Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: More pixels, Better pixels, Bigger Sensor, Cheaper  (Read 6086 times)
peterhorsley
Newbie
*
Offline Offline

Posts: 23


« on: October 21, 2006, 11:02:18 AM »
ReplyReply

Bare with me on this.

It seems that the manufactures want to sell us more pixels.  We probably only need 20MPx to 30Px for commercial work.  A bigger sensor (56mm x 56mm) would work better with our old-time favourite lenses and cameras. We'd all like digi backs to cost less.

More pixels means finer pixel pitch and less chance of moire.

Bigger sensors are expensive because of the production limitations of silicon-based sensors: more area means more chance of a fault.

We hear that manufactures are talking about 50Mpx-60MPx sensors.

Lastly, we've seen some sensor manufactuers use software and a high pixel-count sensor to reduce noise and increase ISO: pixel binning.

Here's the idea.  Daldak (that's what you get when the big sensor people meet) produce the 'ultimate' 60 MPx chip. The 'perfects' go into the P60 or A100.  The slightly less than perfects go into the 'full frame' wedding, fashion, lower-cost product: the Pete15.  

The Pete15 is a 60MPx sensor that achieves low noise, high sensitivity, no moire by hard wiring binning to give a 15 Mpx sensor. The sensor imperfections dissappear with the binning and averaging of pixels.  Production yields go up.  Costs go down.

It is the same sensor as the P60/A100, packaged and firm-wired differently.

And then Daldak realise: if we go from a 60 MPx, 48mmx36mm sensor to a 48mm x 48mm sensor we get a 'full frame' 80Mpx sensor which fits on every 6x6 camera, the Hy6, V 'Blad etc and the 'Pete' version is 20Mpx.  Killer; every medium format camera and lens made for the last 50 years comes into play.

'Hassecon' go wild because their integrated lens-body-back cameras are set up to use every software trick to get the the best out of a 80Mpx/20 MPx sensor.

At this point almost every meduim format customer buys a Pete20 if only because nobody wants to process 1000s of 80Mpx files.  (Except for product, architectural and luminous-landscape shooters; wink)

With a firm grasp on the credit card, students and enthusiasts can buy a Pete20.

Of course the marketing people get all muddled because by now 'Cankon', the 35mm people, have a 20Mpx camera. It takes a while, but we explain that in the old days 35mm and 120 cameras used the 'same' film; Plus X, Tmax, HP5, Delta, Velvia, 'chrome etc; but the look of the finished image was different.  

Photographers get on with the job of making pictures and wondering why graphics people crop the best parts of their images.

Clients are ecstatic and pay-up punctually.

And then I wake up...

Peter
Logged
heinrichvoelkel
Sr. Member
****
Online Online

Posts: 370


« Reply #1 on: October 21, 2006, 11:26:35 AM »
ReplyReply

Thank you so much. I buy everything you propose.

As I'm shooting editorial 99% of my time and my customers will never pay me back what I invest into MFDG, even in 24 months time ( I would love to swap with Mr. Russell, just for the invoice writing stuff, okay for the shooting superb like he does as well, ...), but I would just love to invest my hard earned money to get away from disscussions around digital.

I did play and used the Leica M8. As I mainly shoot with a Mamiya m7 I thought I would like to have a professionel digital rangefinder. Looking at the Leica user forum I just get freaked out by the quality of discussion there. Those guys are talking about the camera bags they have to buy for their digital M( which is not even delivered yet.

Because I use Rollei a lot for portrait stuff as well, I will end up robbing a bank and buying one of this HY6 sinar or leaf machines, just to carry on with what I'm doing and loving to do. Actually I would love to use a digital back as siplictic as aroll of 120 film. Not becaise I'm lazy, just because I like to keep it simple.

Sorry for this post, which doesn't belong to this thread or may be does.

