Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
03-18-2020, 10:42 AM   #16
Otis Memorial Pentaxian
stevebrot's Avatar

Join Date: Mar 2007
Location: Vancouver (USA)
Photos: Gallery | Albums
Posts: 42,007
QuoteOriginally posted by ShawnH Quote
The problem with the AV mode is that shutter speeds can drop too low. That's why I am asking Ricoh top do something about it rather that we trying to put together a work around. I have set up a user mode now for night photography. It goes like this: Mode: ATV > Shutter 1/160 > Fstop: 2.0> ISO 100~3200> Bracketing +/- 2 stops 0 > Exposure Compensation -2 stops. This is a setup so that ISO is bracketed from 3200 through 100. The ISO 3200 is used for preview and ISO 100 is used for actual processing. However, I would like a mode that brackets the shutter speed from my predefined value down say 1/30 & 1/15. If those images turn out sharp enough then I would gain in lower noise and better image quality.
Green button M mode with EV comp.


Steve

03-19-2020, 11:32 AM   #17
Forum Member




Join Date: May 2017
Posts: 65
Original Poster
QuoteOriginally posted by MrB1 Quote
That is what Av Mode with bracketing allows you to do. In the Menu (C2 on a KP) the Bracketing Order 2 or 3 are probably the better options - the displayed shutter speed is the one that's flashing on the EV scale on the rear screen. You can then select the range of shutter speeds you want by moving the set (of e.g. 5 shots) along the scale using the EC control.

Philip
Thank you! I was wondering about that too. In any case, Ricoh would be wise to introduce a mode that would organize this method under a single setting.
03-19-2020, 12:16 PM   #18
Pentaxian
normhead's Avatar

Join Date: Jun 2007
Location: Near Algonquin Park
Photos: Gallery | Albums
Posts: 40,442
QuoteOriginally posted by ShawnH Quote
Thank you! I was wondering about that too. In any case, Ricoh would be wise to introduce a mode that would organize this method under a single setting.
Or would it be better to just understand that the lowest ISO is the best Dynamic range.

What happens when companies start building in these presets, is people lose track of what they can accomplish.
Say you have an bird , you're already shooting pretty wide with a 200mm lens, a camera company preset would insist on at least 1/400s just to help compensate for FL and motion blur.

Guess what happens?
The camera sets the shutter speed to 1/400 instead of 1/30s to be safe, because to a computer this shot is not advisable, and I don't get this image.

Instead, I shoot at 400 ISO, and get this shot, because I understand what I'm doing and how my camera works, and I know I will have a much better image if I go against convention and shoot 200mm at 1/30s.

There is just no excuse for not understanding how your gear works, or trusting a camera company to get it right for you by inventing a one setting to work in every shooting senerio type situation.



It's a mistake to think a camera preset will compensate for what every photographer should know. Guys like me who are always willing to take risks to get the best possible image would suffer if they used it.

No doubt you are seeing the reasons why they should have a maximize DR preset. But there are also reasons why they shouldn't'.

This shot is what it is because

I understood how much DR was required,
I understood how much I could "cheat" on the exposure.
I was willing to gamble to get a great shot, not a correct but un-impressive shot.

Those presets don't let you gamble, and if they did use the same settings as I used on this image, there would be other images where they would mess things up for you.

Presets are convenient, but they don't always make the best use of you camera's abilities.
There's no getting away from learning how your camera works if you wish to maximize the potential. The thing with people who get comfortable with presets is, they often think "that's it". They don't realize how much better their image could have been.

