Pages: 1 2 [3] 4 5 ... 8   Go Down

Author Topic: Why are the tests of IQ250 against CCD-backs taking so long?  (Read 35292 times)

Ken R

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 849
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #40 on: February 07, 2014, 10:26:07 pm »

Good news is that we got the article up (and all the various small typos and issues changing from final draft to the blog platform and formatting and the LL team being swamped during and after their trip to Antarctica).

IQ250, CMOS Fully Realized

Bad news is I'm still organizing and uploading sample files and writing up our findings. Great news is I think they were be very useful in evaluating the real-world usable image circle for each of the tested lenses for each of the tested backs.

Awesome article! Thx for posting. You really clear up a lot of points that have been talked about in these and other forums, but in a much more clear and concise manner. Awesome.

For me the most telling image in the article is the couple with the puppy sitting on the white sofa. Red, pink and skin on the same image are usually a nightmare for most CMOS DSLRs. The skin tones look very pleasing in the IQ250 image and while the color is saturated and the reds vibrant they are not overly so. Most DSLRs clip the reds easily and you loose detail in those areas. Also, Color differentiation on the IQ250 seems outstanding, again, another area where DSLRs struggle.
Logged

Theodoros

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2454
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #41 on: February 08, 2014, 04:08:46 am »

Out of all articles I've ever seen about MF sensor quality, this is the first one that the pictures presented and the development of them reminds so much of a common CMos DSLR… If I didn't know that the article is about an MFDB, I would be confident that this is a next step Dslr that claims to have surpassed other high end DSLRs in DR and colour stability, by sacrificing a bit of its higher Iso performance… Hence (in my mind) the question still remains: "Why should one buy an MFDB (in ten times the price) to only improve a little than a current high end Dslr"? …especially if he can wait for some time and purchase a next generation Dslr that will (inevitably) carry the improvements. To my mind, MFDBs are not there to provide higher resolution than Dslrs, but to supplement what Dslrs can't do or are unsatisfactory in performing…. and this is colour accuracy at lower Iso, Tech and view camera compatibility, multishot, ability to "dig" deeper in the shadows with colour still present, and colour stability with respect to colour tonality. Sorry, but I don't see the special project which the new back can carry out and that alternative (and cheaper) equipment won't be able to perform to the same or better level…  :(
Logged

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3265
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #42 on: February 08, 2014, 04:31:56 am »

I think just knowing that there is a CMOS in this back will make some disregard from it and prefer CCD, it's going to be a HiFi vinyl vs CD thing. And as such, not all arguments will be sane  ;)
Logged

peterv

  • Full Member
  • ***
  • Offline Offline
  • Posts: 160
    • facebook
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #43 on: February 08, 2014, 04:39:01 am »

To my eye, om my MBP and Eizo, the skin colors aren't very pleasing, actually in the neighbourhood of what I have come to expect of "CMOS-colors". I think the skin tones in the portraits are too yellowish and the lady in the river, her legs look red/magenta.

Anyway, it is still early in the testing/fine-tuning phase I guess, so who knows? Thanks for the enjoyable article.
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11288
    • Echophoto
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #44 on: February 08, 2014, 04:39:41 am »

Hi,

The first part of the article is about the development of the IQ platform. I would say that the IQ platform has been met by positive response.

The other part discusses the integration of the Sony CMOS sensor into the IQ-series. Here there is a significant difference. Obviously the sensor vendor offers a choice of CFAs, either more optimized for high ISO or more orthogonal ones. I would suggest that different choices are also made for DSLRs, and maybe even suggest that the choice is not simply between ISO and fidelity.

The big difference between Sony CMOS and Kodak/DALSA CCDs is that the Sony CMOS delivers a digital output while the CCDs deliver an analogue output. With CCDs Phase One can do a lot to optimize signal paths, but with CMOS they just read the digital signals of the chip. There are probably quite a few tunable parameters on the CMOS sensor.

Once you have the signals coming out of the chip, and in case of CCD sensor amplified and converted to digital the image is written to a file. The next step to convert that file to an image, and here is the work of the "Image Professor" comes into play, creating camera profiles to reproduce colour.

Please note, nothing of this is about MF, many compact cameras had CCDs for a long time. Many modern DSLRs have CMOS with on sensor column converters, and that is the technology delivering best shadow detail, but some still have off sensor ADCs (all Canons, Nikon D700 and Nikon D4).

There may be some advantage for Phase One:

1) Better calibration data. Each camera is probably independently measured, and calibration data added to raw file
2) The profiles may be tweaked for skin reproduction under strobe light. While more generic cameras may be optimized for a wider set of conditions.

An interesting question may be why Phase One doesn't make multishot cameras. I guess that the main reason is that the customers they are addressing don't use multishot.

Another point is that DSLR vendors could very well design a camera optimised for studio portrait, but I guess that sector may not be their main sector. There are plenty of studio shooters perfectly happy with Canons and Nikons, so working with Canon or Nikon is probably OK for quite many photographers. I also guess it is a bit about learning the equipment.


Best regards
Erik


Out of all articles I've ever seen about MF sensor quality, this is the first one that the pictures presented and the development of them reminds so much of a common CMos DSLR… If I didn't know that the article is about an MFDB, I would be confident that this is a next step Dslr that claims to have surpassed other high end DSLRs in DR and colour stability, by sacrificing a bit of its higher Iso performance… Hence (in my mind) the question still remains: "Why should one buy an MFDB (in ten times the price) to only improve a little than a current high end Dslr"? …especially if he can wait for some time and purchase a next generation Dslr that will (inevitably) carry the improvements. To my mind, MFDBs are not there to provide higher resolution than Dslrs, but to supplement what Dslrs can't do or are unsatisfactory in performing…. and this is colour accuracy at lower Iso, Tech and view camera compatibility, multishot, ability to "dig" deeper in the shadows with colour still present, and colour stability with respect to colour tonality. Sorry, but I don't see the special project which the new back can carry out and that alternative (and cheaper) equipment won't be able to perform to the same or better level…  :(
Logged
Erik Kaffehr
 

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8815
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #45 on: February 08, 2014, 04:49:52 am »

I think just knowing that there is a CMOS in this back will make some disregard from it and prefer CCD, it's going to be a HiFi vinyl vs CD thing. And as such, not all arguments will be sane  ;)

Correct (except for that there is a difference between vinyl and CDs). Try a double blind test on a couple of images, and the prejudiced will miserably fail in identifying which is which, CMOS or CCD. Because there is no difference, other than the Bayer CFA used.

Apparently, from Doug's article, Phase One had some influence on Sony to choose a different trade-off between color accuracy and overall sensitivity compared to regular DSLRs built for speed. I don't know how close the match is between the current CCD CFAs and the final choice of the CMOS CFAs, so there might still be some difference, but I doubt it would be easily detected in real images. The higher DR of the CMOS is probably the telltale sign that identifies it, not the color accuracy, especially at higher ISO settings (or underexposure).

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Christoph C. Feldhaim

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2509
  • There is no rule! No - wait ...
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #46 on: February 08, 2014, 04:56:42 am »

Correct (except for that there is a difference between vinyl and CDs). Try a double blind test on a couple of images, and the prejudiced will miserably fail in identifying which is which, CMOS or CCD. Because there is no difference, other than the Bayer CFA used.

Apparently, from Doug's article, Phase One had some influence on Sony to choose a different trade-off between color accuracy and overall sensitivity compared to regular DSLRs built for speed. I don't know how close the match is between the current CCD CFAs and the final choice of the CMOS CFAs, so there might still be some difference, but I doubt it would be easily detected in real images. The higher DR of the CMOS is probably the telltale sign that identifies it, not the color accuracy, especially at higher ISO settings (or underexposure).

Cheers,
Bart

Concerning the conflict between speed and color accuracy I don't understand why they don't just sacrify one of the green pixels in the Bayer pattern for an unfiltered one and make the ramaining 3 more specific to get better color separation.

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8815
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #47 on: February 08, 2014, 05:03:40 am »

Concerning the conflict between speed and color accuracy I don't understand why they don't just sacrify one of the green pixels in the Bayer pattern for an unfiltered one and make the ramaining 3 more specific to get better color separation.

Hi Chris,

Probably because that reduces the accurate demosaicing of the green pass-band. It would help if a 4x (or 16x) multistep capture was used, because that samples each pixel position with all filters, and reduction of exposure time would allow to reduce the risk of uneven lighting between sub-exposures.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Christoph C. Feldhaim

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2509
  • There is no rule! No - wait ...
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #48 on: February 08, 2014, 05:11:23 am »

Hi Chris,

Probably because that reduces the accurate demosaicing of the green pass-band. It would help if a 4x (or 16x) multistep capture was used, because that samples each pixel position with all filters, and reduction of exposure time would allow to reduce the risk of uneven lighting between sub-exposures.

Cheers,
Bart

This appears to me as my example being not a trade between speed and color accuracy then, but a trade between spatial resolution and color accuracy.
With the current high resolution solutions this should be a viable option.
Cheers
~Chris

eronald

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6642
    • My gallery on Instagram
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #49 on: February 08, 2014, 05:25:02 am »

This appears to me as my example being not a trade between speed and color accuracy then, but a trade between spatial resolution and color accuracy.
With the current high resolution solutions this should be a viable option.
Cheers
~Chris

I think Kodak had a design like this.

Edmund
Logged
If you appreciate my blog posts help me by following on https://instagram.com/edmundronald

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11288
    • Echophoto
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #50 on: February 08, 2014, 08:26:51 am »

Hi,

I would think that the issue is a bit overrated. For some cameras high ISO is important for marketing, but I guess that high ISO is more about noise reduction than orthogonality of colour filters.

I am not so sure orthogonal filters are optimal. I am pretty sure some overlap is needed to be able to reproduce subtle colours.

CCD vendors normally publish data for spectral sensivity but I have not seen similar data from CMOS vendors.

Best regards
Erik

This appears to me as my example being not a trade between speed and color accuracy then, but a trade between spatial resolution and color accuracy.
With the current high resolution solutions this should be a viable option.
Cheers
~Chris
« Last Edit: February 08, 2014, 08:59:21 am by ErikKaffehr »
Logged
Erik Kaffehr
 

Christoph C. Feldhaim

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2509
  • There is no rule! No - wait ...
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #51 on: February 08, 2014, 10:53:01 am »

Hi,

I would think that the issue is a bit overrated. For some cameras high ISO is important for marketing, but I guess that high ISO is more about noise reduction than orthogonality of colour filters.

I am not so sure orthogonal filters are optimal. I am pretty sure some overlap is needed to be able to reproduce subtle colours.

CCD vendors normally publish data for spectral sensivity but I have not seen similar data from CMOS vendors.

Best regards
Erik


I assume you are thinking in terms of orthogonal vectors describing a vector space.
I am not sure how an orthogonal design could work at all here even if possible to construct and after all the human eye has overlap as well.
You know that a set of points in a 2 dimensional space (the spectral power distribution) is converted into a single point in a 3-dimensional RGB colorspace -
this is an irreversible data reduction - maybe not mathematically, but surely from an engineering point of view.
This happens in a camera and it happens in the human eye.
I can't imagine how orthogonality would apply in this kind of transformation.

If the CFA would be able to exactly mimic that process in a way how it is done in the system of trichromatic vision in the human retina we'd have at least
something equivalently powerful as human trichromatic colorvision, I think.
I don't know if anyone has ever designed a CFA with a spectral response identical to the response of the cones in the human eye or if there are maybe even reasons against that though.

Cheers
~Chris

Vladimirovich

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1311
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #52 on: February 08, 2014, 11:18:16 am »

Awesome article!
except for tiresome mantra about "CMOS colors" with the word CMOS in it... it is the time to really drop it... it is a blood libel really in camera's world  ;D
Logged

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3265
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #53 on: February 08, 2014, 12:00:29 pm »

Phase One has before this back was introduced gained from exaggerating the differences between CMOS and CCD to distance them from DSLRs. So my guess is that the difference has not been that large in actuality, it's all about the CFA, and once you have that you don't need to develop your profiles with any different methodology. So I'm sure color from this back is good.
Logged

Theodoros

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2454
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #54 on: February 08, 2014, 12:30:22 pm »

Phase One has before this back was introduced gained from exaggerating the differences between CMOS and CCD to distance them from DSLRs. So my guess is that the difference has not been that large in actuality, it's all about the CFA, and once you have that you don't need to develop your profiles with any different methodology. So I'm sure color from this back is good.
Lets put it this way… judging from the article's posted pictures, "colour from this back looks good (better) and (more) stable with respect to what we are used to call as "Cmos colour looks" when it's created by another Sony sensor Dslr…"
Logged

eronald

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6642
    • My gallery on Instagram
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #55 on: February 08, 2014, 02:43:32 pm »

This is known as the Luther-Ives condition.
I think Jenoptik had some very good filters on one of their colorimeters.

Edmund

I assume you are thinking in terms of orthogonal vectors describing a vector space.
I am not sure how an orthogonal design could work at all here even if possible to construct and after all the human eye has overlap as well.
You know that a set of points in a 2 dimensional space (the spectral power distribution) is converted into a single point in a 3-dimensional RGB colorspace -
this is an irreversible data reduction - maybe not mathematically, but surely from an engineering point of view.
This happens in a camera and it happens in the human eye.
I can't imagine how orthogonality would apply in this kind of transformation.

If the CFA would be able to exactly mimic that process in a way how it is done in the system of trichromatic vision in the human retina we'd have at least
something equivalently powerful as human trichromatic colorvision, I think.
I don't know if anyone has ever designed a CFA with a spectral response identical to the response of the cones in the human eye or if there are maybe even reasons against that though.

Cheers
~Chris
Logged
If you appreciate my blog posts help me by following on https://instagram.com/edmundronald

eronald

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6642
    • My gallery on Instagram
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #56 on: February 08, 2014, 02:48:59 pm »

except for tiresome mantra about "CMOS colors" with the word CMOS in it... it is the time to really drop it... it is a blood libel really in camera's world  ;D

In recent years, Canon made their reputation on the applications of the CMOS processes. CMOS! was their battle-cry, and their MF adversaries shrewdly de-legitimised the label to fight back, associating it with the dynamic range you get from too many small pixels, and the color delivered by pale CFAs .

Edmund
« Last Edit: February 08, 2014, 02:54:40 pm by eronald »
Logged
If you appreciate my blog posts help me by following on https://instagram.com/edmundronald

Christoph C. Feldhaim

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2509
  • There is no rule! No - wait ...
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #57 on: February 08, 2014, 02:58:46 pm »

I don't understand why it should be so difficult to create a color filter with a predefined spectral response, like mimicking a cone type of the human eye.
There are gazillions of pigments out there with defined spektral transmission - it should be possible to make a mix which matches a given spectral response curve closely.
A computer could calculate that from a database.
Why do we still don't have that?
/ me scratches head ...

Telecaster

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3686
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #58 on: February 08, 2014, 03:09:01 pm »

I don't know if anyone has ever designed a CFA with a spectral response identical to the response of the cones in the human eye or if there are maybe even reasons against that though.

My guess is that since our "red" & "green" cones are so close in their respective responses, the degree of processing needed to emulate human color vision with a more "accurate" CFA would be formidable. Probably just easier to move the green response curve down to shorter wavelengths for more distinct red/green separation.

-Dave-
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11288
    • Echophoto
Re: Why are the tests of IQ250 against CCD-backs taking so long?
« Reply #59 on: February 08, 2014, 03:47:55 pm »

That is my guess, too.

Erik


My guess is that since our "red" & "green" cones are so close in their respective responses, the degree of processing needed to emulate human color vision with a more "accurate" CFA would be formidable. Probably just easier to move the green response curve down to shorter wavelengths for more distinct red/green separation.

-Dave-
Logged
Erik Kaffehr
 
Pages: 1 2 [3] 4 5 ... 8   Go Up