br>

<br>

Thanks for posting this link.<br>

<br>

Best,<br>

<br>

Dave Hartman.

[tom_burke3]

A further issue is that there are very few users of electronics who are interested in having

(physically) large components. The electronics industry's principal solution to yield issues

caused by 'wafer flaw' has always been the reduction in size of the components - given a

constant size wafer with a constant number of defects, if you can pack 2,000 component on

it then the unit cost per component is down to just cents. Therefore there is no great impetus

in the electronics industry as a whole to solve the 'wafer flaw' problem; as long as they can

maintain high volume, low cost yields by miniaturisation, then they're happy, as are their

customers. Digital photographers, who are interested in having low-cost large components,

are definitely swimming against the tide in this respect.

[janvanlaethem]

I had heard the argument before, but this explanation is indeed very clear and accurate. Thanks Jeffrey for posting this link.

 

Camera manufacturers are definitely making more money on smaller sensor cameras than on full frame. A full frame sensor camera will probably yield very poor benefits, but it still is a good showcase for the manufacturer. It's a bit like a car manufacturer that produces a sports car and wins an important race. Sales of family cars will benefit from this, although few people will actually buy the sports car. Likewise, I think most people will buy a 1.3x or 1.6x Canon and only a handfull of professionals will go for the full frame. Unless prices come down significantly, FF will remain a niche market.

[ShunCheung]

Thanks Jeffrey. That is a nicely written article, but none of the points made there should be new to those who have been reading this forum. Some of us have been making those same points repeatedly for 3, 4 years that sensor cost goes up exponentially with size (area). That is why the samll sensor is already the stadard and will continue to be so for years to come, and the so called "full frame" 24x36mm DSLR will mainly be for the high-end, niche market. However, E.J. Peiker explains things in a simple, easy-to-understand way with nice diagrams.

 

Thom Hogan has been suggesting that his "best estimate" is that 18x24mm type DX sensors cost about $50 each and 24x36mm FF ones cost about $500. Those figures are in line with E.J. Peiker's. Given that the retail price for a product is about 4 (or 3.5) times the manufacturing cost, for a Canon 5D, just the FF sensor alone should account for roughly $1500 of the final cost. That is why Canon has to really cut cost in other components to keep the overall price down, and their profit margin for the 5D is, at best, very low.

 

It remains to be interesting to see how Nikon approaches the "full frame" DSLR issue.

[charles_miller]

Thanks for the link. The article is especially clear and lucid.

 

In the case of digital integrated circuits, many fabrication defects merely slow things down. But since there is usually a demand for slower speed CPUs, the yield per wafer is not so severely impacted as would otherwise be the case.

 

In the case of camera sensors, similar defects cause weak cell output. That in turn causes color abberations in the associated pixel. For pro's who make large prints, these abberations are unacceptable.

 

If the chorus of consumer bleating gets loud enough, some camera manufacturer could start shipping full frame sensors that have a few strange pixels. I suspect Canon could do this at any time, but I hope things never get down to that level

 

And by the way, to say that semiconductor manufacturers don't take extensive precautions in order to lower the defect rate is contrary to my experience. Sensors will probably get cheaper and better, but not at the "Moore's law" rate that applies to mictoprosessor chips.

 

In the meantime, appreciate the excellent image sensor choices that are available now.

[joseph_wisniewski]

"best thing about the article is its brevity"

 

And the worst thing about it is assumptions and oversimplifications.

 

No disrespect to the author of that paper, but he's got a few things a bit off.

 

First, his areas are a bit low. He has the area of a 24x36 sensor at 864mm2, which is just the active pixel area, not including surrounding circuitry or bonding area. I don't have Canon figures handy, but I'd assume they're close to Sony. An ICX413AQ has a 23.6x15.6mm active area on a 25.1x17.6mm chip. The 24x36mm Canon should be about a 25.5x37mm chip.

 

Second, I doubt their yield, on a low (by semiconductor industry standards) quantity operation is anywhere near "state of the art".

