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!

The P45+ and DR. And noise. And...

T

thsinar

Guest
Gabor, for your information,

both eXposure and the Brumbaer DNG Converter have a highlight recovery, at all "ISOs". Brumbaer went actually famous and praised by those using it exactly because of this unique "highlight recovery", beside the "white shadings" being applied automatically and in a batch.

I had light situations where I could "recover" more than 3 f-stops in hightlights, thanks to Stefan's unique way to recover.

Don't forget: DNG applications like ACR, LR, Aperture, Raw Developer", etc ... can only recover what is still there. In other words, if the files used in these applications are already "clipped", there is not much to recover.

Best regards,
Thierry

I don't know, how it works with Aperture, but what I see in the DNG file generated by Brumbaer's and eXposure indicates, that the same will happen there.

Capture One may treat the original raw files differently, more sensibly, I don't know.
 

Rethmeier

New member
This is one of the reasons,why I went with the Sinar eMotion-75LV.
Stefan Hess"Brumbear"
No other backs(Phase,Leaf,Blad) can use that option.
Excellent for high contrast scenes!
Cheers,
Willem.
 

fotografz

Well-known member
Gabor, for your information,

both eXposure and the Brumbaer DNG Converter have a highlight recovery, at all "ISOs". Brumbaer went actually famous and praised by those using it exactly because of this unique "highlight recovery", beside the "white shadings" being applied automatically and in a batch.

I had light situations where I could "recover" more than 3 f-stops in highlights, thanks to Stefan's unique way to recover.

Don't forget: DNG applications like ACR, LR, Aperture, Raw Developer", etc ... can only recover what is still there. In other words, if the files used in these applications are already "clipped", there is not much to recover.

Best regards,
Thierry
How can any post program "recover" what is not there?

Frankly, I don't see the practical application of high ISO torture test shots done in relatively good light. It's the deep shadows and high contrast of lower ambient light situations where one is forced to a higher ISO ... and where the problem of noise rears it's head.

Speaking of ISO ... I'm curious as to how much plus or minus tolerance is officially allowed in reporting photographic ISO numbers? I noted that the ISO 800 of the Aptus 75s back I used seemed a bit ambitious, where the ISO 640 of the Leica M8 seemed conservative, and closer to 800 than 640.
I've also been told that there is a fair tolerance in claiming the ISO standards for flash sync leaf shutter speeds.
 
T

thsinar

Guest
Marc,

NONE, can do this, and that's exactly what is my point and what some seem to forget: you can't recover details when they are not there in the first hand.

That is why Brumbaer (AND Sinar eXposure) is unique: it does recover from the very ORIGINAL Sinar RAWS, the ".IA" and ".BR" files, and THERE the information is. And those details are then in the DNGs which one is using further in all these DNG compatible applications.

Yes, I agree, that it makes no sense to "torture" ourselves with well exposed files in optimal light conditions. And that is why I have shot and post mine in less than optimal light, with DEEP shadows, with strong (burnt-out) highlights and with a contrast above the normal.

ISOs: there is no norm specified for this (digital). I have given some information on how Sinar does come to its ISOs. It is very much dependent on the contrast curve which you are using. So if you find the Leaf ISO to be a bit exagerated, it might be because it applies not to the linear curve but to some other default curve.

here what I wrote in another thread:
--------------------------------------------------------------------------------------
The first question one has to ask, is to what the actual (nominal) ISO is corresponding and related. There are of course ISO norms, but those are rather thought for the consumer field and leave open a big "playing gap". Therefore, the manufacturer has to decide by himself to what this ISO should be related.

When doing a lightmetering, the result should lead to a medium grey (Lab 50/0/0) rendered as medium grey. One has to be aware here, that a RGB medium grey does not necessarily correspond to 128/128/128, respectively a centered histogram does not necessarily lead to a correct exposure.

This is strongly depending on the Gamma value of the chosen Colour Working Space: in sRGB, a RGB value of 119/119/119 correspond to a medium grey. In ECI RGB, a medium grey is given with a value of 101/101/101.

The goal of the ISO sensitivity "finding" should therefore be to reach the above value for a medium grey, with ALSO taking in count the used "contrast curve". This has led to the value of ISO 100 as nominal sensitivity for the eMotion 75, and ISO 50 for the eMotion 54 (resp. eMotion 22), with a "default" or "standard" contrast curve of "3" in Captureshop/eXposure.
--------------------------------------------------------------------------------------

Best regards,
THierry

How can any post program "recover" what is not there?

Frankly, I don't see the practical application of high ISO torture test shots done in relatively good light. It's the deep shadows and high contrast of lower ambient light situations where one is forced to a higher ISO ... and where the problem of noise rears it's head.

Speaking of ISO ... I'm curious as to how much plus or minus tolerance is officially allowed in reporting photographic ISO numbers? I noted that the ISO 800 of the Aptus 75s back I used seemed a bit ambitious, where the ISO 640 of the Leica M8 seemed conservative, and closer to 800 than 640.
I've also been told that there is a fair tolerance in claiming the ISO standards for flash sync leaf shutter speeds.
 
Last edited:

fotografz

Well-known member
Marc,

NONE, can do this, and that's exactly what is my point and what some seem to forget: you can't recover details when they are not there in the first hand.

That is why Brumbaer (AND Sinar eXposure) is unique: it does recover from the very ORIGINAL Sinar RAWS, the ".IA" and ".BR" files, and THERE the information is. And those details are then in the DNGs which one is using further in all these DNG compatible applications.

Yes, I agree, that it makes no sense to "torture" ourselves with well exposed files in optimal light conditions. And that is why I have shot and post mine in less than optimal light, with DEEP shadows, with strong (burnt-out) highlights and with a contrast above the normal.

ISOs: there is no norm specified for this (digital). I have given some information on how Sinar does come to its ISOs. It is very much dependent on the contrast curve which you are using. So if you find the Leaf ISO to be a bit exagerated, it might be because it applies not to the linear curve but to some other default curve.

here what I wrote in another thread:
--------------------------------------------------------------------------------------
The first question one has to ask, is to what the actual (nominal) ISO is corresponding and related. There are of course ISO norms, but those are rather thought for the consumer field and leave open a big "playing gap". Therefore, the manufacturer has to decide by himself to what this ISO should be related.

When doing a lightmetering, the result should lead to a medium grey (Lab 50/0/0) rendered as medium grey. One has to be aware here, that a RGB medium grey does not necessarily correspond to 128/128/128, respectively a centered histogram does not necessarily lead to a correct exposure.

This is strongly depending on the Gamma value of the chosen Colour Working Space: in sRGB, a RGB value of 119/119/119 correspond to a medium grey. In ECI RGB, a medium grey is given with a value of 101/101/101.

The goal of the ISO sensitivity "finding" should therefore be to reach the above value for a medium grey, with ALSO taking in count the used "contrast curve". This has led to the value of ISO 100 as nominal sensitivity for the eMotion 75, and ISO 50 for the eMotion 54 (resp. eMotion 22), with a "default" or "standard" contrast curve of "3" in Captureshop/eXposure.
--------------------------------------------------------------------------------------

Best regards,
THierry
Thanks. Interesting info. When I get a chance to discuss this with a knowledgable Hasselblad tech, I'd like to know their POV. I have had little trouble with H3D-II files in terms of recovery.

I do not agree that your ISO 800 shot was a torture test. For ISO 800 to be useful for me, it's usually in some dim conditions that involve motion and the need to get the shutter speed up.

For example, I did a shoot for Muddy Creek, a horse back riding gear manufacturer ... where the job was to document a typical riding trek demonstrating their product in the real world. The product was a unique rain coat that covers the saddle. The shoot started before the sun had risen enough ... so I had to deal with the riding motion as they started out in the early AM.

My torture test:

Girl in water:Hasselblad H3D/31, ISO 800 which was pushed 1.5 stops in post, 255mm @ f/5.4 (this was wide open using a HC150/3.2 with a 1.7X), 1/180th shutter (which was as low as I dare go with the subject in motion).

Cowboy fastening the coat, and 100% detail. This one was @ ISO 800 also, but it only needed about 1/2 to 3/4 stops of "Torture" in post.
 
Last edited:
T

thsinar

Guest
Marc,

I agree with you on the use of high ISOs.

if I read your shooting data correctly, it was ISO 800 at f5.4 at 1/180th.

Mine was f5.6 at 1/40th. I believe that my files is clean enough to go and push it an additional 2 stops, may be even more, and still getting a decent result while capturing something in motion.

I have said it: either way you are doing it, somebody will tell you that it would have been better to test this or that. I have simply made this test in my garden, rushing because I had to "prepare" my daughter to go to school while my wife was looking for my other baby daughter. I have no time to look for moving subjects and only did this test because I was asked to do so by somebody else, and to show that MFDBs (ALL brands, to be clear and to not start another arguing) can easily compete with slr cameras. If somebody is not happy with the way I did my test and the sample presented, he is free to do it under his own conditions. As soon I had started this thread I have been criticized and many have felt the necessity to jump in to show their own ISO 800, ISO 1600, and so on .... It was not my intention to start a brand war nor a comparison with MF brands, but it turned out like this.

Thanks and best regards,
Thierry

My torture test:

Girl in water:Hasselblad H3D/31, ISO 800 which was pushed 1.5 stops in post, 255mm @ f/5.4 (this was wide open using a HC150/3.2 with a 1.7X), 1/180th shutter (which was as low as I dare go with the subject in motion).

Cowboy fastening the coat, and 100% detail. This one was @ ISO 800 also, but it only needed about 1/2 to 3/4 stops of "Torture" in post.
 
T

thsinar

Guest
Welcome.

I have no knowledge how Phocus deals with the files and how it recovers the highlights.
I can speak about Flexcolor and say that (like Captureshop before, and like the other brands' own software) above a certain saturation the highlights are simply and purely clipped and cut away. The reason why this is done is because there are very few situations where all 3 colour channels are fully saturated and without information. Not clipping them would lead to some colour casts in the image. One is loosing an important part of the DR by doing so.

Brumbaer and eXposure do handle this completely differently: as said, there are very few situations were all the 3 colour channels are burnt out completely and at the same time. Most of the time there is some information left in at least one of the 3 channels. It is this information left in only 1 (or sometimes even 2) which is used to recover the 2 other channels.

But again, I do not know how the new Phocus is working here: it would be worth to do a test.

Best regards,
Thierry

Thanks. Interesting info. When I get a chance to discuss this with a knowledgable Hasselblad tech, I'd like to know their POV. I have had little trouble with H3D-II files in terms of recovery.
 

Greg Seitz

New member
Thanks. Interesting info. When I get a chance to discuss this with a knowledgable Hasselblad tech, I'd like to know their POV. I have had little trouble with H3D-II files in terms of recovery.

I do not agree that your ISO 800 shot was a torture test. For ISO 800 to be useful for me, it's usually in some dim conditions that involve motion and the need to get the shutter speed up.

For example, I did a shoot for Muddy Creek, a horse back riding gear manufacturer ... where the job was to document a typical riding trek demonstrating their product in the real world. The product was a unique rain coat that covers the saddle. The shoot started before the sun had risen enough ... so I had to deal with the riding motion as they started out in the early AM.

My torture test:

Girl in water:Hasselblad H3D/31, ISO 800 which was pushed 1.5 stops in post, 255mm @ f/5.4 (this was wide open using a HC150/3.2 with a 1.7X), 1/180th shutter (which was as low as I dare go with the subject in motion).

Cowboy fastening the coat, and 100% detail. This one was @ ISO 800 also, but it only needed about 1/2 to 3/4 stops of "Torture" in post.
Marc,

The Cowboy crop that you show is actually only a 50% crop of the overall shot you showed which effectively means it's a crop of your original downsized to 8 megapixels. Do you have the full 100% crop? Also curious in this circumstance if you needed high ISO why not pull out the D3 or 1DMIII since they seem likely to produce better end results in this case even with their lower number of pixels. Just trying to get my head around the capabilities of these backs.

Thanks,

Greg
 
Last edited:

fotografz

Well-known member
Marc,

The Cowboy crop that you show is actually only a 50% crop which effectively means it's a crop of your original downsized to 8 megapixels. Do you have the full 100% crop? Also curious in this circumstance if you needed high ISO why not pull out the D3 or 1DMIII since they seem likely to produce better end results in this case even with their lower number of pixels. Just trying to get my head around the capabilities of these backs.

Thanks,

Greg
I took the image to 100% (actual pixels) Greg, and cropped a section from that.

My partner shot back-up with a 1DsMKII and the files were not selected by the art director client, in fact my partner trashed most of them after seeing the MFD shots of the same scenes ... the final file use was for laminated trade show murals at 6 feet wide and up ... therefore the Canon nor the Nikon would and could not NOT produce better end results as you presumed without knowing what they were for.
 

LJL

New member
After reading through this thread, I have to admit that I was not quite sure what some of the other torture tests were really trying to show....good DR at high ISO? noise level at high ISO? detail preservation at high ISO? To me, shooting a very bright scene at high ISO, even one with deep shadows, does not seem like a practical or more realistic approach. The push of ISO is going to reduce the DR in most shots, so whatever is in the deep shadows is already going to be suffering, and probably out of range of a more proper exposure for that situation than an exposure where one is also trying to keep the highlights from blowing out.

I realize somewhat the test objectives, but am just not sure the tests are really providing all that much useful data. If you need shadow details, you expose for them, even if that means high ISO and possibly blowing out the highlights. The concept of DR may be supported, but the quality of both the darks and brights are being compromised at the higher ISO. I think that is what Thierry and Marc are talking about with respect to the shooting they do, both for the earlier testing, and for the actual shot delivery to a client. About the only reason you want to push the ISO is if you really need higher shutter speeds, as in Marc's case. I shoot this sort of stuff all the time (high ISO) with DSLRs, because I need high shutter speeds to stop action. The highlights and the deep shadows always suffer, but I get the midtones I need for most subjects.

Thierry's comments about how the Brumbaer and eXposure conversions deal with the data are supported, in my view, and based upon what he has shown. I am not as convinced about some of the other postings, and go back to my original point, remade by Thierry in his ISO discussion, that all algorithms and conversions are not equal. While we may use C1 or ACR or Aperture most often, they are usually NOT providing as good a conversion for higher ISO files. They are being "generalists", by nature, and not taking the pains to deliver the details in the same way as Brumbaer appears to be doing for the Sinar files, and possibly Phocus for the Hasselblad files. I would think that C1 would be doing the same for the Phase back files, but I am not sure at this point.

Sorry for the ramble. This is an interesting thread. What I think may be missing is some real objectivity about just what some of these backs can deliver, or more importantly, which software is able to extract the most from the files they deliver. I am pretty sure that software is not always the most common stuff we want to use for everything else, like ACR/LR, etc. Just my opinion.

As for Marc's cowboy attire shots....I fully agree that he needed to shoot this with MF, as the client wanted a very large output print. This is exactly the thing that I keep fighting myself with some of my work.....the DSLRs will let you work with higher ISOs for shooting, but they are not going to hold up as well for really large prints, posters and banners that a client may want/need. They can do pretty well, but they still are not able to deliver the details as well to handle significant enlargement, from what I am experiencing.

LJ
 

jlm

Workshop Member
what i found interesting was the comment that higher ISO than native simple underexposes and attaches a tag for the processing sw to work with
 

LJL

New member
Here is a link to a rather mathematical explanation and treatment of sensor noise, DR, etc. I am not claiming to be able to understand all of this, but it is worth reading, even past all the equations, to get some understanding of what is going on when we push the ISO from 100 to 400 or 800. Although the actual tests were done with Canon and Nikon DSLRs in this case, and may not seem appropriate in this thread, the theory is very germane and does apply. The more interesting thing to me is that the shot that Thierry took and posted here seems to completely underscore and support what this guy is talking about.....shoot closer to base ISO and under/overexpose to get the best results with processing.

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html

LJ
 

LJL

New member
what i found interesting was the comment that higher ISO than native simple underexposes and attaches a tag for the processing sw to work with
That is precisely the point that the article I linked to makes. The difference comes in with how the software is able to read those tags, and the suggestion is do not rely on the generalist treatment, but manage that yourself in how you process the images. In other words, shoot closer to base ISO and push/pull things with the adjustments in the various converters. In the case of Brumbaer, I think Stefan has gone to some extra lengths to handle each channel much better, and thus produce what may be an optimal conversion. I just do not see that happening with the more generalist conversions that are trying to handle every RAW file produced. I could be very wrong here, but my gut, and experience keeps telling me to look more closely at that software supplied by the camera makers, as it may have a lot more value for better conversions of their files. Unfortunately, until very recently, most of it has been too obtuse for us to seriously consider in our workflows (DPP comes to mind with Canon files for me). Let's hope that starts to change, or that any "universal" files, such as DNG actually do contain all the information and are done with a top notch conversion from the native files BEFORE they get written as DNG files for use in applications that have better workflow.

LJ
 

Greg Seitz

New member
I took the image to 100% (actual pixels) Greg, and cropped a section from that.

My partner shot back-up with a 1DsMKII and the files were not selected by the art director client, in fact my partner trashed most of them after seeing the MFD shots of the same scenes ... the final file use was for laminated trade show murals at 6 feet wide and up ... therefore the Canon nor the Nikon would and could not NOT produce better end results as you presumed without knowing what they were for.
Marc,

Good info regarding the final usage, I was curious to know if the loss of detail from the noise offset the size difference. The D3 would deliver much nicer ISO 800 shots than the 1DsMII but point taken.

Regarding the crop, it's not a 100% crop of the scene you are showing. I upressed your whole scene to the H3D-31 dimensions of 4872 x 6496 and then took a 100% crop of that and get the following which is a much larger area than your crop shows.
 

Graham Mitchell

New member
Whether the correct ISO response is created in hardware or software is a moot point, imo. The results are all that matter.

Either way, when your expose for an ISO 800 image you are capturing 1/8 as many photons as you would if exposing for ISO 100, so the range of possible discrete signal values is smaller. We'll come back to this later.

Afaik, with hardware gain the small signal taken from the sensor is amplified electrically before the AD (analogue-to-digital) conversion. This may or may not result in more noise. Perhaps someone else can chime in. It certainly risks causing clipping unnecessarily.

With software conversion the lower-level data from the AD conversion is multiplied so that the range of data values matches the selected sensitivity. The disadvantage of doing it in software is that the low level signal fed to the AD converter will not take advantage of the full bit depth. In other words, if you underexpose by 3 stops, your signal will only be expressed over 13 bits rather than 16.

However, if there is no meaningful information in the least significant bits (due to the limited number of possible discrete signal values) then the difference becomes academic and insignificant. I suspect that this is the case with MFDBs, and the results seem to support that conjecture. This is why I don't see the point of Panopeeper's analyses, unless of course he is posting merely on behalf of the intellectually curious.
 
P

Panopeeper

Guest
what i found interesting was the comment that higher ISO than native simple underexposes and attaches a tag for the processing sw to work with
ISO settings without associated A/D gain are common among DSLRs as well. For example the highest ISO gain of all Canon's is 1600, with the exception of the 1DMkIII (larger pixels sites!); some are even lower. The top ISO of the Nikon D300 is 1600 and of the D3 is 6400 (very large pixel sites). Many other DSLRs don't go over 800.

Higher ISOs are manufactured numerically in the camera, i.e. the digital values of the highest true gain get multiplied by two or four. This approach is important for JPEG created in-camera, but it reduces the dynamic range by one respecively two stops, so these should not be used when recording raw data. MFDBs are designed for recording raw, thus they don't fake the ISO in the camera, instead the raw processor pushes the intensity. Thus the difference is only formalistic between shooting with ISO 800 and pushing the result by one stop, and shooting with ISO 100 and pushing the result by four stops.

Note: I don't know if there are other MFDBs beside the Phase One P45+ with true ISO gain.
 
P

Panopeeper

Guest
if there is no meaningful information in the least significant bits (due to the limited number of possible discrete signal values) then the difference becomes academic and insignificant. I suspect that this is the case with MFDBs, and the results seem to support that conjecture
Another way to express this is: the 16bit depth allows to store the total usuful gain, as if always the highest useful ISO were selected, in contrast to DSLRs, which store a "window" of the total gain.

In light of this, it is questionable if the 16bit depth of the P45+ is reasonable at all.

This is why I don't see the point of Panopeeper's analyses, unless of course he is posting merely on behalf of the intellectually curious.
There are masses of DSLR owners, who don't know how to make the most of their camera; they are wasting only a few hunder to a few thousand bucks.

However, if someone paid a few tens of thousands bucks for a camera and back, then some intellectual curiosity about how to use it the most effectively should not be the exception.
 

Jack

Sr. Administrator
Staff member
I don't have time to drive around and look for a "blue" subject, as suggested by Jack (for the blue channel).
Thierry, if you are going to quote me, please quote my post directly, not paraphrased as you happen to "remember" it! Here is my actual comment for clarification:

Frankly, I think smooth green flora isn't the best subject to show noise either. As I indicated earlier in this thread, I think the real test is a McBeth card under controlled lighting so we can see how each color responds to each ISO.
So yes, if we want a thread on ISO performance, let's do it in a controlled way -- and we can have panopeeper analyze each cameras capture to his hearts content and report back to us.

In the meantime, I'm more interested in making useful images :D,
 

robsteve

Subscriber
Marc,

The Cowboy crop that you show is actually only a 50% crop of the overall shot you showed which effectively means it's a crop of your original downsized to 8 megapixels. Do you have the full 100% crop? Also curious in this circumstance if you needed high ISO why not pull out the D3 or 1DMIII since they seem likely to produce better end results in this case even with their lower number of pixels. Just trying to get my head around the capabilities of these backs.

Thanks,

Greg
Greg:

Doesn't your method only work assuming the original shot was the full frame image resized for the web and not cropped before posting?

Robert
 
T

thsinar

Guest
Sorry Jack, I was too lazy on this one and to search. All I want to say is that one can as well and even very well check noise behaviour with a green subject.

Best regards,
Thierry

Thierry, if you are going to quote me, please quote my post directly, not paraphrased as you remember it! Here is my actual comment for clarification:



So yes, if we want a thread on ISO performance, let's do it in a controlled way -- and we can have panopeeper analyze each cameras capture to his hearts content and report back to us.

In the meantime, I'm more interested in making useful images :D,
 
Top