Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
11-02-2019, 02:50 PM   #16
Pentaxian




Join Date: Jan 2018
Photos: Gallery
Posts: 1,239
Binary operations

QuoteOriginally posted by jpipg Quote
The rightmost bits contain less information than the even the single least significant of the leftmost bits. In the case that you have a subject with narrow dynamic range, placing the values on the left side of the register gives far larger tonal resolution because those bits contain more data.
I'm hoping I'm completely misunderstanding this, because I'm thinking binary. Individual binary bits have a value of 1 or 0 (whereas decimal digits have a value from 0 to 9) and the leftmost bits can contain no more information than that. Their numerical significance may represent a decimal value of 64, or 128, or 256 etc but it's either that or zero.

Here's the number 167 in 14-bit binary:– 0000000010100111‬ (14-bit dynamic range is what we have to play with, at best)


Shift the bits left by 4 positions and you get:– 0010100111‬0000 which is 2672 in decimal, a much bigger number, so a pixel which originally was a brightness value of 167 jumps up to 2672 by the 4-position shift operation. Great, much brighter!

What happens to a pixel whose original value was zero; which registered no brightness? It goes from 00000000000000 (zero) to 00000000000000. Which is to say, zero. No data is added, no increased brightness is recorded. Damn.

I hope I'm making this clear. It has nothing to do with ETTR, which is about fitting as much of the brightness range of the subject as possible into the actual dynamic range of the camera.


Last edited by StiffLegged; 11-02-2019 at 03:14 PM.
11-02-2019, 03:10 PM - 1 Like   #17
Pentaxian
photoptimist's Avatar

Join Date: Jul 2016
Photos: Albums
Posts: 4,496
QuoteOriginally posted by biz-engineer Quote
Higher ISO as opposed to under-exposure flushes more shadow detail in the RAW files. For example, ISO1600 is 4 stops above ISO100. Shooting at ISO100 underexposed 4 stops, data goes in the RAW file as is. When shooting at ISO1600 with 0 ev exposure compensation, the sensor data bits are shifted 4 positions to the left (MSB on left side notation) and some noise reduction is applied by image processor, then written in RAW file. That's why using ISO1600 offer more dynamic range then ISO100 underexposed by the same amount. That's why ISO100 image pulled shadow contain more details and more noise (minimal or no noise reduction applied to data prior to writting data into RAW format). Unless stacking underexposed exposures, the advantage goes to using a higher ISO setting.
Hmmm... I'm very skeptical of this "bit shifting" explanation. It would imply that the RAW values for an ISO 204800 image (shifted by 11 bits) only have 8 possible values in each channel which create a very spiky histogram.

I would think that high ISO would be implemented either through analog amplification of the signal or by changing the full-scale voltage range on the ADC. Those methods would produce 14-bit outputs although the noise levels would be huge.
11-02-2019, 04:09 PM - 1 Like   #18
Site Supporter
Site Supporter




Join Date: Jun 2008
Location: Idaho
Photos: Gallery
Posts: 1,089
If you had an ideal sensor, no light would translate to 0000 0000 0000 00 with a 14 bit D to A conversion, but in the real world, there is no zero output from a sensor due to noise (except for the occasional random zero value in that noise). At the lowest ISO value for a given camera (let's say that's 100), all 14 bits have worthwhile information and the dynamic range is theoretically 16384:1 or about 84 dB.

If the lens is stopped down 2 f-stops from the previous exposure, the sensor receives one-fourth the light (all pixels) and the upper 2 bits of the D to A conversion revert to 0s. To compensate, the ISO is changed to 400 (which amplifies or rescales the output of the sensor prior to D to A conversion) and now the two uppermost bits regain information. We again have 14 bits of data, however, the noise which previously didn't affect any of those 14 bits, is amplified by 4 and now affects the lowest two bits making them somewhat random (we are considering an ideal system). If those two lower bits are thrown away because they really aren't useful, we have a 12 bit image with a dynamic range of 4096:1 or about 72 dB. Those two bits can still be useful if they are time averaged since that will average out some noise, but that has to be done in camera and before the image is captured. The shorter the shutter speed, the fewer samples to average so there's a conundrum to be considered.

Bottom line is dynamic range is decreased when ISO is raised. When that becomes apparent in a photo depends on a number of things and in many cases, a one (or even two) stop change in ISO may produce no visible changes, but it always has the potential for doing so, and the dynamic range of the system is decreasing.

