Worth moving into a tech camera system for wide angles at this stage?

[Bryan Stephens]

That is what I worry about most - if they announce something new, it's an instant loss of $$$

I own three Rodie HR lenses and I am not too concerned about the newer better lenses as the increment would be so minimal it would take an extreme examination to see where the image quality or sharpness would be improved. BTW the image quality of all three of the lenses I own is beyond amazing.

If they were to announce new lenses, then the older lenses would be a "steal"

 

[Shashin]

Are you sure about this? I think I used to see an analysis of the RAW files of the IQ260 somewhere before, and the range of levels is 0-65535 (16-bit) instead of the 0-16383 (14-bit, like the Nikon NEF). Do you mean that the Phase One digital backs actually samples with 14-bit DAC then does some interpolation to 16-bit when writing data into RAW files?

Just look at DxO Mark measurements. Sticking a 16-bit A/D converter in a camera does not make the sensor actually record 16-bits of information. I shoot with "16-bit" Phase p25+ back and scientific cameras. Basically, there are no 16-bit cameras on the market.

Chalk 16-bit claims up to marketing. It is a pity the dealers don't come clean, although my Phase dealer did when I questioned it.

 

[voidshatter]

Just look at DxO Mark measurements. Sticking a 16-bit A/D converter in a camera does not make the sensor actually record 16-bits of information. I shoot with "16-bit" Phase p25+ back and scientific cameras. Basically, there are no 16-bit cameras on the market.

Chalk 16-bit claims up to marketing. It is a pity the dealers don't come clean, although my Phase dealer did when I questioned it.

Where did you see evidence of 16-bit being marketing?

I have just checked with Raw Digger and confirmed that there are 65536 (2^16) levels for my IQ260 IIQ file. Do you mean these are fake values with holes and gaps?

 

[ErikKaffehr]

Hi,

The story is that CCD readout is noisy. It is reasonable to assume that a CCD cell can hold about 60000 electrons and readout noise may be 12-16 electrons.

So you take 60000 and divide by say 12 and that gives you a signal to noise ratio of 5000. 5000 is 2^12.29 so you need 12.29 bits to represent the signal. Now, bits don't come in parts, so 13 bit would be needed to represent signal.

The rest is simply noise.

One of the interesting points is that the IQ-250, which has a modern CMOS sensor is said to have a DR of 14EV by Phase One. 14EV DR corresponds to exactly 14 bits, and the IQ-250 is a 14 bit back. The IQ-280 is said to have a DR of 13 stops, but uses 16 bits to represent 13 bits worth of data.

Check this page, and look at "image quality": IQ2 series digital backs | Specifications

With the CCD backs they talk about 16-bit opticolor and 13 F-stops, with the IQ-250 just 14 F-stops. And the F-stops are essentially the same as bits.

Best regards
Erik

Where did you see evidence of 16-bit being marketing?

I have just checked with Raw Digger and confirmed that there are 65536 (2^16) levels for my IQ260 IIQ file. Do you mean these are fake values with holes and gaps?

 

[Shashin]

Where did you see evidence of 16-bit being marketing?

I have just checked with Raw Digger and confirmed that there are 65536 (2^16) levels for my IQ260 IIQ file. Do you mean these are fake values with holes and gaps?

What are in those registers? Are they empty? Because if this was a 16-bit camera, it should be leaving the 14-bit competition in the dust, rather than just keeping up with them:

Nikon D800E versus Phase One IQ180 Digital Back versus Pentax 645D - Side by side camera comparison - DxOMark

Yes, you can put a 16-bit A/D converter in a camera. But if you are not actually getting 16-bits of information, what is the point? If you think Phase and Hasselblad engineers don't knows this, then they are just bad at math and are wasting money (and probably should not be doing this work). Personally, I think it is a marketing choice. Numbers sell.

Don't get me wrong. Phase backs are great. Buy them. Use them. But I would not buy them based on a claim of 16-bit. Just like I would not drink Bud based on the promise it would make me handsome and irresistible to women. (Actually, even if it did, it would be too high a price to pay--I like beer.)

