Will Canon Answer the D4s? [CR2]

[Stu_bert]

Personally, I've taken about 55,000 photos at ISO 800 - 3200. I've taken about 15,000 at ISO 400, and less than 10,000 at ISO 100 and 200. Of the ISO 100 photos, I have needed more dynamic range than my 7D offers in about 2000 shots, however I am usually short by maybe one stop (and that is more because of the 7D pixel size...if I had a 5D III, I would have what I need for pretty much everything I've shot before.) In the cases where slight vertical banding noise did show up in the shadows (maybe a couple hundred at most)...I used Topaz DeNoise 5, and was not only able to remove the banding, but I also gained more dynamic range (that's what happens when you reduce noise anyway...you gain DR, but Topaz has a feature that attempts to further recover DR that was lost to shadow noise due to a loss of tonal fidelity, which gains me even more.)

I use GND filtration for my landscape photography, so dynamic range is actually something I have a lot of control over in the field. I would actually greatly appreciate more native sensor DR, as it would reduce my need to use GND filters. It would also help me avoid that unsightly GND artifact where mountaintops end up dark or even black when you need to use more than two to three stops of filtration. That is the single situation where I think having more dynamic range would actually be the most important factor for IQ...ONE situation. I also suspect that tonemapping 14 stops into 8-10 stops without ending up with quirky shifts in contrast and color fidelity would still be very challenging, and I highly doubt I would stop using GND filters even if I had a D800. I still doubt I would push shadows around more than 2-3 stops....but it would be 2-3 stops along with fewer GND filters, which still makes the job easier in the end.

I've been used GNDs for 20+ years, but that's because I have to. However I would not say that gives me control over DR. It allows me to compress it into a range which the sensor can capture. But if you didn't have to use a GND, wouldn't that be better?

Re low ISO - surely the point is that where possible, post processing should be kept to a minimum. It's not the fact that you can push, it's the fact that if you could avoid it, then you can spend more time doing the things you want to, and less effort after.

Third, the more you can see the picture as is, in the field then the less you have to visualise what post processing will do for you.

As a landscape photographer, I would like lower ISO (<100, ideally ISO 12 as I used to shoot in slides), and yes I would like the ability to have better DR and less noise in the picture full stop (but there are also a lot of times where you don't want greater DR and you will reduce it to focus the viewer where you want to)

When I do wildlife, urban or sports the same is true. But that does depend on the shots you are trying to take.

However, does the 1Ds III still take amazingly good photos? Sure does. As does every camera in the last 5-10 years. But is not the point that we still want better/improvements, and aside from the UX/Software features, where else can Canon improve (AF for movies, DR, tracking algorithms)? Like you, I'm not saying it's the end of the world for my photos if I don't have these features - you work with what you have, it's just that it would make it easier.

As for stats on ISO ranges - surely that changes with time? With the 10D I hated going to ISO 200, let alone 400. With a camera today, I'm not bothered about ISO 1600. So I'm not sure that's too helpful.

 

[Don Haines]

As for stats on ISO ranges - surely that changes with time? With the 10D I hated going to ISO 200, let alone 400. With a camera today, I'm not bothered about ISO 1600. So I'm not sure that's too helpful.

Good point! Before I shot Canon, I shot Olympus. I can honestly state that I NEVER shot above ISO 1600... because that was as high as it would go... ISO 1600 was absolutely unusable.... even ISO 800 was so bad that it took heroic effort post-processing it to get something that wasn't embarrassing to show. realistically, ISO 400 was as high as you would ever go. ISO 12,800 on a Canon crop sensor is cleaner than the Oly was at ISO 800, and a 5D2 at ISO 25,600 easily beat ISO800 from the Oly...

 

[Stu_bert]

I'm quite content with the DR of my 5D3 ;D

Lovely shot. But is it only me who feels that a bit more detail in the burnt out sun area would be nicer? A grad filter or change in lighting. Just wondering... I know the hot spot is interesting but JUST A BIT MORE DETAIL perhaps?

Personally, I like it how it is. I might actually increase the glare just a bit. Not every region of a photo needs more detail, sometimes lower detail and less contrast is exactly what you want.

It's a lovely shot and one I really like...but that sky is blown out. An ND grad would render the sky darker and probably lost detail in the darker sky areas. It would have increased contrast where it wasn't wanted. The only way to have fixed this here is to have taken a 2nd photograph but at a 2-3 stop darker exposure and then blended the highlight areas carefully in photoshop using a layer. Shadows can be pulled but clipped highlights are not recoverable. It's also important to render the sky brighter than the foreground, another error I regularly see where ND grads are employed. If an ND grad was used with the above photo, the sky would have been darker than the foreground and wouldn't look right.

Many people here are talking about the D800's extra DR, but the truth is that it's only in the shadows...or rather it's the push-ability in the shadows during post production with low iso noise is really what is being talked about. Highlight clipping / blown highlights occurs at pretty much the same between the 5DIII and D800. So it's not really any extra DR, just better iso thresholds in the deep black areas.

+1 on the picture but it is personal preference. I remember a Pro photographer giving a talk where he describes the best photos as where your eye is drawn into the photo to where you want it. You can take it on a "journey" around the picture, but it never leaves the shot and always comes back to where you want.

I love the car, I get the positioning of the sun, like how the lighting has been done on the car, the slight glare on the roof (don't disagree more might work also) and the composition / low vantage point (tbh far better than I could take) but my eyes still oscillate between the car and the sun, (sorry just personal affliction aka "pet hate").

 

[ewg963]

3 new highend bodies? So one is supposed to be a 7D2, another is the high MP 1D body, and possibly the 1Dx successor is the third? Makes sense if we're having a 5D4 in feb/mar 2015.

A 5DIV in 2015 that could be possible...

 

[Dick]

Canon doesn't have to answer the D4s. It needs to answer the D800(e).

No, they really don't. The 5DIII was the answer, and it has outsold the D800 and far outsold the D800E.

I still think it's possible we'll see a high MP sensor from Canon in the future, not as an 'answer' but rather to capture a part of the market they feel may be untapped. But we'll only see that if Canon feels there's enough of a market for it. I wonder if the D800 sales might be pointing them in another direction...

This is a sensible, yet very stupid, way to look at it. Fans of music artists also waste time on caring about the sales of their own favorite artists. In the end though, all that should be irrelevant to the fans. It doesn't matter if plenty of others bump the same music. Bad sales might actually force something better out the next time.

Canon's sales don't really matter to me. Or actually... Had the 5D3 flopped, Canon would most likely work hard to make improvements and maybe they'd even unlock some of the features that have been artificially hidden to justify the 1DX price tag.

From a sales point of view you can say that they don't have to do S___ now, but instead of viewing the whole situation as a competition agains Nikon, there are other aspects. Besides, they need to beat the crap out of the 5D3 with the upcoming replacement to get money from the 5D3 purchasers again.

The bottom line is though that I don't care how well Canon does on the markets. Nikon can outsell Canon and I just don't care. For me the thing that matters here is the equipment I can afford to use. Surely it's heavily linked to the markets, but that's none of my business.

Lots of contradictions there! Music fans shouldn't care about sales, bad sales might force the artist to do better. Same with the 5DIII. So we shouldn't care about sales, even though you acknowledge that sales will alter what comes next?

Some people like to keep their heads in the sand, that's ok for them. Since I'm happy with some aspects of my Canon gear, and dissatisfied with others, I care about the business drivers that are going to determine what Canon does next.

Maybe I chose the words a bit badly. Music fans want platinum sales, which is ridiculous as they pretty much guarantee a bad album the next time. For some weird reason it matters a lot that others like the same music, like it'd be hard to come up with own opinions without support. As a Mac user I really liked the fact that Macs weren't that common - 0 risk of viruses and so on. Now that everyone has a Mac, it's most likely just a matter of time...

Canon sucking at sales would benefit me more than them outselling everyone. "Canon has to do this and that" is the common argument, when in reality Canon doesn't have to do anything if they get the sales with their current stuff. You might be right that business drivers control things (Damn I hate that wording... I'm a management consultant & I really work hard not to clutter my work with bullshit bingo words like that.), but that is a really sad reality. That means that instead of great stuff, they sell cost competitive mediocrity at the highest price they can. Well, at least I can somewhat afford the hobby, but the perfect lens will never be released.

 

[jrista]

For ISO 160, Canon does a downstream third-stop "pull", which again costs you a third of a stop dynamic range (crushing blacks). Since it is a post-read push, it reduces read noise by a third of a stop (hence, the notion that ISO 160 is "cleaner" than ISO 100...it is, by a minuscule amount.)

Bill Claff's data show the 'jagged' relationship of read noise vs. ISO, where the 'valleys' of noise are 160-320-640-. But, his data show the same jagged plot for DR vs. ISO, except the 'peaks' are the inverse of noise, with the greatest DR at 160-320-640. Your statement of a loss of 1/3-stop ISO at 'pulled' settings doesn't jive with his data that ISO 320 has greater DR than ISO 400 as well as less noise.

You express certainty that ISO 100 is the true base, but I think a model where the real base ISO was somewhat lower than 100, such that multiples of 100 are a light push, multiples of 125 are a harder push, and multiples of 160 are a light pull, would explain Claff's data.

Any idea what's going on there?

