DRones vs. anti-DRones: how to resolve the controversy

[Lee Jay]

I wonder if Chipworks will dissect the 7D II sensor. It's been a long time since they dissected a Canon sensor...

Not so.

https://chipworks.secure.force.com/catalog/ProductDetails?sku=CAN-EOS-70D_Pri-Camera&viewState=DetailView

For $16k you can find out everything you want to know.

 

[LetTheRightLensIn]

Lots, since it turns out that we get a lot of wind around here and so do many other places, water flows and forest scenes happen to just be about at the range that exmor DR can just handle with single shot.

I'm not sure I buy it...but if true then shoot an Exmor camera or load up Magic Lantern. Not sure what else to say. Posting about it here certainly won't push Canon to change their ADC arrangement.

Maybe not. Then again they have to be taking notice how it's getting brought up more and more and how more and more people go crazy over every little thing like the ad campaign. Not signs of well content userbase. Maybe they get afraid enough to decide to move forward and to not cripple the silliest little things in other cases either?

 

[LetTheRightLensIn]

Anyway, at the end of the day if all Canon sees and hears is that it's not that big of a deal, nobody but a few will care, we will be stuck with this old 500nm fab for another decade or two, literally. So I don't see that it does anyone any good to minimize it. Even if you don't need it, it won't hurt you and a new fab might bring stuff that you do care about more too. Plus at least once in a blue moon you must mess up the exposure on a one of shot and at least you'll be able to rescue that better. And for those who do care more, we'll it would be great. Less money to get all your gear from one brand than a mix. Less to carry and drag around which can be a pain, literally. Canon does make awesome lenses, has a very nice UI and so on so it is nicer if Canon improves their sensors to go to a different system.

The thing is, we don't know yet what fabrication line the 70D and 7D2 sensors are made on... but we do know that going to the 20.2Mpixel design from the 18Mpixel design, the ISO performance increased slightly... The more complex lithography required for DPAF should have meant a reduction in high ISO performance, so they must have done something to counter it, and using their existing 180nM line (P/S sensors) seems like the most likely scenario... Also, it costs a lot more money to keep 2 fabrication lines open than one, so my bet is that the death of the 500nM fabrication run is already happening.

Perhaps, but then it's curious that after the long time between 7D models that they didn't start to implement some of their dual ISO read or column on sensor ADC patents and stuff at all yet.

 

[LetTheRightLensIn]

Anyway, at the end of the day if all Canon sees and hears is that it's not that big of a deal, nobody but a few will care, we will be stuck with this old 500nm fab for another decade or two, literally. So I don't see that it does anyone any good to minimize it. Even if you don't need it, it won't hurt you and a new fab might bring stuff that you do care about more too. Plus at least once in a blue moon you must mess up the exposure on a one of shot and at least you'll be able to rescue that better. And for those who do care more, we'll it would be great. Less money to get all your gear from one brand than a mix. Less to carry and drag around which can be a pain, literally. Canon does make awesome lenses, has a very nice UI and so on so it is nicer if Canon improves their sensors to go to a different system.

The thing is, we don't know yet what fabrication line the 70D and 7D2 sensors are made on... but we do know that going to the 20.2Mpixel design from the 18Mpixel design, the ISO performance increased slightly... The more complex lithography required for DPAF should have meant a reduction in high ISO performance, so they must have done something to counter it, and using their existing 180nM line (P/S sensors) seems like the most likely scenario... Also, it costs a lot more money to keep 2 fabrication lines open than one, so my bet is that the death of the 500nM fabrication run is already happening.

I wonder if Chipworks will dissect the 7D II sensor. It's been a long time since they dissected a Canon sensor... Would be nice to know what process Canon is using.

I believe they just did the 70D. The report costs $200 though.

 

[Loswr]

Are the anti-DR guys the only one hurling insults? From my perspective, the insults are originating from both sides, but maybe I'm just crazy.

