The rift between the vocal "ISOers" and the "MPers"...

[thatcherk1]

Someone please tell me if I'm missing something. I've always understood pixel density and noise to be related in this way:

The story with noise is not quite as simple as "more density means more noise". There are a multitude of factors that contribute to noise, and many of them we can control. The 1DX 18mp sensor is considerably better than the 16mp 1D IV sensor, even though its higher density. I think that lends some strong evidence in support of the idea that we can get both better ISO/lower noise and more MP at the same time.

Remember, the context of what I said was a direct comparison of pixel density vs. ISO noise performance. The given in my example is comparing two sensors of identical technology, but with small pixels vs big pixels. It was theoretical.

 

[thatcherk1]

1. "If you were to compare a 1DX as designed now with 18MP with a 1DX that has 36MP, I think the ISO would be better in the 18MP version."

Wait a second. Have you seen a 36MP 1DX sensor? It doesn't exist. The 1DX has an 18MP sensor. You can't argue the quality of a sensor that is a figment of your imagination. Because in my imagination there is a 200MP 1DX sensor that is capable of an ISO of 234585848438493000.

Also, are we comparing noise intensity at 1:1 crop? Because as soon as you down-sample a 36MP image to equal the resolution of an 18MP image, much of the noise goes away.

When we say that an old camera has only a little worse ISO sensitivity than the latest camera, we are always comparing in 1:1 crop. As soon as you match resolutions to the older camera, it goes away. When you compare the two images side by side in a 12x18 print, you'll see the higher res image is much better quality, and the noise is much less noticable because the pixels that make up the noise are much smaller.

I'm surprised so many people complain about higher resolution being contrary to lower noise. Doesn't it make sense that if you have a massive resolution image, you can down-sample it, thus killing significant amounts of noise. At least that's what I observe when I shoot 6400ISO stuff on my 5D, then give it a lower res output.

 

[thatcherk1]

The problem with topics like this is that someone is assuming that those that want more MP don't want better sensor performance when it comes to noise at a given ISO. That's false.

I think there is a thought that although overall sensor performance goes up from generation to generation, there is still a degree that higher MP hinders ISO performance. I think everyone acknowledges that we all want better everything. But there are those who think that to have a lot better MP means only a little better ISO, and they would rather have a lot better ISO and a little better MP. And then there are those who believe that we have maxed out MP and it's pointless to go bigger and it's only a marketing thing. See Ken Rockwell's link on page 2.

Now I don't necessarily subscribe to either of the above ideas. I don't think that higher MP inherently hinders ISO (unless someone can explain correctly how this is the case). I also think that people see things with their own experience. A landscape photographer with a tripod might look at ISO and say "why the heck would you need such high ISO?!?! There is no need for that! I shoot everything at 100, and I use ND filters. But boy do i like printing 40x60 prints with lots of detail". A concert photographer might say "Megapixel? who needs that? photos only end up on websites, and at worst put on 5" CD covers or desktop backgrounds. But boy do I need that high ISO to capture wild rockers in a dark room".

Of course these are the two extremes. But some photographers only do one type of photography, and are blind to the needs of others. And many of us either do several types of photography, or are just educated on more aspects of photography than our own.

The silliest thing which even I am susceptible to is thinking it is productive to engage in this kind of conversation. I have a suspicion that Canon doesn't surf forums to look at "wishlists" or see whether photographers want ISO performance vs. pixel count in their next camera. Maybe they do. But if they do, they probably don't put a ton of weight into it. I figure they have various well-known photographers who test their gear, give input etc. But also all these Japanese electronics manufacturers are notorious for ignoring what users desire and need in equipment and end up going with what nerds in lab coats randomly decide to do. This is especially true in the cinema world (which is why Sony, Panasonic and Canon are losing to Red[US] and Arri[German]), and I think it's true in photography too.

But alas, when we photographers need a break from shooting and it's a rainy day with no jobs to prepare for, then we can come and entertain ourselves and each other talking about what's wrong with our current gear and what we wish we could have exist, or afford. And then argue about it.

I love the internet.

 

[jrista]

1. "If you were to compare a 1DX as designed now with 18MP with a 1DX that has 36MP, I think the ISO would be better in the 18MP version."

Wait a second. Have you seen a 36MP 1DX sensor? It doesn't exist. The 1DX has an 18MP sensor. You can't argue the quality of a sensor that is a figment of your imagination. Because in my imagination there is a 200MP 1DX sensor that is capable of an ISO of 234585848438493000.

Also, are we comparing noise intensity at 1:1 crop? Because as soon as you down-sample a 36MP image to equal the resolution of an 18MP image, much of the noise goes away.

When we say that an old camera has only a little worse ISO sensitivity than the latest camera, we are always comparing in 1:1 crop. As soon as you match resolutions to the older camera, it goes away. When you compare the two images side by side in a 12x18 print, you'll see the higher res image is much better quality, and the noise is much less noticable because the pixels that make up the noise are much smaller.

I'm surprised so many people complain about higher resolution being contrary to lower noise. Doesn't it make sense that if you have a massive resolution image, you can down-sample it, thus killing significant amounts of noise. At least that's what I observe when I shoot 6400ISO stuff on my 5D, then give it a lower res output.

You aren't reading all of my posts. I'm comparing the new higher-resolution 18mp 1DX sensor with the relatively new but still current-gen 16mp sensor of the 1D IV. The 1DX and 1D IV are very similar in their target audience, if not flat out identical. Adjusting for 1.3x crop, they are about the same from a density standpoint, however the 1DX sensor FAR outpaces the 1D IV sensor in every other way. In the case of ISO and noise, its SIGNIFICANTLY better.

Assuming we don't need an extremely high readout speed or native ISO52100 (the 1DX locks its mirror up at 14fps, so mirror flap isn't the limiting factor there, the sensor readout rate is the logical bottleneck), taking all of the other advantages from the 1DX sensor...do you honestly believe that 18mp is the highest we can go with lower noise? Photo shot, thermal, electronic, and readout mechanics...those are the factors that affect noise. Only one of them is beyond our control, the other three are entirely within our control. Given the trend with sensors over the last five years, which has seen 2-3 fold increases in pixel density with marginal increases in noise without significant refactoring of readout mechanics until recently... Well, logically, given both current Sony and 1DX noise floors, I don't see any reason why we can't increase pixel density in a 5D III to 30mp (which, btw, is still FAR less tense than the 18mp 7D or 24mp A77) and ALSO improve noise characteristics.

 

[thepancakeman]

Wait a second. Have you seen a 36MP 1DX sensor? It doesn't exist. The 1DX has an 18MP sensor. You can't argue the quality of a sensor that is a figment of your imagination.

First of all, theoretical is not the same as imaginary. At this point the entire existence of a 1DX is theoretical, because no one owns a production copy of it. So to have a theoretical 36MP (which is the rumored sensor in the Nikon D800) is a perfectly valid concept for discussions of an idea.

And secondly, as this is a rumors site, a lot of what we discuss here aboslutely is the figment of someone's imagination.

 

[thatcherk1]

First of all, theoretical is not the same as imaginary. At this point the entire existence of a 1DX is theoretical, because no one owns a production copy of it. So to have a theoretical 36MP (which is the rumored sensor in the Nikon D800) is a perfectly valid concept for discussions of an idea.

And secondly, as this is a rumors site, a lot of what we discuss here aboslutely is the figment of someone's imagination.

True, no one has seen a production copy of 1DX. But there is an existent 1Dx 18MP sensor. There isn't a 36MP 1Dx sensor. Maybe there is a 36MP Nikon sensor. There are many factors that cause a sensor to have certain properties. You can't take one existing sensor that is 18MP and theorize that a sensor that is identical in every aspect except the number of photo-sites is going to perform in a predictable way.

But I'll play your little game. If you take a 36MP image that presumably has higher noise, and compare it to an 18MP image at 1:1 you will see more noise in the 36MP image. But practically speaking, as soon as you down-sample the image to match the resolution of the 18MP image, I am willing to bet there wouldn't be much difference between the two. One of the biggest effects that down-sampling has on an image is it averages out extremes in pixel variation, which (pixel variation) is exactly what noise turns out to be.

 

[jrista]

First of all, theoretical is not the same as imaginary. At this point the entire existence of a 1DX is theoretical, because no one owns a production copy of it. So to have a theoretical 36MP (which is the rumored sensor in the Nikon D800) is a perfectly valid concept for discussions of an idea.

And secondly, as this is a rumors site, a lot of what we discuss here aboslutely is the figment of someone's imagination.

True, no one has seen a production copy of 1DX. But there is an existent 1Dx 18MP sensor. There isn't a 36MP 1Dx sensor. Maybe there is a 36MP Nikon sensor. There are many factors that cause a sensor to have certain properties. You can't take one existing sensor that is 18MP and theorize that a sensor that is identical in every aspect except the number of photo-sites is going to perform in a predictable way.

But I'll play your little game. If you take a 36MP image that presumably has higher noise, and compare it to an 18MP image at 1:1 you will see more noise in the 36MP image. But practically speaking, as soon as you down-sample the image to match the resolution of the 18MP image, I am willing to bet there wouldn't be much difference between the two. One of the biggest effects that down-sampling has on an image is it averages out extremes in pixel variation, which (pixel variation) is exactly what noise turns out to be.