I am not exactly sure how Bill Claff derives his data. Logically, I am unsure how one could gain dynamic range by shifting the signal after the pixels are read. Think of what Canon does as a digital post-process exposure boost. If you utilize 100% of the sensor's available dynamic range at ISO 400, you have information at every level from 0 through 65535. If you "push", you shift the whole histogram to the right...and clip the highlights. If you "pull", you shift the whole histogram to the left...and clip the shadows. Now, in post, your probably working with a tool like Lightroom, in which case your actual RAW data is never changed, and you can shift the digital signal around to your hearts content without loss of precision or data. However with third-stop ISO settings in a Canon camera, the results after the push or pull are baked in. If you push, anything that clipped past pure white is permanently pure white. You cannot recover it. Similarly, if you pull, anything clipped past pure black is permanently pure black.

I'm wondering if the apparent reduction in the read noise floor is what gives Claff's results this third-stop "pull" boost to DR. If he is ignoring the fact that the highlights shift down along with the rest of the exposure, and simply uses a the 5D III FWC as a constant for the upper limit for the signal, then I can see how dynamic range would theoretically increase. The read noise floor is indeed lower, however because of the pull, you actually destroyed data in the shadows, and the camera itself is not actually utilizing the headroom gained by shifting highlights down (since this all occurs AFTER the sensor has been read)...hence the "loss" of a third of a stop dynamic range.

 

[altenae]

Nice topic (not)

Reminds me on the iphone IOS vs Android topic.
What is the answer from Apple to or what is the answer from Android to......etc.

It's very easy you don't like Canon because they don't have an answer to the D800 or D4s ( what ever the specs will be). Buy the Nikon and stop complaning.

Rather share usefull information on this forum then these never ending stories.

Some people need very fast and good AF and less need for DR

http://www.wildlife-photos.net/gallery/displayimage.php?pid=664
http://www.wildlife-photos.net/gallery/displayimage.php?pid=692

 

[jrista]

High ISO...I'd call that ISO 800 or above. ISO 100, 200, and 400 I consider low ISO, although ISO 400 is kind of in the middle there, and others might have a different opinion.

I don't have a specific study. It's a simple observation, however on that I have been making over the last four years or so. (FYI, I've moderated photo.stackexchange.com since 2010, and have encountered and chatted with quite a number of photographers over the last four years from a wide range of photographic endeavors.) How many white Canon lenses do you see at pretty much every sporting event around the globe? Hundreds to thousands at each and every event. Canon dominates sports, hands down, no question. They really dominate action, not just sports. I spend a lot of time out in nature, and meet a fair number of nature photographers. The very vast majority of the people I've met out in the wilderness, including both wildlife and bird photographers as well as landscape photographers, overwhelmingly have Canon equipment. Canon 1D IV, Canon 5D III, and Canon 1D X are becoming almost ubiquitous in the wildlife and bird world. Canon great white lenses, 300s, 500s, and 600s, are extremely common (particularly the 500 Mark Is...lot of wildlifers and birders use that lens, guess it's at a sweet spot of weight and cost). I've met a few who have Nikon equipment, two of whom use D800's for bird photography. I know of one (now a good friend) who uses Pentax and Nikon. I also know and have encountered/chatted with a decent number of wedding & portrait photographers. Most use the Canon 5D II. A few still use the 5DC (they don't seem to care about resolution). Some use the 5D III (and all of the 5D line wedding photographers had one consistent complaint before the 5D III: Sucky AF.) I know of several wedding photographers who use Nikon and other brands (some have gone to mirrorless as of late, with a variety of brands.) I know two wedding and portrait photographers who use Nikon exclusively. One uses a D800 and D3, the other uses a D7000, with a D800 planned for very soon. I would say that Nikon seems to have a growing following in the strait portraiture arena...not so much for DR, but for the sheer amount of detail the D800 or D600 bring to the table...seems that ridiculous, razor-sharp detail that brings out every single pore is really "in" right now, and there is no question that the D800 offers that in spades.

So, sorry, I don't have an official study for you, but it really isn't a difficult observation to make. Just look around. It's a very well-educated guess. The number of cameras and lenses that you can spot in the world that say "Canon" on them vastly outnumber any other brand. Of those, the biggest group that uses the most cohesive set of camera features are the action shooters. Sports/Olympics, Wildlife, Birds...and you can throw in car racing, air shows (know a few guys who do this, damn good at it too), kayaking, boat racing, pretty much anything you could remotely call a sport, or has moving subjects...the camera is going to be at a higher ISO setting, and is probably a 5D III or a 1D X. The next two biggest groups would be Wedding & Portrait, and Landscapes. Not sure which is bigger...seems pretty evenly split here in Colorado, but if you hit larger metropolitan areas, I would make the educated guess that Wedding and Portrait photographers would end up significantly out-pacing the Landscape photographers (and I mean real landscape photographers...I know more people than I can count who use entry level cameras, Canon, Nikon, Sony, Pentax, everything...and call themselves landscape photographers, but their work wouldn't land on anyones walls (no offense to anyone like this, but)...blown clouds, random people in the frame, lack of interesting composition, effectively point-and-shoot mountain peaks and a few scattered rocks or stubby evergreen trees here and there, never any post processing, thrown up on Imgur, PhotoBucket, or Facebook.)