 

[Bryan Stephens]

Just like I would not drink Bud based on the promise it would make me handsome and irresistible to women. (Actually, even if it did, it would be too high a price to pay--I like beer.)

Is that what my problem is??? :ROTFL:

 

[dougpeterson]

Like I said several posts ago - it's a rabbit hole and will side track any useful conversation.

Focus on evaluating final image quality in real world situations which closely mirror your own (or better yet do your own actual tests). A dealer can help with raws, tests, demos or rentals to facilitate this. Phase One raws produce great color across a surprising variety of lighting scenarios and take a lot of post-processing abuse before showing deleterious artifacts - the nuances of analogue-digital converters (i.e. that not all X bit converters are equally well suited to a particular set of goals, the impact of a dark cal process on the requirements for the analogue path, the shortcomings of dXo's methodology in evaluating the complete-chain etc) will be of interest to some engineers, but will not help you find a camera that works well for your needs.

 

[torger]

I have just checked with Raw Digger and confirmed that there are 65536 (2^16) levels for my IQ260 IIQ file. Do you mean these are fake values with holes and gaps?

I worked with the IIQ format as a programmer so I know the format quite well. During decoding of the format the samples are expanded from 14 to 16 bit, so yes the 16 bit values are "fake". The "holes and gaps" are sort of filled in with a black level offset, but it will not produce real values anyway (as those black levels is not one per pixel). One of the few MF formats that actually store true 16 bit values is the Hasselblad format. If I remember correctly the Leaf .mos format store 14 bit values.

This does not matter though, as the last few bits are pure noise. The reason IIQ format doesn't store the last 2 bits is of course because it would not matter, it's just noise. So it's a good engineering decision.

That is, the pixels are sampled with a 16 bit ADC, but as the last few bits is only noise these are thrown away inside the digital back before stored in the raw file, so the raw converter never gets to see these values.

In other words, it's true that 16 bit ADCs are being used, but it's false those values are stored in full in the raw file, and you should not worry about that as the bits that are thrown away are pure noise.

So "16 bit color" is just a false marketing message. However, the MF CCD way of "high signal + high noise" vs the 135 CMOS of "low signal + low noise" may lead to a difference in look, CCD being more "film-like" and CMOS more "artificial", at least a number of years ago so if you're nice you could say that "16 bit color" was used as a tag to point out that difference. Today it's not so simple though, some CMOS sensors have both higher signal (higher full well capacity) and lower noise than the larger pixels in CCDs due to more effective pixels.

That said the IQ260 has great dynamic range, not the best you can get in a camera currently, but not far from it. It would be silly to pick a different camera because IQ260 would have "too poor dynamic range". An exception to this might be if you specialize in long exposures. Unfortunately there's not much data around how backs and cameras perform in long exposures, all become more noisy that's all I know. I have no idea if say a D810 will be less noisy in a 10 minute exposure than an IQ260 (it would be interesting to know though).

 

[torger]

Thanks for the reply! I understand that I can take an LCC shot with the Grad ND, but I would still get noise in local areas where the Grad ND hurts the foreground. This is still less than ideal.



Would you mind sharing the actual trading prices of the 28XL when the DALSA CCD became popular? Do you still remember the inductive price of the 28XL? I would be interested to know how much has depreciated. The 23HR and the 40HR were both introduced in 2008 and they are 6 years old now. Buying these only to find them out-of-date within a year or two might not justify the performance.



I just have a feeling that I am at the worst break points of technology breakthroughs. There was one from the KODAK to the DALSA, and now it is another one from DALSA to SONY. It's just like that I worry about buying an iPhone 5S right before the announcement of iPhone 6 plus.