Last edited by Bob 256; 11-04-2019 at 03:23 PM.
11-03-2019, 12:25 AM   #19
Veteran Member




Join Date: Feb 2015
Photos: Gallery | Albums
Posts: 7,990
QuoteOriginally posted by photoptimist Quote
Hmmm... I'm very skeptical of this "bit shifting" explanation. It would imply that the RAW values for an ISO 204800 image (shifted by 11 bits) only have 8 possible values in each channel which create a very spiky histogram.I would think that high ISO would be implemented either through analog amplification of the signal or by changing the full-scale voltage range on the ADC. Those methods would produce 14-bit outputs although the noise levels would be huge.
I don't know what method is used, either digital processing or anaIlog gain, but I know that underexposure and higher ISO don't produce the same effect in RAW files, my experience was that higher ISO deliver better images overall, with some loss of detail (while making sure the RAW processing parameters are the same in both cases, with only a difference on the exposure). I'd say, underexposing 1 or 2 stops could be a solution to get more details for low DR scenes, otherwise it'd be better to use a higher ISO.

---------- Post added 03-11-19 at 08:33 ----------

QuoteOriginally posted by photoptimist Quote
Hmmm... I'm very skeptical of this "bit shifting" explanation.
Note: if the gain was analog, how would high light compression curve be applied in case of high light protection enabed (ISO200 = -1 stop underexposure via double shutter speed or aperture)?


Last edited by biz-engineer; 11-03-2019 at 12:33 AM.
11-03-2019, 01:54 AM   #20
Pentaxian




Join Date: May 2015
Photos: Gallery
Posts: 609
QuoteOriginally posted by biz-engineer Quote
I don't know what method is used, either digital processing or anaIlog gain, but I know that underexposure and higher ISO don't produce the same effect in RAW files, my experience was that higher ISO deliver better images overall, with some loss of detail (while making sure the RAW processing parameters are the same in both cases, with only a difference on the exposure). I'd say, underexposing 1 or 2 stops could be a solution to get more details for low DR scenes, otherwise it'd be better to use a higher ISO.

---------- Post added 03-11-19 at 08:33 ----------


Note: if the gain was analog, how would high light compression curve be applied in case of high light protection enabed (ISO200 = -1 stop underexposure via double shutter speed or aperture)?
I seem to remember reading that analog amplification was used up to 1600 ISO and digital gain beyond that for K-5 era cameras.

11-03-2019, 07:55 AM   #21
Pentaxian
photoptimist's Avatar

Join Date: Jul 2016
Photos: Albums
Posts: 4,496
QuoteOriginally posted by biz-engineer Quote
I don't know what method is used, either digital processing or anaIlog gain, but I know that underexposure and higher ISO don't produce the same effect in RAW files, my experience was that higher ISO deliver better images overall, with some loss of detail (while making sure the RAW processing parameters are the same in both cases, with only a difference on the exposure). I'd say, underexposing 1 or 2 stops could be a solution to get more details for low DR scenes, otherwise it'd be better to use a higher ISO.
Interesting. I thought that all the tests of ISO invariance that compared ISO 100, 200, 400, 800, & 1600 shots to a shot taken at ISO 100 with 0, 1, 2, 3, & 4 boost in post showed no differences in noise or lost details.

QuoteOriginally posted by biz-engineer Quote
Note: if the gain was analog, how would high light compression curve be applied in case of high light protection enabed (ISO200 = -1 stop underexposure via double shutter speed or aperture)?
My understanding is that ISO 200 with highlight protection enabled runs the exposure metering logic at ISO 200 but runs the sensor at ISO 100. Thus, the RAW output of the sensor is 1 stop underexposed. It then uses a tone curve change to bring up the shadows and mid-tones by one stop but rolls the tone curve in the highlights to keep them.
11-03-2019, 10:12 AM   #22
Veteran Member




Join Date: Feb 2015
Photos: Gallery | Albums
Posts: 7,990
QuoteOriginally posted by photoptimist Quote
Interesting. I thought that all the tests of ISO invariance that compared ISO 100, 200, 400, 800, & 1600 shots to a shot taken at ISO 100 with 0, 1, 2, 3, & 4 boost in post showed no differences in noise or lost details.
The difference is visible when comparing ISO100 (underexposed 4 stops) to ISO1600 (0 ev exp. comp.), and making sure the raw conversion software has all NR parameters set to 0. The difference won't be visible is only underexposing 1 or 2 stops.

QuoteOriginally posted by photoptimist Quote
My understanding is that ISO 200 with highlight protection enabled runs the exposure metering logic at ISO 200 but runs the sensor at ISO 100. Thus, the RAW output of the sensor is 1 stop underexposed. It then uses a tone curve change to bring up the shadows and mid-tones by one stop but rolls the tone curve in the highlights to keep them.
I don't see how we can know if the sensor is run at ISO100 or ISO200. The camera indicates ISO200, there is no way to know how it done internally, except asking Ricoh how they implement ISO in their cameras. The camera used get the image output, however it is implemented, what to be aware of is that underexposure isn't the same as increasing the ISO setting.
11-03-2019, 11:38 PM   #23
Pentaxian




Join Date: May 2009
Location: Somewhere over the rainbow
Photos: Gallery
Posts: 960
QuoteOriginally posted by biz-engineer Quote
I don't see how we can know if the sensor is run at ISO100 or ISO200. The camera indicates ISO200, there is no way to know how it done internally, except asking Ricoh how they implement ISO in their cameras. The camera used get the image output, however it is implemented, what to be aware of is that underexposure isn't the same as increasing the ISO setting.
We should be able to look at the raw data and see if it was used at iso 100 ( this is from memory ) what I seen was that there was a different baseline exposure stored in the raw data file telling the converter how and what tone curve to use for that particular exposure, this was true with HLP also what we would see is that there was a different BLE used to communicate to the raw converter what tone curve to use for that HLP

Normally Pentax stores a baseline exposure of -0.5

if we were to look at the raw data and take a shot of something grey with highlight protection on then see how far it was from clipping. Then take the same shot using iso 100 and the HLP off and see how it also falls from clipping then we could assume how it works.

---------- Post added 11-04-2019 at 01:00 AM ----------

Here I have taken 2 photos using the same exposure one with highlight protection on and one with it off at base iso


https://photos.smugmug.com/Temp/Temp/i-mJSFVxH/0/c4a23e23/O/HLP.jpg
Here is the one using HLP you can see that it is using a BLE of 0.5 telling the converter what tone curve to use.


https://photos.smugmug.com/Temp/Temp/i-ZwWWSkM/0/7b3e444c/O/nonHLP.jpg
Here it is using the same exposure but at iso 100, First you will see that it is now telling the converter to use a BLE of -0.5

We can also look at the raw histogram and this is telling us that they are using the sensor much the same. If it was using the sensor differently we should see a very different raw histogram but as we see it here they are almost identical. this tell me that they are using the sensor with HLP on as they would use the sensor at iso 100 and just telling the meter to use -1ev and apply a different tonal curve to the raw file

If a person wanted to mimic HLP you could set your camera to -1ev shoot it at base iso 100 and adjust the BLE in the raw file from -0.5 to +0.5 or just use +1ev in your raw converter


Last edited by Ian Stuart Forsyth; 11-04-2019 at 12:08 AM.
11-04-2019, 11:30 PM   #24
Veteran Member




Join Date: Feb 2015
Photos: Gallery | Albums
Posts: 7,990
QuoteOriginally posted by Ian Stuart Forsyth Quote
If a person wanted to mimic HLP you could set your camera to -1ev shoot it at base iso 100 and adjust the BLE in the raw file from -0.5 to +0.5 or just use +1ev in your raw converter
Sure. The question remains is ISO100 -1ev the same as ISO200? I might be very hard to see with only 1 stop difference between the two exposures. Applying 5 stops shows that the raw data is not a direct copy of sensor data unless there is an analog gain applied before A/D conversion on the sensor itself.
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!
analog, aperture, author, blog, blur, camera, converter, curve, data, depth, explanation, exposure, gain, increase, iso, light, movement, photography, sensitivity, sensor, technique, tone, underexposure, values
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Keep the overview on the increasing tethering options for Pentax cameras angerdan Pentax DSLR Discussion 13 11-24-2019 01:53 PM
Increasing weight of Pentax lenses OhioAngie Pentax SLR Lens Discussion 19 08-28-2018 10:35 AM
Nature Finally the activity is increasing. charliezap Post Your Photos! 11 05-17-2014 06:18 PM
thinking out loud about my k30 image settings and increasing contrast Iksobarg Troubleshooting and Beginner Help 13 12-11-2013 05:48 PM
Henrys in Canada increasing Q profile Zaphodmonster Pentax Q 2 03-29-2013 05:02 PM



All times are GMT -7. The time now is 11:41 PM. | 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