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!

P25 Plus noise test/ Very Boring sorry

Guy Mancuso

Administrator, Instructor
Guy,

your CF000040 was three stops underexposed, CF000064 even more. I mean with "underexposed" compared to the maximum exposure without clipping (except for the lamps, their center has been clipped ic CF000062).

This makes me think that the P25 Plus does not have different ISOs, just like the Sinar 54 (unlike the P45 Plus, which does have different ISO gains.

I can tell this with certainty only if I see raw files with a serie of ISOs; same scenery, same lighting.

Thierry,

those TIFs are pretty raw. You can open them with ACR (in Photoshop or Lightroom).

Gabor
So how are you measuring the underexposure . If there were that dark don't you think i would see that in C1 , LR or on the LCD screen or the histogram. If that is the case than every P25 file out there is 3 stops under. This makes no sense. You can't go by some of the highly reflective metal either
 
P

Panopeeper

Guest
yes, I have seen that one can open them, and noticed the same under-exposure
Thierry,

you can see the magnitude of underexposure in ACR only by applying -2 EV "exposure" correction, because ACR applies automatically +2 EV. Note, that this does not show up on the slider.

Here are a ton from the other day . Look in the jack download here
Guy,

all those are ISO100.

Gabor
 
T

thsinar

Guest
Yes Gabor, that is my understanding. I have one question, related to this "automatic +2" applied by ACR: how is this managed/controlled in ACR, based on what criteria?

Thanks and best regards,
Thierry

Thierry,

you can see the magnitude of underexposure in ACR only by applying -2 EV "exposure" correction, because ACR applies automatically +2 EV. Note, that this does not show up on the slider.

Gabor
 

Guy Mancuso

Administrator, Instructor
I will run a new test today of the bathroom. My one issue is i have a old AFD loaner body that is not very well behaved and some images are a stop under. I will shoot the bathroom over. The Kitchen is a tough nut to shoot without lights because of reflections and highlights. I will post the raws to my public folder later today and let you know when they load , takes forever to load.
 

Guy Mancuso

Administrator, Instructor
Did this over again this time right on the money for expousre on the macbeth and right side of frame left side just a touch of blowing out from the window , did not adjust anything. Luminance at 10 color at 51 and lowered the sharpening down

100
200
400
800

Full shot ISO 100 . I will post the 4 raws in a little bit here
 

Guy Mancuso

Administrator, Instructor
Those settings are pretty much to achieve the best sharpness with Luminance at 10. But I would most likely go a touch higher to smooth a little more
 
P

Panopeeper

Guest
So how are you measuring the underexposure
From the maximum possible exposure backwards, based on the non-demosaiced raw data. "Underexposure" is certainly not the correct term. If the exposure is for example 1/2 EV from the very right, that does not mean that the shot is in fact underexposed.

Anyway, I don't know any better term for this. "Perfect ETTR minus 3 EV" is correct, but a little bit too long.

If there were that dark don't you think i would see that in C1 , LR or on the LCD screen or the histogram
ACR is automatically adjusting the exposure of the ISO 800 shots by +2 EV, the ISO 400 by +1 EV and the ISO 100 by -1 EV. I guess C1 is doing that as well, as is the camera's firmware.

If that is the case than every P25 file out there is 3 stops under
Only the ISO 800 shots, if you meter correctly. I went through your shots CF000127-CF000145. All are ISO 100, and most of them are well exposed:

CF000127: -2 2/3 EV
CF000128: -1 1/3 EV
CF000129: -2/3 EV
CF000130: -1/3 EV
CF000131: -1 EV
CF000132: -1 EV
CF000133: -2/3 EV
CF000134: perfect ETTR
CF000135: perfect ETTR
CF000136: perfect ETTR
CF000137: -1/3 EV
CF000138: -1 EV
CF000139: -1/3 EV
CF000140: -1 EV
CF000141: -1/3 EV
CF000142: -1/3 EV
CF000143: -1/3 EV
CF000144: -1/3 EV

(Some *tiny* clippings due to reflections and turned on light bulbs occured in several shots; I qualified those too as "perfect ETTR".)

Now, let's see the exposure of the ISO serie 100-400, CF0000251-CF000254:









The ISO 100 shot is almost perfectly exposed (only 1/3 EV down from the max). However, the pixel values went down with the exposure, stop by stop (the thick white bars under the graphs mark one stop, the thin ones mark 1/3 stop). With ISO gain the pixel values would remain roughly on the same level, i.e. there was no ISO gain.

Using higher ISO is equivalent to a negative exposure bias regarding the raw data. The reason to use different ISOs with this camera is only to see an adjusted preview in-camera and the embedded JPEG image.

In other words, you must not use ISO as a compensation for exposure.

Btw, I do not know, why ACR regards ISO 200 as "native" (i.e. it does not adjust the exposure automatically).

Another aspect: ACR increases the noise reduction as the ISO increases. This too remains hidden from you.
 

Guy Mancuso

Administrator, Instructor
Did you check the new raws from the bathroom with 100,200,400 and 800. Now i have to say from a shooters point of view , I think all we care about is ISO 400 and 800 going to look good. But i understand we should know the science behind it but at the end of the day that would be the bottom line. Now i don't know about what ACR and C1 are doing as these images come in the door. Have you tried bringing these in C1
 
P

Panopeeper

Guest
Did you check the new raws from the bathroom with 100,200,400 and 800
CF0000251-CF000254 are the bathroom shots (I posted their raw histograms above).

Now i have to say from a shooters point of view , I think all we care about is ISO 400 and 800 going to look good. But i understand we should know the science behind it but at the end of the day that would be the bottom line
Guy, there is no scienece here (until now). There is no ISO 400 and 800 (not even 200) in this camera, that's all I am saying. When using higher ISO than 100, you are simply underexposing.

The difference between plain underexposure and underexposure with ISO gain is, that the noise would be lower with ISO gain. I could demonstrate that, but I don't feel anyone here is keen to see that.

You may be happy with the noise after noise removal, but that costs fine details.

Now i don't know about what ACR and C1 are doing as these images come in the door. Have you tried bringing these in C1
I don't have C1; I looked at them in raw format and in ACR.
 

David K

Workshop Member
It sounds interesting but I don't think I'm picking up what you're putting down... What do you mean when you say the camera (presumably back) doesn't have ISO above 100 and that the higher settings are simply "underexposures". How does this differ from a camera that, for lack of a better vocabulary, has a "true" ISO 200 (or higher).
 

Jack

Sr. Administrator
Staff member
He's saying that back manufacturers apply a gain tag to any ISO shot over the back's native ISO, which is applied during the raw conversion stage. Since pumping exposure also increases noise, the raw converter also applies some proprietary noise reduction when it sees the tag. While this may in fact be the case, it is a phenomena of ALL digital sensors and cameras, not just MF DB's; The Leica M8 does this, the Nikon D3 does this and the Canon 1Ds3 does this.

ALL sensors have a single, native ISO and it is raw processing that allows for USABLE higher ISO functionality. So the important factor in my mind, is simple: How good does the processing software do this job? If I get an ISO 800 exposure that looks as low noise as Guy's bathroom shot above while maintaining that much detail, I am a very happy camper! Moreover, when sized to match and compared as prints, the latest generation of MF DB offers a significantly better "noise" signature than any DSLR at similar ISO, at least in my opinion and for my uses.

Cheers,
 

David K

Workshop Member
Thanks for clearing that up... it's kind of what I thought but couldn't express. Seems like we're back where we should be... if the image looks good than how it got there is only of theoretical importance. Would I be correct in assuming that a back with a higher native ISO is more likely to generate a better high ISO image. If so, then this might be a question that prospective back purchasers should ask.
 

Guy Mancuso

Administrator, Instructor
Interesting that the P25 plus came in at the native ISO at 200. i wonder if the P30 plus is actually at ISO 400 there native ISO. Finally i got my arms around this a little, been working too hard on this book project and too much espresso. LOL
 

Jack

Sr. Administrator
Staff member
Seems like we're back where we should be... if the image looks good than how it got there is only of theoretical importance.
Precisely :)

Would I be correct in assuming that a back with a higher native ISO is more likely to generate a better high ISO image.
Sure. (BUT...) It will depend a lot on the physics of the native sensor. For example, how hot the sensor was at capture will have a significant effect on total noise, and total noise gets amplified exponentially as gain is applied. This is where I believe the technological advances in higher ISO performance are occurring. Manufacturers are figuring ways of capturing images using less power, therefore keeping sensors cooler and thus generating less noise -- along of course with obtaining improved battery life. Other active cooling methods are likely in play too, especially for sensors that allow long capture times...

Cheers,
 
P

Panopeeper

Guest
It sounds interesting but I don't think I'm picking up what you're putting down... What do you mean when you say the camera (presumably back) doesn't have ISO above 100 and that the higher settings are simply "underexposures". How does this differ from a camera that, for lack of a better vocabulary, has a "true" ISO 200 (or higher).
Digital sensors do not have different light sensitivities; they capture photons and "count" them. Unfortunately, this "counting" is not an accurate process. I tried to describe this in Post #55. In short, ISO gain means, that the sensor tries to get more detailed data from the same electric charge, though this is inherently less accurate, i.e. more noisy.

When you underexpose a shot and "correct" it in raw processing, for example by the "Exposure" slider of ACR, the raw processor simply multiplies the original raw values. +1 EV causes a multiplication by two. The noise (irregularity) gets multiplied by two as well; thus, areas, which were too dark, now appear brighter - but noisier.

The ISO gain is noisy as well, but to a lesser degree. When increasing the ISO by one stop, the pixel values get roughly doubled, but the noise is less than doubled. With higher and higher ISO, the "noise gain" increases, to a point, where the ISO gain is not better than simply multiplying all pixel values. Some cameras offer even higher ISO settings, but they simulate the effect by multiplying the values in-camera, just like in raw processing these are "fake ISOs".

it is a phenomena of ALL digital sensors and cameras, not just MF DB's; The Leica M8 does this, the Nikon D3 does this and the Canon 1Ds3 does this
It is not a common phenomenon for all digital sensors, not even for MFDBs. I know definitively, that the Phase One P45 Plus does have real ISO gain.

Let's see the difference between the ISO effects of these two cameras.

First, two captures from the P45+; the first is with ISO 400, shutter 1/80s, aperture f/5; the second with ISO 800, shutter 1/80s, aperture f/7.1.

I selected a small area (marked by the orangy rectangle), which is uniformly colored and smooth, thus the noise can be easily measured in term of standard deviation per color.

The red, green and blue pixel averages in the seelcted area are marked by yellow, the standard deviations by magenta.

The average raw pixel values are very close despite the one stop lower exposure of the ISO 800 shot, i.e. the values are roughly doubled (the f/7.1 shot captured only half the photons compared to the f/5 shot). However, the noise is much less than doubled.





Now let's see the comparable P25+ shots. The first is ISO 400, 1/8s, f/13, the second is ISO 800, 1/15s, f/13. I applied +3 EV respectively +4 EV brightness adjustment in the display; this does not affect the raw pixel values, only the RGB.

As the exposure was halved without ISO gain, the second shot shows much lower pixel values (somewhat less than the half of those in the ISO 400 shot). However, the noise is much more than half of that in the ISO 800 shot due to the low exposure (this shot is almost 3 EV from the right edge, as shown by the histogram above).

When the pixel values get doubled in raw processing, the noise gets doubled as well, and that will be much more, than it was in the ISO 400 shot.





Please ignore the appearance of noise on these images, they are GIFs. This layered TIFF file contains the sceen captures without the deterrioration caused by GIF.

An important note: one might think based on these shots, that the noise of the P25+ at ISO 400 is less, than that of the P45+. This comparison is not valid; the P45+ shot can be compared only to the other P25+ shot, because the P45+ ISO 400 shot is already with two stops ISO gain!

the Nikon D3 does this and the Canon 1Ds3 does this
All present DSLRs I have analyzed (two-three dozens) have real ISO gains, but many, perhaps most of them support even higher settings, and those settings are numerical dervatives, done in-camera, which is useful for JPEG only. For example the Nikon D3's highest true ISO is 6400.

the latest generation of MF DB offers a significantly better "noise" signature than any DSLR at similar ISO, at least in my opinion and for my uses
I can hardly debate your uses, but from the present shots is clear: the P25+ has a very good noise characteristic at ISO 100, but the lack of real ISO gain is visible when underexposed.

The consequence of all the above is, that "higher ISO" should be used with the P25+ only when the exposure can not be increased at all, not as a substitute for higher exposure.
 

Jack

Sr. Administrator
Staff member
Maybe I am misunderstanding you as well... The only difference is where (when and how) in the process the gain is applied. You either crank up chip voltage or amplify the chip's base signal after capture, right? Bottom line is either method generates more noise than native and needs to be dealt with somewhere. At the end of the day, if whatever system being used generates an acceptable result for the end user's needs, what difference does it really make?

The little bit of noise showing in Guy's P25+ ISO 800 example would be acceptable to me in many situations, and is certainly better than my 5D or 1Ds3 I tested was at 800. At 400, I can hardly see any objectionable noise in the P25+ example above. I should have my P45+ in a few days and I will post similar examples for us all to compare. My impression in Puerto Rico after the casual comparison Guy and I did there was that the P25+ did a little BETTER job on noise at 800 than the P45+ on the same image.

What I think the more relevant discussion would be, is how this affects the total DR...
 
P

Panopeeper

Guest
The only difference is where (when and how) in the process the gain is applied
Somewhat oversimplified, but one can think of it this way (the kind of digital representation of numbers has nothing to do with their meanings):

if you have a number and multiply it by two, the low-order bit becomes zero in binary representation; i.e. the information does not increase by the multiplication. This is, what is happening in the raw processing, when applying +1 EV exposure adjustment.

On the other hand, with ISO gain, the low order bit becomes meaningful; it carries the differences between different shades, making dark details distinguishable.

This gain may be close to perfect at low ISOs with a good camera; at higher ISO it becomes more and more often unreliable, i.e. noise. Some cameras offer high ISO gain, but it is virtually useless, not better than simply multiplying the values gained by a lower ISO.

Bottom line is either method generates more noise than native and needs to be dealt with somewhere
Yes, but it is not irrelevant, if the noise doubles or inclreases to a smaller degree when reducing the exposure by one stop.

At the end of the day, if whatever system being used generates an acceptable result for the end user's needs, what difference does it really make?
Of course; I did not mean to communicate, that the P25+ does not have its place. If you are shooting in settings, where the illumination is usually not a problem, then this camera may be perfect for you. However, users have to be aware of the fact, that cranking up the ISO with this camera does not yield the same result as with a camera, which offers true ISO gain.

I do not know, which other MFDBs (beside the P45+) offer real ISO gain; perhaps the fake ISO is the norm with MFDBs and the P45+ is the exception. I can determine this only based on comparable shots with different ISOs, like the Guy's bathroom set; I don't have such from any other MFDB.
 
T

thsinar

Guest
Hi Jack,

No, that is not so with Sinar: I have explained and confirmed a few times that there is ABSOLUTELY NO noise reduction filtering taking place before or during the DNG conversion in either the Sinar eXposure or the Brumbaer Converter.

I cannot speak for other brands or other MFDBs, but I can assure you about Sinar's DNGs: no noise reduction applied at this stage at all.

He's saying that back manufacturers apply a gain tag to any ISO shot over the back's native ISO, which is applied during the raw conversion stage. Since pumping exposure also increases noise, the raw converter also applies some proprietary noise reduction when it sees the tag. While this may in fact be the case, it is a phenomena of ALL digital sensors and cameras, not just MF DB's; The Leica M8 does this, the Nikon D3 does this and the Canon 1Ds3 does this.

Cheers,
 
My impression in Puerto Rico after the casual comparison Guy and I did there was that the P25+ did a little BETTER job on noise at 800 than the P45+ on the same image.
Is there any way we can get the same shot from each of these backs and Gabor can show us whether this is technically true or not?
 
Top