I honestly don't have all that much knowledge about studio photographers. I can't really say how big a customer segment studio photographers might be for Canon...but I guess big enough for them to create the 1Ds line in the past. What I DO know about studio photography, it seems to lean medium format (or maybe Leica S-system) a lot more than it leans Canon, Nikon or Sony. Phase One also seems to be the brand I hear about most from the studio photogs I do know or have crossed paths with.

If you present something like some sort of statistical fact, you need to prove it. If it’s just an educated guess based on personal observation, it’s not a very strong base to build an argument on because it has no more value than any other personal observation claiming the opposite unless you have some proven authority on the subject. I have no way of knowing how well-educated your guess is. Please point me to your scientific publications in this particular field of research because moderating a website about photography doesn’t make someone a statistical expert. Ask yourself if the people visiting the website you moderate are even close to a cross section of camera users worldwide (that needs to be true if you want to extrapolate).

Alright, fair enough.

Anyway, if iso 800 and above is high iso, just look at the facts as they are presented by DxO mark. I’m not looking for (yet another) debate about DxO sensor scores. Just look at their measurement.

ISO 800 (high iso by your standard) on the Nikon D800 is as good as ISO 100 on the 5D3 (that’s a 3 stop ISO difference).
At ISO 3200 and above the 5D3 gets me marginally better results.

The reason I think “playing catchup” in the DR department is true for Canon is because they are not purposely crippling the 5D3. It’s not like the autofocus or fps on the 5D2. Canon doesn’t have a senor in production with the low iso DR capabilities of other manufacturers and at some point they need to catchup.

But don’t get me wrong, I’m not arguing the D800 is a better camera or Nikon has better camera systems. I know there is more to photography than dynamic range and sensor resolution. I’m happy with my 5D3 and my Canon TS-E 24mm f/3.5L II is much better than the Nikon 24mm PC-E but as Canon users we probably will be better off if we encourage Canon to catchup in the DR department.

Sure, I don't disagree that Canon lags behind in the low ISO arena. I just don't think having that extra two stops of dynamic range is actually quite as important as may photographers thing. It improves editing latitude, that's all it really does. One way or another, the things we view our photographs on...screen and print...have considerably less dynamic range. Screens average 8 stops, print averages 6-7 stops (or even as little as 5 stops). If anyone has tried to print on a low dMax paper with a natural white L*, they would understand how difficult it can be to compress even 11 stops of dynamic range into less than half that. It is not an easy task, and one has to be careful to avoid posterization and color fidelity issues in the shadows, where the available data (even after you remove noise and recover DR in the case of Canon sensors) is "sparse" compared to the midtones and highlights.

Compressing 13 stops of native dynamic range into 8 stops of screen DR is similar. You can lift the shadows, but beyond a certain point (around 3-4 stops) you run into the same issues...information in those low signal strength areas is sparse...discrete level transitions are often harsh, and ironically, the addition of noise is often the only way to combat posterization. Don't get me wrong, having more DR is useful, but until we have more available dynamic range on screen and in print (both of which I think are coming), being able to push shadows around by 4-6 stops doesn't buy you as much as you might think.

Pushing around exposure with more dynamic range than your screen is like tonemapping a 32-bit HDR image into 16-bit. You have twice the dynamic range (really, more than that, since 32-bit is floating point...so you have MANY times the dynamic range) as your output target. If you have a lot of shadow tones bunched up at the left end of your histogram, and have done much HDR, you would know how difficult it can be to tonemap all that extra dynamic range into the mere few shadow stops you have in a 16-bit scalar image. It's DIFFICULT! The results usually end up having that "HDR Look", where you end up with funky noise issues, posterization, odd tonal gradients, poor color fidelity, and even tonal inversions in the shadows.



I totally agree, though. I am not saying Canon shouldn't continue to push the envelope with their sensors. I just think dynamic range gets FAR more importance put on it by photographers than it really deserves. I think other in-camera factors are more important for most forms of photography, particularly any form of action photography (hence my earlier arguments about my observations of Canon's distribution of customers...honestly, observe a little yourself, and I think you'll come to the same conclusion...sports/action shooters are Canon Photography's bread and butter! ) Autofocus is probably the most important factor for action shooters. After AF, I would say frame rate. If you don't capture the right frame in focus, it really doesn't matter a wit how much dynamic range you have...you either have the wrong frame where that basketball players arm is half out of the frame, or the image is miss-focused. You might have 14 stops of missfocused arm-clipping wonder...but it ain't usable!