Last edited by normhead; 03-19-2020 at 01:49 PM.
03-19-2020, 04:35 PM   #19
Pentaxian
photoptimist's Avatar

Join Date: Jul 2016
Photos: Albums
Posts: 5,113
QuoteOriginally posted by WorksAsIntended Quote
And this is more or less a coincidence, not forced by 14bit by any means. The numbers look very similar and the dynamic increase was pretty linear to bit increase in the past decade, but there is absolutely no proportionality or linearity between those two quantities, unlike some people keep telling in forums until it is considered common sense.[COLOR="Silver"]
The synchronized increase in sensor DR and ADC bits during the history of digital camera development is no coincidence. It's the best engineering solution given a linear photon counting sensor with a certain level of read noise. There's no point in putting a 14 bit ADC on a 12 EV DR sensor, the lowest two bits will be noise. And a camera company would be stupid to put a 12 bit ADC on a 14 EV DR sensor -- more ADC bits would definitely improve the image quality. Thus, the two have improved almost in lock-step.

Sure, for nonlinear sensors and data representations, the DR and the bit-depth might be totally different from each other although given a known nonlinearity of the sensor and known noise characteristics, its possible to find the match between DR and bit-depth and find that improvements in DR merit proportional improvements in bit-depth.

03-19-2020, 05:25 PM   #20
Veteran Member




Join Date: Apr 2018
Posts: 639
QuoteOriginally posted by photoptimist Quote
The synchronized increase in sensor DR and ADC bits during the history of digital camera development is no coincidence. It's the best engineering solution given a linear photon counting sensor with a certain level of read noise. There's no point in putting a 14 bit ADC on a 12 EV DR sensor, the lowest two bits will be noise. And a camera company would be stupid to put a 12 bit ADC on a 14 EV DR sensor -- more ADC bits would definitely improve the image quality. Thus, the two have improved almost in lock-step.

Sure, for nonlinear sensors and data representations, the DR and the bit-depth might be totally different from each other although given a known nonlinearity of the sensor and known noise characteristics, its possible to find the match between DR and bit-depth and find that improvements in DR merit proportional improvements in bit-depth.
I understand where you are comming from, as the EV scale is a log2 scale, but let me make an example: We count photons. The amount of photons in 1ev can be much bigger than 2*14bit can store for a single cell on many pictures. So if we got a sensor that is perfectly linear and detects every single photon with no noise there is the possibilities to fully utelize a 14 bit sensor with zero noise and 1EV only. The perfect resolution is achieved when the exact number of photons is detected and stored inside 1EV the number of photons doubles, so to count 2*2^14 photons you need a 2^14 digit where the 00 0000 0000 0000 stores the 0EV (2^14 photons) and 11 1111 1111 1111 stores the 1EV (2*2^14 photons).

The ADC in the end is nothing else but a electron counter that in a perfectly build sensor is nothing else as a photon counter.


If you look at labority equipment, that is much more linear than a camera sensor, at least for specific wavelengths, you will find 8-16bit equipment for many different EV scales.


Sure, if you found out that 12EV fit 12bit very well, you naturally want to go 14bit with 14EV. But there is absolutly no reason why there is only 14 bit of noise when I build a 28bit sensor for 14EV as the EV do not predict any absolut number of photons/electrons. The quantum parameter in this system is the photon or respectily in the electronical part the electron and this is the only bottleneck of the resolution that is theoretically achievable.


In the very first post about this topic it even sounded like a 14bit sensor will automatically be able to detect 14EV just because it is 14bit.
03-19-2020, 06:16 PM - 1 Like   #21
Pentaxian
photoptimist's Avatar

Join Date: Jul 2016
Photos: Albums
Posts: 5,113
QuoteOriginally posted by WorksAsIntended Quote
I understand where you are comming from, as the EV scale is a log2 scale, but let me make an example: We count photons. The amount of photons in 1ev can be much bigger than 2*14bit can store for a single cell on many pictures. So if we got a sensor that is perfectly linear and detects every single photon with no noise there is the possibilities to fully utelize a 14 bit sensor with zero noise and 1EV only. The perfect resolution is achieved when the exact number of photons is detected and stored inside 1EV the number of photons doubles, so to count 2*2^14 photons you need a 2^14 digit where the 00 0000 0000 0000 stores the 0EV (2^14 photons) and 11 1111 1111 1111 stores the 1EV (2*2^14 photons).

The ADC in the end is nothing else but a electron counter that in a perfectly build sensor is nothing else as a photon counter.