Personally, I'd consider being labeled a "thug" insulting.

thug
THəɡ/ noun
1. a violent person, especially a criminal.

 

[I_Miss_Minolta]

Photography is an artistic endeavor and there have always been two sides to any artistry: on one hand you the creative types and on the other data types.

Look at music:
The creative types are concerned with conveying tones that excite the emotions of the listener.
The data types are concerned with repeating notes patterns on a page, you know, the "underlying data?"

And literature...poetry:
The creative types are concerned with exciting emotion and thought in the reader.
The data types are concerned with the texture of the page and the quality of the type-face under a magnifying glass, you know, the "underlying data?"

And photography is no different; we have the same two, special types:
Our creative types are concerned with composition, lighting, focus...on capturing the wonderment of what we call, "existence."
Our data types are all about dynamic range and shadow detail--the ones and zeroes of the, "underlying data."

So there you go...two wonderful sides to an artistic coin. Love it. Embrace it.

Note: I made up that silly B.S. regarding the music and literature data types--they don't exist.

 

[David Hull]

Anyway, at the end of the day if all Canon sees and hears is that it's not that big of a deal, nobody but a few will care, we will be stuck with this old 500nm fab for another decade or two, literally. So I don't see that it does anyone any good to minimize it. Even if you don't need it, it won't hurt you and a new fab might bring stuff that you do care about more too. Plus at least once in a blue moon you must mess up the exposure on a one of shot and at least you'll be able to rescue that better. And for those who do care more, we'll it would be great. Less money to get all your gear from one brand than a mix. Less to carry and drag around which can be a pain, literally. Canon does make awesome lenses, has a very nice UI and so on so it is nicer if Canon improves their sensors to go to a different system.

The thing is, we don't know yet what fabrication line the 70D and 7D2 sensors are made on... but we do know that going to the 20.2Mpixel design from the 18Mpixel design, the ISO performance increased slightly... The more complex lithography required for DPAF should have meant a reduction in high ISO performance, so they must have done something to counter it, and using their existing 180nM line (P/S sensors) seems like the most likely scenario... Also, it costs a lot more money to keep 2 fabrication lines open than one, so my bet is that the death of the 500nM fabrication run is already happening.

I wonder if Chipworks will dissect the 7D II sensor. It's been a long time since they dissected a Canon sensor... Would be nice to know what process Canon is using.

I would like to know what ADC they are using, I suspect that it is something like the Analog Devices ADDI 7004. Their sensors seem to be just fine regardless of what geometry they are making them in. They seem to be getting in excess of 15 stops of DR out of the latest ones (6D for example), if you believe Sensorgen. They just toss it away on the bottom end due to an implementation that is not optimized performance at the bottom end of the ISO range like the Sony stuff is. In terms of the sensor itself, it appears to be every bit as good as anything that Sony has produced (except for whatever they put in the A7s, or whatever that one is that can see in the dark).

 

[David Hull]

Anyway, at the end of the day if all Canon sees and hears is that it's not that big of a deal, nobody but a few will care, we will be stuck with this old 500nm fab for another decade or two, literally. So I don't see that it does anyone any good to minimize it. Even if you don't need it, it won't hurt you and a new fab might bring stuff that you do care about more too. Plus at least once in a blue moon you must mess up the exposure on a one of shot and at least you'll be able to rescue that better. And for those who do care more, we'll it would be great. Less money to get all your gear from one brand than a mix. Less to carry and drag around which can be a pain, literally. Canon does make awesome lenses, has a very nice UI and so on so it is nicer if Canon improves their sensors to go to a different system.

The thing is, we don't know yet what fabrication line the 70D and 7D2 sensors are made on... but we do know that going to the 20.2Mpixel design from the 18Mpixel design, the ISO performance increased slightly... The more complex lithography required for DPAF should have meant a reduction in high ISO performance, so they must have done something to counter it, and using their existing 180nM line (P/S sensors) seems like the most likely scenario... Also, it costs a lot more money to keep 2 fabrication lines open than one, so my bet is that the death of the 500nM fabrication run is already happening.

I wonder if Chipworks will dissect the 7D II sensor. It's been a long time since they dissected a Canon sensor... Would be nice to know what process Canon is using.

I would like to know what ADC they are using, I suspect that it is something like the Analog Devices ADDI 7004. Their sensors seem to be just fine regardless of what geometry they are making them in. They seem to be getting in excess of 15 stops of DR out of the latest ones (6D for example), if you believe Sensorgen. They just toss it away on the bottom end due to an implementation that is not optimized performance at the bottom end of the ISO range like the Sony stuff is. In terms of the sensor itself, it appears to be every bit as good as anything that Sony has produced (except for whatever they put in the A7s, or whatever that one is that can see in the dark).

I agree, I think the primary source of noise is downstream of the sensors. I think Canon sensors have a lot more dark current (based on my experience with 7D, 5D III, D5300 and D800 astrophotography subs)...when my Canon sensors are very cool (i.e. during winter), such as -8°C, they don't have any visible dark current even after several minutes worth of exposure. Much warmer than that, they do. D5300 (and D5100 files too, I guess) files, on the other hand, fare FAR better at much higher temperatures, they don't seem to have much visible dark up to around 10°C.

I still think a transistor shrink would benefit Canon, as well as a move to a more advanced fab and higher Q.E. design. Overall, though, I agree. I think the primary source of noise is down stream, probably the ADC.

Someone linked a paper about per-pixel ADC recently. Apparently it's fairly difficult to do, but if you do it right, you can dramatically lower the frequency of the ADC units, and increase ADC parallelism to 1/4 the pixel count (one ADC per four pixels, capable of simultaneous output for each attached pixel...so effectively 1/1 parallelism as far as output DU's go). Pretty amazing.

I think that is what is doing the job for Sony. They have a very low speed SAR style ADC so they get the full benefit of their sensor noise figure all the way down to minimum gain. Canon can't do that since their high speed ADC (pipeline architecture, I suspect) can only give about 12.5 effective bits even though it is a 14 bit converter -- that's just all those things do.

 

[fragilesi]

I would encourage you to get more personal experience with topics before writing about them in a manner that reeks of someone that has read up on something on the Internet and decided that this then makes them an expert on the topic. Being an Internet arm-chair expert is easy, problem is that such expertise is doesn't understand where reality disrupts the theory. And there are a few arm-chair experts on here, not just you.

I'm curious. What impression of you do you think replies like that give?

 

[dtaylor]

The proper exposure is one where you don't clip anything that you want to retain and where you put enough light on to minimize noise as best as you can without clipping (or going quite so far as to make processing tricky and leaving too few highlight tones).

I would agree. But whether you increase exposure to minimize noise or decrease exposure to preserve more highlight detail, you are shifting tones away from where you want them to be in print. Hence the reference to middle gray.

Calling it like "people going around underexposing 3 stops" makes it sound like they are making mistaken exposures. You may not have meant to imply that, but many of those who post like that do, since they then say stuff, like learn how to set a proper exposure [insult insult].

I did not mean to imply that, but how else should I describe it? We are over and under exposing to achieve certain things.

 

[Don Haines]

I saw this.... and for some strange reason it made me think of this thread...

https://www.youtube.com/watch?v=AhDG_WBIQgc

 

[dtaylor]

Maybe not. Then again they have to be taking notice how it's getting brought up more and more and how more and more people go crazy over every little thing like the ad campaign. Not signs of well content userbase. Maybe they get afraid enough to decide to move forward and to not cripple the silliest little things in other cases either?

In all honesty I hope so. I certainly would not mind less low ISO shadow noise even though I do not consider it as important as some do.

 

[Sporgon]

I saw this.... and for some strange reason it made me think of this thread...

https://www.youtube.com/watch?v=AhDG_WBIQgc

Ah, Don ! If only DRones were knocked down so easily. I wonder how long it was before the drone was back up and flying. If it was a CR DRone it'be about one minute

 

[Don Haines]

The proper exposure is one where you don't clip anything that you want to retain and where you put enough light on to minimize noise as best as you can without clipping (or going quite so far as to make processing tricky and leaving too few highlight tones).

I would agree. But whether you increase exposure to minimize noise or decrease exposure to preserve more highlight detail, you are shifting tones away from where you want them to be in print. Hence the reference to middle gray.

Calling it like "people going around underexposing 3 stops" makes it sound like they are making mistaken exposures. You may not have meant to imply that, but many of those who post like that do, since they then say stuff, like learn how to set a proper exposure [insult insult].

I did not mean to imply that, but how else should I describe it? We are over and under exposing to achieve certain things.

for most of my shots I am able to expose in the middle. The histogram looks good and nothing runs off of either end... but for many shots (10 percent ?) I could use more range. 2 stops more DR would change that percentage from 10 percent down to about 1 percent... so yes, you can count me as one of those people who would like more DR out of their camera.

And the thing is, If I had those two extra stops, I would still expect more in the next camera... It is natural to expect improvements, just as it is natural to expect technical/scientific people to evaluate performance and identify weaknesses and strengths.. but why attack the messenger? If it doesn't matter to you, then say "that's nice" and ignore the whole debate. If it does mater to you, then debate the facts, not the person.

 

[Don Haines]

I saw this.... and for some strange reason it made me think of this thread...

https://www.youtube.com/watch?v=AhDG_WBIQgc

Ah, Don ! If only DRones were knocked down so easily. I wonder how long it was before the drone was back up and flying. If it was a CR DRone it'be about one minute

This one probably needed more DR (Damage Repair).

 

[dtaylor]

What part of the world do you live in? Because by the sounds of it, wherever you do live, the wind don't blow there and the earth doesn't rotate either.

You can't HDR an interior because of the wind or the rotation of the Earth? ???

As for landscapes...neither GND nor manual blends have any issue with movement as long as there's not a large moving section that crosses the line or mask. You're not going to HDR a sprinter, but wind is seldom an issue in a landscape. I can hand hold a 3 frame bracket for crying out loud. Just how hard is this wind blowing that things radically change in <0.5s?

HDR tools also have features to compensate for motion.

I would encourage you to get more personal experience with topics before writing about them in a manner that reeks of someone that has read up on something on the Internet and decided that this then makes them an expert on the topic. Being an Internet arm-chair expert is easy, problem is that such expertise is doesn't understand where reality disrupts the theory. And there are a few arm-chair experts on here, not just you.

Who urinated in your Cheerios? :

I have literally thousands of brackets shot for blending...most of which were hand held...and have processed probably into the mid-hundreds out of the total number shot. The majority of processed shots have been 2-3 frame manual blends made with layer masks. A handful were done with various HDR tools. Motion between frames simply hasn't been an issue with the manual blends, though I can think of a few HDR software attempts I ended up doing manually specifically because of motion.

But feel free to make more ignorant posts about other forum members if it helps you get over whatever set you off today

 

[Don Haines]

What part of the world do you live in? Because by the sounds of it, wherever you do live, the wind don't blow there and the earth doesn't rotate either.

You can't HDR an interior because of the wind or the rotation of the Earth? ???

As for landscapes...neither GND nor manual blends have any issue with movement as long as there's not a large moving section that crosses the line or mask. You're not going to HDR a sprinter, but wind is seldom an issue in a landscape. I can hand hold a 3 frame bracket for crying out loud. Just how hard is this wind blowing that things radically change in <0.5s?

HDR tools also have features to compensate for motion.

In general I would agree, but I have had lots of problems where there are significant waves.... but then again, that is the large moving section that you mentioned as an exception, so that means that I really agree.