Regrds
Logged
pss
Sr. Member
****
Offline Offline

Posts: 960


WWW
« Reply #2 on: October 21, 2006, 01:01:24 PM »
ReplyReply

someone very adverturous could start a company, secure a couple of billions in funding to produce a 6x6 back with 80mpix 800asa, 1frame/sec., mass produce it and sell it for 5990.-.....how many people would buy it? 1000s..10000s..100000s? i have no clue...if porsche would build 100times as many cars, they would be a lot cheaper too...the question is if 100times more people would buy porsches if they were cheaper...i know i still would not...
there is a post on open photography forums
the guy financed a p25 for 325/month....how any working photographer can shoot film for that amount and stay in business, i don't know......
when i went to photoschool in the early 90s, we were told to allocate 500-1000/month in film/processing/paper....i don't even want to think about all the money i could have saved...
Logged

peterhorsley
Newbie
*
Offline Offline

Posts: 23


« Reply #3 on: October 22, 2006, 08:27:15 AM »
ReplyReply

Quote
Sorry for this post, which doesn't belong to this thread or may be does.

[a href=\"index.php?act=findpost&pid=81518\"][{POST_SNAPBACK}][/a]

Hey Heinrich.  I think you post kind of reflect the some of the same state of mind that prompted me to start this subject.  

On one hand we seem to be spending a lot time talking about the small differences between digital products.  Douglas Adams made a comment once about consumer goods with superficial design flaws hiding their fundamental design flaws. For example the almost full-frame digital sensor has made many thousands of medium format camera outfits obsolete because they are 'almost', but not, full frame.

I was also indulging in a little armchair business planning for the sensor manufacturers.  We are told that MF sensors/backs are costly because of low production yields and small sales volumes.  But we currently have at least 6 different sensors in MF backs (39Mpx, 33Mpx, 30Mpx, 22MPx, 21MPx, 17MPx).  It would seem smart to make one sensor in bulk, then use firmware and software to make 'different' sensors.  I thought one of the many knowledgable people on the forum might comment of the viability of the idea.

Then I realised I'd spent another half hour sitting in front of computer thinking about technology when I could have been doing something enjoyable.

Peter
Logged
eronald
Sr. Member
****
Offline Offline

Posts: 4218



« Reply #4 on: October 22, 2006, 10:13:19 AM »
ReplyReply

Quote
I was also indulging in a little armchair business planning for the sensor manufacturers.  We are told that MF sensors/backs are costly because of low production yields and small sales volumes.  But we currently have at least 6 different sensors in MF backs (39Mpx, 33Mpx, 30Mpx, 22MPx, 21MPx, 17MPx).  It would seem smart to make one sensor in bulk, then use firmware and software to make 'different' sensors.  I thought one of the many knowledgable people on the forum might comment of the viability of the idea.
[a href=\"index.php?act=findpost&pid=81612\"][{POST_SNAPBACK}][/a]

I call this the Mercedes argument - if all car companies standardised and made Mercedes cars, the resulting economies of scale would mean that everyone could afford a Mercedes.

It doesn't work for large cost differentials; I leave the reasons as an exercise for the reader.

Edmund
« Last Edit: October 22, 2006, 10:14:21 AM by eronald » Logged
peterhorsley
Newbie
*
Offline Offline

Posts: 23


« Reply #5 on: October 22, 2006, 08:37:11 PM »
ReplyReply

Quote
I call this the Mercedes argument ...
[a href=\"index.php?act=findpost&pid=81616\"][{POST_SNAPBACK}][/a]

Hi Edmund, I don't know much about cars, and even less about Mercedes, but the automotive industry has some relavance.  Thanks.

Developing the floor pan of cars is expensive, $100's millions, even $billions.

The VW small/medium car; 'Golf' in Europe, 'Rabbit' in the US (I think);  shares it's floor pan with a whole range of cars some of which are sold by Audi even maybe even Fiat.  The selling price ratio is around 1:4; the cheapest costs four times less than the most expensive. Every mass production car maker uses a floor pan as the basis for a range of model variants where the selling price ranges over at least 1:2.  

Different products can be based on a shared component. High volume gives lower per-unit cost: economies of scale.

But as I understand it the big issue in MF sensor manufacturing is not just economies of scale, it is about low yield. The bigger the sensor the more significant the yield issue becomes.  

So my speculative point was really that with the coming 60Mpx+ sensors there is the opportunity to deal with yield issues using firmware and software.  This already happens with the 'mapping' of pixels to deal with response variation to give some improvement to yield.

My main point is about taking a user-centric approach to product development.  The pixel count of a 60Mpx+ sensor of is limited value to most commercial photographers. No moire, low noise, faster file handling, wide-angle images and lower cost benefits a lot of us; the pixel count just has to be adequate for the job.  Think about photographers' discussions about the Aptus 75(33Mpx) and the P45(39Mpx).  Photographers talk about colour and tonal rendering, ISO, useablity, service support, malfunctions, and post-camera work flow.  There is no hot thread for: "Need 6Mpx more".

