cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

6D MKII a disappointment???

skyking
Contributor

I did order the 6D MKII from B&H - arrives Monday. This is an upgrade from my 6D. I am a little concerned about the recent tests showing, at lower ISO's, poorer dynamic range. Apparently the 6D MKII showed very good dynamic range at higher ISO's. Apparently the 80D had better dynamic range at lower ISO"s then the 6D MKII. I'll know a lot better when I get the camera but is that is the truth its a little disappointing for what I'm paying for this camera.

 

Any comments??

 

James

108 REPLIES 108

@TTMartin, The 6DII has 9 stops of dynamic range. Even by your own admissions, that is worse than the technology that Ansel Adams was using in the 1940s.

I am glad that you think that 1940s technology is the benchmark for dynamic range. Some of us would disagree with that...

Honestly I am done talking to you. Feel free to keep on trolling my comments. I have better things to do...

Now, where is that coffee?



@TTMartin wrote:

@KlausEnrique wrote:

Kodak VISION3 has 13 stops of dynamic range. That is film. Look it up...


Ansel Adams didn't use Kodak VISION3 film.

 

And according to Kodak's Vision3 film brochure 'SENSITOMETRIC CURVES “0” on the x-axis represents normal exposure of an 18-percent gray card in the red, green, and blue layers of this film. A white card is 2 1/3 stops higher than normal exposure, and there are at least 3 1/2 stops above that for capturing specular highlight detail. A 3-percent black card is 2 2/3 stops below normal exposure. There are at least 2 1/2 stops of latitude below that for capturing shadow detail.'

 

2 1/3 + 3 1/2 + 2 2/3 + 2 1/2 = 11 stops of dynamic range

 

edit: someone (marketing, internet blogger, ???) must have added pure black and pure white to the 11 actual stops of dynamic range to come up with a 13 stop number. But, as Ansel Adams make very clear in his book 'The Negative', The Zone System, Dynamic Range and Textuaral Ranges that is not the correct definition of 'Dynamic Range'.

 


 


@KlausEnrique wrote:
@TTMartin, The 6DII has 9 stops of dynamic range. Even by your own admissions, that is worse than the technology that Ansel Adams was using in the 1940s.




You obviously haven't read what I wrote. Ansel Adams Zone System is based on 9 stops of dynamic range. You might want to read his book 'The Negative' because he addresses dynamic range outside of that represented by the Zone System of having 'no significance in practical work'.

 

Also, your 9 stop figure comes from a Nikon blogger and is not backed up by other sources. Just because he puts a headline that says he is a 'trusted source for independent sensor data' doesn't make it true. In my opinion those kinds of self proclamations are red flags that it probably means just the opposite. 

 

Even DXOMark that doesn't take into account the black point differences in Canon and other RAW files (*1), nor the baked in Noise Reduction Nikon uses (*2), says the 6D Mk II has 11.9 stops of dynamic range.

dxo 6d vs 6d mk ii.PNG

*1 - Earlier in this thread I addressed how to properly post process to negate the black point differences.

 

*2 - As for documentation on Nikon baking in noise reduction just Google 'Nikon Astrophotography Mode 3 Work Around'. It should be noted that the Nikon D810A has a minimum of ISO 200 presumably because it lacks the pre-RAW processing used to create ISO 64 and ISO 100 on the D810.

edit: This post prior to the release of the 80D by charlyw at FM points out that the math just doesn't add up for the dynamic range being reported on some tests and that processing done to create those artificially high test results comes at a cost of color fidelity.

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

'I am not upgrading, until I can have 14 stops of DR from my 70D or 5d3.'


Since that is physically impossible for a current 14-bit A/D conversion sensor even if the sensor had a 100% quantum efficiency, a ludicrously high fill factor and an absolutely perfect A/D converter - since shot noise already is the dominant factor that limits the dynamic range such a sensor can capture to below 13 stops... You could fool indiscriminate testers by applying some processing to the RAW before it's written or use a non linear A/D conversion curve artificially inflating the DR measurements but IMHO that will mean that other areas like color fidelity will suffer badly - as witnessed by the below par color fidelity shown by both Sony and Nikon cameras in recent years:

 
For example (delta-C should be small, which means that there is little color error, all results taken from imaging-resource reviews): 


Sony A7RII: Average delta-C 5.45, 
Sony A77II: Average delta-C 5.59, 

Nikon D750: Average delta-C 6.19,
Nikon D4s: Average delta-C 7.01,
Nikon D810: Average delta-C 6.71,
Nikon D7200: Average delta-C 6.34,

Canon 5DsR: Average delta-C 3.90,
Canon 7DII: Average delta-C 3.45,
Canon 5DIII: Average delta-C 3.92

But you could pick any of the cameras from each manufacturer and there are clear groupings visible: Canon leading the pack with a big margin ahead of the Sony which in turn has an equally large margin over the Nikon cameras...

Thus I hope that Canon didn't cave in into the demands of those DR "trump card" freaks if that meant that any of the other areas their sensors do exceedingly well suffered...


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

 

Yes, the Canon 6D Mk II and the 7D Mk II have RAWer RAW files than the 5D Mk IV and the 80D that doesn't mean with proper processing they have less usable dynamic range. All it means is the 5D Mk IV and the 80D score better on 'the test'. It will be interesting to see how the quantum efficiencies of the sensors on those cameras compare once that data is available. 



@KlausEnrique wrote:

It is not funny. It is a real Canon deficeincy and the infrared example was not because "I had to" but because it was the latest example I encoutered of Canon's dynamic range being subpar.

 

Like I said, I don't really care if any of you is convinced that this is an issue, but people who look at forums should be aware that it exists. Attached is a straight 30 sec exposure (ISO 100 with a 5DII) of a pretty dark nightclub. Again, the highlights are all blown out, and the shadows all show heavy banding.

 

You will have to excuse me if I don't reply any more to your "I don't think this is an issue" type comments.

 

PS Ansel Adams shot film, which has more dynamic range than the best that Canon can offer, i.e. the 5DIV. Just FYI...

 

 

Banding.JPG



The banding (such as it is) in this photo is not confined to the shadows; it shows pretty much everywhere. So I strongly suspect that it's actually due to the impact of a lot of noise (from using an unnecessarily long exposure at an absurdly low ISO setting) on the conversion to JPEG. Did you use long-exposure noise reduction? (Does the 5D2 even have it?) And do the "bands" show in the RAW image or just in the converted JPEG?

 

And why are we looking at a picture from a 5D2 anyway? Isn't this discussion about a far more recent camera, the 6D2?

Bob
Philadelphia, Pennsylvania USA


@RobertTheFat wrote:

 

And why are we looking at a picture from a 5D2 anyway?


Maybe because he's trying to mislead us?

Want to see how far behind Canon's sensors are? Compare the 5D Mk IV with it's fixed black point RAW file, and the D810A which since it is designed for Astrophotography can't bake in noise reduction. Using his own 'trusted source'.

photons 2.JPG

OMG, can't believe how far behind Canon's sensor tech is!!!

@Bob

The banding is confined to the shadows as the attached photo shows. 

 

"I strongly suspect that it's actually due to the impact of a lot of noise (from using an unnecessarily long exposure at an absurdly low ISO setting)"

 

Wow, you really don't know what you are saying. The banding is due to underexposure, and you are saying that 30 seconds is an "unnecessarily long exposure". Let me say that again: you really think 30 seconds is an "unnecessarily long exposure"????? Have you ever studied photography???

 

I will admit that Canon can bring in weird artifacts in very long exposures. That starts at around 30 minutes, not 30 seconds!

 

As shown in the previous image, the highlights are already blown out. Can you please explain in your great wisdom how I would be able to expose the image in a single shot so that the highlights are not blown out and at the same time have the shadows not show this crazy ammount of banding? Really, please explain that to me, because the only way I know on how to do this is to have a camera with a decent amount of dynamic range. And given that at ISO 100 there was not enough DR, pushing the ISO up does not really increase the dynamic range and it would only increase the amount of blown out highlights!

 

And since you complained about my 5DS example and my 5DII example, I have attached an image from DPReview that shows the 6D, the 6DII and your beloved 5DIV with very obvious banding...

 

I am glad that the 6DII covers ALL your hobby needs, but please do not pretend to be professionals...

 

I feel sorry for all the people who may have read your posts and decided to follow your mistaken advice. 

 

 


@RobertTheFat wrote:

The banding (such as it is) in this photo is not confined to the shadows; it shows pretty much everywhere. So I strongly suspect that it's actually due to the impact of a lot of noise (from using an unnecessarily long exposure at an absurdly low ISO setting) on the conversion to JPEG. Did you use long-exposure noise reduction? (Does the 5D2 even have it?) And do the "bands" show in the RAW image or just in the converted JPEG?

 

And why are we looking at a picture from a 5D2 anyway? Isn't this discussion about a far more recent camera, the 6D2?


No Banding.JPG
DPReview.JPG

 


@KlausEnrique wrote:

DPReview.JPGDPReview.JPG

 

I think you may have fallen for the deception of ISO-invariance.  Canon doesn't make ISO-invariant sensors.  Also, this is not a measure of dynamic range.

 

Dynamic range is measured at ISO 100 +0EV ... not +6EV.  BTW... at +6EV you lose a ton of dynamic range (that's just basic math but I'll explain a bit later).

 

I think Ricoh (Pentax is just a marketing brand, not a company) uses the Sony sensors and Sony... cooks their RAWs in the sensor.  Go look up the Sony "star eater" problem.

 

I first learned about this from one of the astrophotographers in my club.  Most astrophotographers who use DSLRs are using Canon DSLRs (the overwhelming majority -- not just slight edge -- and for a lot of reasons that would deserve its own thread.)  Anyway, one of the astrophotographers in the group started using Nikon (who also uses Sony sensors in most cameras) and members started noticing that the images just looked a little TOO clean.  Stars that we knew were really stars... were completely missing.

 

The camera was cooking the stars out of the image because it thought these faint stars were actually noise.  So much for an honest RAW image.

 

 

So meanwhile, from what I can tell by doing a bit of searching... Ricoh is also using the Sony sensors and your'e seeing the results of "cooked RAW" images... not true uncooked RAW images.

 

DxO seem to be completely fooled by the gimmickery.  

 

Meanwhile back to Canon (at this point I should mention... unlike Canon, Nikon and Ricoh don't make camera sensors anymore).  Canon use both "upstream" and "downstream" amplication to boost ISO.

 

"upstream" and "downstream" are referring to whether the amplification occurs before or after the analog-to-digital conversion of the image.  "upstream" is an analog amplification applied to the signal.  "downstream" is a digital multipliciation of the siganl value.

 

The camera sensor is measuring analog data (light is not digital).  Canon has the ability to boost the gain on this ... to a point, then convert to digital, then apply digital multiplication to boost it further (if needed).   Sony sensors don't do "upstream" amplification (I'm not aware of any current Sony sensor that does this) but Canon sensors do.

 

The trade-off here (and if you are worried about dynamic range it's time for you to pay attention) is that every stop of ISO boost by "downstream" amplification results in a noticeable loss in dynamic range.  Canon applies "upstream" amplification which does not result in a direct loss of DR.  This allows them to boost ISO (to several stops) while protecting the DR.  At some point there's a limit to how effective upstream gain will help and they have to convert to downstream gain (which is a digital manipulation and does result in a loss of DR... but they successfully fight that off.)

 

So let me short-circuit your brain for a moment.... have a look at this:

 

Screen Shot 2017-10-31 at 9.57.20 AM.png

 

 

The image on the left is just a tad cleaner than the image on the right.  Did I photoshop this?  Why should such an image exist?  What's going on here?

 

What's going on is...  nothing.   This is an honest comparison of what happens when you dial a Canon 5D IV to ISO 6400 and then dial a Pentax K-1 to ISO 6400 and don't shoot at ISO 100 and do a +EV boost in software on your computer.  

 

But that's not all... the ISO invariant sensor would have been losing DR for each stop of ISO gain... but the Canon sensor woud resist the loss of DR until the boost exceeds the amount of upstream gain they apply.   If you boost an ISO-invariant camera by +6EV then you lose 6 stops of DR.  A Canon camera such as a 5D III or 5D IV would have lost about 1.5 stops... maybe just 1 stop.

 

PopPhoto (tragically no more) used to show the DR for every stop of ISO (DxO only shows it at ISO 100) and it became evident that as soon as you start boosting ISO, the ISO-invariant sensors immediately start losing DR.  The Canon cameras always trade-places on the DR rank when you do that ... they have more DR at high ISO (because they apply "upstream" gain).

  

I mentioned we'd get to the math... well it's time to talk math.

 

In binary, if I have a 1 bit processor, I can store only two values... 0 or 1... (think of this as not just off and on... but also as black and white). 

 

If I have a 2 bit processor, I can store only 4 values... 0, 1, 2, or 3 (represented by 00, 01, 10, and 11).  You can think of this is as "black", "dark gray", "light gray", and "white" if you were using these values to represent tonality.

 

(Side note:  If you're not familiar with counting in binary... each time you add another bit to the register you double the number of possible values that can be stored.  With 3 bits you can store 8 possible values (0-7).   With 4 bits you can store 16 possible values (0-15 but when written out they use 0-9 then use the letters 'a' through 'f' to represent 10 through 15)).

 

Each time you increase ISO by a full stop (assuming it really is a FULL stop and not some gimmickery again) you have to double the value of light.  You can't double the value of light on a 1 bit processor... there's only 1 bit.  But with a 2 bit processor if you had a value of binary 01 (decimal 1)  you could double it to make it binary 10 (decimal 2).  So 1 becomes 2 because you doubled it.  That makes sense.   Trouble happens if you had a binary 10 already (decimal 2) because a 2 bit processor can't store the result... 2 doubled becomes 4 and a 2 bit processor can only store values from 0 to 3... trying to store a 4 results in an overflow.  If this was presenting tonality in an image... the image would be clipped, over-exposed, blown out, etc.

 

These camera processors have 14 bit sensors.  You can store values from 0 to 65535.  So if some pixel had a value of, say... 30,000 and I needed to double that... then it would double to become 60,000.    That's fine because 60,000 is less than 65535.  But what if we had a pixel that was 40,000.  If we double that we get 80,000 and now we're in trouble because we can't store any values greater than 65535.  This results in clipping.

 

If a camera has 10 stops of native dynamic range and it's an ISO-invariant sensor (which means it really only ever shoots at ISO 100 and uses math to manipulate the output) then when we go to +1 EV, we just dropped the camera's DR down to 9 stops.   Every stop we boost ISO is traded for a loss of 1 stop of DR.

 

But wait... lots of ISO invariant cameras claim to have more DR (well... they do if you believe DxO -- I am not a beliver in DxO because I can do math).

 

The trick is to compress the data.  So instead of boosting ISO straight across the board (e.g. think of Ansel Adams' Zone System... but lets reduce it to just 5 zones).  We can't really double the top zone (zone 5) because we'd get clipping... so we pretty much do nothing.  Then we take zone 4 and we only try to fractionally boost it... zone 3 gets a slightly larger fraction of a boost and so on... zone 1 (the darkest zone) gets a full doubling.  

 

Really we've just "compressed" the DR... so  that 5 stops of true DR are squeezed to fit into 4 stops of DR.  But that doesn't avoid the fact that it's really just 4 stops of DR.  You could do this in Lightroom by adjusting the black point (instead of adjusting the exposure value).

 

Lots of cameras have features that try to do this for JPEG photos ... Canon calls it "highlight tone priority" (squeeze the DR to avoid blowing the highlights... but still bring up the blacks).

 

By the way... these tricks apparently fool DxO.  They don't fool me because... well, I can do math.  

 

A Canon 5D III (and I haven't tested my 5D IV) is going to boost about 5 stops of ISO before it really starts losing any significant DR ... BECAUSE... Canon does "upstream" amplification.  This causes a boost to ISO at the analog level so that it doesn't have to start the digital multiplication.

 

This is the stuff that DxO just doesn't seem to understand.

 

I ran into this on another forum where someone was trying to explain how much better than Nikon D810 was by showing ISO-invariance tests (completely unaware that this kills DR).  I pulled the following samples from DPReview tests.

 

Screen Shot 2017-05-21 at 2.02.45 PM.png

 Screen Shot 2017-05-22 at 1.30.16 PM.png

 

His response was to scoff at these because one is ISO 25600 and the other is ISO 12800 and he felt that was unrealistic and "nobody shoots at those ISOs".

 

Fair enough... let's pull data from something with less ISO gain... I pulled another test comparison and this one at ISO 3200.

 

Screen Shot 2017-05-22 at 1.29.47 PM.png

 

These are just honest results that you get when you operate these cameras at these ISO values and pull the images straight out the camera and don't manipulate them and don't use ISO 100 images and try to boost it by many levels of EV (I don't know photographer who does that... so why do they keep showing those tests)?

 

Lastly... did you happen to notice that the last batch of images ... the Canon camera was actually a 5D III (the 5D IV is better... but he was ranting on about the 5D III).

 

This is why I don't read DxO reviews.

This is why i don't look at ISO-invariance tests.

 

Tim Campbell
5D III, 5D IV, 60Da

@TCampbell

 

You are correct in saying that dynamic range is measured at +0EV. But then, if you do have dynamic range, it won't look like the cr*p that the 6D II produces...

 

It is funny to hear you guys say that Nikon and Sony "cook" their RAW files, but Canon doesn't. That is just such nonsense. They ALL "cook" their RAW files. I am sure you know what a Bayer Filter array looks like and what it does, right? Do your RAW images look like the image below? They don't, and that is because your camera interprets the data and extrapolates the information from the pixels around it (i.e what you call "cooking"). So please stop saying that Nikon cooks the raw files but Canon doesn't.

 

bayer-simulation2.png

 

Now, since you know so much, please explain to me, where does the banding, that one see in underexposed Canon images, come from? I mean, if it is not "cooked", then where does it come from??? Are you even aware that banding does occur in Canon cameras???

 

Now, you say that Pentax is a marketing brand, but that brand has pixel shift technology that gives the best resolution of ANY full frame camera...

 

I don't belong to any astro photography clubs, but I have no reason to doubt you. Nikon, Sony and Pentax do produce cleaner RAW files, and if you just have one pixel of data coming in, it can be a star, but it can also be a hot pixel. I am not surprised that some get cleaned out... So? How many people worry about a star disappearing from their night sky, and how many people push the shadows only to realize that they are useless...

 

I keep on reading your "true uncooked images" references, and it is just SO MISTAKEN; so untrue... And Canon's banding is the most obvious example of how it also cooks its files, just really, really badly...

 

OK, your ISO 6400 Canon 5DIV vs PentaxK1 comparison is accurate, but honestly I don't see how this moves the discussion at all. Canon is not ISO invariant (which is BAD), and other brands are way more invariant than Canon (which is GOOD!)

 

Now you presented that as if were going to be a huge surprise to me, but did you see the "Photons to photos" graph that I posted earlier on. That graph basically says the same thing... Just more succinctly. The K1 has more dynamic range until about ISO 320 - 500. From ISO 500 onwards the 5DIV has more dynamic range. But between the K1 and the 6DII, the K1 has better dynamic range all the way until ISO 2000. And again, if you think ISO 6400 is where one should be shooting photos, then good luck to you!

 

You mentioned the "upstream" amplification of the signal by Canon. Well, guess what, the banding that half of you pretend does not exist is directly the result of that! Canon keeps the electric analog input much longer than all the other camera manufacturers and that is exactly why some rows of pixels get hotter than others, thus producing the banding!!!

 

PS Your comment that at ISO 6400 Pentax has lost 6EVs of dynamic range is true but misleading. By that point the Canon 6DII has lost 3EVs, but that is not because Canon is better. The Canon 6DII started with only 9EVs of dynamic range, whereas the Pentax starts with 12. And again, shooting at ISO 6400 is terrible... Your images are going to look like cr*p no matter what!

 

Comparison.JPG

 

Thank you for your lengthy explanation about the "upstream" amplification and what it does to dynamic range at very high ISOs. You are right! Yes, you are completely right in ALL of that. What you don't seem to understand is what that does at the low end of the ISO range. Because Canon keeps the analog electric input longer it creates the banding artifact that I keep on bringing up, which all of you guys are ignoring! Seriously, can you please give me an equally lengthy reply on why banding happens?

 

Again, if you think ISO 6,400 is the place to take pictures at, then heck the 6DII is pretty good, and at ISO 10,000 it is as good as the 5DIV... What a bargain! Those images are useless. Did you see the image that you posted? Can you read the text on those images??? 

 

Anyhow, I am tired of all of you guys... Good luck shooting your photos at ISO 20,000,000. I just hope that some young kid reading this thread does not fall for all your BS! The 6DII is a TERRIBLE camera for this day and age! Well, except if you shoot at ISO 10,000 or higher!

 


@TCampbell wrote:

@KlausEnrique wrote:

DPReview.JPGDPReview.JPG

 

I think you may have fallen for the deception of ISO-invariance.  Canon doesn't make ISO-invariant sensors.  Also, this is not a measure of dynamic range.

 

Dynamic range is measured at ISO 100 +0EV ... not +6EV.  BTW... at +6EV you lose a ton of dynamic range (that's just basic math but I'll explain a bit later).

 

I think Ricoh (Pentax is just a marketing brand, not a company) uses the Sony sensors and Sony... cooks their RAWs in the sensor.  Go look up the Sony "star eater" problem.

 

I first learned about this from one of the astrophotographers in my club.  Most astrophotographers who use DSLRs are using Canon DSLRs (the overwhelming majority -- not just slight edge -- and for a lot of reasons that would deserve its own thread.)  Anyway, one of the astrophotographers in the group started using Nikon (who also uses Sony sensors in most cameras) and members started noticing that the images just looked a little TOO clean.  Stars that we knew were really stars... were completely missing.

 

The camera was cooking the stars out of the image because it thought these faint stars were actually noise.  So much for an honest RAW image.

 

 

So meanwhile, from what I can tell by doing a bit of searching... Ricoh is also using the Sony sensors and your'e seeing the results of "cooked RAW" images... not true uncooked RAW images.

 

DxO seem to be completely fooled by the gimmickery.  

 

Meanwhile back to Canon (at this point I should mention... unlike Canon, Nikon and Ricoh don't make camera sensors anymore).  Canon use both "upstream" and "downstream" amplication to boost ISO.

 

"upstream" and "downstream" are referring to whether the amplification occurs before or after the analog-to-digital conversion of the image.  "upstream" is an analog amplification applied to the signal.  "downstream" is a digital multipliciation of the siganl value.

 

The camera sensor is measuring analog data (light is not digital).  Canon has the ability to boost the gain on this ... to a point, then convert to digital, then apply digital multiplication to boost it further (if needed).   Sony sensors don't do "upstream" amplification (I'm not aware of any current Sony sensor that does this) but Canon sensors do.

 

The trade-off here (and if you are worried about dynamic range it's time for you to pay attention) is that every stop of ISO boost by "downstream" amplification results in a noticeable loss in dynamic range.  Canon applies "upstream" amplification which does not result in a direct loss of DR.  This allows them to boost ISO (to several stops) while protecting the DR.  At some point there's a limit to how effective upstream gain will help and they have to convert to downstream gain (which is a digital manipulation and does result in a loss of DR... but they successfully fight that off.)

 

So let me short-circuit your brain for a moment.... have a look at this:

 

Screen Shot 2017-10-31 at 9.57.20 AM.png

 

 

The image on the left is just a tad cleaner than the image on the right.  Did I photoshop this?  Why should such an image exist?  What's going on here?

 

What's going on is...  nothing.   This is an honest comparison of what happens when you dial a Canon 5D IV to ISO 6400 and then dial a Pentax K-1 to ISO 6400 and don't shoot at ISO 100 and do a +EV boost in software on your computer.  

 

But that's not all... the ISO invariant sensor would have been losing DR for each stop of ISO gain... but the Canon sensor woud resist the loss of DR until the boost exceeds the amount of upstream gain they apply.   If you boost an ISO-invariant camera by +6EV then you lose 6 stops of DR.  A Canon camera such as a 5D III or 5D IV would have lost about 1.5 stops... maybe just 1 stop.

 

PopPhoto (tragically no more) used to show the DR for every stop of ISO (DxO only shows it at ISO 100) and it became evident that as soon as you start boosting ISO, the ISO-invariant sensors immediately start losing DR.  The Canon cameras always trade-places on the DR rank when you do that ... they have more DR at high ISO (because they apply "upstream" gain).

  

I mentioned we'd get to the math... well it's time to talk math.

 

In binary, if I have a 1 bit processor, I can store only two values... 0 or 1... (think of this as not just off and on... but also as black and white). 

 

If I have a 2 bit processor, I can store only 4 values... 0, 1, 2, or 3 (represented by 00, 01, 10, and 11).  You can think of this is as "black", "dark gray", "light gray", and "white" if you were using these values to represent tonality.

 

(Side note:  If you're not familiar with counting in binary... each time you add another bit to the register you double the number of possible values that can be stored.  With 3 bits you can store 8 possible values (0-7).   With 4 bits you can store 16 possible values (0-15 but when written out they use 0-9 then use the letters 'a' through 'f' to represent 10 through 15)).

 

Each time you increase ISO by a full stop (assuming it really is a FULL stop and not some gimmickery again) you have to double the value of light.  You can't double the value of light on a 1 bit processor... there's only 1 bit.  But with a 2 bit processor if you had a value of binary 01 (decimal 1)  you could double it to make it binary 10 (decimal 2).  So 1 becomes 2 because you doubled it.  That makes sense.   Trouble happens if you had a binary 10 already (decimal 2) because a 2 bit processor can't store the result... 2 doubled becomes 4 and a 2 bit processor can only store values from 0 to 3... trying to store a 4 results in an overflow.  If this was presenting tonality in an image... the image would be clipped, over-exposed, blown out, etc.

 

These camera processors have 14 bit sensors.  You can store values from 0 to 65535.  So if some pixel had a value of, say... 30,000 and I needed to double that... then it would double to become 60,000.    That's fine because 60,000 is less than 65535.  But what if we had a pixel that was 40,000.  If we double that we get 80,000 and now we're in trouble because we can't store any values greater than 65535.  This results in clipping.

 

If a camera has 10 stops of native dynamic range and it's an ISO-invariant sensor (which means it really only ever shoots at ISO 100 and uses math to manipulate the output) then when we go to +1 EV, we just dropped the camera's DR down to 9 stops.   Every stop we boost ISO is traded for a loss of 1 stop of DR.

 

But wait... lots of ISO invariant cameras claim to have more DR (well... they do if you believe DxO -- I am not a beliver in DxO because I can do math).

 

The trick is to compress the data.  So instead of boosting ISO straight across the board (e.g. think of Ansel Adams' Zone System... but lets reduce it to just 5 zones).  We can't really double the top zone (zone 5) because we'd get clipping... so we pretty much do nothing.  Then we take zone 4 and we only try to fractionally boost it... zone 3 gets a slightly larger fraction of a boost and so on... zone 1 (the darkest zone) gets a full doubling.  

 

Really we've just "compressed" the DR... so  that 5 stops of true DR are squeezed to fit into 4 stops of DR.  But that doesn't avoid the fact that it's really just 4 stops of DR.  You could do this in Lightroom by adjusting the black point (instead of adjusting the exposure value).

 

Lots of cameras have features that try to do this for JPEG photos ... Canon calls it "highlight tone priority" (squeeze the DR to avoid blowing the highlights... but still bring up the blacks).

 

By the way... these tricks apparently fool DxO.  They don't fool me because... well, I can do math.  

 

A Canon 5D III (and I haven't tested my 5D IV) is going to boost about 5 stops of ISO before it really starts losing any significant DR ... BECAUSE... Canon does "upstream" amplification.  This causes a boost to ISO at the analog level so that it doesn't have to start the digital multiplication.

 

This is the stuff that DxO just doesn't seem to understand.

 

I ran into this on another forum where someone was trying to explain how much better than Nikon D810 was by showing ISO-invariance tests (completely unaware that this kills DR).  I pulled the following samples from DPReview tests.

 

Screen Shot 2017-05-21 at 2.02.45 PM.png

 Screen Shot 2017-05-22 at 1.30.16 PM.png

 

His response was to scoff at these because one is ISO 25600 and the other is ISO 12800 and he felt that was unrealistic and "nobody shoots at those ISOs".

 

Fair enough... let's pull data from something with less ISO gain... I pulled another test comparison and this one at ISO 3200.

 

Screen Shot 2017-05-22 at 1.29.47 PM.png

 

These are just honest results that you get when you operate these cameras at these ISO values and pull the images straight out the camera and don't manipulate them and don't use ISO 100 images and try to boost it by many levels of EV (I don't know photographer who does that... so why do they keep showing those tests)?

 

Lastly... did you happen to notice that the last batch of images ... the Canon camera was actually a 5D III (the 5D IV is better... but he was ranting on about the 5D III).

 

This is why I don't read DxO reviews.

This is why i don't look at ISO-invariance tests.

 


 


@KlausEnrique wrote:

 

It is funny to hear you guys say that Nikon and Sony "cook" their RAW files, but Canon doesn't. That is just such nonsense. They ALL "cook" their RAW files. I am sure you know what a Bayer Filter array looks like and what it does, right? Do your RAW images look like the image below? They don't, and that is because your camera interprets the data and extrapolates the information from the pixels around it (i.e what you call "cooking"). So please stop saying that Nikon cooks the raw files but Canon doesn't.

 

bayer-simulation2.png

  


Umm, a true RAW image would be a black and white greyscale image. The color is added because the RAW converter knows which pixel has which color of the Bayer Filter over it.

 

You are confusing RAW conversion with the RAW file itself. Yes, Canon DPP does much of the same type of cooking when it converts the RAW file that Nikon and Sony bake into the RAW file.

 

Again the difference is Nikon and Sony bake it into the RAW file so you are stuck with it. With Canon that processing is done at the time of RAW conversion.

 

That has some ramifications, one Canon won't do as well on 'tests' that use a RAW converter other than DPP, the other is if you need a RAWer RAW file like Astrophotographers do then either you have to buy a specialized D810A from Nikon (which has the same dynamic range as the Canon 5D Mk IV) or hack your Nikon camera to prevent it from baking in processing in the RAW file.

I am not confusing anything... If manufacturers did not "cook" their images, that is what the RAW files would look like!

 

Seriously, everything that you said below is incorrect. I thought photography was your hobby. I am beginning to think it is trolling...

 


@TTMartin wrote:

@KlausEnrique wrote:

 

It is funny to hear you guys say that Nikon and Sony "cook" their RAW files, but Canon doesn't. That is just such nonsense. They ALL "cook" their RAW files. I am sure you know what a Bayer Filter array looks like and what it does, right? Do your RAW images look like the image below? They don't, and that is because your camera interprets the data and extrapolates the information from the pixels around it (i.e what you call "cooking"). So please stop saying that Nikon cooks the raw files but Canon doesn't.

 

bayer-simulation2.png

  


Umm, a true RAW image would be a black and white greyscale image. The color is added because the RAW converter knows which pixel has which color of the Bayer Filter over it.

 

You are confusing RAW conversion with the RAW file itself. Yes, Canon DPP does much of the same type of cooking when it converts the RAW file that Nikon and Sony bake into the RAW file.

 

Again the difference is Nikon and Sony bake it into the RAW file so you are stuck with it. With Canon that processing is done at the time of RAW conversion.

 

That has some ramifications, one Canon won't do as well on 'tests' that use a RAW converter other than DPP, the other is if you need a RAWer RAW file like Astrophotographers do then either you have to buy a specialized D810A from Nikon (which has the same dynamic range as the Canon 5D Mk IV) or hack your Nikon camera to prevent it from baking in processing in the RAW file.


 


@KlausEnrique wrote:
I am not confusing anything... If manufacturers did not "cook" their images, that is what the RAW files would look like!

 

Seriously, everything that you said below is incorrect. I thought photography was your hobby. I am beginning to think it is trolling...

  

bayer-simulation2.png

  


@TTMartin wrote:

Umm, a true RAW image would be a black and white greyscale image. The color is added because the RAW converter knows which pixel has which color of the Bayer Filter over it.

 

You are confusing RAW conversion with the RAW file itself. Yes, Canon DPP does much of the same type of cooking when it converts the RAW file that Nikon and Sony bake into the RAW file.

 

Again the difference is Nikon and Sony bake it into the RAW file so you are stuck with it. With Canon that processing is done at the time of RAW conversion.

 

That has some ramifications, one Canon won't do as well on 'tests' that use a RAW converter other than DPP, the other is if you need a RAWer RAW file like Astrophotographers do then either you have to buy a specialized D810A from Nikon (which has the same dynamic range as the Canon 5D Mk IV) or hack your Nikon camera to prevent it from baking in processing in the RAW file.


 

As for a RAW file having color in the pattern of the Bayer filter you are just flat out wrong. I can't help you with that. The RAW file will contain a black and white grayscale image. The RAW conversion software then needs to know which color of the Bayer filter corresponds to which pixel.

That is why when Canon comes out with a new camera we get post with photos like this when people use a RAW converter that doesn't know which pixel to assign to which color.

 

14686i3328BA9A22568DA1.jpg

 

As for who is trolling, I know it is Politics 101 to accuse others of exactly what you are doing, then when I respond it sounds like **bleep** for tat. Well, you've got 17 posts here all of which are complaining about Canon's DR. Instead I'm here helping people, in fact one my replies have been selected as the solution to a person's problem 76 different times. So I'm not the troll here, but, you certainly seem to be.

 



 

 

Announcements