Finally...I do believe Canon will deliver more DR. How and when, and at what cost, I honestly don't know. However having been a Canon customer for almost five years now, I've only grown more fond of them. They make a good product, they put in the effort to ensure their product, whatever it is (even if it is perceptually inferior to the competitions) is solid, reliable, and always backed by the best customer support in the industry (and I speak from experience on a few occasions where I needed to send my lenses in for repair.) Canon, in my opinion, very carefully listened to the most important and broadly held demands of their customers with the designs of both the 1D X and 5D III. I mean, people here on CR and on many other forums like DPR seemed surprised that Canon chose to release the 1D X with "only" 18 megapixels, and were surprised that the 5D III "only" got a 1mp boost. Personally, I felt that Canon delivered exactly what their customers were asking for...as I'd heard, and even called for myself, the following on photo.stackexchange.com (that site I moderate...which actually has a very broad worldwide participation, so I think it is a decent source of information like this):

"I'm so sick and tired of the megapixel wars! I want better pixels, not more pixels!"
"I don't care about pixel count. I just wish Canon would make less noise at high ISO."
"Fewer megapixels! All I need is 10mp. I can print very large with just 10mp. Give me low noise ISO 204,800!! I want to photograph the aurora as I see them with my own eyes."
"The 1D III AF system really needs to be replaced with something much better. Something like Nikon's D3 AF system." (The guy was talking about a denser, reticular AF design, which Nikon actually had first.)
"The 5D II 9pt AF is horrible. I just don't have the points I need to focus where I need. I wish they would make the hidden AF assist points selectable." (That guy got a lot more than he asked for!)

I heard a lot of this kind of stuff from 2010 (and maybe late 2009) through 2012. It was only AFTER the release of the D800 and the 5D III that the things people complained about regarding Canon's equipment changed. Once they actually had fewer megapixels and better high ISO, they stopped complaining about it. Now, they complain about not having more DR (although I know for a fact that many of them are action shooters, and rarely use ISO settings below 800...so more DR wouldn't do them a bit of good. Fewer megapixls, bigger pixels, and less noise at high ISO, however, would do them a LOT of good! )

Personally, I was one of those hoping the 5D III would get more DR and more megapixels. I wanted a landscape camera. I still do. Ironically, the 5D III will still fill a role as a bird and wildlife photography camera given its feature set, and it is one of the top two things on my list of camera gear to buy this year. So I still find the 5D III to be a great camera. Regardless, I still hope Canon releases a landscape camera with lots of megapixels and lots of DR at some point in the future, because that is the one area of my work where more megapixels and more DR are literally the most useful things (AF and frame rate don't matter for jack when it comes to landscapes...but I also admit that as a landscape photographer, I become part of a much smaller minority of Canon customers, so I don't expect them to release anything with tons of MP and more dynamic range in the price range I want...I suspect the camera will be a 1Ds X at $5000. :\)

 

[jrista]

I'm quite content with the DR of my 5D3 ;D

Lovely shot. But is it only me who feels that a bit more detail in the burnt out sun area would be nicer? A grad filter or change in lighting. Just wondering... I know the hot spot is interesting but JUST A BIT MORE DETAIL perhaps?

Personally, I like it how it is. I might actually increase the glare just a bit. Not every region of a photo needs more detail, sometimes lower detail and less contrast is exactly what you want.

It's a lovely shot and one I really like...but that sky is blown out. An ND grad would render the sky darker and probably lost detail in the darker sky areas. It would have increased contrast where it wasn't wanted. The only way to have fixed this here is to have taken a 2nd photograph but at a 2-3 stop darker exposure and then blended the highlight areas carefully in photoshop using a layer. Shadows can be pulled but clipped highlights are not recoverable. It's also important to render the sky brighter than the foreground, another error I regularly see where ND grads are employed. If an ND grad was used with the above photo, the sky would have been darker than the foreground and wouldn't look right.

Many people here are talking about the D800's extra DR, but the truth is that it's only in the shadows...or rather it's the push-ability in the shadows during post production with low iso noise is really what is being talked about. Highlight clipping / blown highlights occurs at pretty much the same between the 5DIII and D800. So it's not really any extra DR, just better iso thresholds in the deep black areas.

You wouldn't consider the blown out sky to be artistically desirable? Personally, I think the sky is exactly how it should be (maybe even more blown out, covering more of the corner, would be even better.) With our own eyes, we couldn't see this scene without the sky being much brighter than it is depicted here in this photo. I think it speaks to realism that V8 captured it the way he did, and I think it maintains a certain artistic flare.

Not everything in photography is nor should be about recovering every last scrap of detail, so much so that you could see sunspots in the glaring sun...I think that would result in a grave imbalance in a shot like this.