I don't think anyone disagrees with you on that point. You always have to normalize samples for comparison before comparing. I think that was the point dtaylor and myself tried to get across on the "Earthshatteringly Disappointed" thread, but it seems to be lost on most people.

 

[thatcherk1]

You aren't reading all of my posts. I'm comparing the new higher-resolution 18mp 1DX sensor with the relatively new but still current-gen 16mp sensor of the 1D IV. The 1DX and 1D IV are very similar in their target audience, if not flat out identical. Adjusting for 1.3x crop, they are about the same from a density standpoint, however the 1DX sensor FAR outpaces the 1D IV sensor in every other way. In the case of ISO and noise, its SIGNIFICANTLY better.

Assuming we don't need an extremely high readout speed or native ISO52100 (the 1DX locks its mirror up at 14fps, so mirror flap isn't the limiting factor there, the sensor readout rate is the logical bottleneck), taking all of the other advantages from the 1DX sensor...do you honestly believe that 18mp is the highest we can go with lower noise? Photo shot, thermal, electronic, and readout mechanics...those are the factors that affect noise. Only one of them is beyond our control, the other three are entirely within our control. Given the trend with sensors over the last five years, which has seen 2-3 fold increases in pixel density with marginal increases in noise without significant refactoring of readout mechanics until recently... Well, logically, given both current Sony and 1DX noise floors, I don't see any reason why we can't increase pixel density in a 5D III to 30mp (which, btw, is still FAR less tense than the 18mp 7D or 24mp A77) and ALSO improve noise characteristics.

You apparently aren't reading the post that you are quoting of me. Notice how I only quoted one sentence in your post. That means I'm only commenting on that sentence. I don't remember mention of a 1Div being a part of that sentence. What I do remember is a statement about a hypothetical 36MP sensor being noisier than a real 18MP sensor.

I also don't remember saying anything about the 1Dx not outpacing the 1Div sensor. I don't recall saying anything about 18MP being the highest we can go with lower noise.

In fact I would agree with you so much so that I disagree with you. I think there is a long way we can go in fighting noise performance. You speak of Photo shots which aren't controllable by us mere mortals. Sure we can't destroy them, but we can tame them by increasing pixel density to capture all the variations, then downsample, evening them out (hypothetically of course).

I don't know where you get the impression from my post that I think we can't improve ISO performance and pixel density.

 

[Mt Spokane Photography]

If Nikon puts out a 36mp camera in a pro level body, then I'd expect it to be good! I think that 36mp is approaching the area where with todays technology noise will be acceptable. If not, they won't make it.

I'm always in favor of seeing different manufacturers develop new products and offer options. It would really be boring if they were all the same, photographers need a choice.

Still, as long as I get enough pixels to crop my FF image, I'd go for low noise and better light sensitivity if that was a option. I did not get a D3S because 12 mp did not meet that criteria, but 18mp does.

 

[jrista]

@thatcherk1: Perhaps I misread, I'm not sure. I've been trying to make the point that I don't think there is anything physically that would prevent us from increasing MP and maintaining or even reducing noise...without the need to downsample. That at a 100% crop comparison, given some of the read mechanics improvements made by Sony in some of their sensors, and some of the known problems with the read mechanics in Canon's sensors from the last several years, and the apparent improvements with the 1DX sensor, we should be able to improve both MP and noise at the same time. It sounded like you were saying that wasn't possible, but apparently that it isn't necessary in apples-to-apples (normalized, i.e. resized) comparisons.

I completely agree that in a normalized context, even if they appear noisier at 100%, more MP will rarely if ever be worse than a lower MP image when properly normalized. So...I guess we agree.

 

[jrista]

I'm always in favor of seeing different manufacturers develop new products and offer options. It would really be boring if they were all the same, photographers need a choice.

Yup! Thats exactly my point, too!

 

[thatcherk1]

Well jrista, I guess we'll just have to agree to agree.

 

[AprilForever]

If Nikon puts out a 36mp camera in a pro level body, then I'd expect it to be good! I think that 36mp is approaching the area where with todays technology noise will be acceptable. If not, they won't make it.

I'm always in favor of seeing different manufacturers develop new products and offer options. It would really be boring if they were all the same, photographers need a choice.

Still, as long as I get enough pixels to crop my FF image, I'd go for low noise and better light sensitivity if that was a option. I did not get a D3S because 12 mp did not meet that criteria, but 18mp does.

Ahhh... the joys of competition in the free market world... If they do make one, I can't wait to see the "I'm switching to Nikon" threads popping up! ;D Seriously, though, tech progresses, and perhaps, our grandchildren's grandchildren may not even know what SLR means...