Yes with grad ND you will get noise in local areas. In architecture subjects this can quite often be a problem (as the local area is often a building) and then bracketing/HDR merge would be more suitable, but in landscape I find that those local areas is often narrow trees and such which is not hurt as much by noise. Additionally, it will depend on your post-processing techniques, I usually keep a virtual grad as a part of the tonemapping (generally a bit weaker than the one used in the shot though) as I think it yields a more natural look. If you have a back-lit scene it's good to keep some of the back-lit feel, meaning that the tree sticking up will be quite dark which makes noise much less visible, otherwise you'll get a "grunge HDR" look which does not look particularly natural. That's my view anyway. Your mileage may vary

The 28XL can still be bought new for about $4.7k, and it still makes sense for Kodak sensors (like the Hasselblad H5D-50) or large pixel dalsa sensors like my own Aptus 75. Searching for sales prices I find more "want to buy" than actual sales. The 28XL is a bit special case, as it was released at a time the 6um Dalsa was already popular so not many were bought. Some were still bought, and note that some do use them on 6um Dalsa, it works okay if you don't shift much, and the only other alternative the Rodenstock Digaron-S 28 has a hard limit on 70mm image circle. I'd guess that there are more users on suitable backs interested in this lens than there have been sold new (which WTB indicates) so the deprication from new price would be about the same as any other lens. The most recent sales price I found was from march 2013 and then it was sold for about 70% of the purchase price, ie a typical sales price for any lens in good condition.

For a lens to totally lose value it must be so old that there are no users left shooting with backs compatible with the lens, and there's still a number of years before that will happen.

If you want to stay yourself on the latest and greatest you may need to upgrade in a few years yes, but you will be able to sell your lenses to a good price. I don't think it's too likely that we'll see a new tech lens line soon. One problem is the central shutter mount. To make lenses more retrofocus than today the lenses will be even more complex and heavy than the Rodenstock 32mm, which already today is at the limit what a Copal 0 shutter can handle. New lens designs would require a different type of shutter than Copal 0, and I don't think the current Copal 0 compatible electronic shutter is any better. A much wider shutter thread is required than today to support heavier lenses. I still hope for that we'll see wide-angle compatible CMOS sensors before we see a new tech lens line. While MF-SLR lenses are designed with some residual chromatic abberations etc that is corrected in the raw converter (=simpler optical designs possible), tech lenses are designed to work well without lens corrections. Thanks to not having to be much retrofocus (or completely symmetrical) the complexity can be kept low and performance high, but the more retrofocus you need the more complex the optical design must be to work well without lens corrections. The Rodenstock 32 might then end up looking as a quite simple, light and cheap lens in comparison. That would not be a great development of tech cameras I think.

It's true that lots of things is happening right now, the future of tech cameras is very fuzzy, it's not easy to predict what direction it will go. Due to the high performance of current systems the second hand value will be pretty good, in terms of percent, but things like digital back will lose many dollars anyway due to the high purchase price.

As you're shooting a D800E today you don't necessarily need to buy a IQ260 and Rodenstock lenses to get an improvement. You could go for some lower end. If that DR metric is very important to you then it will be more difficult to use older CCDs though... The P65+ should still be rather good in terms of DR though, all 6um Dalsa-based digital backs are pretty good in terms of DR. Kodaks and larger pixel sensors is one step behind in terms of DR. I don't think it's a big issue, but it will depend on shooting and post-processing style of course.

For me personally the current dream system is a H5D-50 digital back and a complete Schneider Digitar lens line on a Linhof Techno. It's not the "best" considering all metrics, but it suits my creative style the best, and it also fits my "engineering values" how a tech cam should be. I don't believe that lenses should be as heavy and complex as possible with utmost sharpness as only metric. I think symmetrical simple designs is the right way to go for "large format style" creativity (that I can actually hike with the gear and have many lenses without breaking my back is important to me too), and the Kodak 50 megapixel sensor is the best at handling that currently, it also has a nice balanced size for the 90mm image circles.

For a different user there's a different dream system, the choice of tech camera system is very personal. In a way, that's part of the fun. You'll end up with a very personal camera.