It seems to me that radically remapping the massive amount of data that comes from a 'defective' 60MPx+ sensor has the potential to result is a very useful 15Mpx to 20Mpx photographic tool with advantages over current 15Mpx to 20MPx sensors.  

The same approach could, in my dreams, give us a marketable old-school-full-frame MF sensor.

It could also make business sense.

Peter
Logged
eronald
Sr. Member
****
Offline Offline

Posts: 4218



« Reply #6 on: October 23, 2006, 05:55:28 AM »
ReplyReply

Yes, software binning would allow the sale of chips with dead pixels or rows, or pixels, if they work at all. I agree that the sale of such would be neat. I assume the back makers have thought about it.

This does not solve the issues of big surface image chips.

In the end, you need to get *some* image data off the sensor, which means a working system of "data pipes". The problem is that many defects will knock out not only cells but the the simple piping between them which runs across the chip, killing a good part of the whole thing.

A fault-tolerance scheme for a cell array needs to incorporate a fail-safe routing scheme which will add considerably to the design complexity of each cell. Then you need to reconfigure the routing at some point to handle your individual chip sample.

Redundancy is a can  of worms, and as far as I know unsolved it although everybody realizes that it' an interesting problem. Think of it as going from a uni-cellular to a multi-cellular organism. A  solutio, probably requires modified processes and "base layers" under each and every chip. We can wait for Intel and Canon to solve it; Kodak and Dalsa don't have the resources.

Quote
Hi Edmund, I don't know much about cars, and even less about Mercedes, but the automotive industry has some relavance.  Thanks.

Developing the floor pan of cars is expensive, $100's millions, even $billions.

The VW small/medium car; 'Golf' in Europe, 'Rabbit' in the US (I think);  shares it's floor pan with a whole range of cars some of which are sold by Audi even maybe even Fiat.  The selling price ratio is around 1:4; the cheapest costs four times less than the most expensive. Every mass production car maker uses a floor pan as the basis for a range of model variants where the selling price ranges over at least 1:2. 

Different products can be based on a shared component. High volume gives lower per-unit cost: economies of scale.

But as I understand it the big issue in MF sensor manufacturing is not just economies of scale, it is about low yield. The bigger the sensor the more significant the yield issue becomes. 

So my speculative point was really that with the coming 60Mpx+ sensors there is the opportunity to deal with yield issues using firmware and software.  This already happens with the 'mapping' of pixels to deal with response variation to give some improvement to yield.

My main point is about taking a user-centric approach to product development.  The pixel count of a 60Mpx+ sensor of is limited value to most commercial photographers. No moire, low noise, faster file handling, wide-angle images and lower cost benefits a lot of us; the pixel count just has to be adequate for the job.  Think about photographers' discussions about the Aptus 75(33Mpx) and the P45(39Mpx).  Photographers talk about colour and tonal rendering, ISO, useablity, service support, malfunctions, and post-camera work flow.  There is no hot thread for: "Need 6Mpx more".

It seems to me that radically remapping the massive amount of data that comes from a 'defective' 60MPx+ sensor has the potential to result is a very useful 15Mpx to 20Mpx photographic tool with advantages over current 15Mpx to 20MPx sensors. 

The same approach could, in my dreams, give us a marketable old-school-full-frame MF sensor.

It could also make business sense.

Peter
[a href=\"index.php?act=findpost&pid=81675\"][{POST_SNAPBACK}][/a]
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5171


« Reply #7 on: October 23, 2006, 10:34:49 AM »
ReplyReply

Dalsa has now shown how to do color binning on the chip during read-out, eliminating much need to hard-wire this binning: instead, the future sensors might as well all offer both the 60MP and 15MP options might as well both be there. Cheaper sensors could just have higher defect rates in 60MP mode, handled with slight resolution loss to bad pixel mapping and such.

60MP in 36x48mm will still have easily large enough photosites to have very good dynamic range and low shadow noise for a great amount of work at low to moderate ISO speeds, so it makes sense to keep the leaving binning as an option for cases (a distinct minority of all medium format work I would think) where higher ISO speed or very high DR are needed more than the extra resolution.

The idea that there is a market for a mere 15MP in a large and always rather expensive sensor is very unlikely.

The medium format market has mostly abandoned 6x6 (56x56mm) and larger formats in favor of 645 (42.5x56mm) along with somewhat smaller digital formats. The Hasselblad V series is down to one model (503CW) from about eight a few years ago, and the 503CW might be down to slowly selling off remaining stocks. Anyway, the V series has never been updated to auto-focus. Rollei has been in desperate situations for many years. So I greatly doubt that any sensor maker is going to produce a product only useful in those two fading glories, the Hasselbad V and Rollei 6x6 format bodies.
Logged
PhaqueName
Newbie
*
Offline Offline

Posts: 9


« Reply #8 on: October 23, 2006, 10:50:12 AM »
ReplyReply

Quote
The medium format market has mostly abandoned 6x6 (56x56mm) and larger formats in favor of 645 (42.5x56mm) along with somewhat smaller digital formats. The Hasselblad V series is down to one model (503CW) from about eight a few years ago, and the 503CW might be down to slowly selling off remaining stocks. Anyway, the V series has never been updated to auto-focus. Rollei has been in desperate situations for many years. So I greatly doubt that any sensor maker is going to produce a product only useful in those two fading glories, the Hasselbad V and Rollei 6x6 format bodies.
[a href=\"index.php?act=findpost&pid=81783\"][{POST_SNAPBACK}][/a]

the rumors about this rotating back on the new hiya6 scare me. it appears now that the back will have be pulled back from the body, in order to rotate, thus leaving ample opportunity for dirt/dust/grit to land on the sensor during a job. plus, the friction of constantly rotating the thing just seems to be asking for trouble. i would envision a square chip, but let there be a button on the camera or viewfinder, to indicate vertical or horizontal. by pressing the button, the viewfinder would electronically black out the out-of-frame areas, and then the software would tag the file, thus auto-cropping it into a horizontal or vertical, once the file was opened in the software. camera designers should be forced to design rotating backs in the same room conditions where the camera will eventually be used -- on the beach, under a waterfall, on a windy open landscape. simply, the camera should be sealed, or in the least, not force the photographer to expose the chip during a job.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5171


« Reply #9 on: October 23, 2006, 11:09:30 AM »
ReplyReply

Quote
the rumors about this rotating back on the new hiya6 scare me.
[a href=\"index.php?act=findpost&pid=81787\"][{POST_SNAPBACK}][/a]
I believe that there are DMF backs that allow sensor rotation "in place" by rotating a knob on the back of the back; surely that is the best approach, so long as you can tolerate the need for the unused larger square coverage of the mirror, lenses, internal anti-flare baffles, hoods, and VF.

A question: do or could any DMF backs have "dark slides", even maybe the nice built in ones that I believe Rollei uses in film backs? Rename them as "dust curtains" or "sensor protectors".
Logged
James Russell
Guest
« Reply #10 on: October 23, 2006, 12:57:18 PM »
ReplyReply

Quote
the rumors about this rotating back on the new hiya6 scare me. it appears now that the back will have be pulled back from the body, in order to rotate, thus leaving ample opportunity for dirt/dust/grit to land on the sensor during a job. plus, the friction of constantly rotating the thing just seems to be asking for trouble. i would envision a square chip, but let there be a button on the camera or viewfinder, to indicate vertical or horizontal. by pressing the button, the viewfinder would electronically black out the out-of-frame areas, and then the software would tag the file, thus auto-cropping it into a horizontal or vertical, once the file was opened in the software. camera designers should be forced to design rotating backs in the same room conditions where the camera will eventually be used -- on the beach, under a waterfall, on a windy open landscape. simply, the camera should be sealed, or in the least, not force the photographer to expose the chip during a job.
[{POST_SNAPBACK}][/a]


Mr. Name,

I doubt if your going to see sensors get larger, probably just smaller.

I have a friend that reminds me of you.  He's a good photographer but for some reason he keeps looking for the perfect camera.

It's kind of strange because my friend has already tried, owned or used the digital backs from all three of the top manufacturers.  In his film days he also shot with Fuji 680's, blads, Nikons, even tried a range finder or two, but now he just keeps searching.

In all actuality there is no perfect camera, never has been and the answer is most of us just find the camera, the camera back and develope our own workarounds for it.


