DPReview's turn to beat up the 5DS/5DSr results

[Hillsilly]

You might be right. I might be reading more into people's comments than is really there.

BTW - shadow aren't the image destroying evil that the Sonikon corporation makes them out to be.

 

[bluenoser1993]

The amount of light a camera (should be) is capable of responding to at ISO 100 is double that for ISO 200. Similarly each stop of DR represents a doubling in the amount of light recorded. Thus the relationship between ISO and DR should be linear.

School is roughly 20 years in my past, but I believe this relationship is logarithmic. The graph only forms a straight line because the X axis has been manipulated.

 

[sanj]

I will point out that it wasn't me that started this debate about what's normal, rather it was someone suggesting that "ISO 400" wasn't normal (if fact it would appear to be almost normal):

This graph reminded me why I don't give a (insert your favorite phrase here) about dynamic range. The difference between sensors only shows up at ISOs that I seldom shoot at. Once you get into a normal range the differences are so tiny as to be totally irrelevant.
...

So here we have it that "unfocused" is suggesting that the ISO levels where there is a measurable difference (100-400) are somehow "not normal."

First, I wasn't aware that we were in sensor court and taking depositions. But, if that's the case, then please quit distorting and misquoting me.

As I have said, I absolutely consider ISO 400 to be "normal" for me.

On the chart, at 400 and above (and to some degree at 200 and above) the differences may be measurable, but they are insignificant. "Measurable" is your term. Not mine.

Many things are measurable, but still irrelevant (which is the term I used). The difference between six inches and six and an eighth inches may be measurable, but for the purpose of procreating, it's not relevant.

As I patiently explained before, 1/3 of a stop difference at ISO 400, is not significant for me. I know how to properly process an image and I routinely adjust various areas of a photograph by more than 1/3 stop.

Hahahahaa

 

[bluenoser1993]

The amount of light a camera (should be) is capable of responding to at ISO 100 is double that for ISO 200. Similarly each stop of DR represents a doubling in the amount of light recorded. Thus the relationship between ISO and DR should be linear.

School is roughly 20 years in my past, but I believe this relationship is logarithmic. The graph only forms a straight line because the X axis has been manipulated.

May as well correct myself, I realized after posting that both axis are manipulated in the same manner, so my comments are not correct. Sorry :-[

 

[Famateur]

The problem with the graph isn't that the red line is above the blue but rather that one line is almost straight and the other is not.

That the Canon line isn't straight indicates that Canon has not yet fixed the problems with their sensor.

So, if I understand what you've said, the problem with Canon is that it holds its max dynamic range for a few stops of ISO before falling? If Canon "solves" its problems, its max dynamic range would start falling immediately as you increase ISO, just like the Sony/Nikon? Hmm...interesting logic. That would be a nice straight line, but...

Yes, that is correct.

What if the graph started at the same max dynamic range as Sony/Nikon and kept the same non-straight shape? Wouldn't that be much better? You could boost the ISO a few stops before it falls off. I happen to like the shape of Canon's graph.

No.

It looks good until you understand that the shape of the graph for an "ideal" (or "perfect") sensor is a straight line like Nikon/Sony - only higher up (further away from the axes.)

The amount of light a camera (should be) is capable of responding to at ISO 100 is double that for ISO 200. Similarly each stop of DR represents a doubling in the amount of light recorded. Thus the relationship between ISO and DR should be linear.

If you look at this page:
http://sensorgen.info/CanonEOS-6D.html
and pay attention to the "Saturation" column, each row is roughly twice the one below or half the one above.

In an ideal world, each time that number doubles, the number in the "DR" column should go up by 1. The Nikon D810 is perhaps the best example of this:
http://sensorgen.info/NikonD810.html

That the Nikon sensor doesn't reflect reality vs ideal.

Comparing these two web pages is perhaps the best way to illustrate the problem with Canon's sensors.

Maybe I'm just suffering from lack of sleep, but when it comes to the graph, wouldn't the "ideal" graph be a horizontal line? In other words, wouldn't we all want a sensor that can maintain steady/maximum dynamic range throughout its ISO range? Obviously, that's not currently possible with known technology, but who knows what breakthroughs we'll see in the next 20 years. Again, we're talking about the shape of a graph representing a relationship between dynamic range and ISO sensitivity, not technology.

