Dynamic Range - Try it for yourself, conclude for yourself: 5D III vs. A7r

[DominoDude]

Maybe Canon is addressing this thread here:
http://www.seeimpossible.usa.canon.com/

Hmm. Canon only sees "impossible", eh? Does that mean, they see that making a better sensor is impossible for them, so they aren't bothering? Or, they see impossible, and intend to make the impossible possible?

Eh, I'll believe Canon has improved their sensors when I see it.

Could it be that it relates to how they see the chance of, finally, convincing you that they can make good sensors to put in their already more than good bodies? (just teasing you and pulling your leg, jrista)

 

[Marsu42]

Check out this real world use of the D750 and it's dynamic range... Don't think Canon can do that unfortunately...

Think again, I'm shooting these sunset scenes all the time using Magic Lantern - your Canon has greater dynamic range than any Nikon or Sony and outputs 16bit raw files. It's a usability hassle and Canon is certainly behind Nikon in terms of on-sensor dr, but if people are so set upon not using the available software fix the need cannot be that bad.

 

[Eldar]

Think again, I'm shooting these sunset scenes all the time using Magic Lantern - your Canon has greater dynamic range than any Nikon or Sony and outputs 16bit raw files. It's a usability hassle and Canon is certainly behind Nikon in terms of on-sensor dr, but if people are so set upon not using the available software fix the need cannot be that bad.

I have never tried ML´s software/frimware before, but I think I would like to. But I do not want more problems than necessary. When I go to ML´s web site, I do not find any support for the 5DIII, but I believe that is the camera you´re using. Is that correct? If so, how do you go about getting the thing installed?

 

[DominoDude]

Think again, I'm shooting these sunset scenes all the time using Magic Lantern - your Canon has greater dynamic range than any Nikon or Sony and outputs 16bit raw files. It's a usability hassle and Canon is certainly behind Nikon in terms of on-sensor dr, but if people are so set upon not using the available software fix the need cannot be that bad.

I have never tried ML´s software/frimware before, but I think I would like to. But I do not want more problems than necessary. When I go to ML´s web site, I do not find any support for the 5DIII, but I believe that is the camera you´re using. Is that correct? If so, how do you go about getting the thing installed?

Jumping into debate:
http://builds.magiclantern.fm/#/ In "Select platform" you should find your body.
Download latest zip and unpack it. Format a card in your camera, and then transfer the files to the root of the card. (2 files + 1 directory)

 

[Marsu42]

I have never tried ML´s software/frimware before, but I think I would like to. But I do not want more problems than necessary. When I go to ML´s web site, I do not find any support for the 5DIII, but I believe that is the camera you´re using. Is that correct? If so, how do you go about getting the thing installed?

Nope, I've got a 6d, but the 5d3 is much better supported by ML. The website is outdated, look at their forum, there's a good installation thread: http://magiclantern.fm/forum/

Basically it's downloading ML "rolling release" for your camera (http://builds.magiclantern.fm/#/), put the files on your cf card, update the firmware with Magic Lanter's mini-firmware, done. ML runs from the cf card, the fw update only tells the camera to enable loading it.

Beware though, ML has many features and it'll take you some time to sort through them - what you want is the "dual_iso" module.

 

[Sporgon]

I have never tried ML´s software/frimware before, but I think I would like to. But I do not want more problems than necessary. When I go to ML´s web site, I do not find any support for the 5DIII, but I believe that is the camera you´re using. Is that correct? If so, how do you go about getting the thing installed?

Nope, I've got a 6d, but the 5d3 is much better supported by ML. The website is outdated, look at their forum, there's a good installation thread: http://magiclantern.fm/forum/

Basically it's downloading ML "rolling release" for your camera (http://builds.magiclantern.fm/#/), put the files on your cf card, update the firmware with Magic Lanter's mini-firmware, done. ML runs from the cf card, the fw update only tells the camera to enable loading it.

Beware though, ML has many features and it'll take you some time to sort through them - what you want is the "dual_iso" module.

If ML are doing this with normal sensors what would Canon be able to do with dual pixel technology ?

 

[Marsu42]

If ML are doing this with normal sensors what would Canon be able to do with dual pixel technology ?

The ML people are wondering, too, if dual pixel can be harnessed for a similar feature.

