Resolutions champs vs. high ISO kings a small experiment…

[ErikKaffehr]

Hi,

Raw images from DPREview test site, all at 3200 ISO. Processing is "identical" and the images were scaled to 40x60 cm size at 180PPI.

The image size was chosen to be around A2. It is usually considered that 180 PPI is needed for a really good print.

So, what do I see?

Noise on grey patches is remarkably close between all. This is not really surprising as shot noise dominates and that should be near identical for all of these cameras.

On the white text on black background the Sony A7sII falls apart, 12 MP is simply not enough and OLP filtering is far to weak. Some of the resolution advantages of the high res cameras are visible although the images are scaled down.

[ErikKaffehr]

The DPReview test image also has an etching that is good to show colour aliasing artefacts. On this the Sony A7sII did really well, while the A7rII clearly fails.

All the non OLP filtered images show some colour artefacts, but it is only the A7rII that goes over the edge.

Best regards
Erik

[Telecaster]

I also find that region of DP Review's test image to be very informative based on my own needs/interests.

-Dave-

[BernardLanguillier]

Lack of AA filter is clearly a huge issue for such scenes, but those are rare in the real world.

The pentax K1 in pixel shift mode should put everybody else very far hehind.

Cheers,
Bernard

[Paul2660]

There has been a lot of talk of game changers over on the MF form. Basically a 4 year old chip moved to a mirrorless environment. (Let the hate begin).

Yet Hasselblad really had a game changer years ( multi shot) an excellent way to defeat the known limitations of the Bayer pattern.

Pentax with both the K1 and before the K3ii offer a different methodology with pixel shift. No you don't get the Hasselblad 50 to 200 resolution increase but the Hasselblad multi shot was static subjects only.

Pentax with pixel shift is a game changer at least to me. They are weak in lenses but may work on that in the future. And the fact that the flange distance is so close to Canon and a Nikon 1mm or less means no lens adaptation.

But the images I have taken are just impressive very. Color, detail, noise etc are class leading.  Enough for me to switch or at least add this excellent camera to my pack.

If you get time Eric, go back to the dpreview comparisons and add the K1 and look at the higher ISO shots in pixel shift mode. I dare say there is not another camera that can touch 6400 even the high end Nikon and Canon offerings. Also LR does a very good job with the pixel shift files unless you have too much motion. Silkypix does better but their software lacks the LR or C1 quality to me. The results are impressive enough to me that it's worth using the pixel shift on most of my setups for landscapes.

This type of tech will probably never come to Nikon or Canon unless they are forced and the volume from Pentax sakes will not be enough.

If Hasselblad ever adds this or similar tech to their new mirrorless camera I will sit up and take notice, it's just that much better.

Paul C

[BernardLanguillier]

Indeed Paul, the K1 is a game changer.

In MF I never quite understood why Hasselblad hasn't been getting more recognition for their MS offerings.

Cheers,
Bernard

[henrikfoto]

Does this sensor-shift system need a scene with no movements at all like a multishot camera
or is it more forgiving?

Seems like the big problem with both the 645Z and the K1 is the lacking of high quality lenses.
Maybe we will see new fuji or zeiss lenses for these systems at Photokina?

[ErikKaffehr]

Hi Paul,

Thanks for the suggestion to check out the Pentax! I have not done any "deep" comparison yet, but just doing a quick and dirty comparison in DPReviews image comparison tool impresses.

Yes, Hasselblad has been doing multishot for a long time and it may be that they didn't get the credit they may have deserved. Kudos to Pentax for introducing the technology in an affordable and usable package.

Regarding your remark on the Hasselblad using a four year old sensor in mirrorless:

• Hasselblad making a relatively affordable mirrorless is a good thing, at least IMHO.
• If that Sony sensor really is four years old than it is time for a new one.
• But, MFD may not move at the speed of lesser sensor sizes.
• I am not so sure about the advantage of the 44x33 mm sensor over 34x36. It is larger, by something like 70%, so it should perform at 170 ISO as a 24x36 sensor at 100 ISO. It probably has some resolution advantage over 24x36.
• The K1 multishot examples shows that good sampling may be more interesting than fine sampling.