If Canon's graph was the same shape as its current graph, dropping off at the same rate and at the same ISO settings, but the max dynamic range at ISO 100 was equal to Sony/Nikon (i.e. they both start at the same point on the Y Axis), wouldn't that be better than immediately falling off? What makes the shape of the graph flawed in that scenario?

It seems to me that the more the graph approximates horizontal slope, the better, but again, 6 hours of sleep in the last 60.

 

[eml58]

Well, I've owned the 5DsR for 3 weeks now, 1500 images later I'm just impressed & very happy with what I have.

It's not a Phase One, but it wasn't meant to be, I've owned a D810, and this is better in every respect, for me.

I can quite honestly say I have never taken any notice of DXO, Dilbert or DP Review, I tend towards making purchasing decisions based on what Photographers have to say about their experience with a particular piece of equipment, then work with the equipment to reduce operator error, so far it's worked Ok for me.

And quite honestly in my view all this discussion of "graphs" is nonsense, and I agree with "Famatuer" regards the extra 1/8", I sincerely hope my wife does as well.

 

[ahsanford]

I can't recall anyone here suggesting otherwise. Certainly might have missed it, but most people have no problem admitting that sony sensors, toshiba sensors, aptina sensors, samsung sensors, etc, have a wider DR at low sensitivities than canon. The only recurring contention is how much it matters.

+1. I don't really think brand loyalty or hatred of DXO's shenanigans is at the point we're putting our fingers in our ears and denying anything. Canon's sensors are certainly second best for some applications. I just don't think it's remotely large enough a gap to walk out of my very comfortable, comprehensive first-party Canon ecosystem.

But as a rank amateur, my livelihood doesn't depend on getting the best / latest tech. Pros may have a different read on this.

- A

 

[traveller]

...
Maybe I'm just suffering from lack of sleep, but when it comes to the graph, wouldn't the "ideal" graph be a horizontal line? In other words, wouldn't we all want a sensor that can maintain steady/maximum dynamic range throughout its ISO range? Obviously, that's not currently possible with known technology, but who knows what breakthroughs we'll see in the next 20 years. Again, we're talking about the shape of a graph representing a relationship between dynamic range and ISO sensitivity, not technology.

I know what you're thinking - why isn't the DR the same for all ISOs? Each stop of DR represents twice the level of light as the previous.

Assume that I can measure light in ml. With 1ml I can get 1 stop of DR. With 2 ml I can get 2 stops of DR. With 16ml of light, I can get 4 stops. Each time I double the amount of light that I can collect, I can get another stop of DR. At 1ml, it simply isn't possible to get 4 stops of DR because there isn't enough light.

If Canon's graph started at "14 stops" at ISO 100 and was 13 stops by ISO 400 (i.e. relatively flat), the question wouldn't be "wow, even response from ISO 100 to ISO 400" but "Why isn't ISO 100 giving us 15 stops of DR?"

Whilst Dilbert's explanations aren't the clearest, for once I have to back him up -the curve on the Canon graph shows the component of total noise that is from read noise, which is proportionally greater at lower ISO values, because read noise is a fixed component (i.e. isn't ISO dependent). If I remember correctly, the older version of the DXO Mark site used to have a version of the graphs that allowed you to see this effect better.

 

[Orangutan]

Assume that I can measure light in ml. With 1ml I can get 1 stop of DR. With 2 ml I can get 2 stops of DR. With 16ml of light, I can get 4 stops. Each time I double the amount of light that I can collect, I can get another stop of DR. At 1ml, it simply isn't possible to get 4 stops of DR because there isn't enough light.

I don't think this is correct, or else I don't follow your explanation. Are you saying it has to do with full-well capacity? Could you explain further? As an example, consider shooting at ISO6400 at noon on a very sunny day. so there's plenty of light.

My assumption (though I'll defer to those who are trained engineers) has been that it has to do with amplification noise.