Most likely Canon doesn't do it because it's enterprising and admittedly rather hackish - the shots in-camera are interlaced & have screwed up wb, then you have to run a cpu-intensive intermediary step cr2->dng before postprocessing ... probably cannot be done in camera with sufficient quality.

 

[lintoni]

Think again, I'm shooting these sunset scenes all the time using Magic Lantern - your Canon has greater dynamic range than any Nikon or Sony and outputs 16bit raw files. It's a usability hassle and Canon is certainly behind Nikon in terms of on-sensor dr, but if people are so set upon not using the available software fix the need cannot be that bad.

;D There's a nice English idiom that goes something like "Now that's set the cat among the pigeons!"

 

[gruhl28]

This thread is probably too long already, and I'm not sure if anyone has already made this point (I haven't seen it made in any other threads on this topic). I'm not denying that having extra dynamic range is nice, but think about all the beautiful pictures taken on Fuji Velvia and the fact that Velvia has only a few stops of dynamic range, far less than even the "worst" Canon available. Again, I'm not saying that there aren't times when I'd like more dynamic range or that it isn't an advantage, just pointing out that some of the most fantastic photos ever taken were done with something that had far less dynamic range than any current DSLR.

Thanks, jrista, for the photos.

A similar point has been made. The standard reply is not to deny it, but to say that auto-focus, smart metering, high ISO IQ and high framerate have all contributed to higher "keeper rates." Also, high pixel count allows a bit more cropping, and PP software has made the job of "developing" your capture quicker and easier.

It's entirely true that great shots have been taken with simpler gear, but that does not at all detract from the legitimate desire for even better tech to make success more likely and less work to achieve.

I agree, just making the point that one can get carried away with always wanting something better and forget that terrific stuff was done with equipment that wasn't near the standards available today.

 

[gruhl28]

This thread is probably too long already, and I'm not sure if anyone has already made this point (I haven't seen it made in any other threads on this topic). I'm not denying that having extra dynamic range is nice, but think about all the beautiful pictures taken on Fuji Velvia and the fact that Velvia has only a few stops of dynamic range, far less than even the "worst" Canon available. Again, I'm not saying that there aren't times when I'd like more dynamic range or that it isn't an advantage, just pointing out that some of the most fantastic photos ever taken were done with something that had far less dynamic range than any current DSLR.

Thanks, jrista, for the photos.

Welcome.

One thing to note about Velvia 50 (which is one of the the key ISO used by 4x5 positive (slide) film landscape photographers) is it actually has more dynamic range than it's given credit for. When visually observed on a light table, it might exhibit about 5 stops or so of dynamic range. A lot of 4x5 shooters drum scan their Velvia 50 slides these days, and when they do, they get well over 110mp worth of resolution out of them. Drum scans also pull out FAR more detail from the shadows than you generally see with the naked eye, and a lot more than what a flat bed scanner can pull out as well. Downsample a 110mp image to 22mp, and the dynamic range would trounce the 5D III (and probably even trounce the D810). Any visible noise is going to be very clean and random as well, and Velvia (as well as Provia, which from what I understand from those who still regularly shoot 4x5 film landscapes, is actually better for drum scanning as it has about a stop more DR.)

Hi jrista,
If you don't mind, can you explain the relationship between downsampling and dynamic range? I don't recall ever hearing that downsampling can improve dynamic range, not sure I understand how that works. I could imagine that scanning with different intensities of light could pull more DR out of a slide than you would see looking at it on a light table or projected at one intensity, but I don't understand how the downsampling helps. I guess it would hide some grain, but anything that's blown out would be gone.

 

[Loswr]

This thread is probably too long already, and I'm not sure if anyone has already made this point (I haven't seen it made in any other threads on this topic). I'm not denying that having extra dynamic range is nice, but think about all the beautiful pictures taken on Fuji Velvia and the fact that Velvia has only a few stops of dynamic range, far less than even the "worst" Canon available. Again, I'm not saying that there aren't times when I'd like more dynamic range or that it isn't an advantage, just pointing out that some of the most fantastic photos ever taken were done with something that had far less dynamic range than any current DSLR.

Thanks, jrista, for the photos.

Welcome.

One thing to note about Velvia 50 (which is one of the the key ISO used by 4x5 positive (slide) film landscape photographers) is it actually has more dynamic range than it's given credit for. When visually observed on a light table, it might exhibit about 5 stops or so of dynamic range. A lot of 4x5 shooters drum scan their Velvia 50 slides these days, and when they do, they get well over 110mp worth of resolution out of them. Drum scans also pull out FAR more detail from the shadows than you generally see with the naked eye, and a lot more than what a flat bed scanner can pull out as well. Downsample a 110mp image to 22mp, and the dynamic range would trounce the 5D III (and probably even trounce the D810). Any visible noise is going to be very clean and random as well, and Velvia (as well as Provia, which from what I understand from those who still regularly shoot 4x5 film landscapes, is actually better for drum scanning as it has about a stop more DR.)

Hi jrista,
If you don't mind, can you explain the relationship between downsampling and dynamic range? I don't recall ever hearing that downsampling can improve dynamic range, not sure I understand how that works. I could imagine that scanning with different intensities of light could pull more DR out of a slide than you would see looking at it on a light table or projected at one intensity, but I don't understand how the downsampling helps. I guess it would hide some grain, but anything that's blown out would be gone.

Yes, please do explain. It sounds suspiciously like what DxOMark does to calculate a DR Score that exceeds the bit depth of the ADC. How does downsampling increase the DR beyond the limits of the film emulsion used to capture the scene?

 

[Vgramatikov]

i was Nikon shooter with d90,d300 and now 7d,70d. Used all pro canon body gear too. Included 1dx and 5d3.
As a wildlife and nature photographer DR is really important but not as AF and speed.

So if you looking closely you will found that even Nikon d300 and especially d90 have better DR than latest Canon body`s. Before 4-5 years i have pulled much more shadows via my Nikon d90 NEFF file than now in my 70d raw`s.
But these days`s Nikon files have to wrong colors, tints, bad overall fine detail and so one. Not as good long optics and i switch to Canon. Now... i fell lucky cause for me Canon have greatest AF and speed cameras as 7d,70d,5d3 and now 7d2 and 1dx. So i feel good and on the right course.

DR is big deal indeed for landscape and portrait photography and available light photography.

ISO performance seems equal and the 7d2 may be again slightly better than d7100 for example. As my 70d is better or equal to d7100 on 800 nad 1600 iso`s. DR on that iso`s is equal to if not better in Canon`s.

So huge DR at base iso is indeed big deal for those who are enable to use these settings. I almost never shoot below 400 iso on actions scenes and if i do 100 iso is usually static and i have time to get proper exposure. So i really don`t care for DR at base iso`s.

I think if i was generally landscape photographer for sure 610,d800/E/810 are best sensors and system available.(Made by Sony).

So i do not thing that 3 EV or much much less shadow noise are not important. Indeed this will save your time ,improve your final quality, make you flexible and sometimes may be save your ass from the field mistakes.

We speak about 2-3 EV difference here witch is already not minor difference. I say this is major one of somebody really needed.

Peace. There are better sensors, better systems, better lenses but only if you know how to use them.

 

[LetTheRightLensIn]

Well I have to say I sure as heck could've used more DR the last few days. On a fast pace, with others some of whom were not into photograph, tons of ground to cover, often fast changing and soon fading light, so much to see, even when multi-shot tripod HDR would've worked only had time to set up the tripod once. So it's probably gonna be mostly a mess for the extreme DR woods scenes I took. Knew I shoulda rented the A7R for this week. Maybe some never need it (other than to help fix the rare mistake and maybe a few are in a studio and work at a slow pace and never have a mistake since even the mistakes can be re-shot) and some only rarely, but it is simply disingenuous to claim that nobody can ever need it or that such scenarios are nearly impossible to find or that 2-3 more stops wouldn't do them any good anyway. I could've made great use of Exmor a few dozen times today alone. People don't just bring this up over some lab tests or for DxO crowing, people bring it up for REAL WORLD SHOOTING.
Of course my 5D3 handles tons of shots well I'm sure, but man having that Exmor performance for ISO100-400 sure as heck would be nice at times. I mean I really, truly could've made good use and hit quite a few shots the last few days where just 3 more stops at low ISO would've made a realistic difference for sure.

 

[LetTheRightLensIn]

So, dynamic range is relative to noise. Downsampling averages pixels together, thus reducing noise by the square root of the number of source pixels averaged into each output pixel. This isn't exactly going to improve your editing latitude...as to downsample you have to convert to a non-RAW format like TIFF. That takes totally independent source channels and combines them together into RGB subpixels for each output pixel. Once the independent source channels are combined like that, you lose a fair amount of editing latitude. The primary benefit of downsampling is generally going to be after you've made your exposure adjustments. The more you downsample, the lower the noise in your final output.

Downsampling a 36.3mp image to 22.3mp has a pretty big impact on the smoothness of shadow tones (you can see this for yourself with the four RAW images I've shared in the first post of this thread.) The improvement is around 30% (SQRT(36.3)/SQRT(22.3) = 1.28x). If you scan a 4x5" film slide such that you ended up with a 90mp image, you could reduce noise by a factor of 2x (or by 100%, reduce it in half.) A 4x5" slide can be scanned as a higher resolution than that...up to maybe 130mp, which is about 6x the megapixel count of a 5D III. Donwsample that image, and you reduce noise by a factor of 2.45x.

You wouldn't be able to lift shadows by another six stops after downsampling like that (not without some serious technique...you might be able to if you used a tool like PixInsight, but it would be a LOT of work, which kind of negates the benefit as far as editing latitude goes), however the expectation is that you do all your lifting before downsampling. The noise improvement you get from downsampling is really just a bonus. I figure there is about a 30% improvement in noise when downsampling A7r images to 5D III size. The A7r looks a little bit better than the 5D III before downsampling, but after downsampling is when you really obviously SEE the differences. A 145% improvement would be nothing short of RADICAL.

Hah, so finally after like 30 decades of my trying to get this across and your calling DxO Print DR and my claims and everyone else's so much nonsense or worse, finally you get the camera and rent and see DxO and some of us were right all along. ;D

Whew, at last.

Anyway at last you have fully come around on all points I and a few others have been trying to make for years. ;D

 

[LetTheRightLensIn]

This thread is probably too long already, and I'm not sure if anyone has already made this point (I haven't seen it made in any other threads on this topic). I'm not denying that having extra dynamic range is nice, but think about all the beautiful pictures taken on Fuji Velvia and the fact that Velvia has only a few stops of dynamic range, far less than even the "worst" Canon available. Again, I'm not saying that there aren't times when I'd like more dynamic range or that it isn't an advantage, just pointing out that some of the most fantastic photos ever taken were done with something that had far less dynamic range than any current DSLR.

Thanks, jrista, for the photos.

Welcome.

One thing to note about Velvia 50 (which is one of the the key ISO used by 4x5 positive (slide) film landscape photographers) is it actually has more dynamic range than it's given credit for. When visually observed on a light table, it might exhibit about 5 stops or so of dynamic range. A lot of 4x5 shooters drum scan their Velvia 50 slides these days, and when they do, they get well over 110mp worth of resolution out of them. Drum scans also pull out FAR more detail from the shadows than you generally see with the naked eye, and a lot more than what a flat bed scanner can pull out as well. Downsample a 110mp image to 22mp, and the dynamic range would trounce the 5D III (and probably even trounce the D810). Any visible noise is going to be very clean and random as well, and Velvia (as well as Provia, which from what I understand from those who still regularly shoot 4x5 film landscapes, is actually better for drum scanning as it has about a stop more DR.)

Hi jrista,
If you don't mind, can you explain the relationship between downsampling and dynamic range? I don't recall ever hearing that downsampling can improve dynamic range, not sure I understand how that works. I could imagine that scanning with different intensities of light could pull more DR out of a slide than you would see looking at it on a light table or projected at one intensity, but I don't understand how the downsampling helps. I guess it would hide some grain, but anything that's blown out would be gone.

Yes, please do explain. It sounds suspiciously like what DxOMark does to calculate a DR Score that exceeds the bit depth of the ADC. How does downsampling increase the DR beyond the limits of the film emulsion used to capture the scene?

Because it is calculated from max well capacity value/max brightness level (which stays constant during downsampling (or upsampling for that matter) and read noise and we are just trying to compare on a normalized basis comparing detail and noise at the same energy scale. So the max value stays the same and the read noise lowers (of course as you filter away some of the read noise, you filter away higher frequency signal down there.)

When they measure mid-tone SNR they basically just measure noise about a solid block of gray (or maybe green for one channel) and the color is what is should be and that is scale invariant and as you downsample and combine the noise about the scale invariant number you want goes down.

In the most general case, we filter away high frequency noise so noise goes down, of course high frequency signal also gets filtered away bt a 12MP sensor would've even collect any of the higher frequency signal that a 36MP sensor could collect so we are filtering away actual detail down to a 12MP signal and also reducing noise to that scale and then we can fairly compare the noise between them at the same scale.

It's just normalizing to the same scale so you assume 8MP level of detail and then compare how much noise at that signal scale and you filter all signal detail and noise down to the same 8MP level and then compare.

 

[Loswr]

This thread is probably too long already, and I'm not sure if anyone has already made this point (I haven't seen it made in any other threads on this topic). I'm not denying that having extra dynamic range is nice, but think about all the beautiful pictures taken on Fuji Velvia and the fact that Velvia has only a few stops of dynamic range, far less than even the "worst" Canon available. Again, I'm not saying that there aren't times when I'd like more dynamic range or that it isn't an advantage, just pointing out that some of the most fantastic photos ever taken were done with something that had far less dynamic range than any current DSLR.

Thanks, jrista, for the photos.

Welcome.

One thing to note about Velvia 50 (which is one of the the key ISO used by 4x5 positive (slide) film landscape photographers) is it actually has more dynamic range than it's given credit for. When visually observed on a light table, it might exhibit about 5 stops or so of dynamic range. A lot of 4x5 shooters drum scan their Velvia 50 slides these days, and when they do, they get well over 110mp worth of resolution out of them. Drum scans also pull out FAR more detail from the shadows than you generally see with the naked eye, and a lot more than what a flat bed scanner can pull out as well. Downsample a 110mp image to 22mp, and the dynamic range would trounce the 5D III (and probably even trounce the D810). Any visible noise is going to be very clean and random as well, and Velvia (as well as Provia, which from what I understand from those who still regularly shoot 4x5 film landscapes, is actually better for drum scanning as it has about a stop more DR.)

Hi jrista,
If you don't mind, can you explain the relationship between downsampling and dynamic range? I don't recall ever hearing that downsampling can improve dynamic range, not sure I understand how that works. I could imagine that scanning with different intensities of light could pull more DR out of a slide than you would see looking at it on a light table or projected at one intensity, but I don't understand how the downsampling helps. I guess it would hide some grain, but anything that's blown out would be gone.

Yes, please do explain. It sounds suspiciously like what DxOMark does to calculate a DR Score that exceeds the bit depth of the ADC. How does downsampling increase the DR beyond the limits of the film emulsion used to capture the scene?

Because it is calculated from max well capacity value/max brightness level (which stays constant during downsampling (or upsampling for that matter) and read noise and we are just trying to compare on a normalized basis comparing detail and noise at the same energy scale. So the max value stays the same and the read noise lowers (of course as you filter away some of the read noise, you filter away higher frequency signal down there.)

When they measure mid-tone SNR they basically just measure noise about a solid block of gray (or maybe green for one channel) and the color is what is should be and that is scale invariant and as you downsample and combine the noise about the scale invariant number you want goes down.

In the most general case, we filter away high frequency noise so noise goes down, of course high frequency signal also gets filtered away bt a 12MP sensor would've even collect any of the higher frequency signal that a 36MP sensor could collect so we are filtering away actual detail down to a 12MP signal and also reducing noise to that scale and then we can fairly compare the noise between them at the same scale.

It's just normalizing to the same scale so you assume 8MP level of detail and then compare how much noise at that signal scale and you filter all signal detail and noise down to the same 8MP level and then compare.

I get the normalize to compare idea. But if you take a 'sensor' with 5 stops of DR (Velvia was being discussed), and take a picture of a scene with 15 stops of DR, you've lost 10 stops of DR at capture - you have parts of the scene with saturated data, and parts with zero data outside of noise/gain - clipped highlights and blocked shadows. Same idea if you have a sensor with 13 stops of DR and a 15 stop scene - you've lost 2 stops of information. Now, when you downsample that 130 MP drum scan or that 36 MP image file to 22 MP, do you get the data from the blocked shadows and clipped highlights back?

If you buy a stalk of celery at the grocery store and cut it into 5 pieces, you can later cut it into 20 pieces...but that won't get you the celery root or the leafy greens for your stock.

 

[Lee Jay]

Well I have to say I sure as heck could've used more DR the last few days. On a fast pace, with others some of whom were not into photograph, tons of ground to cover, often fast changing and soon fading light, so much to see, even when multi-shot tripod HDR would've worked only had time to set up the tripod once.

You don't need a tripod to shoot an HDR. In fact, you don't even need one second. Just shoot a bracketed burst, and use software to obtain perfect alignment in post.

 

[StatisticsRule]

I get the normalize to compare idea. But if you take a 'sensor' with 5 stops of DR (Velvia was being discussed), and take a picture of a scene with 15 stops of DR, you've lost 10 stops of DR at capture - you have parts of the scene with saturated data, and parts with zero data outside of noise/gain - clipped highlights and blocked shadows. Same idea if you have a sensor with 13 stops of DR and a 15 stop scene - you've lost 2 stops of information. Now, when you downsample that 130 MP drum scan or that 36 MP image file to 22 MP, do you get the data from the blocked shadows and clipped highlights back?

If you buy a stalk of celery at the grocery store and cut it into 5 pieces, you can later cut it into 20 pieces...but that won't get you the celery root or the leafy greens for your stock.

Actually, the data is not completely lost like it is in your celery example.

Consider a set of data divided into three parts. The first part has a (normalized) value of 1. The second part is 11 stops below that, but not exactly zero. The third section has a value exactly zero. Put some random noise (i.e. photon noise or read noise) on top of that. Then "digitize" the result to 10 stops. In other words, truncate the result so you have data with 10 bit numbers for each pixel.

If you average over enough pixels you will find that you can distinguish the low signal section from the section which is exactly zero. This is because the mean values of the two sections will be different. The data is not in any one pixel, but statistically is spread across many pixels.

When you are downsampling, you are averaging, and that helps recover that statistical information. If the noise is random and obeys ergodicity, then your signal-to-noise will improve as the square-root of the number of pixels averaged during downsampling [see signal averaging]. Average 4 pixels, and you gain a factor of two (1 bit) of extra information.

Keep in mind that the total information in the image does not change. In this case, you are trading off bit depth for spatial resolution. If you know your 16 megapixel image only has one bright and one dark region, your knowledge of those bright and dark values can be much more accurate than if you request each pixel in your original image to be a potentially different value. It is the space-bandwidth product which matters. [for very technical discussion, the following paywall-restricted paper is one example.]

In short, if you take a 16 megapixel image with 10 bits per pixel, you can downsample that to a 4 megapixel image with approximately* 11 bits of information per pixel.

[edit]
Just so there is no confusion later, the extra bit in the above example will come from improved shadows. You can't do much about clipped highlights, but there is more information in those noisy shadows than you might think.
[/edit]

---
* The approximation is because it depends on the noise and signal properties. In most cases of interest to this group, it is a reasonable approximation. This approximation also assumes your downsampling algorithm is using more than 11 bits to represent each number.

 

[sanj]

Well I have to say I sure as heck could've used more DR the last few days. On a fast pace, with others some of whom were not into photograph, tons of ground to cover, often fast changing and soon fading light, so much to see, even when multi-shot tripod HDR would've worked only had time to set up the tripod once.

You don't need a tripod to shoot an HDR. In fact, you don't even need one second. Just shoot a bracketed burst, and use software to obtain perfect alignment in post.

Unless of course the light is low or one wants absolute perfection.

 

[privatebydesign]

Well I have to say I sure as heck could've used more DR the last few days. On a fast pace, with others some of whom were not into photograph, tons of ground to cover, often fast changing and soon fading light, so much to see, even when multi-shot tripod HDR would've worked only had time to set up the tripod once.

You don't need a tripod to shoot an HDR. In fact, you don't even need one second. Just shoot a bracketed burst, and use software to obtain perfect alignment in post.

Unless of course the light is low or one wants absolute perfection.

If you want "absolute perfection", well the highest IQ we can currently attain within our personal budgets, you need to make a commitment to that, so many times this argument is brought up and it is a strawman. People spend years taking the time and effort to learn their equipment, post processing techniques, and getting the time and place just right for that "absolute perfection", now the argument is 'well I want to be able to attain IQ comparable to that at a fast pace while on vacation within an organized party with twenty other people', grow up people.

If you want great shots take the time and trouble to create great shots, no excuses.