Dynamic range of Sony sensor used in Pentax 645z et al

[the_ether]

The sensor used in the Pentax 645z and other 'medium format' cameras is supposed to have fantastic dynamic range. I was wondering where that lies in relation to a) photographic film; and b) a C type or giclee print.

Regarding the range of a print, I was wondering whether there was any material benefit to be had from stitching multiple exposures to create an 'HDR' image, or whether an image form say, a Pentax 645z had enough dynamic range.

[voidshatter]

If you look into Tim Parkin's tests you would be able to observe that the IQ180 has better SNR in the shadow when compared against the film grain in the large formats.

It is known by both real pictures as well as scientific numbers that the IQ180 has less dynamic range than the Nikon D800E. The 645Z is an enlarged version of the D800E so it has even better dynamic range.

[ndevlin]

The dynamic range is nuts compared to most of what's gone before. a Lot more than film. You should look at the example Philip Bloom had on his review, where the foreground is basically pitch dark as-shot, and only the brightest sky is correct, and then he recovers back everything in post and it looks like he had 4800 ws of light filling the scene. That's kind of the deal. 

http://philipbloom.net/2014/12/23/pentax645z/  scroll way down.

SAVE THE HIGHLIGHTS. The rest comes back.

HDR? Well, there is an interesting in-camera HDR mode you can use with RAWs. I've played with it little but haven't found it makes a huge difference. Multi-frame blending may still make sense in extreme contrast situations.  But I don't know anyone who uses grads with a 645z, though, if that tells you anything.

I stand to be corrected by the experts, but I don't believe the printed page rivals the DR captured in-camera.

- N.

[BernardLanguillier]

I would expect DxOMark to test the 645Z pretty soon since they had tested the 645D. We can expect it to have more than 15 bits of engineering DR.

But, together with the IQ250, it certainly has more than enough to cover most scenes, certainly landscape. There may be some interior shots where the range between the dark parts of a room and the bright outside is still too large to capture one shot, while the human eye is able to adapt dynamically when moving around the scene and manage these extremes.

The achievable DR in a print being very limited (the darkest black is not that much darker than the brighest white) is in fact not directly related to the need to capture a scene with the maximum possible DR. All current cameras have more DR than a sheet of paper can offer, but it doesn't mean that all the cameras are equally capable at delivering a suitable print of those scenes.

Those cameras with enough DR to capture all the range of illuminations from the brightest to the darkest areas of a scene will give the possibility to map these tones down to the Dr available in the output medium. Those cameras not offering enough DR will not be able to render in a pleasant way (free of objectionable noise) either the brighest or the darkest tones in a scene that will appear as blocked up.

The whole game is about mapping the tones present in the initial file to a narrower DR output medium so as to re-create the perception of the human eye.

Cheers,
Bernard

[hjulenissen]

The whole game is about mapping the tones present in the initial file to a narrower DR output medium so as to re-create the perception of the human eye.

Waiting for the UHD-thing to (possibly) deliver larger CR displays with 10 or 12-bit (gamma) luminance is also an interesting possibility.

http://www.ericsson.com/res/docs/whitepapers/wp-uhd.pdf
"Sport is a potential killer app for HDR, as today audiences cannot easily see specular highlights plus detail in shadows.
...
A typical legacy CRT (for which the old NTSC and PAL TV systems were designed) can go from about 0.1 nit to 100 nits. The dynamic range is thus about 1000, so the intensity can double roughly 10 times, called “10 stops.” Viewers have never been offered HDRTV or cinema services, and are often watching “clipped” pictures, where highlights or details in shadows are lost."

https://www.smpte.org/publications/industry-perspectives/schubin-HPA2014
"HDR can increase picture detail by reducing clipping, and it also can offer the contrast that contributes to sharpness. "

By capturing the scene as accurately as possible (including raw files), you have the option of rendering it as perceptually "good" as possible using todays print/display tech, then re-render it better at a later time when display/print tech improves.