 

[jagsiva]

It's just that I have a feeling like buying an iPhone 5S right before the announcement of iPhone 6 plus. I might be wrong but I have a bad feeling on this, given that the Rodenstock lenses such like the 23HR and the 40HR are already 6 years old and they seem to be at the breakpoints when the design needs to be refreshed to keep up with the most current digital technology (in this case, the SONY CMOS like the IQ250, IQ250, Credo 50 etc).

I don't believe this is a valid comparison when it comes to lenses. On electronics, absolutely, there is an initial depreciation in the fist 2 years of 40-50%. This is primarily driven by the the technology improvements, and customer see the additional value in the new technology.

Glass, on the other hand, I feel is different. Glass like the Rodie HR, 2nd gen SK-XL, Leica S, and most M, and even the OTUS are being developed at the edge of "physics". I am not sure there will be a "needle moving" advancement in optical science that will make these lenses drop in value.

Now, there are lenses with lots of electronics, and high-quality glass like the Canon big whites, that do drop in price when a new generation is introduced, but this again is for the most part driven by electronics technology - IS, weight, etc.

The lenses we are talking about here are the absolute best in design and materials with very high-quality, but simple designs. Take the 32HR for example. It is likely the best lens I have ever used for landscapes. If I decide next week, I want an electronic shutter, I can do that. If I want to go to a view camera, I can swap a board and do that. If I want to use an FP on Alpa or Arca, I can do that too. If CMOS sensors come out in FF (or whatever we are calling 645FF these days), pigs fly, and Phase comes out with a magnesium mirror less body with an EVF, I'm sure a mount could easily be adapted to use this very lens. As for the lens itself, I doubt very much there will be something better in the current context of lens physics.

I have way too many lenses for way too many systems. I would count the Rodie 23/32/40/90SWHR, SK60XL/120ASPH, Coastal Optics 60, and likely the two new OTUS lenses as ones that will outlive several bodies and systems.

Having said this, there is a new and alternate wave of high-quality "system-output" with brilliant output IQ. By this, I mean a tightly integrated chain of lens, sensor, firmware, black-box RAW processing, and post-capture RAW processing. The files out of the Sony A7R with 55.18FE are in this category. Again, there is a lot of black-magic cooking going on here even before I touch the RAW file, do I care, no. But it is a different way of skinning the cat. Take a Leica 50LUX and put on the same camera, and the results are quite different.

I do see more of this happening going forward.

 

[narikin]

If Rodenstock were bringing out new Tech Cam lenses in the next year, they would have announced them at Photokina. They did not. It's clearly not on the horizon.

They also made the error of redesigning lenses around a particular sensor generation many years ago, with the Digaron-S lenses for less than FF CCD sensors: fabulous lenses, fast, and super sharp, but then guess what - tilt/shift tech cams arrived, plus new FF sensors came out. I very much doubt they will repeat that error of redesigning everything just for this generation of Sony CMOS, with all its crosstalk problems - in other words: sensor technology will evolve before the lenses.

A last point is that when they did release the new 90HRSW (Yellow band) it was double the price of the old 90HR (blue band). The old lenses have retained a LOT of their original purchase value, as the new ones are much more expensive. I sold my used blue one for 90% of what I paid for it. The market for MF digital lenses of the HRSW caliber is very small, expectations and quality demand is high. The same reason why an Otus is 4x the price of a regular 'premium quality' lens. I don't think an Otus redesign will happen for 10 years at least. Likewise for new generation Roddy HR's.

 

[torger]

The greatest factor of uncertainity I think is if we will see a return of wide angular response sensors or not. The problem is that technical wide angle photography is unique in requiring this feature, as far as I know, and it's not likely a niche photography need will drive sensor features.

I would not be surprised if the full-frame 54x41mm CMOS when it eventually comes will be say 120 megapixels and have even more narrow angular response -- as high pixel count sells (but smaller pixels tend to have more crosstalk issues than larger) and it does not matter for mirrorbox cameras and indeed not even mirrorless designs which in the 135 world still have quite long flange distance.

In really small sensors like in mobile phones there's a need to enlarge photo diode area so there are things like back-illumination putting the photodiode at the surface and make it larger. However, I don't think there's the same gain in larger sensors, just put a microlens there and you gather all the light you need anyway. So it's not guaranteed that the small sensor tech will find it's way to MF-sized sensors.

If the trend of narrower angular response sensor continues, then the current tech wide angle lenses will live as long as the current CCD backs lives, say 10 years from now.

I don't have enough knowledge in sensor design and sensor markets to know how likely or not it is to see a return of wide angular response. It would be interesting to talk to a Sony or Dalsa engineer about that. I'd like to see a 60 megapixel 6um full-frame CMOS with light-shields, that would be a good start (ie you block the low angle light instead of letting it pass to the next pixel and become crosstalk, it's not as good as capturing it in the right pixel but with high DR CMOS losing a stop or two at the edge would be no disaster) and can probably be made with current technology, if just someone other than Kodak from 2008 thinks that light-shields is a good idea to have in an MF sensor.

As far as I understand and can interpret the cross-section there is no particular light-blocking structures between the pixels in the Sony CMOS sensor, some is caught on the wiring, but no targeted attempt to reduce crosstalk is there, unlike in Kodak sensors that has light-shields. Why? My guess is that it was simply considered unnecessary as none of the typical photo applications need them. Wide angle tech photography is a tiny niche, so tiny that Sony sensor engineers might not even know about its requirements.

Anyway, ever since photography became digital it's no longer possible to plan your system 10-20 years ahead. Noone knows which turns technology will take.

 

[Paul2660]

It's interesting to research the whole Crosstalk issue. It's actually been around as a term much longer than I realized. There are a lot of white papers from various Cell phone companies, that realized this issue a while ago, and they were concerned about it on center obviously. These companies, use the term "crosstalk" in reference to the issues that Torger has referred to. So I believe the term has been around at least in the the development teams for a while.

CMOSIS, I believe is the chip in Lecia, seems to have quite a bit of info out there. They appear to be using a different designed micro lens, one that is to help keep the light rays from having such an angular error and stay straighter.

Lot's of known facts about crosstalk, and interesting to read.

Paul

 

[torger]

In very small sensors where pixels sizes are like 1um, ie mobile phones, you can get crosstalk from diffraction effects, ie even if the light comes straight at the pixel it can scatter around and leak into the next. With these very tiny pixel sizes the relative height also becomes high, a cross-section of the pixel looks like a deep deep well with the photodiode at the bottom. Compared to this type of sensors the Sony MF CMOS has a wide angular response. In other words, the literature on sensor crosstalk is in many cases dealing with solutions to increase an angular response from 7 degrees to 15 degrees in mobile phone sensors, but what symmetrical tech wides need is ideally 120 degrees, but "only" 80 degrees will probably be enough.

Anyway here's a comparison between the cross-section of a KAF-39000 pixel (Kodak 39 megapixel sensor, used in P45+) which is similar in design to the newer KAF-51000 50 megapixel sensor, and then a cross-section of the Sony sensor used in a D800, similar design to the Sony MF sensor. In the Kodak sensor you can see how flat the pixel is, it's wider than it's high and it also have light shields. With the Sony sensor the pixel is higher than it's wide, and there's no dedicated light shields. Sure there's wiring to block some light, but only in one direction (that there's open space in the other direction is not seen from this angle).

CMOS sensors require more components on the chip itself, so I guess it's harder to make CMOS pixels low in height, ie it's easier to design a CCD chip with wide angular response than a CMOS chip. While CMOS brings us live view, it may also break the wide angle designs we love. But say if you enlarge the pixel to 6um, use the most recent technology with the flattest wiring and add some light shields, maybe you could compete with the old Kodak CCDs. We'll see...

