Dynamic Range-Can't they or Won't they?

davesrose wrote in post #17746214
I think you may be confusing jpeg file compression with "tone-mapping" "compression" if you're insisting both systems have "one algorithm". My posts were about 14bpc to 8bpc: there's not just one algorithm involved.

To convert raw data into JPG data, there are multiple different software packages you can use. These libraries are used by both the firmware and DPP, thus the same algorithm is used by both the camera and DPP. It is very quite simple if you understand how software is written, how reusable libraries are created and shared and linked in, etc.

I only brought up JPG compression because the camera has a set compression level it uses for large fine JPG, and I had a different compression level in DPP when I converted, and thus caused a smaller file with less quality, thus wiping out some of the noise that I saw out of camera. That is what caused the 2 JPG files to be different. Nothing more, nothing less. I have no idea where you make these strange mental leaps. We are done, I will have to put us in time out now, sorry about that.

I have moved away from JPG at this point to compare files anyways, I am now looking at the raw content directly, to make comparisons. Sorry I brought up what I was doing to make the comparisons, it started an entirely new chain of thought with somebody. I am trying to "see" the floor noise when I take a very fast black frame. I will just try to adjust exposure in the raw, along with curves to bring out that noise floor.

sploo wrote in post #17746225
I think that, once again, you may be reading something into someone else's post that was neither written nor implied.

So we're now injecting (out of the blue as normal) JPEG file compression. It's like Topics Tourette's in here.

Squirrel!!.....

TeamSpeed wrote in post #17746245
Squirrel!!.....

Indeed. Indeed.

(Sadly, it's not really any more random than some of the interjections we've had in this thread)

sploo wrote in post #17746252
Indeed. Indeed.

(Sadly, it's not really any more random than some of the interjections we've had in this thread)

Sorry for mine, but as I design solutions and write software as a career, including image manipulation using GNU or Imagemagick packages linked into our solutions, when those topics come up about softare where somebody is clearly ignorant of what is going on inside those black boxes, I have a hard time biting my tongue. I have to go back to what I was taught as a kid "when I am right, I don't have to defend it, and if I am wrong or ignorant, I have no right to defend it".

Now back to looking at my raws to see what the noise floor looks like when comparing the 7D2 to the 5D3... I am open to other better programs for looking at the raw data to see the amount of usable DR I might have for the 2 bodies I have currently.

Look guys, there does just seem to be differences in terminology/ confusion about tonal range again. Let me try to explain tone mapping in relation to sensor DR (and how RAW is HDR). With sensor DR, the contrast is saturation point (white) vs "acceptable" noise floor (black). These tests of the black point of RAW are just showing when the last "bit" of information is still a "clean" black (the noise floor). Let's say the Sony sensor gets 14 "stops" of saturation vs Canon's 12 "stops". How does that relate in logarithmic tonal scale? Like this:


2,4,8,16,32,64,128,256​,512,1024,2048,4096,81​92,16384

now the Canon sensor with full exposure would look like this instead:

0,0,8,16,32,64,128,256​,512,1024,2048,4096,81​92,16384

That's what the last 2 (blocked up/clipped/noisy/etc) "stops" look like in a tonal scale ....and why I think these tests of sensors are really asinine!

When the camera "compresses" or "tone-maps" tonal data to 8bpc, it is picking the white and black point, and then adjusting gamma curves in both "shadows" and "highlights" to bring 14bpc data down to 8bpc.

TeamSpeed wrote in post #17745317
... All the talk about needing more DR seemingly has everything to do with workflow more than anything. ...

So in the end, it seems all the DR discussions are more about reduced workflow and faster processing times of images reflecting a high DR scene, correct?

I believe that some arguing for more DR have made that point exactly in this thread. It is about workflow, and it is about being able to capture for the highlights and preserve all that low end and pull it out of the noise.

Interesting correlation about desire for higher DR, Vs. the actual current out put with the Audio world.

Consider that dynamic range in audio is VERY similar to what we are now discussing here.

One big difference is the upper limit is not as much a hard ceiling in Audio as it is in imaging sensors where "Blown out" is always the top end. In audio we can clip the top end, but avoiding it just requires more power, and thus more volume. In Imaging more power will not stop 100% white from being 100% white.

But other than that we still have a a very similar situation where "better" always meant more dynamic range, more information across a broader range of volume, and more detail at the lowest volume getting pulled out of the noise floor.

Another interesting correlation, and the one that may be part of this very discussion, is that despite more DR and volume headroom being the Holy Grail of audio reproduction for decades, with amplifiers constantly being able to be quieter and more powerful with faster response to transient signal, source material having less noise (CD Vs. Vinyl etc)

.. despite all this advance in the ability to play back more dynamic range,..

..the modern "out put media" ( ie: todays music and more specifically, today's studio mixes,) have all but squashed any improvement we had in DR down to next to nothing.

In the mix, all the signals get "compressed", "normalized", "limited" and otherwise have the DR, the difference between the lowest audible volume of music one can hear over noise, and the highest peaks of volume, crushed to death leaving us with a final product with far less DR than any Simon and Garfunkel vinyl of the 1970's had, despite the fact that our equipment is capable of reproducing a much higher DR than any 1970's payback equipment was capable of.

Hmmm..

And yet despite the trend towards killing the DR in audio, there have always been hold outs. Classical recordings , the good ones, are still trying to maintain as much true to life DR as possible, for example.

P.S. when you look up audio amplifier specs, the measurement for DR is called "Signal To Noise"
Output dynamic range is the range, usually given in dB, between the smallest and largest useful output levels. The lowest useful level is limited by output noise, while the largest is limited most often by distortion from lack of power. The ratio of these two is quoted as the amplifier dynamic range. More precisely, if S = maximal allowed signal power and N = noise power, the dynamic range DR is DR = (S + N ) /N

This thread for instance has a very LOW Signal to Noise Ratio

Tareq wrote in post #17746041
Anyone here wants me to upload a RAW file of one nightshot scene that can show some DR from my medium format? just as fun here, many examples and i don't see the point of this thread and so many posts with each trying to defend his point, honestly speaking i lost to understand about DR and RAW from that so many posts here replying each other.

Should i upload the file and you play with it and see if you can get about DR from it? It is a RAW file from medium format camera, if it is useless to upload or not interested in then i understand and i won't upload.

Tareq, you might simply go online to download RawDigger (it is free), do the analysis on your medium format RAW file, then post the resulting screenshot as a JPG, like posting an ordinary photo.

Or you can send me or TeamSpeed a PM with a link to YouSendIt or other file sharing site which allows you to send a CR2, and one of us can put it in RawDigger analysis.

CyberDyneSystems wrote in post #17746403
I believe that some arguing for more DR have made that point exactly in this thread. It is about workflow, and it is about being able to capture for the highlights and preserve all that low end and pull it out of the noise.

Yes indeed! Sorry for potential offenses in what "dynamic range" means for all "systems" (be it the scene, sensor, RAW, and jpeg). From a tonal level, the "DR" between Sony vs Canon is very slim (and why it's complex "pinhead" minutiae ). Since any file over 8bpc is considered "HDR" it needs to be "tone-mapped" (whether through automatic or manual "algorithms") : the differences in sensors just demand a slightly different workflow. To date, a single exposure gets you 14 stops of data. If the Canon doesn't reach full exposure, it will exhibit noise in the "shadows" more quickly then Sony. All RAW tests confirm this. So is this thread about 14 stops of light being enough, or beating a dead horse about Canon not fully getting those last 6 shades of tone in their ADC? As photographers, I think there are many scenes that do demand going up to even 30 stops of light....and traditional camera sensors would have to take many decades in order to support that kind of light exposure. What's nice about digital is that the RAW has a higher range then your output (to "recover" tones) or you can extend DR by merging multiple exposures. At the end of the game, photographers still have situations where they need to decide whether under or over exposure is acceptable.

davesrose wrote in post #17746307
How does that relate in logarithmic tonal scale? Like this:


2,4,8,16,32,64,128,256​,512,1024,2048,4096,81​92,16384

now the Canon sensor with full exposure would look like this instead:

0,0,8,16,32,64,128,256​,512,1024,2048,4096,81​92,16384

That's what the last 2 (blocked up/clipped/noisy/etc) "stops" look like in a tonal scale ....and why I think these tests of sensors are really asinine! .

but, Dave, the camera RAW file does NOT zero out the bottom two zones in the Canon exposure... the signal is still there, but it is mixed in with circuit noise!

Wilt wrote in post #17746439
but, Dave, the camera RAW file does NOT zero out the bottom two zones in the Canon exposure... the signal is still there, but it is mixed in with circuit noise!

That's why I said the 0 could by "noisy" This gets complicated...sorry for the flame war, but I am trying to show the relationship of dynamic range as it relates to "tone-mapping"

CyberDyneSystems wrote in post #17746417
..

This thread for instance has a very LOW Signal to Noise Ratio

Perhaps I should apply to the thread title as per requests for warning..

Wilt wrote in post #17746423
Tareq, you might simply go online to download RawDigger (it is free), do the analysis on your medium format RAW file, then post the resulting screenshot as a JPG, like posting an ordinary photo.

Or you can send me or TeamSpeed a PM with a link to YouSendIt or other file sharing site which allows you to send a CR2, and one of us can put it in RawDigger analysis.

I see

I think it is not necessary, maybe it is irrelevant, thanks for your post.

It didn't in the beginning nor was it the original intention of the thread. Thanks for diminishing its value even further.

WilsonFlyer wrote in post #17746454
It didn't in the beginning nor was it the original intention of the thread. Thanks for diminishing its value even further.

It was a joke, sorry it rubbed you the wrong way.
I often try to lighten things with humor, and often shouldn't.

Be aware that personally I am finding a lot of interesting info in this thread. The Joke was meant to poke fun at the haters.