There has been a lot of talk of game changers over on the MF form. Basically a 4 year old chip moved to a mirrorless environment. (Let the hate begin).

Yet Hasselblad really had a game changer years ( multi shot) an excellent way to defeat the known limitations of the Bayer pattern.

Pentax with both the K1 and before the K3ii offer a different methodology with pixel shift. No you don't get the Hasselblad 50 to 200 resolution increase but the Hasselblad multi shot was static subjects only.

Pentax with pixel shift is a game changer at least to me. They are weak in lenses but may work on that in the future. And the fact that the flange distance is so close to Canon and a Nikon 1mm or less means no lens adaptation.

But the images I have taken are just impressive very. Color, detail, noise etc are class leading.  Enough for me to switch or at least add this excellent camera to my pack.

If you get time Eric, go back to the dpreview comparisons and add the K1 and look at the higher ISO shots in pixel shift mode. I dare say there is not another camera that can touch 6400 even the high end Nikon and Canon offerings. Also LR does a very good job with the pixel shift files unless you have too much motion. Silkypix does better but their software lacks the LR or C1 quality to me. The results are impressive enough to me that it's worth using the pixel shift on most of my setups for landscapes.

This type of tech will probably never come to Nikon or Canon unless they are forced and the volume from Pentax sakes will not be enough.

If Hasselblad ever adds this or similar tech to their new mirrorless camera I will sit up and take notice, it's just that much better.

Paul C

[BernardLanguillier]

Seems like the big problem with both the 645Z and the K1 is the lacking of high quality lenses.
Maybe we will see new fuji or zeiss lenses for these systems at Photokina?

The more promising the sensor the more important the lenses.

Each time I look seriously at MF, I end up not making the jump for 3 key reasons:
- cost and/or cost perfo ratio
- usability
- a deep question whether I will get better results than those I am getting with my triplet of Otus on my current DSLR and what I expect the next gen to be. Yes, many MF lenses are excellent but the Otus are simply sublime.

So this is again the reason why I won't buy a K1, I know I'll end loving the technical excellence of the file are be frustrated by the lack of sublime lenses to put in front of it.

Cheers,
Bernard

[BernardLanguillier]

I think you'll find their MS offerings have far more in the way of recognition than a historically Phase centric forum would have you believe.

Good point!

Cheers,
Bernard

[shadowblade]

Pentax has little traction due to lens selection.

Hasselblad (and Leica S) has less traction than Phase One here, because, as a closed system, it has a far more limited lens selection than Phase One digital backs that can be used with almost any lens. Not a problem for those shooting fashion or portraiture with just one or two lenses (and the AF is useful) but a fatal flaw for those who need a wide range of focal lengths.

Pixel shift is useful, but only for things that don't move. It effectively gives you the same thing as a Foveon sensor - separate RGB values for each pixel - but over three exposures rather than one. It would be much more valuable in a Sony E-mount (which already has the moveable sensor needed to implement it), which can use a wide range of lenses, than on a Pentax body. I'd like to see further development of the one pixel/one exposure/multiple values Foveon concept - it would eliminate colour aliasing for both static and moving scenes, and give better colour accuracy and resolution than a Bayer sensor. And you could still use pixel shift to move the entire sensor, by half a pixel width in each direction, for four exposures and four times the resolution.

[Guillermo Luijk]

Kudos to Pentax for introducing the technology in an affordable and usable package.

More affordable and usable is the Olympus hires multishot mode: mixes eight 16Mpx shots in a row in a single 64Mpx RAW file with an effective resolution of around 40Mpx of information (camera JPEGs produced with this mode are 40Mpx), with zero colour nor monochrome aliasing artifacts and 1,5 stops of increased dynamic range. And Olympus plans to shorten the capture time to allow doing it handheld and with sufficiently slow moving subjects.