 

[jordan.j]

The greatest factor of uncertainity I think is if we will see a return of wide angular response sensors or not. The problem is that technical wide angle photography is unique in requiring this feature, as far as I know, and it's not likely a niche photography need will drive sensor features.

I would not be surprised if the full-frame 54x41mm CMOS when it eventually comes will be say 120 megapixels and have even more narrow angular response -- as high pixel count sells (but smaller pixels tend to have more crosstalk issues than larger) and it does not matter for mirrorbox cameras and indeed not even mirrorless designs which in the 135 world still have quite long flange distance.

In really small sensors like in mobile phones there's a need to enlarge photo diode area so there are things like back-illumination putting the photodiode at the surface and make it larger. However, I don't think there's the same gain in larger sensors, just put a microlens there and you gather all the light you need anyway. So it's not guaranteed that the small sensor tech will find it's way to MF-sized sensors.

If the trend of narrower angular response sensor continues, then the current tech wide angle lenses will live as long as the current CCD backs lives, say 10 years from now.

I don't have enough knowledge in sensor design and sensor markets to know how likely or not it is to see a return of wide angular response. It would be interesting to talk to a Sony or Dalsa engineer about that. I'd like to see a 60 megapixel 6um full-frame CMOS with light-shields, that would be a good start (ie you block the low angle light instead of letting it pass to the next pixel and become crosstalk, it's not as good as capturing it in the right pixel but with high DR CMOS losing a stop or two at the edge would be no disaster) and can probably be made with current technology, if just someone other than Kodak from 2008 thinks that light-shields is a good idea to have in an MF sensor.

As far as I understand and can interpret the cross-section there is no particular light-blocking structures between the pixels in the Sony CMOS sensor, some is caught on the wiring, but no targeted attempt to reduce crosstalk is there, unlike in Kodak sensors that has light-shields. Why? My guess is that it was simply considered unnecessary as none of the typical photo applications need them. Wide angle tech photography is a tiny niche, so tiny that Sony sensor engineers might not even know about its requirements.

Anyway, ever since photography became digital it's no longer possible to plan your system 10-20 years ahead. Noone knows which turns technology will take.

I don't think the back-illuminated CMOS or the curved sensor would come to the medium format digital in the near future.

But I have a simple question:

Why can't Rodenstock or Schneider just design a lens like the Canon 17mm TS-E?

Is it now a safer bet to go for the Canon 17mm TS-E + ALPA FPS + Credo 50 / IQ150 Route and wait for the 645 fullframe SONY CMOS?

 

[ErikKaffehr]

Hi,

Some questions arise:

1) Is there a market for a Canon 17 TS-E like lens from Schneider & Rodenstock?
2) Is that market large enough to generate satisfying return on investment?
3) Can the lens be made at reasonable cost?
4) Does Schneider or Rodenstock have the capabiity to produce that lens?

Regarding future versions of the Sony sensor little is known. It is not even known what market impact the present sensor has. Much of the initial marketing emphasis was on high ISO performance, and it may be that the camera is mainly oriented towards the MDSLR market.

Will the actors migrate to CMOS or stay with CCD? Will other CMOS vendors emerge? Will larger CMOS sensors be developed? Will other CMOS-vendors technologies be compatible with Sony's technology?

I guess that we may see some move to CMOS sensors being compatible with larger beam angles, as this may be beneficial in building smaller cameras.

Best regards
Erik

But I have a simple question:

Why can't Rodenstock or Schneider just design a lens like the Canon 17mm TS-E?

Is it now a safer bet to go for the Canon 17mm TS-E + ALPA FPS + Credo 50 / IQ150 Route and wait for the 645 fullframe SONY CMOS?

 

[ErikKaffehr]

Hi Anders,

Thanks for explaining the issue. Your description is dead on.

I strongly feel that the 16-bit depth is quite a bit of desinformation, and it quite often shows up even in well written articles as an advantage of MFDBs.

It is a pity that so much marketing is focusing on myths, instead of discussing the real benefits of MFD, whatever that may be. Giving some bad information reduces the credibility of all information.

Best regards
Erik

I worked with the IIQ format as a programmer so I know the format quite well. During decoding of the format the samples are expanded from 14 to 16 bit, so yes the 16 bit values are "fake". The "holes and gaps" are sort of filled in with a black level offset, but it will not produce real values anyway (as those black levels is not one per pixel). One of the few MF formats that actually store true 16 bit values is the Hasselblad format. If I remember correctly the Leaf .mos format store 14 bit values.

This does not matter though, as the last few bits are pure noise. The reason IIQ format doesn't store the last 2 bits is of course because it would not matter, it's just noise. So it's a good engineering decision.

That is, the pixels are sampled with a 16 bit ADC, but as the last few bits is only noise these are thrown away inside the digital back before stored in the raw file, so the raw converter never gets to see these values.

In other words, it's true that 16 bit ADCs are being used, but it's false those values are stored in full in the raw file, and you should not worry about that as the bits that are thrown away are pure noise.

So "16 bit color" is just a false marketing message. However, the MF CCD way of "high signal + high noise" vs the 135 CMOS of "low signal + low noise" may lead to a difference in look, CCD being more "film-like" and CMOS more "artificial", at least a number of years ago so if you're nice you could say that "16 bit color" was used as a tag to point out that difference. Today it's not so simple though, some CMOS sensors have both higher signal (higher full well capacity) and lower noise than the larger pixels in CCDs due to more effective pixels.

That said the IQ260 has great dynamic range, not the best you can get in a camera currently, but not far from it. It would be silly to pick a different camera because IQ260 would have "too poor dynamic range". An exception to this might be if you specialize in long exposures. Unfortunately there's not much data around how backs and cameras perform in long exposures, all become more noisy that's all I know. I have no idea if say a D810 will be less noisy in a 10 minute exposure than an IQ260 (it would be interesting to know though).

 

[torger]

But I have a simple question:

Why can't Rodenstock or Schneider just design a lens like the Canon 17mm TS-E?

Is it now a safer bet to go for the Canon 17mm TS-E + ALPA FPS + Credo 50 / IQ150 Route and wait for the 645 fullframe SONY CMOS?

The Canon TS-Es are designed to deliver no steeper angle than 20 degrees at the image circle edge (which actually is a little bit too much for the A7r, but fine for the Credo 50), the Rodenstock Digarons about 35 degrees (my estimate from the data sheets, have not been able to verify it as I don't own any of the wides), and Schneider being symmetrical have the same angle towards the sensor as the field of view, meaning up to almost 60 degrees for the 28XL (to get down to/below 35 degrees with Schneider you must up to the 72mm focal length).

Anyway I think both Rodenstock and Schneider could design such a lens if they wanted to. Maybe not exactly like the Canon (Canon have some unique manufacturing techniques), but the concept of a heavy retrofocus lens. I'm no optical design expert but with today's computer-aided methods I don't think it's that hard to do, what makes optical design difficult is to choose appropriate tradeoffs in the design, to balance conflicting goals.

If you would want to make a lens as sharp as the current Rodenstock Digarons (note: Canon TS-E II lenses while good is not as sharp as the Digarons), but change the angle from 35 to 20 degrees, you would get the problem that you need even more glass elements, and still the barrell distortion may increase, and to counteract that you would need even more glass. You would end up with a lens heavier than 1kg, and a normal Copal shutter would not be able to support that so you would either need to use an oversized Copal (with slower shutter speed etc) or make some new shutter.

Normal wide angle lenses that cannot be shifted has the advantage that you can apply lens corrections in the raw converter, and most modern lenses are designed for that. If you turn off lens corrections the Hasselblad H and Phase One 645DF wides will show quite some chromatic abberations, ie they have made simpler optical designs and correct them in the raw converter. Tilt/shift lenses don't have that luxury, they need to perform without lens corrections, and users expect them to perform better than the corrected MF-DSLR lenses.