 

Third, it's an image sensor, the dopings are as deep as 6um to get good sensitivity. That's a lot deeper than microprocessors or memories (their fractional um geometries wouldn't even work with circuitry that deep) and the deep diffusions and implantations increase processing time (increasing cost) and lower yield considerably.

 

Fourth, there are two organic post processing steps, the application of the Bayer filters and the microlenses, again cost adds.

[ShunCheung]

Joseph, I prefer simple answers. Perhaps it is only 95% accurate, but IMO it beats a 20-page write up that is 99.9% correct but nobody has the patience to read. That is why I still have never read any one review in DPReview from beginning to end.

[mark_chappell]

<I>... their profit margin for the 5D is, at best, very low.</i><P>

 

If Shun could provide real profit margin data, instead of speculation, this authoritative-

sounding statement could

achieve some credibility.

[steven_chan3]

I remember when the 1.5 crop D100 was $2000. Now these crop cameras are 1/2 to 1/4 that price. When a 1.5 crop sensor is $10, and a full frame sensor is $100, I think a lot people will be willing to pay 10X more for the full-frame. It's just a matter of time.

[ShunCheung]

Mark, I already did the calculation earlier based on sensor cost estimates from multiple sources; at least I find those sources credible. Nobody will be able to provide Canon proprietry profit margain information in this forum. So what I provided is probably as good as you can get here. Essentially Thom Hogan has made the same point. If that doesn't satify you, so be it. It is not my objective to convince every single person.

 

If you have such priprietry information to show the same or different, I invite you to share that with us. Otherwise, just accept the fact that estimates is the best we can do here.

[mark_chappell]

<I>Otherwise, just accept the fact that estimates is the best we can do here.</i><P>

 

Then why not clearly label them as estimates/guesses/speculation, instead of stating them

as if they were facts? I have no idea if Canon makes or loses money on the 5D, but I'm

certainly not going to state flatly that it's profitable for them or not, unless I have some

<B>real</b> data -- not extrapolations or guesswork.

[ShunCheung]

Mark, I suggest you read my first post again. It clearly specifies that Thom Hogan's sensor prices are estimates. E.J. Peiker has done the same in his article. Since those numbers are fairly close (within 20% or so), IMO it passes the "two independent sources" requirement, and I drew my conclusions based on them. I stand by everything I wrote there. If that doesn't satisfy you, you are entitled to your opinions.

 

I have no doubt that every time I post, there are people out there who disagree with me. I am not too concerned about that.

[mark_chappell]

For fun I'm going to belabor this unwinnable argument a bit more...<P>

 

<I>Mark, I suggest you read my first post again. It clearly specifies that Thom Hogan's

sensor prices are estimates. E.J. Peiker has done the same in his article. Since those

numbers are fairly close (within 20% or so), IMO it passes the "two independent

sources" requirement,</I><P>

 

And despite the fact that these were estimates, you flatly, without qualifications, stated:

<I>That is why Canon has to really cut

cost in other components to keep the overall price down, and their profit margin for the

5D is, at best, very low.</i> In fact, you do not know what the profit margin is. And

assuming that two independent guesses constitutes verification is a bit of a

stretch, IMO.<P>

 

I would not have launched into this somewhat pointless little rant if you had inserted the

words <I>I'm

guessing</i> into your sentence, i.e. <I>... overall price down, and<b> I'm guessing

</b>their profit margin...

</i>. Perhaps it's a bit over the top but I have a pet peeve when opinions and

guesses and speculation are stated as facts.

[KenPapai]

I'd be willing to make an equally educated guess and say that Canon's profit margin on each 5D is 10% and the retailers' are enjoying another 10% margin, so a $2800 5D makes nice enough money for Canon and B&H, thereby leading the consumer to the high profit items: lenses, filters, bags, batteries, sensor cleaners, and memory cards.

 

The profit margin on those entry level DSLRs must be very small, if not even a loss leader.

[plasma181]

OK, how about this? You have 2 roughly half-frame sensors that cost $34 each. A full frame would cost about $385. Why can't we make a camera that has two small sensors laid side by side? Now the assembly costs only $64.

[joseph_wisniewski]

There would be an annoying 5mm gap through the middle of the image...