DSLR Sensor Economics

[yakiniku]

Technology Insider: DSLR Sensor Economics
Chipworks

[falconeye]

Originally Posted by yakiniku

Technology Insider: DSLR Sensor Economics

@yakiniku. This is a very interesting quote. Thanks a lot.

From this source:
APS-C sensor: ~$70-80
FF sensor: ~$300-400

My own (speculative) estimate as given here
http://www.pentaxforums.com/forums/340522-post30.html
was:
APS-C sensor: ~$50
FF sensor: ~$150

Meanwhile, the turnover Samsung creates with sensors shipped to Pentax was quoted in one of those Photokina threads. If I remember correctly, it translated to 48$ per (APS-C) sensor. So, let's assume the 50$ estimate is pretty accurate.

In the Technology Insider article, the numbers per wafer were assumed to be 20 for FF and 200 for APS-C. This is a flaw in the reasoning.

20 is the number as shown correct for FF on a 200mm wafer.

This would translate to 44 for 1/2 surface sensors on a 200mm wafer, not 200!

On a 300mm wafer, the waste is smaller, and about 160 chips of size 16x24mm would fit. 200 would mean no waste and 15x22mm chips. But a 300mm wafer line is not mature and depreciated and the whole argument wouldn't hold true anymore. So, there is a possible factor 4 in error in the given cost estimate...

So, for the time being, I speculate a FF CMOS sensor is in the 200$ price range.

[mithrandir]

I think part of the problem with going FF is what to do with the inbody SR so the image doesn't slip the dimensions of the sensor in motion.

[deejjjaaaa]

Originally Posted by mithrandir

I think part of the problem with going FF is what to do with the inbody SR so the image doesn't slip the dimensions of the sensor in motion.

it was not a problem for Sony

Super Steady Shot with Full Frame Sensors

should not be a problem for Pentax

[Canada_Rockies]

Originally Posted by falconeye

@yakiniku. This is a very interesting quote. Thanks a lot.

From this source:
APS-C sensor: ~$70-80
FF sensor: ~$300-400

My own (speculative) estimate as given here
http://www.pentaxforums.com/forums/340522-post30.html
was:
APS-C sensor: ~$50
FF sensor: ~$150

Meanwhile, the turnover Samsung creates with sensors shipped to Pentax was quoted in one of those Photokina threads. If I remember correctly, it translated to 48$ per (APS-C) sensor. So, let's assume the 50$ estimate is pretty accurate.

In the Technology Insider article, the numbers per wafer were assumed to be 20 for FF and 200 for APS-C. This is a flaw in the reasoning.

20 is the number as shown correct for FF on a 200mm wafer.

This would translate to 44 for 1/2 surface sensors on a 200mm wafer, not 200!

On a 300mm wafer, the waste is smaller, and about 160 chips of size 16x24mm would fit. 200 would mean no waste and 15x22mm chips. But a 300mm wafer line is not mature and depreciated and the whole argument wouldn't hold true anymore. So, there is a possible factor 4 in error in the given cost estimate...

So, for the time being, I speculate a FF CMOS sensor is in the 200$ price range.

It's not only a problem of fitting the sensor into the wafer and cutting. The real problem is getting the maximum number of sensors that are not flawed enough to prevent their use. The defect rate goes up exponentially (as we statisticians know well) as the size of the sensor increases. To visualize this, take a sheet of paper and put random spots on it to represent fatal flaws. You will see that as the sensor area increases, the flaws take out a proportionately larger percentage of the "good" chips.

[KungPOW]

Originally Posted by deejjjaaaa

it was not a problem for Sony

Super Steady Shot with Full Frame Sensors

should not be a problem for Pentax

It has everything to do with the size of the image circle.

If the lens has a large enough image circle, in-body shake reduction works. If the image circle is too small, or just the size needed to cover the sensor it will not work.

[deejjjaaaa]

Originally Posted by KungPOW

It has everything to do with the size of the image circle.

If the lens has a large enough image circle, in-body shake reduction works. If the image circle is too small, or just the size needed to cover the sensor it will not work.

so did you compare image circle from Sony FF lenses and from Pentax FF lenses

and also who much Sony actually moves the sensor - that is what the link talks about...

[philmorley]

Originally Posted by Canada_Rockies

It's not only a problem of fitting the sensor into the wafer and cutting. The real problem is getting the maximum number of sensors that are not flawed enough to prevent their use. The defect rate goes up exponentially (as we statisticians know well) as the size of the sensor increases. To visualize this, take a sheet of paper and put random spots on it to represent fatal flaws. You will see that as the sensor area increases, the flaws take out a proportionately larger percentage of the "good" chips.

not really allowing for this and very rough & simplified

apsc $50 -> cameras from $500
ff (2.5x area of apsc) say 4x apsc cost at $200 -> cameras from $2000 (canon 5d)
645 (2.6x area of ff) say 4x ff cost = $800 -> camera from $8000 ?? (well mamiya is 10k)

[Jonson PL]

Thanks for the link

[KungPOW]

Originally Posted by deejjjaaaa

so did you compare image circle from Sony FF lenses and from Pentax FF lenses

and also who much Sony actually moves the sensor - that is what the link talks about...

I've read that article before.

I think the sensor in both the Sony, and the Pentax move a great deal more then the article sugests is required.

I have tried to find information about image circle for Pentax lenses, but I have had little luck. I have not tried to find the info for Sony lenses.

A while back I was wondering How far the Pentax sensor can move. From photos of the mechanism, I estimated the total diagonal travel to be between 6 and 8 mm. Going with a 6 mm travel, or +/-3mm, the lens would need an image circle of 50 mm to cover the full range of motion the sensor can travel.

Really the only point is, if the sensor only needs to travel +/- 1.5mm, and the lens can cover Ø46mm, then it will work. If the image circle is Ø43mm, then the corners with not be sharp. Maybe The camera uses digital voodoo to fix the corners. Anyone know?

[deejjjaaaa]

Originally Posted by KungPOW

A while back I was wondering How far the Pentax sensor can move. From photos of the mechanism, I estimated the total diagonal travel to be between 6 and 8 mm. Going with a 6 mm travel, or +/-3mm, the lens would need an image circle of 50 mm to cover the full range of motion the sensor can travel.

that just might be necessary to prevent (if possible) any damage to the stabilization mechanism and sensor itself from some catastrophic impacts while IS is working (and such impacts will ruin the shot of course), but normal IS operation requires just +/- 1.5mm... just my 2c

[Canada_Rockies]

Originally Posted by philmorley

not really allowing for this and very rough & simplified

apsc $50 -> cameras from $500
ff (2.5x area of apsc) say 4x apsc cost at $200 -> cameras from $2000 (canon 5d)
645 (2.6x area of ff) say 4x ff cost = $800 -> camera from $8000 ?? (well mamiya is 10k)

Yes, it was intended as a simplified introduction to the concept. You get a higher failure rate with larger samples from the substrate. It can be googled. The rate is exponential.

[philmorley]

Originally Posted by Canada_Rockies

Yes, it was intended as a simplified introduction to the concept. You get a higher failure rate with larger samples from the substrate. It can be googled. The rate is exponential.

hi, sorry I meant my figures were simplified

[jeffkrol]

Originally Posted by philmorley

hi, sorry I meant my figures were simplified

Here is a quote from a personal inquiry to someone fairly embedded in the DSLR "industry". I see no reason to doubt his estimates:
This is 6 months old BTW:
Camera bodies are probably (even w/ the added electronics) not much more expensive than an old SLR was to make.
With APS, the basic sensor price is estimated in the US$40-50 range in quantity, US$70-100 in low quantity. With FX, the basic sensor price is 10x that (for ~2x the area). You’d have to guess that would scale up exponentially. So if you’re using double FX as many MF are, you could be another 10x, though at some point really low volume starts to help you (you’re going to be using an older fab that is costed out). But there are so many factors involved, it’s difficult to estimate. US$3000 is as good as any figure. Also, remember that for something like a 645D Pentax almost surely would have to buy the full supply of sensors up front (e.g. If the estimated 18-month need is 5000 units, they’d be paying cash for most of that up front; almost certainly a sensor run that small is going to be done in one or two batches; so, figure US$2000 and 2500 units and you’re talking US$5 million in cash up front, and I’m not sure that Pentax currently has that kind of free money floating around at the moment [or it would take Hoya giving it to them]).