 

[Niki]

<div name=\"googleone_share_1\" style=\"position:relative;z-index:5;float: right; /*margin: 70px 0 0 0;*/ top:70px; right:120px; width:0;\"><glusone size=\"tall\" count=\"1\" href=\"http://www.canonrumors.com/?p=15482\"></glusone></div><div style=\"float: right; margin:0 0 70px 70px;\"><a href=\"https://twitter.com/share\" class=\"twitter-share-button\" data-count=\"vertical\" data-url=\"http://www.canonrumors.com/?p=15482\">Tweet</a></div>
<p>Nikon’s extremely vague development announcement of the <a href=\"http://nikonrumors.com/2014/01/06/nikon-announces-the-development-of-the-nikon-d4s-camera.aspx/\" target=\"_blank\">D4s body at CES 2014</a> has sparked the wrath of Canonites wondering when Canon is going to move beyond <a href=\"http://www.canonrumors.com/2014/01/canon-powershot-n100-official/\" target=\"_blank\">selfie technology based cameras</a>.</p>
<p>Lets be clear first, Nikon hasn’t actually released any solid specs or descriptions of the technology in the new body.</p>
<p>We’re told by a longtime source that Canon is indeed still in the game and has some “groundbreaking” camera bodies coming in 2014. Canon will take a different approach at the Sochi games and have test bodies out there without the development announcement. Canon plans to make a “big splash” at the World Cup in Brazil in the spring.</p>
<p>An array of lenses an 3 prosumer/professional DSLR’s are coming in 2014. Along with a host of Cinema EOS products in April.</p>
<p>Patience appears to be the key for the Canon photographer….</p>
<p><strong><span style=\"color: #ff0000;\">c</span>r</strong></p>

D4s…so??

 

[Sporgon]

Surely that picture has been filled, either by flash or some reflector ?

I don't see how it can be an example of DR

 

[Loswr]

For ISO 160, Canon does a downstream third-stop "pull", which again costs you a third of a stop dynamic range (crushing blacks). Since it is a post-read push, it reduces read noise by a third of a stop (hence, the notion that ISO 160 is "cleaner" than ISO 100...it is, by a minuscule amount.)

Bill Claff's data show the 'jagged' relationship of read noise vs. ISO, where the 'valleys' of noise are 160-320-640-. But, his data show the same jagged plot for DR vs. ISO, except the 'peaks' are the inverse of noise, with the greatest DR at 160-320-640. Your statement of a loss of 1/3-stop ISO at 'pulled' settings doesn't jive with his data that ISO 320 has greater DR than ISO 400 as well as less noise.

You express certainty that ISO 100 is the true base, but I think a model where the real base ISO was somewhat lower than 100, such that multiples of 100 are a light push, multiples of 125 are a harder push, and multiples of 160 are a light pull, would explain Claff's data.

Any idea what's going on there?

I am not exactly sure how Bill Claff derives his data. Logically, I am unsure how one could gain dynamic range by shifting the signal after the pixels are read. Think of what Canon does as a digital post-process exposure boost. If you utilize 100% of the sensor's available dynamic range at ISO 400, you have information at every level from 0 through 65535. If you "push", you shift the whole histogram to the right...and clip the highlights. If you "pull", you shift the whole histogram to the left...and clip the shadows. Now, in post, your probably working with a tool like Lightroom, in which case your actual RAW data is never changed, and you can shift the digital signal around to your hearts content without loss of precision or data. However with third-stop ISO settings in a Canon camera, the results after the push or pull are baked in. If you push, anything that clipped past pure white is permanently pure white. You cannot recover it. Similarly, if you pull, anything clipped past pure black is permanently pure black.

I'm wondering if the apparent reduction in the read noise floor is what gives Claff's results this third-stop "pull" boost to DR. If he is ignoring the fact that the highlights shift down along with the rest of the exposure, and simply uses a the 5D III FWC as a constant for the upper limit for the signal, then I can see how dynamic range would theoretically increase. The read noise floor is indeed lower, however because of the pull, you actually destroyed data in the shadows, and the camera itself is not actually utilizing the headroom gained by shifting highlights down (since this all occurs AFTER the sensor has been read)...hence the "loss" of a third of a stop dynamic range.

Yes, I get the effects of in-camera push and pull. Where does your certainty that ISO 100 is the true base come from? As I stated, Claff's data could be explained if base ISO is slightly lower than 100. On the ML forums, one if their devs determined that using ISO 100 multiples with 'a little bit of negative gain' yielded the highest DR, and that's also consistent with base ISO being less than 100.

 

[Loswr]

Music fans want platinum sales, which is ridiculous as they pretty much guarantee a bad album the next time.

Ok, except that Canon has led the market every year for 10 years. So success didn't engender failure, at least from a sales perspective.

As a Mac user I really liked the fact that Macs weren't that common - 0 risk of viruses and so on. Now that everyone has a Mac, it's most likely just a matter of time...

