The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

Phase One IQ3 100mp. Shipping now. Full frame CMOS. HDMI.

cly

Member
This is especially true of Phase One which is a dealer-oriented company.
That's simply not true. They are a dealer-oriented company in your part of the world, in certain other parts of the world the photographer deals directly with P1.

Chris
 

yatlee

Member
Had a quick test with the new back side by side with my IQ260 today. I took some shot with my Alpa Max and Rodenstock 32mm, but I don't have time to do anything with the shot until probably tomorrow.
 

torger

Active member
Can someone comments on the DR test in the LL forum that shows DR ís only 13 stops?
As far as I understand one needs to run the back in 16 bit mode to get the full dynamic range, and then it will write a slightly different file format. The DCraw/Rawdigger guys (or perhaps me...) will probably have to reverse-engineer that before it can be fully analyzed. The files available for analysis now are in the old 14-bit-expanded-to-16-bit format and cannot hold the full dynamic range, if it exists. So the final verdict on this needs to wait.

It depends on how you count DR too, if we look how DxOMark does it they check per pixel, and then normalized to 8 megapixel image. In the latter there's gain the larger surface you have. I think that when Phase One says 15 stops they don't mean per pixel (which would be extremely impressive), but some variant of this scaled down measure.

However at some point there'll be dual gain readouts or those non-linear filters (lots of patents for this exists since some time) and then lots of DR can be had. Remains to be seen if this sensor have some of this new technology or not. Most likely is that it's just an evolution of the current sensors, it's not likely that revolutionary tech would come in a big sensor first.
 

torger

Active member
I've known that for years. And it's precisely my point: why is it left to the dealers to do the essential tests, post-launch?
I think it says something about the size of the tech cam market. It's big on forums like this, but the mass of users use the DF/XF systems and that is what Phase One focuses their limited testing resources on. They're afterall a quite small company.

It seems like the tech cam segment has been growing, probably due to the enthusiast market rather than the pro market. The A-series is a response to that, but conveniently enough it doesn't do any movements.

I also think it as simple as that Sony's technology is simply not geared for this type of lenses, or really not any current sensor technology. Limited angular response has always been a weakness and the requirements tech wides put on the sensor is simply on a different scale and what can be provided without a major redesign. So I think Phase One had the choice to either not get a CMOS full-frame sensor at all or ignore that it won't work that well with tech wides, except for center frames.

To make current sensor tech work with current tech wides I think pixel pitch need to be made larger, the microlenses preferably be removed or at least reduce their height/efficiency, and preferably light shields should be put there (which likely reduce light-sensitive area further). The result of that would be lower pixel count, lower DR and more aliasing. The wiser move to sell cameras is probably to do what's done now, max out image quality for the cameras that sell the most, ie the DF/XF systems.

I had hope for BSI to solve issues, but I think there may be a few years. I think the most realistic way forward for tech cameras is that Rodenstock makes a few new wide angle lenses, or we just learn to ignore the color fidelity / uniformity issues and/or shift less. We saw that with the IQ250 people are prepared to compromise quality and/or flexibility quite a lot to get the CMOS advantages. On the other hand the Dalsas aren't without issues either, and the extreme DR of the CMOS compensates somewhat. It's a slippery slope with gradual acceptance of more and more movement limitations and image quality issues. As long as there's nothing wrong with the sharpness (the 32HR - just wow), people seems to be willing to accept quite some limitations in other factors.

My own heart beats for the large format shooting style, where rich movements are an intergral part of image making. So of course I'm worried when I see that we start sliding on this slippery slope towards tech cam becoming equal to Phase One A-series, ie a very sharp center frame snapper.
 
Last edited:

ondebanks

Member
Of course Phase One has tested this back, quite extensively, including with Tech Cams. But camera manufacturers, as a general group, do not release large volumes of test images especially when third party components (e.g. Rodenstock lenses on an Arca body) are involved. For instance you don't see Canon releasing gigs of raw files with Sigma lenses or Nikon releasing gigs of raws showing skintone when combining their cameras with various brands of LED lighting.
Some counter-arguments:

1) https://www.alpa.ch/en/site/new-phase-one-100-mp-back-on-the-alpa
How is it then that Alpa had no difficulty releasing test images made with 3rd party components? - Alpa don't make the IQ3 back and they don't make the Rodie lenses! What's good enough as a business practice for Alpa is not good enough for Phase One?

2) Isn't it far more likely that Canon don't release gigs of raw files with Sigma lenses, because they want people to buy their Canon lenses rather than Sigma ones? But Phase One don't make tech cameras or tech-cam lenses, so that excuse doesn't arise.

3) If Phase One have genuinely done the tests, why don't they share the results with us? They would not have to release raw files to fill the information vacuum. They could just issue tech statements about what does or does not work, or within what limits.

Not being hard on you, Doug - the work you do is great - I just think they've sent you out front to defend their indefensible policy.

Ray
 

Stefan Steib

Active member
Some counter-arguments:

1) https://www.alpa.ch/en/site/new-phase-one-100-mp-back-on-the-alpa
How is it then that Alpa had no difficulty releasing test images made with 3rd party components? - Alpa don't make the IQ3 back and they don't make the Rodie lenses! What's good enough as a business practice for Alpa is not good enough for Phase One?

.........

Ray
Maybe because Phase wants to sell their new XF system and the lenses also in the first place ?
And second because the tech cam shooters are a tiny fraction of those using it on fixed bodies anyway.

So I also doubt there will be any new techcam lenses AT ALL. Schneider has closed their lens department on 31.December and Rodenstock is probably not sellling enough to devellop another round of new Retrofocus wideangles that will work better with such highres backs.

Why do you think Alpa is also pushing the FPS so hard and releasing adapters for all kind of NON-Tech lenses ?
Copal has ended support for the tech lens shutters and Cambo released a modified 24mm Samyang to get the actus into wide angle.

Wake up and see the truth: classic techcam is a dinosaur and the meteor has already striked.

Regards
Stefan
 

torger

Active member
Part of the problem is that stating the limits is hard. If you define the limits as when sensor crosstalk is triggered you probably have problems also in the center frames. However the visibility of the issues increases gradually, and the acceptance varies from person to person, just as "sharp enough" is highly subjective.

I do agree though that it seems like Phase One doesn't even try, there seems to be no connection at all Rodenstock-Phase One-Sony. Sony does their thing and Rodenstock has to adapt. Lenses are released with much larger intervals than sensors though and unfortunately the current crop of lenses did not foresee the direction sensors would take. Rodenstock Digaron wides now look like the Schneider Digitar wides when Dalsa came with their 6um technology.

I also think there is the problem that if Rodenstock would design lenses that matched Sony properties they would have to be so much retrofocus that the number of lens elements required to get the performance we've come to expect would just be insane. The 32HR today is not exactly a simple lens, and it would either be more complex than that or be less sharp. Rodenstock really can't "lower" themselves to the standards of XF SLR wides, especially since they can't rely on software lens corrections.

I think that the possibility to make lenses with low angle (thanks to larger pixel sizes of MFD and that the users accept LCC, vignetting and small largest aperture) is key to what has made tech cams reach their unique position as wide angle performers. This position is now threatened as this type of lens design may not be possible in the longer term.
 

torger

Active member
Wake up and see the truth: classic techcam is a dinosaur and the meteor has already striked.
It's painful, but I realize that you may be right. The trend is evident, but on the other hand camera systems seem to take decades to die in any case :)

The tech companies should really have formed a consortium to protect their interests, but anyway of Linhof, Alpa, Arca and Cambo it seems like Linhof is at most risk with the current trend. They don't have an DSLR-compatible view camera, no focal plane shutter relies heavily on tech cam lenses. Alpa, Cambo and Arca has adapted to the trends well so far with the FPS, the DSLR-compatible view cameras and adapters to use other than tech cam lenses.

With my Digitar/Kodak-based system I feel like I'm shooting Contax, and I don't really have anything to upgrade to without sacrificing flexibility. I'm hoping for a 44x33mm wide angular response CMOS coming soon, but if the IQ3 100 does not show a step in the right direction concerning sensor technology (like the first indications are) then my hope is reduced.

Currently I'm so totally not interested in giving up large format shooting style that I see it more likely that I'll start shooting large format film after my current setup gives up than going back to mainstream cameras. If someone still makes sheet film by then that is...

I'm not sure Rodenstock is selling that bad, they've had virtual monopoly on new systems for a while and now gets real monopoly, and with a growing enthusiast market I think the tech cam sales grows. Phase One recently launched the A-series collaboration with Alpa, which indicates that they probably sell at least a bit. So I would not be surprised if Rodenstock actually do make a new set of wides, probably one or two would be enough.
 

modator

Member
Can someone comments on the DR test in the LL forum that shows DR ís only 13 stops?

I cant wait for more test, how far is it from 1st gen IQ back.
Hi Dan,
on the advertisement side there is some hints:
P1 state it has achieved 15 stop of dynamic range, then they say also it's a true 16 bit output, but with 16 bit by math, the dynamic range is 16 stop not 15, so who stole the missing stop ?
Until now, there is no datasheet from SONY that says something clear about the sensor itself, for now it's a black box, and to determine the real DR is not an easy task.
Speculation about the technology SONY used on their other sensor models tell's us a more conservative 14 Bit on ADC converter block this can permit in the best case a DR of 14 stop, so assuming 12 to 13 stop in dynamic range could be a good measure of the performance of this device.
This is applying also on the 50Mp sensor, No Datasheet nor any information of the CFA construction / spectral response, the only advantage between the two chip is coming from the size of pixels 5,3 is slight better than 4,3 micron.

Have good Epiphany, Best regards, Domenico.
 

ondebanks

Member
P1 state it has achieved 15 stop of dynamic range, then they say also it's a true 16 bit output, but with 16 bit by math, the dynamic range is 16 stop not 15, so who stole the missing stop ?
Readout noise and well depth stole it. The bit depth sets the maximum possible DR for a pixel, but analog signal to noise is the usual limiting factor.

Ray
 

ondebanks

Member
Maybe because Phase wants to sell their new XF system and the lenses also in the first place ?
And second because the tech cam shooters are a tiny fraction of those using it on fixed bodies anyway.

So I also doubt there will be any new techcam lenses AT ALL. Schneider has closed their lens department on 31.December and Rodenstock is probably not sellling enough to devellop another round of new Retrofocus wideangles that will work better with such highres backs.

Why do you think Alpa is also pushing the FPS so hard and releasing adapters for all kind of NON-Tech lenses ?
Copal has ended support for the tech lens shutters and Cambo released a modified 24mm Samyang to get the actus into wide angle.

Wake up and see the truth: classic techcam is a dinosaur and the meteor has already striked.

Regards
Stefan
You're probably right, Stefan. But I've no skin in the tech-cam game, so I've no waking up to do. My conversation is centred on how Phase One operate.

Ray
 

modator

Member
Readout noise and well depth stole it. The bit depth sets the maximum possible DR for a pixel, but analog signal to noise is the usual limiting factor.

Ray
Ray,
You are right about the readout noise, the CCD have a sort of construction assimilable to a well, the CMOS is working in (sample and add the signal) to a register... maybe this chip use a CDS scheme (correlated double sampling), anyway the aspect of the degradation of DR inside the chip is sure, and sure is the worser homogeneity that the CMOS is known for.

Returning to the pixel size of 4,6 microns, with the available technology a saturation level of circa 12000e- can be a good measure, that value sits well in 2^14 (16384) and again 14 BIT COME IN PLAY, that's the reason why I doubt the implementation of 16 BIT DAC is something useless, even with the larger pixel found in the 50MP chip 14 bit can be sufficient to the job.

Domenico.
 
Readout noise and well depth stole it. The bit depth sets the maximum possible DR for a pixel, but analog signal to noise is the usual limiting factor.

Ray
Right, pure white is the saturation point so all extra DR goes into shadows, also sensor DR is separate from the ADC being able to resolve that DR. Having a 15-stop sensor with 14-bit ADCs will just give you really clean shadows since some of the noise will be below the limits of the ADC's ability to "see" it. conversely, having a 12-stop sensor on a 14-bit ADC gives us the noisy shadows we all knew and loved since before ~2012.

In practice, this will actually amount to slightly more noise being visible compared to the 33x44 sensor, since we'll be seeing into shadows that would otherwise have been just pure black. I'm not entirely sure if that last stop can be really taken advantage of, but I guess it's always nice to have something than not.
 

yatlee

Member
I'm happy to share some 16 RAW with LCC, but not sure if how to set it up with a file sharing site. If someone is willing to put it up on a file sharing service, I'm happy to send the file to you from my google drive. PM me with your email.

They are all shot with Alpa Max with Rodenstock 32mm.
 

yatlee

Member
Rodenstock 32mm at F11, 10 degree drop, 2 shot stitch with LCC applied. The top corners are near the image circle. Full size file on the 4K screen is quite stunning. Still looking through the rest of the files.
 

Attachments

torger

Active member
Thanks for the images. In the comparison image it's clear that the IQ3 100 doesn't manage to hold color after LCC correction. The color is good close to the image circle center, but you can clearly see the incresing cast. This is almost 100% sure a crosstalk effect, that is not microlenses or something else.

The bad news with that is that it's extremely hard to correct in software. I tried to develop a crosstalk cancellation algorithm for some time when the IQ250 was new but never managed to get it stable, and obviously Phase One engineers hasn't suceeded either.

While I may be a perfectionist more than others when it comes to color (and less so when it comes to sharpness) I think the color cast looks so severe in these examples that I find it unlikely that anyone with the IQ260+32HR combo would upgrade the back to the IQ3.

By the way what was the shift settings in the second shot? Due to offset microlenses you can get a sudden performance drop after a certain shift amount, which could be seen with the IQ250. If we're lucky the performance improves a lot if you shift a bit less.

To really find shift limits you need to increase shift with small steps (say 3-5mm) until it falls apart. If you shift way past the limit it looks worse than it is due to the offset microlenses. The shift limit may also be different in portrait and landscape orientation due to how pixel wiring is oriented on the chip.
 

jagsiva

Active member
I see my bubble bursting :(

But will wait for more samples. LCC's behave quite erratically when you hit the hard limit of the lens. If anyone has the RAW files, a cropped (LCC and image file) application of LCC maybe helpful.
 

gerald.d

Well-known member
Rodenstock 32mm at F11, 10 degree drop, 2 shot stitch with LCC applied. The top corners are near the image circle. Full size file on the 4K screen is quite stunning. Still looking through the rest of the files.
Ouch. That's appalling.

Thanks for the doing the test.

Kind regards,


Gerald.
 
Top