 

[traveller]

To be honest, I've stopped reading this site's forum so often because of the "Great DR Debate". Let me offer these "fantasy" choices to prospective 5DS(R) buyers:

[list type=decimal]
[*]Canon offers a 5DS RS with the Sony 42MP BSI Sensor
[*]Sony announce a firmware upgrade that optimises AF on EF Mount lenses: they are now just as fast as on the 5DS(R); on top of this, it implements full 14 bit uncompressed RAW files
[*]Nikon announce the D850 with the new Sony 42MP BSI sensor and offers a like-for-like trade in to swap all your EF Mount glass for F Mount
[/list]

Would you take the existing 5DS(R) or one of options 1-3? Your answer probably shows exactly why you are on this forum either running Canon down or defending them.

 

[3kramd5]

To be honest, I've stopped reading this site's forum so often because of the "Great DR Debate". Let me offer these "fantasy" choices to prospective 5DS(R) buyers:

[list type=decimal]
[*]Canon offers a 5DS RS with the Sony 42MP BSI Sensor
[*]Sony announce a firmware upgrade that optimises AF on EF Mount lenses: they are now just as fast as on the 5DS(R); on top of this, it implements full 14 bit uncompressed RAW files
[*]Nikon announce the D850 with the new Sony 42MP BSI sensor and offers a like-for-like trade in to swap all your EF Mount glass for F Mount
[/list]

Would you take the existing 5DS(R) or one of options 1-3? Your answer probably shows exactly why you are on this forum either running Canon down or defending them.

I have both a 5DSR and A7R2 on order (I previously ordered the 5DS but recently canceled it in favor of the R). If I could have fantasy option 1 in place of both, that would be a no-brainer. Obviously a whole hell of a lot of internals would have to change. It's not merely the sensor subassy, it's the end to end signal chain. But hey, it's fantasy, so I'll instead opt for

4: a version of the Pentax 645Z using whatever sensor architecture is in the Red Epic Dragon with offers for a like for like glass swap from EF to medium format.

 

[rs]

Nikon announce the D850 with the new Sony 42MP BSI sensor and offers a like-for-like trade in to swap all your EF Mount glass for F Mount

There's a huge amount of lens development required before that could happen, not to mention a larger diameter lens mount to allow for the 1.2 AF lenses.

 

[Corneria]

Nikon announce the D850 with the new Sony 42MP BSI sensor and offers a like-for-like trade in to swap all your EF Mount glass for F Mount

There's a huge amount of lens development required before that could happen, not to mention a larger diameter lens mount to allow for the 1.2 AF lenses.

You mean f/1.0 (Canon 50/1.0L for example). Nikon has a 50/1.2 Ais, and the new Mitakon 85/1.2 also shows that f/1.2 is no problem for Nikon. It's a pity though that they don't have any AF f/1.2 lenses.

 

[Sporgon]

Nikon announce the D850 with the new Sony 42MP BSI sensor and offers a like-for-like trade in to swap all your EF Mount glass for F Mount

There's a huge amount of lens development required before that could happen, not to mention a larger diameter lens mount to allow for the 1.2 AF lenses.

You mean f/1.0 (Canon 50/1.0L for example). Nikon has a 50/1.2 Ais, and the new Mitakon 85/1.2 also shows that f/1.2 is no problem for Nikon. It's a pity though that they don't have any AF f/1.2 lenses.

The Nikon F mount itself is capable of taking at least a 50 mm f/1.2 lens, indeed they make an Ais one. However in going fully electronic Nikon took the decision to go for the cheap option (there's a surprise) and position the electronic contacts in such a way as to make it difficult to fit a fully electronic f 1.2 lens. Compare this with Pentax who had the elegant solution of fitting the contacts through the mount flange.

 

[clicstudio]

I, for one can't wait for the day a camera sees what my eyes see. Who cares about 50MP when you can't take a photo of a person next to a window with a view of a bright day and you have to decide if u want the camera to measure the person or the background. If u choose the person, the background will be all white and blown out. If u choose the background. Then the person will be too dark.
I hate that. No matter how expensive the camera or the lens. It's all about that DR.
Video cameras can "see" it all already. So why can't photo cameras?
I rather go back to 6pm with the DR of the human eye.