I guess I'm clinging to the hope that the Unix kernel will protect us...

That means that instead of great stuff, they sell cost competitive mediocrity at the highest price they can. Well, at least I can somewhat afford the hobby, but the perfect lens will never be released.

Sad, but true.

 

[jrista]

For ISO 160, Canon does a downstream third-stop "pull", which again costs you a third of a stop dynamic range (crushing blacks). Since it is a post-read push, it reduces read noise by a third of a stop (hence, the notion that ISO 160 is "cleaner" than ISO 100...it is, by a minuscule amount.)

Bill Claff's data show the 'jagged' relationship of read noise vs. ISO, where the 'valleys' of noise are 160-320-640-. But, his data show the same jagged plot for DR vs. ISO, except the 'peaks' are the inverse of noise, with the greatest DR at 160-320-640. Your statement of a loss of 1/3-stop ISO at 'pulled' settings doesn't jive with his data that ISO 320 has greater DR than ISO 400 as well as less noise.

You express certainty that ISO 100 is the true base, but I think a model where the real base ISO was somewhat lower than 100, such that multiples of 100 are a light push, multiples of 125 are a harder push, and multiples of 160 are a light pull, would explain Claff's data.

Any idea what's going on there?

I am not exactly sure how Bill Claff derives his data. Logically, I am unsure how one could gain dynamic range by shifting the signal after the pixels are read. Think of what Canon does as a digital post-process exposure boost. If you utilize 100% of the sensor's available dynamic range at ISO 400, you have information at every level from 0 through 65535. If you "push", you shift the whole histogram to the right...and clip the highlights. If you "pull", you shift the whole histogram to the left...and clip the shadows. Now, in post, your probably working with a tool like Lightroom, in which case your actual RAW data is never changed, and you can shift the digital signal around to your hearts content without loss of precision or data. However with third-stop ISO settings in a Canon camera, the results after the push or pull are baked in. If you push, anything that clipped past pure white is permanently pure white. You cannot recover it. Similarly, if you pull, anything clipped past pure black is permanently pure black.

I'm wondering if the apparent reduction in the read noise floor is what gives Claff's results this third-stop "pull" boost to DR. If he is ignoring the fact that the highlights shift down along with the rest of the exposure, and simply uses a the 5D III FWC as a constant for the upper limit for the signal, then I can see how dynamic range would theoretically increase. The read noise floor is indeed lower, however because of the pull, you actually destroyed data in the shadows, and the camera itself is not actually utilizing the headroom gained by shifting highlights down (since this all occurs AFTER the sensor has been read)...hence the "loss" of a third of a stop dynamic range.

Yes, I get the effects of in-camera push and pull. Where does your certainty that ISO 100 is the true base come from? As I stated, Claff's data could be explained if base ISO is slightly lower than 100. On the ML forums, one if their devs determined that using ISO 100 multiples with 'a little bit of negative gain' yielded the highest DR, and that's also consistent with base ISO being less than 100.

Well, certainly. Literal ISO is always different. It isn't even consistent within the same brand, though. For example, according to DXO, "measured" ISO 100 on the 5D III is 80, measured ISO 200 is 160, where as "measured" ISO 100 on the 70D is 93 and ISO 200 is 160. The 6D measures 80 for ISO 100, but 153 for ISO 200.

My point was that when you select ISO 100 on a Canon camera, you have selected the cameras actual native base ISO (whether it is 80 or 93 or 100 in actuality under the scenes.) Whether the measured ISO is 80 or 100 doesn't change anything, though. Instead of ISO 125 for the first 1/3rd stop push, you would have a measured ISO 105, and for the first 1/3rd stop pull you would have a measured ISO 140. Same difference in the end...you are clipping either blacks or whites, and baking in that loss of information into the final output that is actually recorded into the RAW file.

 

[Loswr]

Well, certainly. Literal ISO is always different. It isn't even consistent within the same brand, though. For example, according to DXO, "measured" ISO 100 on the 5D III is 80, measured ISO 200 is 160, where as "measured" ISO 100 on the 70D is 93 and ISO 200 is 160. The 6D measures 80 for ISO 100, but 153 for ISO 200.

My point was that when you select ISO 100 on a Canon camera, you have selected the cameras actual native base ISO (whether it is 80 or 93 or 100 in actuality under the scenes.) Whether the measured ISO is 80 or 100 doesn't change anything, though. Instead of ISO 125 for the first 1/3rd stop push, you would have a measured ISO 105, and for the first 1/3rd stop pull you would have a measured ISO 140. Same difference in the end...you are clipping either blacks or whites, and baking in that loss of information into the final output that is actually recorded into the RAW file.

So DxO measures an ISO less than 100 when ISO 100 is selected. But DxO only measures full stops. It would seem they have the capability to measure 1/3-stops, but they don't. Perhaps it's just too much work. Or perhaps they do measure them, use them in Optics Pro, but don't publish them. I wonder, if they actually measured ISO 160, perhaps the real measurement would be closer to the setting. Canon's internal processing isn't necessarily as one would expect, as shown by the clandestine ISO boost with very wide aperture lenses.

 

[jrista]

Well, certainly. Literal ISO is always different. It isn't even consistent within the same brand, though. For example, according to DXO, "measured" ISO 100 on the 5D III is 80, measured ISO 200 is 160, where as "measured" ISO 100 on the 70D is 93 and ISO 200 is 160. The 6D measures 80 for ISO 100, but 153 for ISO 200.

My point was that when you select ISO 100 on a Canon camera, you have selected the cameras actual native base ISO (whether it is 80 or 93 or 100 in actuality under the scenes.) Whether the measured ISO is 80 or 100 doesn't change anything, though. Instead of ISO 125 for the first 1/3rd stop push, you would have a measured ISO 105, and for the first 1/3rd stop pull you would have a measured ISO 140. Same difference in the end...you are clipping either blacks or whites, and baking in that loss of information into the final output that is actually recorded into the RAW file.

So DxO measures an ISO less than 100 when ISO 100 is selected. But DxO only measures full stops. It would seem they have the capability to measure 1/3-stops, but they don't. Perhaps it's just too much work. Or perhaps they do measure them, use them in Optics Pro, but don't publish them. I wonder, if they actually measured ISO 160, perhaps the real measurement would be closer to the setting. Canon's internal processing isn't necessarily as one would expect, as shown by the clandestine ISO boost with very wide aperture lenses.

I don't know why DXO doesn't measure third stops. Or half stops, for that matter. Either way, I don't really think it matters. All I know is that the way Canon manages their non-full-stop ISO settings irks me. I wish they would just use electronic gain at the pixel for every ISO setting (up until those very high ISO settings...seems whatever they do at very very high ISO results in better output than simple gain off the pixel.)

Anyway, I don't suspect Canon will be changing their MO any time soon. Even if they improve DR, I expect to be saddled with quirky third-stops for a long time to come (however they work, whatever their consequences.)

 

[Loswr]

I don't know why DXO doesn't measure third stops. Or half stops, for that matter. Either way, I don't really think it matters. All I know is that the way Canon manages their non-full-stop ISO settings irks me. I wish they would just use electronic gain at the pixel for every ISO setting (up until those very high ISO settings...seems whatever they do at very very high ISO results in better output than simple gain off the pixel.)

Anyway, I don't suspect Canon will be changing their MO any time soon. Even if they improve DR, I expect to be saddled with quirky third-stops for a long time to come (however they work, whatever their consequences.)

If it really bothers you that much, you could always get a 1D X. It would go great with your 600/4L IS II...

 

[jrista]

I don't know why DXO doesn't measure third stops. Or half stops, for that matter. Either way, I don't really think it matters. All I know is that the way Canon manages their non-full-stop ISO settings irks me. I wish they would just use electronic gain at the pixel for every ISO setting (up until those very high ISO settings...seems whatever they do at very very high ISO results in better output than simple gain off the pixel.)

Anyway, I don't suspect Canon will be changing their MO any time soon. Even if they improve DR, I expect to be saddled with quirky third-stops for a long time to come (however they work, whatever their consequences.)

If it really bothers you that much, you could always get a 1D X. It would go great with your 600/4L IS II...

Aye! I know! I wish I could afford it...but I have too many things I want to buy. At the top of my list right now is a Celestron EdgeHD DX 1100 SCT (about $4000) and the 5D III (seems to run under $3000 now). I could get both of those for about the cost of a single 1D X...and both of those would be more useful to me in a much broader range of situations than just the 1D X.

 

[Orangutan]

At the top of my list right now is a Celestron EdgeHD DX 1100 SCT (about $4000) and the 5D III (seems to run under $3000 now). I could get both of those for about the cost of a single 1D X...and both of those would be more useful to me in a much broader range of situations than just the 1D X.

Where do you live? Maybe we could split costs and share.

 

[jrista]

At the top of my list right now is a Celestron EdgeHD DX 1100 SCT (about $4000) and the 5D III (seems to run under $3000 now). I could get both of those for about the cost of a single 1D X...and both of those would be more useful to me in a much broader range of situations than just the 1D X.

Where do you live? Maybe we could split costs and share.

;D I live in Colorado. It's actually a fairly decent place for astronomy/astrophotography, as it is rather arid, so few problems with humidity screwing with a telescope. (I think on the best of days we might have 30% humidity, and on most days it is way down around 16%. There have been times when we've had less than 10% humidity.) We have mountains, too, so you can get up to 11,000 feet or more for thinner, clearer air.