Ad
Ad
Ad
Pages: « 1 2 3 [4] 5 »   Bottom of Page
Print
Author Topic: Leaf's advertising - why the false claims?  (Read 12901 times)
FredBGG
Sr. Member
****
Offline Offline

Posts: 1651


« Reply #60 on: April 21, 2013, 03:09:11 PM »
ReplyReply

Fred,

The scope for adjustment is impressive, but is it not also academic? After all, no competent photographer would have shot your first, grossly under-exposed image. Had the same image been shot exposed to the right, push of shadows, if required at all, would have been marginal.

Please can you provide a real-world example where push of such magnitude was warranted?

While the example I posted is very extreeme it is a demonstration of the capability of the camera. In all situations where shadows are pushed a little or a fair bit there is some quality loss compared to exposing the shadows more. What is importasnt with the example I posted is that even in exteeme cases the results are still very good. This meands that in more commomn practial cases the quality will be even higher.

There are endless exampekls in the real world where this is an enourmous advantage. If this were not the case Kodak would not have strived for years to increase dynamic range of negative film.

A few examples. Fashion runway. Black model wearing white top and black skirt or panrs with dark, but embroiderted boots. Light on the runway is prety much more from above.

Fashion with two models in the sun. One blond model and one model with jet black hair. L'oreal is the suject of the editorial.
Next shot. One model leaning against a white car with the other model inside the car.

Black vocanic rock and vitage white car.

Black vocanic cliffs soaked in water with a nude beauty shot. Jet black hair.

I do a lot of high key lighting with lekos and fresnel spots in studio and on location.
Deep moody shadows. The way the detail falls off into the shadows makes quite a difference.

The most important thing I see in the test I posted is that despite lifing the shadows many stops there is still absolute black in the final image.
This is because there is information deep into the blacks.

Another reson why dynamic range and high quantization (lots of steps... real ones, not just the extent of the file format) is the whole reproduction process. If one looks at what the independant channels of data used to make offset printing plates they are very adjusted. If you stay with modest dynamic range and quantization you end up with banding issues and that requires the offset printer to add some noise to control this.
The cleaner and more robust the file is the better off you are.

If we also add on top of that the dodging and burning required in fashion and beauty retouching the more you start with the better.
This is particularly imposrtant for skin as the tones are subtle and we are wired to read skin tones more than other things.

Now I'm not saying that MFD is not capable of doing a very good job at this, but it is not quite as good as the D800 and D600.
Very close, but the important thing is that while a few years ago 35mm DSLRs lagged behind in this area they are not leading.
It's not nesccessary to invest $20-50,000 anymore to obtain high dynamic range that used to be unique to MFD.

It will be interesting to see how the newer sensor in the IQ260 performs in this area, not that the IQ160 was a slouch at all.

Logged
AreBee
Full Member
***
Offline Offline

Posts: 138



WWW
« Reply #61 on: April 21, 2013, 04:42:24 PM »
ReplyReply

Edmund,

Quote from: eronald
I think you genuinely believe this degree of DR is not of practical relevance; aside from the fact that one could do one-shot HDR, I am pointing out that it IS relevant because in fact pushed DR is what allows modern digital backs to simulate ISO higher than 100...

Relax. Your credentials are safe. I simply was not aware of this information. Thank you for helping me to understand.

Erik,

Quote from: ErikKaffehr
I have recently made a posting where I indicated that DR was overemphasised. I still think it is, but I found several images with DR in the 11-13 EV range. If you include setting sun, for instance and want decent foreground drawing you need all the DR you can get. The reason is that you need to underexpose intentionally to minimize clipping on the highlights. It is not underexposure, technically I would say it is correct ETTR.

Thank you for the link. I have previously read the thread. Smiley

Fred,

Quote from: FredBGG
There are endless exampekls in the real world where this is an enourmous advantage. If this were not the case Kodak would not have strived for years to increase dynamic range of negative film.

I do not doubt it. However, my original post did not refer to film, but to a state of the art digital sensor.

I appreciate that the dynamic range of some compositions will exceed the latitude of the sensor. What I queried was the practical usefulness of having such scope for adjustment in a sensor that natively has the ability to capture significant dynamic range. However...

Quote from: FredBGG
What is importasnt with the example I posted is that even in exteeme cases the results are still very good. This meands that in more commomn practial cases the quality will be even higher.

This makes sense. Smiley Thank you.
Logged

ondebanks
Sr. Member
****
Offline Offline

Posts: 833



« Reply #62 on: April 21, 2013, 04:45:05 PM »
ReplyReply

However, no one I know deliberately shoots images grossly under-exposed and then pushes the file to the extent indicated in Fred's post. If doing so is useful for some then it is news to me, hence my question to Fred.

Deliberately grossly underexposing and pushing is more common than you think: it is the definition of high-ISO shooting with any camera with a flat readnoise-vs-ISO curve. That includes all MFD cameras, and nearly all recent Sony-sensored CMOS cameras, like the D800 itself.

So unless Nikon are screwing around with the RAW data (and they admittedly have poor form in that regard - messing up the stats by subtracting the bias, and median-filtering hot pixels in long exposures), then Fred's photo above - let's say it was at ISO 100 and 5 stops underexposed - would give virtually the exact same RAW file if it were a correctly exposed ISO 3200 shot.

A system with high DR will also have a good high ISO performance and it will be able to to do the FredBG trick.

That is true of most cameras one encounters these days, yet it is not really correct to make it a general rule.
High DR can result from either large full well capacity or low readout noise. In the latter case, the camera will also be particularly good for high ISO. But in the former case, it won't...the CCDs in MFD systems are good examples of this.

So it is better not to explicitly couple DR with high ISO ability: because there are high-DR cameras (like the Dalsa-sensored MFD ones at their native pixel sizes) that are disappointing when it comes to high ISO.

It is possible to achieve good ISO performance with little DR, if the sensor is clean but readout circuitry is weak using variable pre amplification, that is pretty much what Canon does.

Exactly. And that's another reason why it is better not to explicitly couple DR with high ISO ability: because there are high-ISO cameras that are disappointing when it comes to high DR.

The perfect sensor would have both large full well capacity and low readout noise. Sony and Fuji are getting close with their latest CMOS APS/FF sensors, although the smallish pixel sizes favoured these days limit how large the full well capacity can get.

Ray
Logged
AreBee
Full Member
***
Offline Offline

Posts: 138



WWW
« Reply #63 on: April 21, 2013, 04:56:18 PM »
ReplyReply

Ray,

Quote from: ondebanks
Deliberately grossly underexposing and pushing is more common than you think: it is the definition of high-ISO shooting with any camera with a flat readnoise-vs-ISO curve.

Yes, I can see now that I was ignorant of the way in which cameras return their high ISO performance. Thank you for clarifying. Smiley
Logged

ondebanks
Sr. Member
****
Offline Offline

Posts: 833



« Reply #64 on: April 21, 2013, 04:59:26 PM »
ReplyReply

Ray,

Yes, I can see now that I was ignorant of the way in which cameras return their high ISO performance. Thank you for clarifying. Smiley

You're very welcome...glad that my explanation helped.

Ray
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7394


WWW
« Reply #65 on: April 21, 2013, 10:51:49 PM »
ReplyReply

Ray,

You are of course right. I forgot that MFD-s tend to have low base ISO. My reasoning was that if we ETTR the dynamic range will be fully utilized. So if we have large dynamic range we can underexpose a number of steps. But if a sensor has low base ISO, like 50 and we push three stops we still just get 400 ISO with a three exposure push.

One of the great mysteries with MF in my view is the low base ISO, but I guess things add up. No microlenses, perhaps more othogonal CGAs? Pixel size should be ISO neutral, as FWC is proportional to pixel area, but so is the number of incident photons.

Learning in small bits...

Best regards
Erik


That is true of most cameras one encounters these days, yet it is not really correct to make it a general rule.
High DR can result from either large full well capacity or low readout noise. In the latter case, the camera will also be particularly good for high ISO. But in the former case, it won't...the CCDs in MFD systems are good examples of this.

Logged

ondebanks
Sr. Member
****
Offline Offline

Posts: 833



« Reply #66 on: April 22, 2013, 05:02:05 AM »
ReplyReply

One of the great mysteries with MF in my view is the low base ISO, but I guess things add up. No microlenses, perhaps more othogonal CGAs? Pixel size should be ISO neutral, as FWC is proportional to pixel area, but so is the number of incident photons.

"No microlenses, perhaps more othogonal CGAs" - together these add up to lower net quantum efficiency. Yes, that's a factor alright. One notices that the relatively few MFD cameras/backs with microlensed sensors have ~1 stop higher base ISO than their stablemates of the same pixel tech and size but without microlenses. P21+ vs P20+ and P25+; P30+ vs P45+; Hassy 31 vs 39 MP & 40 vs 50 MP...to take a few examples.

"Pixel size should be ISO neutral, as FWC is proportional to pixel area" - that's true in general, but don't forget the 3rd dimension - physical pixel junction depth into the wafer - and the "4th dimension" - the energy barrier set by the anode voltages, which control how densely one can pack the photo-electrons and still hold them in the "energy well". Like charges repel each other of course, so if you want to squeeze tens of thousands of negative electrons into a smaller volume, you need a stronger positive anode voltage centred on the pixel to overcome that repulsion.

So by exploiting at least one of these additional dimensions (I honestly have not checked which), Dalsa have often delivered greater FWCs for a given pixel area than Kodak. The other factor is that Kodak have pretty consistently had chips with higher spectral transmission efficiency than Dalsa. So the Kodaks have often had smaller pixel capacities AND they filled them at a faster rate: two things which raise the base ISO. This explains why hardly any Kodak-based MFD unit has had a base ISO of 25***, but several Dalsa ones have had...and when the Dalsa ones have started at ISO 50 or 80, the equivalent Kodak ones have started at ISO 100. 

Ray

***  AFAIK, only two Kodak-based MFD units have had a base ISO of 25: the Sinarbacks 44M and 54M. But I've never understood how, since all the other MFD manufacturers using these same two sensors have set their base ISO to 50 or even 100. Perhaps as tethered backs aimed at studio use, Sinar were setting them up for more extreme ETTR than would be safe under outdoor light conditions.
This highlights another factor which muddies the waters: manufacturers have some leeway to choose what they like for base ISO. Thus, to take a recent example, Phase One set the Dalsa 40MP sensor to ISO 50 in the P40+, while Leaf set it to ISO 80 in the Aptus-II 8. There can even be inconsistency for the same manufacturer: Leaf then went on to set the same sensor back to ISO 50 in the Credo 40. So perhaps one shouldn't read too much into the base ISO value, since deliberate variations of up to 1-stop for the same sensor are common.
Logged
ondebanks
Sr. Member
****
Offline Offline

Posts: 833



« Reply #67 on: April 22, 2013, 06:02:49 AM »
ReplyReply

"Pixel size should be ISO neutral, as FWC is proportional to pixel area" - that's true in general, but don't forget the 3rd dimension - physical pixel junction depth into the wafer - and the "4th dimension" - the energy barrier set by the anode voltages, which control how densely one can pack the photo-electrons and still hold them in the "energy well".
....So by exploiting at least one of these additional dimensions (I honestly have not checked which), Dalsa have often delivered greater FWCs for a given pixel area than Kodak. 

And doh! - how could I have forgotten? - another likely reason for Kodak's smaller FWCs is that they (always?) have stronger anti-blooming than Dalsa, so they probably devote more pixel area to the overflow drain.

Ray
Logged
Ben Rubinstein
Sr. Member
****
Offline Offline

Posts: 1733


« Reply #68 on: April 22, 2013, 09:47:04 AM »
ReplyReply

I was looking at the sensor of my Aptus II-8 today, you can see the centre fold of the two joined sensors quite clearly. Serious question, why aren't they using a couple of D800 sensors joined together? The software can't be that hard if a small company like Leaf can do it (not knocking their achievement), it would give incredible image quality for significantly less cost would it not? Oh and allow stuff like real live view and incredible high iso. Please pardon me if the question is naive.
Logged

ced
Sr. Member
****
Offline Offline

Posts: 266


« Reply #69 on: April 22, 2013, 10:36:58 AM »
ReplyReply

And what shall they stick it together with superglue??
Logged
Ben Rubinstein
Sr. Member
****
Offline Offline

Posts: 1733


« Reply #70 on: April 22, 2013, 10:44:39 AM »
ReplyReply

Doug has informed me I was mistaken, the sensor in my back is not a dual co-joined sensor such as was in some backs, not sure what I'm seeing, there is a definite horizontal line running down the middle but apparently it's not a join. There were some backs using this method however were there not historians?
« Last Edit: April 22, 2013, 10:52:37 AM by Ben Rubinstein » Logged

ondebanks
Sr. Member
****
Offline Offline

Posts: 833



« Reply #71 on: April 22, 2013, 11:15:36 AM »
ReplyReply

Joining sensors would be wonderful in theory, but you would not be able to get active pixel butted right up against active pixel. There is always a perimeter of dummy and black reference pixels, and usually a lot of output pins...all these would get in the way, so every image would have a thick dark line down the middle. Research astronomers do put up with this in their mosaic-cams, but terrestrial photographers would not.

Ray
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7394


WWW
« Reply #72 on: April 22, 2013, 11:29:23 AM »
ReplyReply

Hi,

I have considered the third dimension but noted that FWCs don't differ a lot between pixels of comparable sizes, so assumed that the third dimension is quite similar for all. My knowledge is not in microelectronics, just that I am interested.

It's nice to hear some good explanations from someone having insight. I learn a lot!

Best regards
Erik


"No microlenses, perhaps more othogonal CGAs" - together these add up to lower net quantum efficiency. Yes, that's a factor alright. One notices that the relatively few MFD cameras/backs with microlensed sensors have ~1 stop higher base ISO than their stablemates of the same pixel tech and size but without microlenses. P21+ vs P20+ and P25+; P30+ vs P45+; Hassy 31 vs 39 MP & 40 vs 50 MP...to take a few examples.

"Pixel size should be ISO neutral, as FWC is proportional to pixel area" - that's true in general, but don't forget the 3rd dimension - physical pixel junction depth into the wafer - and the "4th dimension" - the energy barrier set by the anode voltages, which control how densely one can pack the photo-electrons and still hold them in the "energy well". Like charges repel each other of course, so if you want to squeeze tens of thousands of negative electrons into a smaller volume, you need a stronger positive anode voltage centred on the pixel to overcome that repulsion.

So by exploiting at least one of these additional dimensions (I honestly have not checked which), Dalsa have often delivered greater FWCs for a given pixel area than Kodak. The other factor is that Kodak have pretty consistently had chips with higher spectral transmission efficiency than Dalsa. So the Kodaks have often had smaller pixel capacities AND they filled them at a faster rate: two things which raise the base ISO. This explains why hardly any Kodak-based MFD unit has had a base ISO of 25***, but several Dalsa ones have had...and when the Dalsa ones have started at ISO 50 or 80, the equivalent Kodak ones have started at ISO 100. 

Ray

***  AFAIK, only two Kodak-based MFD units have had a base ISO of 25: the Sinarbacks 44M and 54M. But I've never understood how, since all the other MFD manufacturers using these same two sensors have set their base ISO to 50 or even 100. Perhaps as tethered backs aimed at studio use, Sinar were setting them up for more extreme ETTR than would be safe under outdoor light conditions.
This highlights another factor which muddies the waters: manufacturers have some leeway to choose what they like for base ISO. Thus, to take a recent example, Phase One set the Dalsa 40MP sensor to ISO 50 in the P40+, while Leaf set it to ISO 80 in the Aptus-II 8. There can even be inconsistency for the same manufacturer: Leaf then went on to set the same sensor back to ISO 50 in the Credo 40. So perhaps one shouldn't read too much into the base ISO value, since deliberate variations of up to 1-stop for the same sensor are common.
Logged

ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7394


WWW
« Reply #73 on: April 22, 2013, 11:32:37 AM »
ReplyReply

Hi,

Sensors can be stitched on wafer in manufacture, and indeed it is my understanding that full frame sensors use one dimentional stitching. An MF sensor woud require multidimensional stitching.

Best regards
Erik

Joining sensors would be wonderful in theory, but you would not be able to get active pixel butted right up against active pixel. There is always a perimeter of dummy and black reference pixels, and usually a lot of output pins...all these would get in the way, so every image would have a thick dark line down the middle. Research astronomers do put up with this in their mosaic-cams, but terrestrial photographers would not.

Ray
Logged

ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7394


WWW
« Reply #74 on: April 22, 2013, 11:41:05 AM »
ReplyReply

Hi,

We have many happy MF owners on this forums. Obviously MF sensors would probably have some advantages. I have not seen so many Apples to Apples comparisons between sensors of different size.

Personally, I think that we have some advantage in higher photon count at base ISO, thus having better shot noise. Also I can presume that CGA may be more optimized for color separation than for high ISO. That may give MF an advantage.

Regarding color, I would ask if there would be a major difference in color if we made a comparison with white balance set on a gray card and if we used color profiles genareted by the same software, like Xrite Color Checker Passport or Adobe DNG profile editor?

Best regards
Erik
Logged

Doug Peterson
Sr. Member
****
Offline Offline

Posts: 2776


WWW
« Reply #75 on: April 22, 2013, 11:46:23 AM »
ReplyReply

Doug has informed me I was mistaken, the sensor in my back is not a dual co-joined sensor such as was in some backs, not sure what I'm seeing, there is a definite horizontal line running down the middle but apparently it's not a join. There were some backs using this method however were there not historians?

I guess it's a matter of semantics what you call it (e.g. "stitch" vs "etch" vs "segmented").

Most modern medium format CCDs have more than one read out area, which requires a read-out channel transit the sensor, which creates a visual line when you hold the sensor at an angle to direct light. They are exposed on a single piece of silicon however, and are not "joined" in the sense that they do not create sensors and then glue them together. Such an approach would be extremely hard to use in made-for-standard-photographic-applications camera since, as pointed out above you'd have to have butt them together within one pixel (e.g. 6 microns) apart and then calibrate them to produce nearly identical read-outs across a huge spectrum of possible exposures.
« Last Edit: May 29, 2013, 08:10:12 AM by Doug Peterson » Logged

DOUG PETERSON (dep@digitaltransitions.com), Digital Transitions
Dealer for Phase One, Mamiya Leaf, Arca-Swiss, Cambo, Profoto
Office: 877.367.8537
Cell: 740.707.2183
Phase One IQ250 FAQ
FredBGG
Sr. Member
****
Offline Offline

Posts: 1651


« Reply #76 on: April 22, 2013, 11:53:04 AM »
ReplyReply

Doug has informed me I was mistaken, the sensor in my back is not a dual co-joined sensor such as was in some backs, not sure what I'm seeing, there is a definite horizontal line running down the middle but apparently it's not a join. There were some backs using this method however were there not historians?

The sensor is not made out of two chips physically joined together, however the chip is manufactured one area at a time due to the maximum size
that can be made at on time by the stepper.The chip is moved between tiles. This is why they are visable to the naked eye when looking at the sensor.
Calibration has to be performed to not make the bounderies visable in final images. Not all copies of sensors will have the same precision. It is very complicated to reposition the chip between tiles when tollerences are so infinately small.



many people have had issues with these bouderies showing up. Both calibration and software tricks help resolve the issue.
Logged
FredBGG
Sr. Member
****
Offline Offline

Posts: 1651


« Reply #77 on: April 22, 2013, 12:06:33 PM »
ReplyReply

I guess it's a matter of semantics what you call it (e.g. "stitch" vs "etch" vs "segmented").

Most modern medium format CCDs have more than one read out area, which requires a read-out channel transit the sensor, which creates a visual line when you hold the sensor at an angle to direct light. They are laser etched out of a single piece of silicon however, and are not "joined" in the sense that they do not create sensors and then glue them together. Such an approach would be extremely hard to use in made-for-standard-photographic-applications camera since, as pointed out above you'd have to have butt them together within one pixel (e.g. 6 microns) apart and then calibrate them to produce nearly identical read-outs across a huge spectrum of possible exposures.

Are CCDs made by etching with a laser?
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7394


WWW
« Reply #78 on: April 22, 2013, 02:38:27 PM »
ReplyReply

Not as far as I know. Chips are normally made in a photolitographic processs.

Best regards
Erik

Are CCDs made by etching with a laser?
Logged

ondebanks
Sr. Member
****
Offline Offline

Posts: 833



« Reply #79 on: April 22, 2013, 07:11:38 PM »
ReplyReply




What back is that, Fred? Looks like a 60 or 80 MP full 645 frame?

With this discussion of multiple stepper blocks laid down on a single MFD chip, I think we strayed away from Ben's question, which was about "using a couple of D800 sensors joined together". A system which integrates multiple self-contained sensors could be done, but not gaplessly, for the reason I gave above.

However, such an array of sensors might be good for stitching panos in a very short time, or simulating much larger formats - two slightly displaced shots would be all you need to fill in the gaps. Think of Fotodiox's RhinoCam concept, with not one NEX camera, but several NEX sensors (or D800 sensors) side by side.

Ray
Logged
Pages: « 1 2 3 [4] 5 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad