Medium Format Forum
Register a free account now!
If you are registered, you get access to the members only section, can participate in the buy & sell second hand forum and last but not least you can reserve your preferred username before someone else takes it.
How much bigger can the sensor be for the Hasselblad H system
- Thread starter wang
- Start date
Anyone trust their eyes anymore? Just curious.
I did say "16 bit might have produced a tad better detail." MIGHT HAVE.
One way to determine if it could have, would be to do it. Some people have a need to know, while others have a need to see. Perhaps both are valid paths to the same conclusions?
With RAW files you have that option at anytime. Unfortunately, I don't have the RAW files of the shot.
What I do have is the same shot as 8 bit tiffs done at ISO 100, 200, 400, and 800. I would be curious if the 800 file would benefit from 16 bit processing for a 100% print. If I couldn't tell the subtile differences, then 8 bit is it. However, experience tells me it probably would benefit from 16 bit because the tonal aspects are under assult @ ISO 800.
Meaning, when the ISO 100 file is compared @ 100% to the 800 file @ 100%, there was a loss of very subtile detail in some skin areas @ 800. I'm sure there are other reason for this, yet I also am curious whether 16 bit would've preserved some of those details which I know are there from the lower ISO files.
Mostly, the curiosity is based on results you can actually see ... the degree of which can vary from image to image. So, to play it safe I usually opt for 16 bit processing for as long as I can during the processing stage.
I did say "16 bit might have produced a tad better detail." MIGHT HAVE.
One way to determine if it could have, would be to do it. Some people have a need to know, while others have a need to see. Perhaps both are valid paths to the same conclusions?
With RAW files you have that option at anytime. Unfortunately, I don't have the RAW files of the shot.
What I do have is the same shot as 8 bit tiffs done at ISO 100, 200, 400, and 800. I would be curious if the 800 file would benefit from 16 bit processing for a 100% print. If I couldn't tell the subtile differences, then 8 bit is it. However, experience tells me it probably would benefit from 16 bit because the tonal aspects are under assult @ ISO 800.
Meaning, when the ISO 100 file is compared @ 100% to the 800 file @ 100%, there was a loss of very subtile detail in some skin areas @ 800. I'm sure there are other reason for this, yet I also am curious whether 16 bit would've preserved some of those details which I know are there from the lower ISO files.
Mostly, the curiosity is based on results you can actually see ... the degree of which can vary from image to image. So, to play it safe I usually opt for 16 bit processing for as long as I can during the processing stage.
Hi Marc,
> Anyone trust their eyes anymore? Just curious. > > I did say "16 bit might have produced a tad better detail." MIGHT HAVE.
Certainly, and all I asked was why you believed more bit depth gave you more detail, that's all. I just wanted to see what your understanding of the process was. Not a thing wrong with what you said.
> One way to determine if it could have, would be to do it. Some people > have a need to know, while others have a need to see. Perhaps both are > valid paths to the same conclusions?
I agree %100. Of course you should go with your eyes, but...knowing the theory as well, and seeing if your eyes/experience are even remotely close to the theory is valuable information for some. If you (not you in particular) aren't getting the results you think you should be (or are getting results that don't make sense), then perhaps there is something in the workflow that can help.
> However, experience tells me it > probably would benefit from 16 bit because the tonal aspects are under > assult @ ISO 800.
The number of usable bits is directly related to noise. At the higher ISOs, the noise would eat into the lower bits, and therefore they are less likely to be good. I don't know the characteristics of the particular system you are using, so I can't say for sure...but in all my systems, that is true.
> Meaning, when the ISO 100 file is compared @ 100% to the 800 file @ > 100%, there was a loss of very subtile detail in some skin areas @ > 800. I'm sure there are other reason for this, yet I also am curious > whether 16 bit would've preserved some of those details which I know > are there from the lower ISO files.
I would believe the difference is as I said above, noise, so I'm not sure that 16 bits (which I don't really believe all 16 are "good" BTW, it's just a convenient step up from 8...I'd bet more like 12 good bits which is a dynamic range of 1:4096 or 3.6) would help. It would be interesting to see the results of a test.
> So, to play it safe I > usually opt for 16 bit processing for as long as I can during the > processing stage.
For that reason alone, I believe it's worth doing most everyting in HDR (High Dynamic Range) mode. But in reality, if you shoot it "right" (or scan it right) meaning get the exposure right in the first place, and do little to no tonal manipulation post processing, you'll see little to no difference. So, either way is certainly valid, it depends more on your workflow and types of images you shoot IME.
Regards,
Austin
> Anyone trust their eyes anymore? Just curious. > > I did say "16 bit might have produced a tad better detail." MIGHT HAVE.
Certainly, and all I asked was why you believed more bit depth gave you more detail, that's all. I just wanted to see what your understanding of the process was. Not a thing wrong with what you said.
> One way to determine if it could have, would be to do it. Some people > have a need to know, while others have a need to see. Perhaps both are > valid paths to the same conclusions?
I agree %100. Of course you should go with your eyes, but...knowing the theory as well, and seeing if your eyes/experience are even remotely close to the theory is valuable information for some. If you (not you in particular) aren't getting the results you think you should be (or are getting results that don't make sense), then perhaps there is something in the workflow that can help.
> However, experience tells me it > probably would benefit from 16 bit because the tonal aspects are under > assult @ ISO 800.
The number of usable bits is directly related to noise. At the higher ISOs, the noise would eat into the lower bits, and therefore they are less likely to be good. I don't know the characteristics of the particular system you are using, so I can't say for sure...but in all my systems, that is true.
> Meaning, when the ISO 100 file is compared @ 100% to the 800 file @ > 100%, there was a loss of very subtile detail in some skin areas @ > 800. I'm sure there are other reason for this, yet I also am curious > whether 16 bit would've preserved some of those details which I know > are there from the lower ISO files.
I would believe the difference is as I said above, noise, so I'm not sure that 16 bits (which I don't really believe all 16 are "good" BTW, it's just a convenient step up from 8...I'd bet more like 12 good bits which is a dynamic range of 1:4096 or 3.6) would help. It would be interesting to see the results of a test.
> So, to play it safe I > usually opt for 16 bit processing for as long as I can during the > processing stage.
For that reason alone, I believe it's worth doing most everyting in HDR (High Dynamic Range) mode. But in reality, if you shoot it "right" (or scan it right) meaning get the exposure right in the first place, and do little to no tonal manipulation post processing, you'll see little to no difference. So, either way is certainly valid, it depends more on your workflow and types of images you shoot IME.
Regards,
Austin
Q.G.,
> Posted by Q.G. de Bakker (Qnu) on Saturday, March 24, 2007 - 1:26 am: > > Still the same old Austin...
Of course. I am civil, try to understand what people are saying, don't engage in personal attacks, and offer my experience. Exactly what should be done on this list. Unlike you. You're looking for a "fight", trying to point out something wrong (without adding anything significant to the conversation), making personal attacks, and are typically unable to actually carry out a civil discourse.
> Have fun.
I always do, thanks for your concern. But unlike you, not at other people's expense.
You've once again turned a perfectly civil and interesting discourse into a personal attack. I, personally, have had enough of your petty and juvenile pot-shots. If you don't have anything to add that isn't rife with personal comments, simply keep those responses to your self.
Austin
> Posted by Q.G. de Bakker (Qnu) on Saturday, March 24, 2007 - 1:26 am: > > Still the same old Austin...
Of course. I am civil, try to understand what people are saying, don't engage in personal attacks, and offer my experience. Exactly what should be done on this list. Unlike you. You're looking for a "fight", trying to point out something wrong (without adding anything significant to the conversation), making personal attacks, and are typically unable to actually carry out a civil discourse.
> Have fun.
I always do, thanks for your concern. But unlike you, not at other people's expense.
You've once again turned a perfectly civil and interesting discourse into a personal attack. I, personally, have had enough of your petty and juvenile pot-shots. If you don't have anything to add that isn't rife with personal comments, simply keep those responses to your self.
Austin
IMO, Austin and Q.G. are vital information resources to many on this forum. Please don't let a verbal squabble rob us of your collective knowledge.
"I would believe the difference is as I said above, noise, so I'm not sure that 16 bits (which I don't really believe all 16 are "good" BTW, it's just a convenient step up from 8...I'd bet more like 12 good bits which is a dynamic range of 1:4096 or 3.6) would help. It would be interesting to see the results of a test."
Not quite grasping this concept. Are you saying that because of noise at higher ISOs the bit depth is effectively lessened?
Well, actually I can answer my own question. It doesn't matter since the digital back files are processed as either 8 bit or 16 bit tiffs. 16 bit provides the best bet. While you can alter modes from 16 bit to 8 bit you can't go back ... unless you go back to the original RAW file.
"I would believe the difference is as I said above, noise, so I'm not sure that 16 bits (which I don't really believe all 16 are "good" BTW, it's just a convenient step up from 8...I'd bet more like 12 good bits which is a dynamic range of 1:4096 or 3.6) would help. It would be interesting to see the results of a test."
Not quite grasping this concept. Are you saying that because of noise at higher ISOs the bit depth is effectively lessened?
Well, actually I can answer my own question. It doesn't matter since the digital back files are processed as either 8 bit or 16 bit tiffs. 16 bit provides the best bet. While you can alter modes from 16 bit to 8 bit you can't go back ... unless you go back to the original RAW file.
polypal
Active Member
Please gentlemen keep the forum limited to a usefull discussion on matters concerning photography.
Any personal remarks especially from those who prefer to dip their pens in acid gets us nowhere.
Forum etiquette is quite simple:
Do not engage in personal attacks on others.
Paul
Any personal remarks especially from those who prefer to dip their pens in acid gets us nowhere.
Forum etiquette is quite simple:
Do not engage in personal attacks on others.
Paul
Hi Marc,
> "I would believe the difference is as I said above, noise, so I'm not > sure that 16 bits (which I don't really believe all 16 are "good" BTW, > it's just a convenient step up from 8...I'd bet more like 12 good bits > which is a dynamic range of 1:4096 or 3.6) would help. It would be > interesting to see the results of a test." > > Not quite grasping this concept. Are you saying that because of noise > at higher ISOs the bit depth is effectively lessened?
Correct! If you want more of a technical explanation as to why this is, I am more than happy to do so.
Regards,
Austin
> "I would believe the difference is as I said above, noise, so I'm not > sure that 16 bits (which I don't really believe all 16 are "good" BTW, > it's just a convenient step up from 8...I'd bet more like 12 good bits > which is a dynamic range of 1:4096 or 3.6) would help. It would be > interesting to see the results of a test." > > Not quite grasping this concept. Are you saying that because of noise > at higher ISOs the bit depth is effectively lessened?
Correct! If you want more of a technical explanation as to why this is, I am more than happy to do so.
Regards,
Austin
That would be good of you Austin.
While I normally evaluate final prints by eye under an Ott Light, it is as you say "prudent" to know some of the reasons behind those results should you need to figure out how to improve less than stellar prints.
Much appreciated in advance, we can all learn together.
While I normally evaluate final prints by eye under an Ott Light, it is as you say "prudent" to know some of the reasons behind those results should you need to figure out how to improve less than stellar prints.
Much appreciated in advance, we can all learn together.
Hi Marc,
The "bits" measure voltage from the sensor cells. If you are measuring a total voltage swing from the sensor cells of, say, +3V to -3V, you have a total voltage swing of 6V. If you have 8 bits, or 256 possible values, each bit represents 6V / 256 or ~0.023 Volts.
Noise is a variance in the voltage for which the same input "s&le" will give a range of voltage values from the sensor & "analog front end" (AFE). If the noise in the sensor/AFE is, say, +/- .01V, for the above conditions (6V swing, 8 bits), the noise is less than what 1 bit can represent. So, the noise doesn't "eat into the lower bits". But, as a note, you always have an inaccuracy in the lowest bit anyway due to "quantization error", which is another discussion, and is not necessary to explain for this discussion.
But, when the noise increases (as it does when going to higher ISOs), say to +/- .03V, then the noise is greater than the value one bit can represent. This means that a range (for ex&le 100 - 104) of the 256 values will represent the same "s&le". In this case, it will be (.06V - 0.023) / .023V or ~3 bits will be lost, 2 to noise, one to quantization error. So, you effectively have 6 bits, instead of 8.
So, as noise increases, you get less "usable" bits. Does that make sense now?
Regards,
Austin
The "bits" measure voltage from the sensor cells. If you are measuring a total voltage swing from the sensor cells of, say, +3V to -3V, you have a total voltage swing of 6V. If you have 8 bits, or 256 possible values, each bit represents 6V / 256 or ~0.023 Volts.
Noise is a variance in the voltage for which the same input "s&le" will give a range of voltage values from the sensor & "analog front end" (AFE). If the noise in the sensor/AFE is, say, +/- .01V, for the above conditions (6V swing, 8 bits), the noise is less than what 1 bit can represent. So, the noise doesn't "eat into the lower bits". But, as a note, you always have an inaccuracy in the lowest bit anyway due to "quantization error", which is another discussion, and is not necessary to explain for this discussion.
But, when the noise increases (as it does when going to higher ISOs), say to +/- .03V, then the noise is greater than the value one bit can represent. This means that a range (for ex&le 100 - 104) of the 256 values will represent the same "s&le". In this case, it will be (.06V - 0.023) / .023V or ~3 bits will be lost, 2 to noise, one to quantization error. So, you effectively have 6 bits, instead of 8.
So, as noise increases, you get less "usable" bits. Does that make sense now?
Regards,
Austin
Hi Marc,
> If you have 8 bits, or 256 possible values, each bit represents 6V / 256 or ~0.023 Volts.
This isn't quite phrased right. Each value from 0-255 represents a delta of ~0.023 volts from the previous or subsequent value. For ex&le, 1 represents ~0.023V, and 10 represents ~0.230V, and 100 represents ~2.300V.
Regards,
Austin
> If you have 8 bits, or 256 possible values, each bit represents 6V / 256 or ~0.023 Volts.
This isn't quite phrased right. Each value from 0-255 represents a delta of ~0.023 volts from the previous or subsequent value. For ex&le, 1 represents ~0.023V, and 10 represents ~0.230V, and 100 represents ~2.300V.
Regards,
Austin
Thus the "effective" bit depth per different ISOs as opposed to a universal bit depth across all ISOs, right?
So, the question still is: does it make sense to process higher ISOs (such as those at the high end of any given camera ... in this case 800 ), at 16 bit verses 8 bit? Is there still something to gain? This said, I realize it may depend on how accurate the exposure is, or is that incorrect, and it is a non-discriminating effect and exposure play little or no role?
So, the question still is: does it make sense to process higher ISOs (such as those at the high end of any given camera ... in this case 800 ), at 16 bit verses 8 bit? Is there still something to gain? This said, I realize it may depend on how accurate the exposure is, or is that incorrect, and it is a non-discriminating effect and exposure play little or no role?
Hi Marc,
> Thus the "effective" bit depth per different ISOs as opposed to a > universal bit depth across all ISOs, right?
Right. Lower ISO will typically give more bits of "clean" data, higher ISO will typically give less bits of "clean" data, effectively reducing the number of "usable" bits.
> So, the question still is: does it make sense to process higher ISOs > (such as those at the high end of any given camera ... in this case > 800 ), at 16 bit verses 8 bit? Is there still something to gain? This > said, I realize it may depend on how accurate the exposure is, or is > that incorrect, and it is a non-discriminating effect and exposure > play little or no role?
The exposure may be significant. The bottom line is you need to have the data you want to capture placed within the range of the sensor, exactly the same as with scanning film. But, my guess is, for most scenes, the limit will be the camera's sensor, not the scene.
I can't say if it makes sense to process higher ISOs at a reduced bit depth or not. I'd need some actual data, and it will be different from system to system. But, instead of theorizing, I'd suggest experimenting with your own system and see what you see. The most demanding scenes with respect to bit depth, are flat fields, like blue sky. Though a scene you typically won't be using high ISO with ;-)
Regards,
Austin
> Thus the "effective" bit depth per different ISOs as opposed to a > universal bit depth across all ISOs, right?
Right. Lower ISO will typically give more bits of "clean" data, higher ISO will typically give less bits of "clean" data, effectively reducing the number of "usable" bits.
> So, the question still is: does it make sense to process higher ISOs > (such as those at the high end of any given camera ... in this case > 800 ), at 16 bit verses 8 bit? Is there still something to gain? This > said, I realize it may depend on how accurate the exposure is, or is > that incorrect, and it is a non-discriminating effect and exposure > play little or no role?
The exposure may be significant. The bottom line is you need to have the data you want to capture placed within the range of the sensor, exactly the same as with scanning film. But, my guess is, for most scenes, the limit will be the camera's sensor, not the scene.
I can't say if it makes sense to process higher ISOs at a reduced bit depth or not. I'd need some actual data, and it will be different from system to system. But, instead of theorizing, I'd suggest experimenting with your own system and see what you see. The most demanding scenes with respect to bit depth, are flat fields, like blue sky. Though a scene you typically won't be using high ISO with ;-)
Regards,
Austin
wbulte
Active Member
Hi Austin,
As a side note: your ex&le uses base-10. Your 0.023V corresponds to 1 LSB (base-2).
Can you comment on the 3/1000't of a Volt? I'm thinking of thermal drift here, is the ADC functionality stable enough to always get this minute voltage difference right? I would imagine that the actual sensor can have distinctly different temperatures, due to high and low light intensities entering it from the lens. Say the sun sitting in a corner of the picture or something similar.
Wilko
As a side note: your ex&le uses base-10. Your 0.023V corresponds to 1 LSB (base-2).
Can you comment on the 3/1000't of a Volt? I'm thinking of thermal drift here, is the ADC functionality stable enough to always get this minute voltage difference right? I would imagine that the actual sensor can have distinctly different temperatures, due to high and low light intensities entering it from the lens. Say the sun sitting in a corner of the picture or something similar.
Wilko
Thanks Austin,. Well, when I get an opportunity to shoot a clear sky, I can shoot the Leaf Aptus @ 800 on the focal plane shutter Mamiya 645 AFD-II, and use Zeiss glass stopped down. The Mamiya offers shutter speeds to 1/4000th coupled with say f/32 of a 180/CFE. That should get us there : -)
Hi Wilco,
> As a side note: your ex&le uses base-10. Your 0.023V corresponds to > 1 LSB (base-2).
Not sure what you're getting at here... 1 base 2 is the same as 1 base 10.
> Can you comment on the 3/1000't of a Volt? I'm thinking of thermal > drift here, is the ADC functionality stable enough to always get this > minute voltage difference right?
Probably not. In the ex&le I gave, the range you'll get for 8 bits is +/- 1/2 of the lsb, or +/- ~.01V.
> I would imagine that the actual > sensor can have distinctly different temperatures, due to high and low > light intensities entering it from the lens. Say the sun sitting in a > corner of the picture or something similar.
The sensor will only see the light for the duration of the speed of the shutter. I don't see this as an issue, unless you're shooting VERY long exposures, which will be way out of the range of the sensor if you are shooting the sun.
Regards,
Austin
> As a side note: your ex&le uses base-10. Your 0.023V corresponds to > 1 LSB (base-2).
Not sure what you're getting at here... 1 base 2 is the same as 1 base 10.
> Can you comment on the 3/1000't of a Volt? I'm thinking of thermal > drift here, is the ADC functionality stable enough to always get this > minute voltage difference right?
Probably not. In the ex&le I gave, the range you'll get for 8 bits is +/- 1/2 of the lsb, or +/- ~.01V.
> I would imagine that the actual > sensor can have distinctly different temperatures, due to high and low > light intensities entering it from the lens. Say the sun sitting in a > corner of the picture or something similar.
The sensor will only see the light for the duration of the speed of the shutter. I don't see this as an issue, unless you're shooting VERY long exposures, which will be way out of the range of the sensor if you are shooting the sun.
Regards,
Austin
wbulte
Active Member
Hi Austin,
Eh, um, of course
I was mixing my line of thought with a P&S camera with an electronic LCD viewfinder. Heating the sensor with a SLR during the typically short exposure would call for some really weird subject.. (now where did I leave that supernova again? ;-)
All that said: I think I remember that for really long exposures (astrophotography) they cool down the sensor to well below freezing. With a Peltier element if I remember right. This helps to improve the signal to noise ratio. I am not sure (but curious) if cooling the sensor is also done for "normal" photography?
Wilko
Eh, um, of course
I was mixing my line of thought with a P&S camera with an electronic LCD viewfinder. Heating the sensor with a SLR during the typically short exposure would call for some really weird subject.. (now where did I leave that supernova again? ;-) All that said: I think I remember that for really long exposures (astrophotography) they cool down the sensor to well below freezing. With a Peltier element if I remember right. This helps to improve the signal to noise ratio. I am not sure (but curious) if cooling the sensor is also done for "normal" photography?
Wilko
Yeah, NDs. Actually, I've a whole set for the 203FE because I like shooting the 110/2 FE wide open at noon ... in the summer ... at the beach ... LOL.
Well, if 1/4000th at f/32 doesn't get it then either the sun went Nova on us, or it's a Nuclear blast I'm shooting : -)
Yes, cooling is done. Even the CFV has a fan. I once had my old Kodak back go down on a 100F day while shooting an event. Had to take in the house and set it by the air-conditioning vent. Worked fine after that, I just gave it a glass of ice-cold lemonade once in awhile : -)
Well, if 1/4000th at f/32 doesn't get it then either the sun went Nova on us, or it's a Nuclear blast I'm shooting : -)
Yes, cooling is done. Even the CFV has a fan. I once had my old Kodak back go down on a 100F day while shooting an event. Had to take in the house and set it by the air-conditioning vent. Worked fine after that, I just gave it a glass of ice-cold lemonade once in awhile : -)