+1

 

[privatebydesign]

The proper exposure is one where you don't clip anything that you want to retain and where you put enough light on to minimize noise as best as you can without clipping (or going quite so far as to make processing tricky and leaving too few highlight tones).

I would agree. But whether you increase exposure to minimize noise or decrease exposure to preserve more highlight detail, you are shifting tones away from where you want them to be in print. Hence the reference to middle gray.

Calling it like "people going around underexposing 3 stops" makes it sound like they are making mistaken exposures. You may not have meant to imply that, but many of those who post like that do, since they then say stuff, like learn how to set a proper exposure [insult insult].

I did not mean to imply that, but how else should I describe it? We are over and under exposing to achieve certain things.

for most of my shots I am able to expose in the middle. The histogram looks good and nothing runs off of either end... but for many shots (10 percent ?) I could use more range. 2 stops more DR would change that percentage from 10 percent down to about 1 percent... so yes, you can count me as one of those people who would like more DR out of their camera.

And the thing is, If I had those two extra stops, I would still expect more in the next camera... It is natural to expect improvements, just as it is natural to expect technical/scientific people to evaluate performance and identify weaknesses and strengths.. but why attack the messenger? If it doesn't matter to you, then say "that's nice" and ignore the whole debate. If it does mater to you, then debate the facts, not the person.

Don, the problem seems to me to be that people give an opinion that is personal or state a spurious "fact", and then get defensive when that opinion is questioned, they take it personally so the cycle begins.

For instance, I agree with you, more will be very welcome and even when it gets here yet more will be expected, but I could take issue with your numbers, which might sound personal to some, I suspect very few people have "issues" anywhere near 10% of the time (and in a subtle way you set yourself up for what might appear to some a personal attack, initially you put a question mark next to the 10% but then dropped it), if they did then all the film shooters ever, and every digital camera up to now would be found wanting an unacceptable amount of the time, and in general, my experience is, that just isn't true. Of course there will be people who shoot a specific type of scene where those numbers might be accurate, and you could very well be one of them, but to suggest that camera DR capabilities fail 10% of the time is not true for me, or for many of the photographers I speak to regularly and for whom I print.

People are very quick to take rebuttals of their personal opinions personally, they are unwilling or emotionally unable to accept that the comment they made to invoke the rebuttal wasn't a soundly based fact they can back up with supporting independent evidence.

 

[Lee Jay]

I have DR problems a lot but that's because I get only get 7-9 stops of DR in many of my shots because so many are at high ISO.

If I had my choice, which I don't because of the quantum nature of light, I'd rather have two more stops of DR at high ISO than at low ISO.

 

[Loswr]

Don, the problem seems to me to be that people give an opinion that is personal or state a spurious "fact", and then get defensive when that opinion is questioned, they take it personally so the cycle begins.

I'd say 'and' instead of 'or'. The crux of the 'problem' is when people give an opinion that is personal and state that opinion as a spurious fact. Consider the following sets of examples:

[list type=decimal]
[*]"Canon sells more dSLRs than any other manufacturer."
[*]"A Canon dSLR best meets my needs."
[*]"Canon dSLRs are the best."
[/list]

[list type=decimal]
[*]"Current Canon sensors have less low ISO DR that their competitors."
[*]"The IQ of current Canon sensors doesn't meet my needs."
[*]"Canon sensors have poor IQ."
[/list]


In each case, statement (1) is simple fact, generally not disputed. Statement (2) is a personal opinion, and completely valid whether or not others share that opinion. Statement (3) is that personal opinion stated as a spurious fact. Statements like #3 are generally untenable, yet people who make statements like that tend to get defensive when their statements are rightfully questioned.

Stating your opinion as opinion is fine...stating your opinion as fact will draw a challenge. Defending that opinion as fact just makes you look foolish. Restating that spurious fact over and over in thread after thread will not only draw a challenge, it will piss people off.