-h

[voidshatter]

I would expect DxOMark to test the 645Z pretty soon since they had tested the 645D. We can expect it to have more than 15 bits of engineering DR.

I wouldn't think so. The sensor of the 645Z/IQ250 is virtually an enlarged version of that of the D800E with about the same SNR in shadow. To estimate the dxomark DR of the 645Z/IQ250:

14.3+ln((51.4/36)^.5)/ln(2)
= 14.6 EVs

which is still slightly below that of the D810 with native base ISO 64.

When the 135 format version of the D7200 is out, we would expect the dxomark DR to be:

14.6+ln((2.25)^.5)/ln(2)
= 15.2 Evs

The lag of technology for larger sensors is more and more difficult to be offset by sensor size. See how the IQ180 got surpassed by the D800 within just one year. We live in the digital era so expect the exotic gear to become electronic waste at lightning speed.

[Kolor-Pikker]

I wouldn't think so. The sensor of the 645Z/IQ250 is virtually an enlarged version of that of the D800E with about the same SNR in shadow.

I wouldn't say it's even a "virtually" enlarged version, since it features a lower pixel density than the D8xx series cameras (30.5mp at same crop level), plus it natively comes with no AA filter, while the D800E cancels out the effect with a second convolution filter. It's a unique sensor design that was built from the ground up, but is based on the techniques and technology that Sony has developed over the years on its small-format sensors.
The D810 and 33x44 sensor came out a couple months apart, so I doubt any groundbreaking changes were implemented in the meantime, but going by simple math/physics, the larger pixel cells should deliver slightly higher DR, assuming all other factors are equal.

In any case, I too am curious how the 645Z would score in DxOmark, even though I hardly see it as a practical resource.

[ndevlin]

More to the point, who actually cares? I am yet to see a great photograph that I was revulsed from by lack of that last stop of dynamic range.  I've never met a serious photographic artist or high-end commercial professional who is anything but satisfied with the performance of their current state of the art tools in this respect.

The DR of the Sony 50MP MF chip - certainly as realized in the 645z - is "adequate".  Adequate for what? Pretty damn near everything.

- N.

[voidshatter]

More to the point, who actually cares? I am yet to see a great photograph that I was revulsed from by lack of that last stop of dynamic range.  I've never met a serious photographic artist or high-end commercial professional who is anything but satisfied with the performance of their current state of the art tools in this respect.

The DR of the Sony 50MP MF chip - certainly as realized in the 645z - is "adequate".  Adequate for what? Pretty damn near everything.

- N.

CCD backs have less dynamic range:





Exposure bracketing is not ideal because there can be alignment issues. I have never seen anyone who can align my IQ260 files perfectly with reasonable effort: https://www.dropbox.com/sh/5i9dgi756x4yije/AABKcpNvQov8SFzbZuupv3E1a?dl=0





Misalignment between layers (due to vibration or filter etc) would cause drop of sharpness, hence defeating the purpose of pixel peeping with expensive digital backs. Why not just settle with a Nikon D800E/Canon 5DSR instead then? That's the main reason I dumped the IQ260.

With great dynamic range you would be able to capture pictures with real sharpness at pixel peeping level (without alignment issues).



You don't need gear with the best performance to make "a great picture" or become a serious photographic artist or high-end commercial professional. The extra performance is only for those who are willing to challenge what cannot be easily done by others before.

[Kolor-Pikker]

More to the point, who actually cares? I am yet to see a great photograph that I was revulsed from by lack of that last stop of dynamic range.  I've never met a serious photographic artist or high-end commercial professional who is anything but satisfied with the performance of their current state of the art tools in this respect.

The DR of the Sony 50MP MF chip - certainly as realized in the 645z - is "adequate".  Adequate for what? Pretty damn near everything.