BTW:  Love your flower photography [a href=\"http://www.pbase.com/canyonlu/image/364292]http://www.pbase.com/canyonlu/image/364292[/url]

Best,

JR
Logged
peterhorsley
Newbie
*
Offline Offline

Posts: 23


« Reply #11 on: October 24, 2006, 04:33:20 PM »
ReplyReply

Quote
The idea that there is a market for a mere 15MP in a large and always rather expensive sensor is very unlikely.
[a href=\"index.php?act=findpost&pid=81783\"][{POST_SNAPBACK}][/a]

15Mpx doesn't sound like much does it?  But then 17Mpx doesn't sound like much.  Photographers seem to be choosing to buy Aptus 17/65's (17/28Mpx), why?  It's not just $/pixel, a 1Ds Canon (16.7Mpx) is cheaper.

Maybe it's because a certain sized Mpx MF solution is a better 'photographic' solution for some photographers.

It's kind of lazy marketing to characterise the merits of a product by a single number.  Photographers are mostly interested in how their work ends up looking in it's finished form; print, magazine, projection, display.  Photographers wouldn't be considering using an MF camera if a single statistic was the only relavent factor in selecting a camera.  I think we are smarter than that.

Quote
In all actuality there is no perfect camera, never has been and the answer is most of us just find the camera, the camera back and develope our own workarounds for it.
[a href=\"index.php?act=findpost&pid=81783\"][{POST_SNAPBACK}][/a]

And what drives forums like these is that we are not just choosing the best available camera for our work.  It's more like lens range - camera body - digi back - file transfer interface - RAW software - computer O/S - support software - storage hardware: the camera is part of the workflow. It's a complex choice.

That's why a digi back with a modest number of better pixels is often going to be the right tool for fashion, weddings, and any other high volume, short turn around photographic work.

Quote
I doubt if your going to see sensors get larger, probably just smaller.
[a href=\"index.php?act=findpost&pid=81783\"][{POST_SNAPBACK}][/a]

...and photographers are staying the same size. Do we want to use smaller sensors?

Best

Peter
« Last Edit: October 24, 2006, 07:08:27 PM by peterhorsley » Logged
Steve Kerman
Full Member
***
Offline Offline

Posts: 142


« Reply #12 on: October 24, 2006, 07:13:59 PM »
ReplyReply

As I understand it, the manufacturers are already using the partially-bad dies.  They simply map out the bad pixels, and interpolate data from the neighboring good pixels.

Thus, one would expect that there would be rather few die available for the suggested "15 MPix" usage, as they would already be in use as "60 MPix" sensors.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5171


« Reply #13 on: October 25, 2006, 08:49:01 AM »
ReplyReply

Quote
15Mpx doesn't sound like much does it?  But then 17Mpx doesn't sound like much.  Photographers seem to be choosing to buy Aptus 17/65's (17/28Mpx), why?  It's not just $/pixel, a 1Ds Canon (16.7Mpx) is cheaper.
[a href=\"index.php?act=findpost&pid=82080\"][{POST_SNAPBACK}][/a]
Those 17MP sensors are also in the smaller and thus significantly less expensive 33x44mm format, instead of 36x48 or 37x49.  (What is the price difference between the 17MP, 33x44mm models and the ones with larger sensors?) I was talking about sensors in the larger "48mm wide" formats, and sensors that are capable of 60MP by simply not doing the "2x2" binning. And as I said before, I see no advantage to always binning down to 15MP over offering the choice of 15 and 60, even if the 60MP requires a bit of pixel mapping.

By the way, 60MP mode adds another option: heavy cropping when the lenses you have do not have the telephoto or macro reach to fill the frame with a certain subject. Even if you only need 15MP or less in the end, you have an extra 2x cropping latitude available.
Logged
peterhorsley
Newbie
*
Offline Offline

Posts: 23


« Reply #14 on: October 29, 2006, 07:47:07 AM »
ReplyReply

Better pixels, cheaper.

It seems kinda dumb but the point of this thread was: Better pixels.  
For commercial use 30 Mpx is enough.
Lets move on.
Photography is where technology meets craft and art.
Cameras are tools.
Digi-backs are fantastic, and flawed.

2008?

In the mean-time I'm makng the best pics I can...

Peter
Logged
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad