Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Why not 6x4.5 CMOS sensors?  (Read 6797 times)
Gel
Full Member
***
Offline Offline

Posts: 161


Excuse me while I bust a cap


« on: April 20, 2012, 09:31:32 AM »
ReplyReply

Is there a reason they are all CCD?

Forgive me for sounding obvious but why don't we have CMOS sensor technology (with it's high ISO capability) put into a larger sensor on a digital back?

I can understand Canon / Nikon not thinking it's a big enough market but why not one of the big boys like Phase/Leaf/Hassy?
Logged

Chris Giles Photography
ErikKaffehr
Sr. Member
****
Online Online

Posts: 8011


WWW
« Reply #1 on: April 20, 2012, 11:44:52 AM »
ReplyReply

Hi,

Because Phase/Leaf/Hassy don't make sensors. They buy sensors from Dalsa and what used to be Kodak. Also, I guess that Nikon and Canon sell more full frame sensors in a week than Phase/Leaf/Hassy in a year.

Best regards
Erik


Is there a reason they are all CCD?

Forgive me for sounding obvious but why don't we have CMOS sensor technology (with it's high ISO capability) put into a larger sensor on a digital back?

I can understand Canon / Nikon not thinking it's a big enough market but why not one of the big boys like Phase/Leaf/Hassy?
Logged

shadowblade
Sr. Member
****
Online Online

Posts: 749


« Reply #2 on: April 20, 2012, 06:42:14 PM »
ReplyReply

This is an obvious target area for Sony.

They can't compete with Canon/Nikon on DSLRs, because of lens selection - may as well just sell sensors to both of them.

Where they can compete is in mirrorless technology - stick two D800e sensors side by side for a 24x72mm sensor with live view, work with Zeiss to bring out a range of suitable lenses (starting with a 24mm or 28mm tilt-shift, for the equivalent horizontal angle of view of 12 or 14mm on full-frame - MF users often want perspective and focal plane control), sell the camera body for $5k (because it doesn't need all the fancy mirrors and viewfinder assemblies) and they'd practically wipe out Phase One, Mamiya and Hasselblad overnight.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5182


« Reply #3 on: April 20, 2012, 07:18:16 PM »
ReplyReply

The main issue seems to be that there is a far bigger overhead in designing and putting an active pixel CMOS sensor into production compared to a CCD, and the far lower sales volume of formats larger than 36x24mm mean that this investment would not be cost effective. Or at least so far, and in the judgement of potential customers like Phase One, Hasselblad, Pentax and Leica. I mention them, not the various possible sensor makers, because it is the willingness of these potential customers to guarantee some minimum order at some acceptable price that dictates whether a company like Teledyne-Dalsa goes ahead with such a low volume project.

I think of this like new models of commercial aircraft, where the decision to go ahead is based on whether enough deposits are put down by airlines. I am skeptical of the idea that sensor makers just design and produce specialty products like this "on spec" and then go looking for customers. For example, it seems that the direction of Sony's development of sensors for DSLRs has often been heavily influenced by the wishes of its dominant customer, Nikon, to the point that Nikon often gets to use a new Sony DSLR sensor before Sony's far smaller DSLR division does.


P. S. you cannot simply stick two sensors side-by-side to get something usable in a high quality MF camera. The resulting seam is tolerable for some uses like X-rays, and Teledyne-Dalsa does make some large CMOS sensors that way, but nor for "photographic" cameras. Instead, any sensor larger than about 33x26mm must be fabricated by "stitching", which means that it is "etched" onto the silicon part at a time, moving the stepper field between each part and being very careful with alignment, which lads to a high rate of rejects. Both Canon and Dalsa have described processes involving each piece being only 24x12mm. The Canon example involved a three piece stitch for the 1Ds sensor. As the size goes up and so the number of "joins" needed in this process goes up, the failure rate increases rapidly, which is a major reason why camera costs rise rapidly once the size goes beyond about the mainstream "APS-C" sizes.
« Last Edit: April 20, 2012, 07:20:14 PM by BJL » Logged
ErikKaffehr
Sr. Member
****
Online Online