The ability to do that has been based on the optical design formula tradeoff that you don't need large aperture and you can have steep angle towards the sensor. Making the lenses more and more retrofocus take away some of that, and the only way around that is to make an extremely complex design, which of course would mean a very expensive lens, and a very heavy lens, and making it unfeasible to use the traditional shutters.

So it's just way better to get a CMOS that can handle steeper angle... if I were Schneider or Rodenstock I would sit tight for a while and hope that it will happen.

A comparison example: the retrofocus Rodenstock Digaron-W 32 has 14 lens elements and weights 800 grams, the symmetrical Schneider Digitar 35 has 8 lens elements and weighs 240 grams. A new 32 with more retrofocus would have even more lens elements and weigh even more (and cost more) than the current Digaron-W. Already the 32 has problems with loading the Copal shutter too much so it's very delicate, more than one have got their lens bent at the shutter just by setting down the tripod a bit too hard.

Zeiss Otus series is another example of what happens when you take the optical design to an extreme, the 55mm lens weighs 970 grams and has 12 lens elements which is a lot for a 55mm lens and its small format image circle. I'm very curious how a 24mm wide angle in the Otus series would be designed...

Keep in mind that 17mm and even 24mm on a 44x33mm sensor will be extremely wide angle. Some indeed like to shoot with ultra-wides, and in architecture you may have no other choice in some situations, but I think for landscape photography you get nicer perspectives if you don't go wider than "24mm 135 FOV", ie about 31mm for a 44x33mm sensor and 38 for full-frame 645, with shift available you can go a bit narrower still ie Digaron-W 32 for the 44x33 and Digaron-W 40 for the full-frame 645. For the 48x36/49x37mm sensor size 35mm is quite nice, and indeed that's my widest lens currently.

As things look today, Alpa is probably the best tech brand to go for when it comes to adapting to future changes in lens/sensor combinations. Arca-Swiss is also pretty strong (the Photokina 2014 product releases showed that they can do electronics too), while Cambo lags and Linhof even more so, as they don't have the same ability to make electronics, at least not yet. I have still have no problem with using Linhof "old school" design, as I like that type of design and if tech cams will change direction into a boring oversized A7r type of design I will most likely drop out of tech cams when it's no longer possible to continue the traditional way. That dark scenario would still mean that I have 10 years of Linhof shooting ahead, only when my Copal shutters fails and I can't have them repaired or replaced and my digital back fails and I can't get it repaired or replaced with a something compatible with symmetrical lens design that journey will end.

 

[jerome_m]

Normal wide angle lenses that cannot be shifted has the advantage that you can apply lens corrections in the raw converter, and most modern lenses are designed for that.

Actually, the Hasselblad converter (Phocus) will apply corrections even when the lenses are shifted using the HTS system. For this to work, the tilt and shift values are stored in the exifs.

If you turn off lens corrections the Hasselblad H and Phase One 645DF wides will show quite some chromatic aberrations, ie they have made simpler optical designs and correct them in the raw converter.

Not as much as one would think for the H lenses, chromatic aberrations are actually very small. What Phocus mostly corrects is distortion, which is not low enough for reproduction work without digital correction.

 

[jerome_m]

Hi everyone, I mostly photograph landscapes and architecture, and after lurking in this forum for a while, I decide to register a forum account and ask your ideas about joining this club.

I have been shooting with a Nikon D800E. I am not satisfied with the corner sharpness and the chromatic aberration of the Nikon 14-24mm lens. When I saw the pictures taken by the Rodenstock 23mm and 35mm lens I was astonished by the corner sharpness. As a fanatic pixel peeper of wide angles I am now deeply fascinated by the HR lenses.

As somebody else noted, you should have your zoom checked. Corner sharpness is not perfect on the 14-24 but chromatic aberration is relatively low. You may also want to test the 16-35 VR.

An alternative would be a H5D-50 with a HCD 28 and HC 50-II lens. These two lenses are real gems. The camera can also bracket automatically, which would solve dynamic range problems.