If you look at labority equipment, that is much more linear than a camera sensor, at least for specific wavelengths, you will find 8-16bit equipment for many different EV scales.


Sure, if you found out that 12EV fit 12bit very well, you naturally want to go 14bit with 14EV. But there is absolutly no reason why there is only 14 bit of noise when I build a 28bit sensor for 14EV as the EV do not predict any absolut number of photons/electrons. The quantum parameter in this system is the photon or respectily in the electronical part the electron and this is the only bottleneck of the resolution that is theoretically achievable.


In the very first post about this topic it even sounded like a 14bit sensor will automatically be able to detect 14EV just because it is 14bit.
Current digital cameras are photon counters in principle but that does not mean that an input signal of 1 photon per pixel induces an output digital value of 00 0000 0000 0001. There's a numerical proportionality constant that's tied to the quantum efficiency of the color channel; the natural or ISO-amplified analog gain of the read circuit; and the voltage range settings of the ADC. Nor does a digital value of 00 0000 0000 0001 correspond to any particular EV value for the scene. That's determined by the exposure settings.

The maximum sensor DR (before ADC) is determined by the full well capacity of the pixel and the combined effects of read noise and dark current of the sensor which add a stochastic amount of electrons to the signal. Those noise sources define the darkest signal the sensor can register.

What matters is that the sensor is linear in output and the stored digital value is also linear, too. That implies that an average reading of 11 1111 1111 1111 is almost +14EV brighter than an average reading of 00 0000 0000 0001.
03-19-2020, 11:46 PM   #22
Site Supporter
Site Supporter
BruceBanner's Avatar

Join Date: Dec 2015
Posts: 5,404
QuoteOriginally posted by ShawnH Quote
The problem with the AV mode is that shutter speeds can drop too low. That's why I am asking Ricoh top do something about it rather that we trying to put together a work around. I have set up a user mode now for night photography. It goes like this: Mode: ATV > Shutter 1/160 > Fstop: 2.0> ISO 100~3200> Bracketing +/- 2 stops 0 > Exposure Compensation -2 stops. This is a setup so that ISO is bracketed from 3200 through 100. The ISO 3200 is used for preview and ISO 100 is used for actual processing. However, I would like a mode that brackets the shutter speed from my predefined value down say 1/30 & 1/15. If those images turn out sharp enough then I would gain in lower noise and better image quality.
You have multiple ways to control Av modes shutter speeds;

1) Change AUTO ISO Settings values. Select 200 or 400 as the Minimum ISO.
2) Change AUTO ISO Settings values Shutter Speed values, Fast, Normal or Slow.

03-19-2020, 11:59 PM - 1 Like   #23
Veteran Member




Join Date: Apr 2018
Posts: 639
QuoteOriginally posted by photoptimist Quote
Current digital cameras are photon counters in principle but that does not mean that an input signal of 1 photon per pixel induces an output digital value of 00 0000 0000 0001. There's a numerical proportionality constant that's tied to the quantum efficiency of the color channel; the natural or ISO-amplified analog gain of the read circuit; and the voltage range settings of the ADC. Nor does a digital value of 00 0000 0000 0001 correspond to any particular EV value for the scene. That's determined by the exposure settings.

The maximum sensor DR (before ADC) is determined by the full well capacity of the pixel and the combined effects of read noise and dark current of the sensor which add a stochastic amount of electrons to the signal. Those noise sources define the darkest signal the sensor can register.

What matters is that the sensor is linear in output and the stored digital value is also linear, too. That implies that an average reading of 11 1111 1111 1111 is almost +14EV brighter than an average reading of 00 0000 0000 0001.
No, linearity does not mean a 1 to 1 transition, there can be any linearity factor. y = photoncount is as linear as y = 0001 * photon count. The last thing is closer to what happens. The photon count gets rounded to quantities of mutiple thousands because the "bad" counting resolution of the the single pixxel. So with a better sensor at same dynamic range (!) you need a higher bandwidth to store the data without loosing any.
As long as the data resolution is not higher than the quantum parameter there is real data to store. The very basic of digital processing.
The example of course was a very specific situation how a range of 1EV for an example amount of photons (and exposure) could be stored in 14bit, completly linear. With another amount of photons the storage would be different. Of course this is only a single example and not exactly achievable as exposure.
Of course in the processing the quantum parameter is the amount of different levels of current measurable. Current is electrons per time and the amount of electrons is x*photons on a linear cell.

Last edited by WorksAsIntended; 03-20-2020 at 12:12 AM.
03-20-2020, 01:41 AM   #24
Veteran Member




Join Date: Apr 2018
Posts: 639
This graphs shows two possible linear 14bit sensors for different dynamic range.

In a single EV step the number of photons doubles, as does the number of values possible with every bit bandwidth added.

The linearity is possible for every dynamic range and you can make full use of the bits, as long as there are enough values to put into this range. The number of values that are in a range are limited by the quantum parameter. In this case the number of current levels that can be measured, but on a perfect sensor it is the number of photons. So with a better sensor it does make sense to increase the bit depth without increasing the dynamic range.
Attached Images
 

Last edited by WorksAsIntended; 03-20-2020 at 03:30 AM.
03-20-2020, 07:08 AM   #25
Pentaxian
photoptimist's Avatar

Join Date: Jul 2016
Photos: Albums
Posts: 5,113
QuoteOriginally posted by WorksAsIntended Quote
This graphs shows two possible linear 14bit sensors for different dynamic range.

In a single EV step the number of photons doubles, as does the number of values possible with every bit bandwidth added.

The linearity is possible for every dynamic range and you can make full use of the bits, as long as there are enough values to put into this range. The number of values that are in a range are limited by the quantum parameter. In this case the number of current levels that can be measured, but on a perfect sensor it is the number of photons. So with a better sensor it does make sense to increase the bit depth without increasing the dynamic range.
An 8EV, 14bit sensor makes no sense.

The minimum detectable signal for a silicon photodiode is n≥1 electrons (1 is the minimum due to quantization of photons, but noise pushes n higher than that). Thus, a 14bit sensor must be able to register a signal across a range from n electrons to 16384*n electrons. (Note, that implies that the minimum well depth for a 14 bit sensor is 16384 electrons but they are typically much deeper than that because n>1). And the only way the sensor sees a signal ranging from n to 16384*n electrons is if the light levels are ranging over 14 EV.

Sure you can make an 8EV sensor which would be something on a spectrum of designs from 1) a very shallow well-depth sensor (tiny pixels) paired with an ultra-low-noise read circuit to 2) a deep well-depth sensor (big pixels) paired with a noisy read circuit. Examples of the first are found in smartphones. Examples of the second are found in larger sensor cameras optimized for the highest-possible frame-rate and videography (upping the pixel clock makes the read circuits noisy). And although you could pair one of these 8EV sensors with a 14 bit ADC, the lower 6 bits of every reading would be just noise.

Note: All of this is in the context of current camera technologies that use silicon photodiodes.

P.S. The green line on the graph looks wrong. If an 8EV sensor produces a digital output of 16384 with a light input of 8 EV (or 256*N photons), it should produce outputs of 4096 with 6 EV (64*n photons), 1024 with 4 EV (16*n photons), 256 with 2 EV (4*n photons).

Last edited by photoptimist; 03-20-2020 at 08:05 AM. Reason: Added PS comment on graph
03-20-2020, 07:42 AM   #26
Veteran Member




Join Date: Apr 2018
Posts: 639
QuoteOriginally posted by photoptimist Quote
An 8EV, 14bit sensor makes no sense.

The minimum detectable signal for a silicon photodiode is n≥1 electrons (1 is the minimum due to quantization of photons, but noise pushes n higher than that). Thus, a 14bit sensor must be able to register a signal across a range from n electrons to 16384*n electrons. (Note, that implies that the minimum well depth for a 14 bit sensor is 16384 electrons but they are typically much deeper than that because n>1). And the only way the sensor sees a signal ranging from n to 16384*n electrons is if the light levels are ranging over 14 EV.
Of course you need at least 16384 electrons. This is why there is a "multitude" in the label. This usually is in the area of e5,but of course depends on the situation heavily. On the abscissa is a linear photon scale, but not a count, there is a factor n>>1 for every value. This is exactly the point, the number of electrons is much higher than the storage resolution and as long this is the case a higher resolution ( in terms of photon count resolution)sensor is possible to build.
The quantum limit is very far from reached, also we are far from a 100% photon efficiency. This means we need many photons on camera sensors to get a single electron.
There are single cell ccds that can do this and can displayan infinite amount of values in a single ev, just by single counting low photon rates for a long time.

---------- Post added 03-20-20 at 08:05 AM ----------

Another easy example. A real photon counter for the range of 16384-32768 photons (which can be build with a single large photo cell)
First exposure, counts 16384 photons, the second one counts 32768 photons.
The range is 1EV, the number of bits needed to store the information is 14bit, 00 0000 0000 0000 being 16384 photons, and 11 1111 1111 1111 being 32768.
Perfect linear scale for 1EV to 14bit.

---------- Post added 03-20-20 at 08:34 AM ----------

Maybe it would be clearer if I put the label as "number of photons / n, n>7 e N"

Last edited by WorksAsIntended; 03-20-2020 at 07:47 AM.
03-20-2020, 01:23 PM   #27
Forum Member




Join Date: May 2017
Posts: 65
Original Poster
QuoteOriginally posted by normhead Quote
Or would it be better to just understand that the lowest ISO is the best Dynamic range.

What happens when companies start building in these presets, is people lose track of what they can accomplish.
Say you have an bird , you're already shooting pretty wide with a 200mm lens, a camera company preset would insist on at least 1/400s just to help compensate for FL and motion blur.

Guess what happens?
The camera sets the shutter speed to 1/400 instead of 1/30s to be safe, because to a computer this shot is not advisable, and I don't get this image.

Instead, I shoot at 400 ISO, and get this shot, because I understand what I'm doing and how my camera works, and I know I will have a much better image if I go against convention and shoot 200mm at 1/30s.

There is just no excuse for not understanding how your gear works, or trusting a camera company to get it right for you by inventing a one setting to work in every shooting senerio type situation.



It's a mistake to think a camera preset will compensate for what every photographer should know. Guys like me who are always willing to take risks to get the best possible image would suffer if they used it.

No doubt you are seeing the reasons why they should have a maximize DR preset. But there are also reasons why they shouldn't'.

This shot is what it is because

I understood how much DR was required,
I understood how much I could "cheat" on the exposure.
I was willing to gamble to get a great shot, not a correct but un-impressive shot.

Those presets don't let you gamble, and if they did use the same settings as I used on this image, there would be other images where they would mess things up for you.

Presets are convenient, but they don't always make the best use of you camera's abilities.
There's no getting away from learning how your camera works if you wish to maximize the potential. The thing with people who get comfortable with presets is, they often think "that's it". They don't realize how much better their image could have been.
You make a good point. Here is a simplified idea. How about if the camera would produced an extra RAW file at ISO 100 no matter what the camera settings were? That way you would have a file with the maximum DR regardless of other settings.
03-20-2020, 04:55 PM   #28
Pentaxian
normhead's Avatar

Join Date: Jun 2007
Location: Near Algonquin Park
Photos: Gallery | Albums
Posts: 40,442
QuoteOriginally posted by ShawnH Quote
You make a good point. Here is a simplified idea. How about if the camera would produced an extra RAW file at ISO 100 no matter what the camera settings were? That way you would have a file with the maximum DR regardless of other settings.
That's actually a good idea. I wonder if that's something you could do isomer camera filters in jpeg, and then trying to match them in raw. The last time I did that Iw as amazed at how hard it is to get an image as good was what the jpeg engine can turn out. That could work and possibly save a lot of effort for up to 90% of your images. The ones that came out as expected you just use the jpeg, for the ones that don't you use the raw.
03-22-2020, 03:41 PM - 1 Like   #29
Site Supporter
Site Supporter
BruceBanner's Avatar

Join Date: Dec 2015
Posts: 5,404
Dunno if this helps also, but with the KP, Menu 1>ISO Auto Setting>ISO Sensitivity Options>put it to 'Tv'. Now you will notice that the Minimum Shutter Speed below changes from Fast, Normal and Slow to actually allowing you to specify a lowest acceptable shutter speed, so if for example 1/15 is as low as you ever want to go, it will give you that and no lower.

I really wish the K-1 got the same firmware as the KP, it just has a few more options as well as the Aperture Bracketing Shooting Mode.

-------------

Anyway, I'm a bit late to the party, I'm just wanting to get my head around the OP's point. Am I to understand that say the following settings were used in a shot and were necessary to correctly expose the image;

Shot A
1/50
f5.6
ISO 3200

The user could not drop shutter speed further without running the risk of motion blur (either from the subject in the frame moving or user shaking the camera+lens)
f5.6 is the preferred aperture used, optimum sharpness and correct DoF for the required shot
ISO 3200 is given because of the available light the scene is given (ie it's no overly bright environment etc).

If the user took the shot at;

Shot B
1/50
f5.6
ISO 100

then the shot would be grossly underexposed, to chimp the back of the camera LCD screen would show a black screen. However, once the user is back at home in front of a computer, taking Shot B into a RAW editor like LR, increasing Exposure slider and thus bringing the shot 'back to life', once increasing the exposure enough to catch up somewhere close to the brightness that Shot A obtained with it's ISO 3200, would the image quality/noise be better, or worse? Would the detail in the shadows be better or worse? Clearly highlights would be better protected?
And so this is 'ISOless'? The idea that underexposed images are not a deal breaker, because just boost the exposure in post and you get something with a similar or even better IQ than what you would have had at the time of the shot with a higher ISO and more correct exposure?

If Shot B is the intention of the OP, but Av is not allowing holding at 1/50 well enough, does TAv then become a viable option? I just tested now with my K-1. I set my camera to use 1/50, f5.6 and it gave ISO 1250. If I then pressed the EV Comp button I could just lower the ISO and thus underexpose the image and get ISO 100 with still having 1/50 and f5.6.
03-23-2020, 04:22 AM   #30
Pentaxian




Join Date: May 2013
Location: Hertfordshire, England
Posts: 845
QuoteOriginally posted by BruceBanner Quote
Dunno if this helps also, but with the KP, Menu 1>ISO Auto Setting>ISO Sensitivity Options>put it to 'Tv'. Now you will notice that the Minimum Shutter Speed below changes from Fast, Normal and Slow to actually allowing you to specify a lowest acceptable shutter speed, so if for example 1/15 is as low as you ever want to go, it will give you that and no lower.
That's a great tip, Bruce - I hadn't discovered that useful setting, so thank you for posting.

Philip
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
aperture, bit, bracket, camera, chance, customer service, dr, exposed, exposure, increase, iso, iso invariance, iso less, light, mode, pentax mirrorless camera, photographer, photography, range, resolution, ricoh, ricoh imaging, sense, sensor, shutter, suggestion, times
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Pentax K-1 Video Quality - Dynamic Range/Matching with other cameras/Best Practices bobcobb Video Recording and Processing 33 01-05-2021 03:03 AM
Pentax K1 Dynamic Range Question Roadboat24 Pentax Full Frame 12 05-28-2019 04:41 AM
CHIAROSCURO — The Dynamic Range Mistake interested_observer Photographic Technique 10 03-31-2019 02:53 PM
ISO vs. Dynamic Range FozzFoster Pentax DSLR Discussion 20 12-13-2018 04:14 PM
Dynamic Range Expansion mode Rush2112 Pentax DSLR Discussion 6 01-17-2009 02:23 AM



All times are GMT -7. The time now is 08:12 AM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top