Posts: 8011


WWW
« Reply #4 on: April 20, 2012, 07:37:08 PM »
ReplyReply

Hi,

In a sense I would say that it may make more sense to build a truly optimal full frame camera than a true 645. It would probably be possible to construct a series of prime lenses that are diffraction limited at say f/4 and use a full frame version of the NEX-7 sensor. That combo would have 54 MP and probably have 14 EV of dynamic range. Such a camera would be quite affordable.

Best regards
Erik


The main issue seems to be that there is a far bigger overhead in designing and putting an active pixel CMOS sensor into production compared to a CCD, and the far lower sales volume of formats larger than 36x24mm mean that this investment would not be cost effective. Or at least so far, and in the judgement of potential customers like Phase One, Hasselblad, Pentax and Leica. I mention them, not the various possible sensor makers, because it is the willingness of these potential customers to guarantee some minimum order at some acceptable price that dictates whether a company like Teledyne-Dalsa goes ahead with such a low volume project.

I think of this like new models of commercial aircraft, where the decision to go ahead is based on whether enough deposits are put down by airlines. I am skeptical of the idea that sensor makers just design and produce specialty products like this "on spec" and then go looking for customers. For example, it seems that the direction of Sony's development of sensors for DSLRs has often been heavily influenced by the wishes of its dominant customer, Nikon, to the point that Nikon often gets to use a new Sony DSLR sensor before Sony's far smaller DSLR division does.


P. S. you cannot simply stick two sensors side-by-side to get something usable in a high quality MF camera. The resulting seam is tolerable for some uses like X-rays, and Teledyne-Dalsa does make some large CMOS sensors that way, but nor for "photographic" cameras. Instead, any sensor larger than about 33x26mm must be fabricated by "stitching", which means that it is "etched" onto the silicon part at a time, moving the stepper field between each part and being very careful with alignment, which lads to a high rate of rejects. Both Canon and Dalsa have described processes involving each piece being only 24x12mm. The Canon example involved a three piece stitch for the 1Ds sensor. As the size goes up and so the number of "joins" needed in this process goes up, the failure rate increases rapidly, which is a major reason why camera costs rise rapidly once the size goes beyond about the mainstream "APS-C" sizes.
Logged

shadowblade
Sr. Member
****
Online Online

Posts: 749


« Reply #5 on: April 20, 2012, 07:50:22 PM »
ReplyReply

Sticking two sensors together would require exactly one more join than making those two sensors in the first place. If two full-frame sensors require 4 joins to make (for 3 pieces each) a 24x72mm sensor would require only 5 joins. Not a huge increase...

True, sales volume is a factor, but there would be a thousand times (or more) customers who would buy a $5k panoramic body for each one who would spend $40k on an IQ180, and they'd also steal many customers from DSLRs.

The main issue seems to be that there is a far bigger overhead in designing and putting an active pixel CMOS sensor into production compared to a CCD, and the far lower sales volume of formats larger than 36x24mm mean that this investment would not be cost effective. Or at least so far, and in the judgement of potential customers like Phase One, Hasselblad, Pentax and Leica. I mention them, not the various possible sensor makers, because it is the willingness of these potential customers to guarantee some minimum order at some acceptable price that dictates whether a company like Teledyne-Dalsa goes ahead with such a low volume project.

I think of this like new models of commercial aircraft, where the decision to go ahead is based on whether enough deposits are put down by airlines. I am skeptical of the idea that sensor makers just design and produce specialty products like this "on spec" and then go looking for customers. For example, it seems that the direction of Sony's development of sensors for DSLRs has often been heavily influenced by the wishes of its dominant customer, Nikon, to the point that Nikon often gets to use a new Sony DSLR sensor before Sony's far smaller DSLR division does.


P. S. you cannot simply stick two sensors side-by-side to get something usable in a high quality MF camera. The resulting seam is tolerable for some uses like X-rays, and Teledyne-Dalsa does make some large CMOS sensors that way, but nor for "photographic" cameras. Instead, any sensor larger than about 33x26mm must be fabricated by "stitching", which means that it is "etched" onto the silicon part at a time, moving the stepper field between each part and being very careful with alignment, which lads to a high rate of rejects. Both Canon and Dalsa have described processes involving each piece being only 24x12mm. The Canon example involved a three piece stitch for the 1Ds sensor. As the size goes up and so the number of "joins" needed in this process goes up, the failure rate increases rapidly, which is a major reason why camera costs rise rapidly once the size goes beyond about the mainstream "APS-C" sizes.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5182


« Reply #6 on: April 20, 2012, 08:14:42 PM »
ReplyReply

Sticking two sensors together would require exactly one more join than making those two sensors in the first place. If two full-frame sensors require 4 joins to make (for 3 pieces each) a 24x72mm sensor would require only 5 joins. Not a huge increase...

...  there would be a thousand times (or more) customers who would buy a $5k panoramic body for each one who would spend $40k on an IQ180, and they'd also steal many customers from DSLRs.
I would instead expect that the unusual 3:1 aspect ratio you propose would be a far lower volume niche product than the dominant 4:3 to 3:2 sensor shapes. One hint is the complete absence of such 72x24mm in CCDs, where the initial cost of product development would be lower than with a new CMOS sensor design. Another factor is that for panoramas, scanning backs using linear sensors are often an acceptable and far more cost effective solution.


And the idea that this could be done for a camera price anywhere close to $5000, far less than even the relatively high volume Pentax 645D with its smaller 44x33mm sensor, is wildly optimistic.


Ah, but the forums are a constant source of ideas for products that would obviously be great successes, if only all the companies who could offer such products were not run by either complete idiots who know less than us forum pundits do about what is and is not commercially viable. Like the fact that Sony has not chosen to pursue what you call an obvious target.

Seriously, you should least try to understand the likely technical and economic reasons why things are not the way you would like them to be, instead of glibly claiming that you know an obvious but neglected solution.
« Last Edit: April 20, 2012, 08:20:14 PM by BJL » Logged
shadowblade
Sr. Member
****
Online Online

Posts: 749


« Reply #7 on: April 20, 2012, 10:07:33 PM »
ReplyReply

I would instead expect that the unusual 3:1 aspect ratio you propose would be a far lower volume niche product than the dominant 4:3 to 3:2 sensor shapes. One hint is the complete absence of such 72x24mm in CCDs, where the initial cost of product development would be lower than with a new CMOS sensor design. Another factor is that for panoramas, scanning backs using linear sensors are often an acceptable and far more cost effective solution.


And the idea that this could be done for a camera price anywhere close to $5000, far less than even the relatively high volume Pentax 645D with its smaller 44x33mm sensor, is wildly optimistic.


Ah, but the forums are a constant source of ideas for products that would obviously be great successes, if only all the companies who could offer such products were not run by either complete idiots who know less than us forum pundits do about what is and is not commercially viable. Like the fact that Sony has not chosen to pursue what you call an obvious target.

Seriously, you should least try to understand the likely technical and economic reasons why things are not the way you would like them to be, instead of glibly claiming that you know an obvious but neglected solution.



612, 617, 624 and X-Pan formats were all widely used in film, so it's hardly niche. Not as widely used as less-elongated formats, perhaps, but still very common. Scanning backs aren't a solution if you need exposure times greater than a few milliseconds - lighting in a scene can change wildly over the course of the hours it takes to make a panoramic exposure at 1 second shutter speed.