Extrapolate this to any larger format and the result will outperform pixel-shift systems from any other brand (Hasselblad, Pentax).

http://www.imaging-resource.com/PRODS/olympus-e-m5-ii/olympus-e-m5-iiTECH2.HTM

Regards



www.guillermoluijk.com

[Jack Hogan]

Folks in this thread may be interested in taking this perceptual test of K-1 pixel-shifted vs unshifted renditions.  Once you have taken your shot at it, check out the answers in the 'Drum Roll' post.

Jack

[Paul2660]

Pentax has little traction due to lens selection.

Hasselblad (and Leica S) has less traction than Phase One here, because, as a closed system, it has a far more limited lens selection than Phase One digital backs that can be used with almost any lens. Not a problem for those shooting fashion or portraiture with just one or two lenses (and the AF is useful) but a fatal flaw for those who need a wide range of focal lengths.

Pixel shift is useful, but only for things that don't move. It effectively gives you the same thing as a Foveon sensor - separate RGB values for each pixel - but over three exposures rather than one. It would be much more valuable in a Sony E-mount (which already has the moveable sensor needed to implement it), which can use a wide range of lenses, than on a Pentax body. I'd like to see further development of the one pixel/one exposure/multiple values Foveon concept - it would eliminate colour aliasing for both static and moving scenes, and give better colour accuracy and resolution than a Bayer sensor. And you could still use pixel shift to move the entire sensor, by half a pixel width in each direction, for four exposures and four times the resolution.

This is a huge misunderstanding, actually pixel shift works for things that move and works quite well.  You do not have to have a totally static subject.  I have now shot hundreds of images in pixel shift mode.  All outdoor landscape, most with some wind.  NET, the files look great unless you have 20mph hour winds blowing. 

The real problem is that Adobe LR, can't process the images well if there is wind.  Sometimes it can other times you will pick up some of the strange green aliasing that the Dpreview article showed.  Phase One, who knows what their plans are, they tend to be more mute.  Currently the multishot is not supported, (I don't know if they support the multi-shot on the Olympus either), but if Phase does come out with support, then I am sure it will be a well thought out conversion.   You can easily see just how well the files can be processed, by using the full version of Silkypix.  Same files that show the aliasing in LR, process out fine in Silkypix, net Adobe did not spend enough time on their raw conversion (similar to the terrible job they do for Fuji Files which are also a bit different, which is a sad fact as Adobe sure has the resources to do better)

But let me say it again, it's not like Hasselblad MS, which would not tolerate movement at all, this will and the results show for themselves. 

I agree Pentax has a limiter, their glass, and they may step to the plate and resolve this, maybe not.  If Sigma would just make their full Art line for Pentax and Zeiss would start making the ZK line again, things would be better. 

However the 15-30, as as good as the 14-24 (Nikon) and that is my main working lens from Nikon in the field.

This morning, I have winds blowing 15-20 miles an hour, would I use Pixel shift, no.   

As many know I prefer to work with water and slower exposures.  Here again LR totally fails on the conversion, yet Silkypix does an excellent job and gets a great conversion out of the same file that LR fails on.

For my work, the Pentax also offered a lot of other excellent features, (astro tracker), (GPS), (HDR with raw), not to mention pixel shift.  So for the price/value it is a great example of where technology can go with an older chip as I am sure this is the same chip that the D810 has (Sony).

I would say this, it's well worth renting a K1 and 15-30 from lens rentals.com to see just what you can gain from the pixel shift technology. 

Paul C

[Paul2660]

Hi Paul,

Thanks for the suggestion to check out the Pentax! I have not done any "deep" comparison yet, but just doing a quick and dirty comparison in DPReviews image comparison tool impresses.

Yes, Hasselblad has been doing multishot for a long time and it may be that they didn't get the credit they may have deserved. Kudos to Pentax for introducing the technology in an affordable and usable package.

Regarding your remark on the Hasselblad using a four year old sensor in mirrorless:

• Hasselblad making a relatively affordable mirrorless is a good thing, at least IMHO.
• If that Sony sensor really is four years old than it is time for a new one.
• But, MFD may not move at the speed of lesser sensor sizes.
• I am not so sure about the advantage of the 44x33 mm sensor over 34x36. It is larger, by something like 70%, so it should perform at 170 ISO as a 24x36 sensor at 100 ISO. It probably has some resolution advantage over 24x36.
• The K1 multishot examples shows that good sampling may be more interesting than fine sampling.

Edit: I was wrong on the date of the 50MP chip, appears to be Feb 2014)

Hi Eric, the 50MP Sony dates way to 2011, as the IQ250 was announced in early 2012, so the chip had to be around prior to the announcement in February.  Obviously the chip does a great job as evidenced by many Phase and Hasselblad cameras that have used it.  It's just when I think mirrorless, I think of the features that Sony and Fuji have in their cameras and continue to add to their cameras.  The Fuji MF solution if it happens at Photokina will be mirrorless for sure, and I am very interested to seeing what they can add to the Sony 50MP sensor as I am also sure it will be the same chip.

Paul C

[BernardLanguillier]

Hi Eric, the 50MP Sony dates way to 2011, as the IQ250 was announced in early 2012,...

Are you sure Paul? I believe it was announced in 2014.

http://www.dpreview.com/articles/4031307198/phase-one-announces-iq250-50mp-cmos-medium-format-back

This makes the 50mp back 2.5 years old, which is still pretty old for such a component.

Cheers,
Bernard

[Paul2660]

Does this sensor-shift system need a scene with no movements at all like a multishot camera
or is it more forgiving?

Seems like the big problem with both the 645Z and the K1 is the lacking of high quality lenses.
Maybe we will see new fuji or zeiss lenses for these systems at Photokina?

Yes, you can shoot moving subject matter to a degree, you will not get a car, person, animal etc. But in landscapes you don't have to have everything totally still, as the files can tolerate wind movement if you shutter speed could, i.e. shooting in the 1/125th range vs 1/25th range to stop motion.  LR current can't handle this very well and creates green edges and strange aliasing around certain parts of the files.  But the Silkypix software (full version) can do a great job on these types of files.  Obviously 15mph winds would be a problem, but I have been pleased to find that Pixel shift works in 85% of my shooting environments and works quite well.

On the lenses, some of this is public perception, gone back.  Pentax does not have the glass range that Nikon and Canon offer, but currently they have several D FA lenses all zooms, that basically cover the range up to 450mm.

15-30mm, rivals the Nikon 14-24 (which I use and love) it's very strong from 15mm to 24mm a bit weak out to 30mm
24-70mm, could be better @24mm to 30mm, but from 30mm to 70mm excellent
70-200mm, as good as any Nikon or Canon equivalent I have shot over the years, but also like Canon and Nikon heavy.
150-450mm, not sure, have not shot this lens, but planning to rent one from Lens Rentals soon. (IMO over priced when you consider the excellent Nikon
    150-500mm, which was around 1.4K last time I checked and is an excellent lens)
200 F 2.8 Based on older film lens called a DA but works in Full Frame mode with no compromise  great lens
300 F4 Also based on older Film lens, called a DA but works in Full frame mode also great lens, just like the Nikon and Canon versions are.

Edit:  Note all of the modern zooms are Weather resistant, as is the camera.  But the term weather resistant is a bit misleading as this camera when used with a WR lens can be totally dunked under water and still work fine.  Something that it shares with the 645z.  As a photographer who uses his equipment mainly in the field and around water, this is a nice feature

100mm D-FA Macro strange looking lens, but great lens for the price
28-105mm D-FA, not sure on this lens but it gets good reviews
40mm F2.8, pancake lens DA lens, but works in Full frame mode with no crop.  Like the old Nikon 45mm 2.8 limited hyperlocal range tack sharp
77mm 1.8  D FA excellent lens which gets very little press

All older K mount glass mounts to the K1, and there is a ton of that out there and some of it is very good. 

There are still Zeiss ZK mounts out on eBay and one pops up everyone in a while.  I personally would love to find the Zeiss 21mm in a ZK mount.  I have located a 25mm F 2.8 but owner is a bit proud of it.

Sure Canon and Nikon have 2 to 3 more versions of each of the lenses i have listed, but with that range you can get the job done. 

I guess my other point is that Pentax has shown just what is capable from a 36MP camera, i.e. files that will appear to be close to 3D in depth, with colors are that are amazing.  When you see the results for the first time, then you look back on your other files shot in normal mode and it's a quick decision, use this technology whenever I can! It's also sad to think that Canon and Nikon could do a similar setup, but never will, as they have both gone down the vibration optimization in the lens design, thus they more than likely don't want to make more lenses without IS or VR.

Paul C

[Paul2660]

Are you sure Paul? I believe it was announced in 2014.

http://www.dpreview.com/articles/4031307198/phase-one-announces-iq250-50mp-cmos-medium-format-back

This makes the 50mp back 2.5 years old, which is still pretty old for such a component.

Cheers,
Bernard

Thanks Bernard,  time does fly as you get older for sure.  I purchased my IQ260 in 2014, and I thought the announced the IQ250 before the 260, but I have those dates confused.  Now that I remember it Phase did not allow a trade in from the 260 to 250 and the "special price" I received was not very "special".  But even then the 1:3 crop factor stopped me from making the move. 

Paul C

[Paul2660]

Folks in this thread may be interested in taking this perceptual test of K-1 pixel-shifted vs unshifted renditions.  Once you have taken your shot at it, check out the answers in the 'Drum Roll' post.

Jack

Best way to see the difference is to rent a camera.  So far a lot of reviewers are still only using LR/ACR even when it's a known fact that in many circumstances Adobe fails to pull the best results from the pixel shift files.  I have hope that C1 will eventually support this mode (pixel shift) but as it did not make it with 9.2's release along with support for the K1, I am not holding much hope.  As for Adobe, it's my feeling that they make one pass on a raw file and file it DONE.  They don't seem to take the time to understand a unique format, instead attempt to force fit the file into a standard process.  Their less than stellar Fuji conversion is just one more example and that has been around for at least 2 years with no change, even though hundreds of articles have been written showing the issues, yet small companies like Iridient get it right the first time. 

So far LR Adobe can handle about 75% of the Pixel shift images I have taken without a huge amount of the green shift aliasing.  LR can tolerate just a bit of motion, but nothing like what Silkypix can resolve.

Paul  C

[Paul2660]

On the subject of Pixel Shift and Pentax: a few more thoughts

1.  Pentax could have handled this technology quite a bit better.  Seeding cameras to Phase One (C1) and Adobe LR ACR and working closer with these
     companies before releasing this camera.  Currently the best raw conversion I have seen is from SilkyPix however their software can't hold a candle to
     LR or C1 in regards to the depth that either of those programs offer a photographer.

2.  The software that Pentax includes with the camera, Digital Camera Utility 5, to me is not a great step forwards.  The software is very limited in use and
     really does not benefit the subject of a photographer gaining insight into the advantages of Pixel shift technology.

3.  The vast majority of reviews I have read, all are critical of Pixel shift, mainly because they are using Adobe's raw conversion, either in ACR or LR. 
     Currently Adobe does a good job on the files as long as there is very little movement, which in my shooting environment would be mainly due to wind.
     The Adobe conversion lacks the ability to really combine the images and account for minor movement which the Silkypix conversion can do.  The difference
     is pretty impressive and I will try to post a side by side later. 

4.  Pixel shift makes a good lens very good and an excellent lens outstanding in resolving power.  Everyone should try this at sometime in their photographic
     journey just to get a better understanding  as to how much detail is lost in the normal Bayer pattern color interpolation process.

5.  There is a lot of confusion on read forums with the raw software that Pentax includes.  Digital Camera Utility 5 is using a subset of Silkypix, but it not
     Silkypix, not close.  The cost of the full version of Silkypix (and I do feel if you are used to LR or C1 you will want to use the full version) is around $295.00
     US, not cheap, but you will only get frustrated with Pentax's own software.

6.  By far the best review I have seen on this camera is one by Nick Devlin, on Youtube, which you can find here:

     https://www.youtube.com/watch?v=87yGmD71nyA


Paul C