I suppose at some point everyone must have said something similar about their craft, but the fact is that as technology advances people dare to take bigger risks that they would have simply written off as undoable before and not even tried. So in some respect you are very much right, the 645Z wouldn't have made previous images better, but long ago images were also shot differently than they are now; If you had told a photographer from even 10 years ago that you intended on shooting a magazine-quality image at ISO6400 you'd be called crazy for sure.

What if you wanted to capture a shot that you would have to bracket exposures for, but the subject matter prevented you from doing so? With these new 14-stop+ cameras, the answer is simple - just push the shadows.

The traditional methods you utilize now say that this is overkill, but as one gets used to the capabilities of the technology, someday you'll look back and wonder how you ever managed with a camera you couldn't recover shadows with and ISO3200 was considered borderline usable.

[Chris Valites]

The traditional methods you utilize now say that this is overkill, but as one gets used to the capabilities of the technology, someday you'll look back and wonder how you ever managed with a camera you couldn't recover shadows with and ISO3200 was considered borderline usable.

I don't know about THAT. I made some photos with my D70 that I fully enjoyed and still think are great portraits, even though the red channel was always clipped straight to hell, the resolution was horrible, and the base ISO of 200 was noisy. Not to mention the innumerable photos shot over the years on film at insanely high ISO that are timeless, classic photographs, or the mistakes made during the processing of Capa photographs from the D-Day invasion that still convey the feeling of the event, etc.

However...

If you don't look at the camera as a way of managing, but rather, conveying what you see the way you WANT it to be conveyed, then it's a different conversation. You're afforded a different toolset; much like a normal person with a Ferrari wouldn't post much better lap times than one with a Ford Focus RS, it's on you to use the tools to maximize your potential.

...

The dynamic range of the backs brings out some amazing qualities these days, but one of the simplest ones for me is the lessening need for a graduated ND filter while I'm hiking. This is greatly appreciated, given that the less gear I need to lug up a mountain, the better...

[Doug Peterson]

CCD backs have less dynamic range:

As a general note these images compare results from a very specific niche usage important to Void's preferred shooting style. They combine strong shadow pushes with long exposure via strong ND filter, which requires using ISO140 on the IQ260. The results from an IQ260 at ISO50 for normal length exposures are excellent, as are the results from an IQ260 at ISO140 for long exposures if you aren't hugely abusing the shadow-recovery-slider.

The results of his tests are very helpful, if like him, you do long exposures during the day with strong ND filters, and very strong shadow pushes and very large prints are a requirement. If this does not describe your needs then the tests are accurate but not relevant.

We maintain a large inventory of example raw files of our customers to use from our own personal shooting, company tests, and images donated by clients. So if your needs are different than Void's feel free to contact us to play with files more relevant to your shooting style.

Unrelated to that... we've seen amazingly good results from the IQ250, the IQ150, and Credo 50 coupled with the industry leading Capture One raw processing algorithms. To get some idea of how much time/effort went into the joint effort between the software and hardware team at Phase One, to maximize the results from the IQ250 you can read my article published here on luminous Landscape on the IQ250 Sensor.

[Joe Towner]

I've done some amazing recovery on shadows with the 645z, base iso, long exposure, and then pulling up the exposure 4 stops in Lightroom.

How that compares to a film exposure has too many variables, plus that whole waiting to find out after the processing.  I'd recommend you rent/hire one and try it out - either you'll believe in it, or you'll call it crap and move on.  The comparison is too personal to be subjective, as what works for me may not work for you.

[ErikKaffehr]

Hi,

I would say "same generation" Sony sensor. There are small improvements all the time.

Pixel size does not really relates to DR, unless you are pixel peeping at actual pixels. For any other scale pixels are bundled. So what matters is the number of electron charges the sensor can hold. That number is proportional to the area of the sensor.

The 645 sensor is 70% larger than a 24x36 sensor, so it should have a small advantage in DR, like half a stop.

Also, both Nikon and Pentax used to make better use of the sensor than Sony.

The great advantage of the Sony and some other CMOS sensors is the low readout noise at base ISO, and also the low base ISO. Which may indicate large full well capacity. A large FWC will result in low ISO and large DR.

Compared with film, which the OP was asking, digital sensors are much more clean and this is coming in part from much higher QE (Quantum Efficiency) of digital sensors compared to film.

Maximum exposure on digital is limited by full well capacity, while film takes a lot of gradual overexposure. With lower exposure both film and digital get more noisy, simply because photon statistics getting worse. In the shadows there will be a lot of noise in film, quite a lot on CCD and very little modern CMOS.

OLP filtering or the lack of it has no relation whatever to dynamic range. OLP filtering just suppresses aliasing, mostly colour aliasing.

Best regards
Erik

I wouldn't say it's even a "virtually" enlarged version, since it features a lower pixel density than the D8xx series cameras (30.5mp at same crop level), plus it natively comes with no AA filter, while the D800E cancels out the effect with a second convolution filter. It's a unique sensor design that was built from the ground up, but is based on the techniques and technology that Sony has developed over the years on its small-format sensors.
The D810 and 33x44 sensor came out a couple months apart, so I doubt any groundbreaking changes were implemented in the meantime, but going by simple math/physics, the larger pixel cells should deliver slightly higher DR, assuming all other factors are equal.

In any case, I too am curious how the 645Z would score in DxOmark, even though I hardly see it as a practical resource.

[voidshatter]

Compared with film, which the OP was asking, digital sensors are much more clean and this is coming in part from much higher QE (Quantum Efficiency) of digital sensors compared to film.

Maximum exposure on digital is limited by full well capacity, while film takes a lot of gradual overexposure. With lower exposure both film and digital get more noisy, simply because photon statistics getting worse. In the shadows there will be a lot of noise in film, quite a lot on CCD and very little modern CMOS.

Here is a comparison a while ago: if the desired threshold for SNR is 0dB then yes film has more dynamic range; if the desired threshld for SNR is 20dB then the D3X has more dynamic range than the GPortra 160NC.



Not sure how the 645Z/IQ250 compares but I am pretty confident that with multi-sampling (temporal noise reduction) a modern Nikon camera can achieve very clean files in the shadow.

[hjulenissen]

OLP filtering or the lack of it has no relation whatever to dynamic range. OLP filtering just suppresses aliasing, mostly colour aliasing.

There is an indirect relation: if you happen to like the crunchy/aliased/... look of OLPF-less sensors, then having an actual OLPF-less sensor means that you get what you are after right out of the box. If you purchase a camera with OLPF-filtering, you would need to do deconvolution/sharpening to get similar degrees of crispyness, thus raising noise levels.

-h

[Chris Livsey]

or the mistakes made during the processing of Capa photographs from the D-Day invasion that still convey the feeling of the event, etc.

Whilst there is no denying the power of those few exposed images made by Capa the recent evidence is that there were no errors in in processing. That story was “invented” to cover the fact that Capa only took those few famous, and rightly so,exposures the rest were never ruined in processing, they never existed: http://www.nearbycafe.com/artandphoto/photocritic/2014/06/10/alternate-history-robert-capa-on-d-day-1/

Be warned this is a long series of posts, and there are entrenched vested interests fighting here. ( But then we are used to that here!!)

[Gel]

It's sexy. One day we'll have 5 stops either side. (I hope).

[Paul2660]

It's sexy. One day we'll have 5 stops either side. (I hope).

Gel

Impressive push, are those at base iso?  I really like the wedding shot.

Thanks
Paul

[Gel]

The building one is ISO500 and my dog was ISO100.

As an idea the exposure slider is like the cameras digital amp. So say you need to push ISO100 4 stops you will get a very similar amount of noise in the image as you would if you'd shot it correctly at ISO1600. At least that's how it feels.