CMOS fabrication is far easier and cheaper than CCD fabrication. Almost any factory capable of making silicon chips, processors, etc. can make a CMOS sensor. CCDs are built differently, and require different production plants. Also, the reason for the price of MF backs isn't manufacturing costs - it's simply because they can charge that much, since theirs is a low-volume product purchased largely by advertising houses and large companies, not by individuals. A CMOS MF sensor, targeted at individual photographers, would be a far higher volume product with a much greater audience than the specialised CCDs. Live view, MF video and high ISO capability alone would ensure that.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5182


« Reply #8 on: April 20, 2012, 11:00:49 PM »
ReplyReply

612, 617, 624 and X-Pan formats were all widely used in film, so it's hardly niche. Not as widely used as less-elongated formats, perhaps, but still very common.
Even with the vastly lower costs of the film versions (where rolls of standard 135 or 120 film are used), panoramic  cameras were and are a tiny fraction of the volume of cameras, so your usage of "very common" is strange. With digital, there is no using the same common films but the need for different, far lower volume sensors, so the costs are far more of a barrier.

Scanning backs aren't a solution if you need exposure times greater than a few milliseconds - lighting in a scene can change wildly over the course of the hours it takes to make a panoramic exposure at 1 second shutter speed.
Read up on scanning digital panoramic cameras: hour long exposures are not at all needed in the format sizes you are talking of. They can even handle some action, since no one part of the scene need be exposed for longer than with a normal "array" sensor. Consider for example the 160MP Seitz 6x17, and imagine the cost of a 6x17cm sensor!
http://www.roundshot.ch/xml_1/internet/de/application/d438/d925/f934.cfm

CMOS fabrication is far easier and cheaper than CCD fabrication. Almost any factory capable of making silicon chips, processors, etc. can make a CMOS sensor.

If that were true, one would have to wonder why Dalsa, Kodak and such have stayed with the supposedly more difficult, more expensive CCD technology, despite its added disadvantages of worse noise and lack of Live View capability. The facts stongly suggest that there are significant barriers of technology and/or cost. As I said earlier, those barriers are not so much the fabrication, but the original design of a good, large activepicel CMOS sensor, with its unique "mixed mode" requirements of analog and digital elements on a single chip, that is difficult. By the way it was a false hope that CMOS sensors would be cheeaper to fabricate than CCDs. Apparently the special requirements of the mixed mode nature and having adequate well depth nullified that hoped-for cost advantage.


I have to ask: if it is so obvious and easy and with such good sales potential as you seem to think, why do you think that neither Sony any other company is providing a CMOS sensor in a format larger than 36x24mm?
« Last Edit: April 20, 2012, 11:05:27 PM by BJL » Logged
Gel
Full Member
***
Offline Offline

Posts: 161


Excuse me while I bust a cap


« Reply #9 on: April 21, 2012, 02:10:43 AM »
ReplyReply

One may suggest the size of the market for these products is based around the goods offered.

I shoot 50 weddings a year, give me a 22mp 9 micron CMOS with clean ISO 128000 at 1 fps and I'd take two.
Logged

Chris Giles Photography
gerald.d
Sr. Member
****
Offline Offline

Posts: 401


« Reply #10 on: April 21, 2012, 03:55:42 AM »
ReplyReply

This is an obvious target area for Sony.

They can't compete with Canon/Nikon on DSLRs, because of lens selection - may as well just sell sensors to both of them.

Where they can compete is in mirrorless technology - stick two D800e sensors side by side for a 24x72mm sensor with live view, work with Zeiss to bring out a range of suitable lenses (starting with a 24mm or 28mm tilt-shift, for the equivalent horizontal angle of view of 12 or 14mm on full-frame - MF users often want perspective and focal plane control), sell the camera body for $5k (because it doesn't need all the fancy mirrors and viewfinder assemblies) and they'd practically wipe out Phase One, Mamiya and Hasselblad overnight.
If anyone were to go down this hypothetical route, I suspect it would be far more likely that they would stick two side by side for a 36x48 sensor.
Logged
yaya
Sr. Member
****
Offline Offline

Posts: 1150



WWW
« Reply #11 on: April 21, 2012, 07:48:30 AM »
ReplyReply

Glad we have BJL here!
Logged

Yair Shahar | Product Manager | Mamiya Leaf |
e: ysh@mamiyaleaf.com | m: +44(0)77 8992 8199 | www.mamiyaleaf.com | yaya's blog
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 8385



WWW
« Reply #12 on: April 21, 2012, 02:41:12 PM »
ReplyReply

This question has been in the air for years. Cmos is a 15 years old tech that has started to show superior imaging potential 8 years ago and delivered big time on this promise 3 years ago.

The extended lead of the D800 is only making more obvious a fact that was true years ago already.

The guys running Hassy and Phase must have known that 5 or 6 ago. Now, is their business model and mode of execution compatible with the needs of leading technological development? In other words, are they providing enough business to the sensor manufacturers? My view remains that the only company that might be close to doing that now is Pentax, and their prices are twice lower.

We will see if there is a 645d mk II and whether it uses a CMOS sensor. If it does not, chances are that the whole MF thing will further spiral down into a niche market targetting the luxury segment with cameras delivering less real world performance than top DSLRs costing 10 times less.

Compare Silicon graphics to nvidia and you'll see why.

Cheers,
Bernard
Logged

A few images online here!
BJL
Sr. Member
****
Offline Offline

Posts: 5182


« Reply #13 on: April 21, 2012, 03:07:03 PM »
ReplyReply

My view remains that the only company that might be close to doing that now is Pentax, and their prices are twice lower. ... We will see if there is a 645d mk II and whether it uses a CMOS sensor.
Agree on many points, in particular that the most likely place for CMOS sensors to arrive in formats larger than 36x24 is the relatively high volume sector tha Pentax is working on.
Quote
If it does not, chances are that the whole MF thing will further spiral down into a niche market targetting the luxury segment with cameras delivering less real world performance than top DSLRs costing 10 times less.

Compare Silicon graphics to nvidia and you'll see why.
That last comparison hits close to home: I was involved in justifying the purchase of some $10,000+ custom SG and Sun workstations that were obsoleted by vastly less expensive, faster, and more user-friendly Macintosh and Windows/Linux systems just a few years later.

Let me add that if Pentax _does_ offer something like a "645D II" with state-of-the-art 44x33 CMOS sensor, to which it can add the most sophisticated AF and metering in the larger than 36x24 world, it would add greatly to the challenges already coming from Nikon FX to the Mamiya and Hasselblad 645 format based systems.
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 8385



WWW
« Reply #14 on: April 21, 2012, 08:52:34 PM »
ReplyReply

Let me add that if Pentax _doe_ offer something like a "645D II" with state-of-the-art 44x33 CMOS sensor, to which it can add the most sophisticated AF and metering in the larger than 36x24 world, it would add greatly to the challenges already coming from Nikon FX to the Mamiya and Hasselblad 645 format based systems.

Yes, but it could fund the development of the sensors Phase and Hassy need.

In essence I am afraid that there will probably be only one MF sensor manufacturer left in 3 years from now and that would be the supplier selected by Pentax.

Cheers,
Bernard
Logged

A few images online here!
BJL
Sr. Member
****
Offline Offline

Posts: 5182


« Reply #15 on: April 21, 2012, 09:02:54 PM »
ReplyReply

Yes, but it could fund the development of the sensors Phase and Hassy need.
Yes, CMOS sensors in the format or formats used by Pentax would likely be available to other camera/back makers too. My concern is that Pentax seems to be adopting 44x33 format exclusively for its DMF system (putting aside the red herring of a couple of lenses intended in part for 645 film camera users, and so still covering the 645 format frame), so that the larger DMF formats like 48x36 and full 645 format would still be limited to CCDs. I suppose that if this happens, Mamiya and Hasselblad/Fuji might just have to refocus on 44x33 format, adding a few lenses and viewfinders for it.
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 8385



WWW
« Reply #16 on: April 21, 2012, 10:13:53 PM »
ReplyReply

Yes, CMOS sensors in the format or formats used by Pentax would likely be available to other camera/back makers too. My concern is that Pentax seems to be adopting 44x33 format exclusively for its DMF system (putting aside the red herring of a couple of lenses intended in part for 645 film camera users, and so still covering the 645 format frame), so that the larger DMF formats like 48x36 and full 645 format would still be limited to CCDs. I suppose that if this happens, Mamiya and Hasselblad/Fuji might just have to refocus on 44x33 format, adding a few lenses and viewfinders for it.

I am not sure that the technology would differ between a 44x33 sensor vs a 56x42 sensor.

The key technological jump is the move to cmos.

So if Dalsa/Kodak manage, thanks to Pentax, to secure the business visibility they need to justify the investement needed to develop a cmos techno matching their expectetions, it could also be applied to larget sensors. Pentax would probably be smart enough to leave the door open for that.

Cheers,
Bernard
Logged

A few images online here!
BJL
Sr. Member
****
Offline Offline

Posts: 5182


« Reply #17 on: April 21, 2012, 10:40:22 PM »
ReplyReply

Bernard,
    I was thinking only of the economies of scale of putting something like a 54x42mm CMOS sensor into production: this is not particularly made easier by having a 44x33 CMOS sensor in production, any more than the current existence of 36x24 sensors makes 44x33 ones readily available. I also think that such a sensor is far more likely to come from an existing maker of good photographically suitable CMOS sensors that only has to scale up to 44x33 (like Sony, Panasonic, Samsung, Fujifilm, or whoever designs and makes the sensors for Nikon One cameras), not from a company that is not currently making such CMOS sensors of any size (like Dalsa or True Sense Imaging, which is the the former Kodak sensor division).
« Last Edit: April 22, 2012, 12:55:34 PM by BJL » Logged
Stefan.Steib
Sr. Member
****
Offline Offline

Posts: 421



WWW
« Reply #18 on: April 22, 2012, 11:43:40 AM »
ReplyReply

The actual devellopment of a mask and a process for a large CMOS alone ,will cost between 10-20 mill.$US. This is without R&D costs for further peripherals and changes in Camera/Back architecture. I think it is the naked and brutal truth that none of the actual makers alone can afford this amount of money to be spent from their actual turnarounds.

The only chance they may have is sitting down to a table and devellop a standard chip for all of them. Something like a Ford 8 cylinder Formula one motor to give base for all the teams which worked for nearly 16 years. If they are smart they will do this. But I fear logics do not work in this connection..........

regards
Stefan
Logged

Because Photography is more than Technology and "as we have done this all the time"
www.hartblei.de     www.hcam.de    www.spectralize.com
paul_jones
Sr. Member
****
Offline Offline

Posts: 565


WWW
« Reply #19 on: April 22, 2012, 03:10:17 PM »
ReplyReply

The actual devellopment of a mask and a process for a large CMOS alone ,will cost between 10-20 mill.$US. This is without R&D costs for further peripherals and changes in Camera/Back architecture. I think it is the naked and brutal truth that none of the actual makers alone can afford this amount of money to be spent from their actual turnarounds.

The only chance they may have is sitting down to a table and devellop a standard chip for all of them. Something like a Ford 8 cylinder Formula one motor to give base for all the teams which worked for nearly 16 years. If they are smart they will do this. But I fear logics do not work in this connection..........

regards
Stefan

really? where did you get a figure like that?

did RED need to spend that much for their chips that they developed? RED also was promising the super large chips a while back, they wouldn't have done that unless it was achievable as a lower run "professional" cine camera.

paul
Logged

check my new website
http://www.paulrossjones.com
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad