80D dynamic range

[Dukai]

Hi,
I am having some issue with my Canon 80d. I bought the camera a few month ago, coming from a 600d (t3i) that got stolen. I am more and more convinced there must be an issue with the metering or sensor of some sort that's constantly blowing out the high lights on pictures where it shouldn't (I think)... At least I'm pretty sure my old 600d did a better job recording dynamic range and I find it weird.
For instance I have tried an experiment, I have just taken a scene's measure in high light and shadow using the spot measure and locking to iso100 at f5,6. The highest speed in the highlights gives me 1/1250 and 1/15 in the shadow, that's a bit less than 6 stops difference. Now if I go slower than 1/250, the highlights are blown out. At the same time, at 1/250 I lose tons of details in the shadow...
I don't have another camera to compare by taking the exact same shot with same lens, settings and scene so I'm not sure if I'm wrong. I'm not an expert so I'm not confident enough either to send the camera to repair or not... What do you guys think?
Could it be a hardware problem? Or am I completely wrong somewhere?
Thanks!
Ps English isn't my mother language, I hope you understand Smiley Happy

[TCampbell]

1/15 to 1/1250 is 7.25 stops.  You could do this as a RAW but it would be a struggle to capture that accurately with a JPEG.

Even so... that's mapping the brightest bits to almost pure white and the darkest bits to almost pure black and you might want more range.  I've done a 7 stop HDR image of an old covered bridge.  The exposure that pleasnatly exposed the outside landscape left the interior too black to see any detail in the internal wood contruction or beams.  The expousre that revealed nice details of the interior left the outside entirely too bright.  Technically the camera was capturing the whole dynamic range (all 7 stops) without clipping and I could verify that by looking at the historgram... it just didn't look as pleasing to the eye.  So I used HDR to capture the subject and let the HDR tone-mapping compress the dynamic range to something that looked more pleasing.

Are you shooting and storing the images as "RAW" files (.CR2)?

If you are shooting JPEG, all kinds of things can be happening.   Even DxO (whom I am suspicious of but that's another thread) claims the 80D has 13.2 stops of dynamic range.  So the camera can handle the range but that doesn't necessarily mean you'll like the result without a little extra tone-mapping.

In-camera spot metering is tricky.  When you select spot metering it only uses the center of the image (occasionally someone will think that it uses the same point as a the focus point when a single focus point is selected -- it does not.)  Also the area of that size depends on the angle of view of the lens you happen to be using.  

I have a high end Sekonic light meter that has a 1º spot meter (which really is 1º - that's a lot tighter than the camera can do).  To shoot a landscape, normally you'd meter the brightest point you can find in the scene, then the darkest point you can find in the scene, then set the exposure to be half-way between to capture the range (the Sekonic meter I own has a profiling tool that will help test the dynamic range of the camera and will alert me if I need to shoot an image that requries more DR then a single shot can provide.)  When you do in-camera metering, the spot metering may be selecting a large piece of image and it's averaging it wen reporting the meter reading.  In other words maybe it's sampling a 10º wide area... in which there are both some darker and brighter points, but when averaged out it reports a reading to you (but that doesn't mean that's really the brightest or darkest point in your scene.) 

BTW, you lose DR if you start bumping up the ISO.  You only get all of the camera's DR at ISO 100.

Have you looked at the histogram for your image?

Do you happen to have a 16-bit TIFF version of the image (created from the .CR2 using Canon's Digital Photo Professional software that came with your camera) that you can share?

[TTMartin]

---

@Dukai wrote:
Hi,
I am having some issue with my Canon 80d. I bought the camera a few month ago, coming from a 600d (t3i) that got stolen. I am more and more convinced there must be an issue with the metering or sensor of some sort that's constantly blowing out the high lights on pictures where it shouldn't (I think)... At least I'm pretty sure my old 600d did a better job recording dynamic range and I find it weird.
For instance I have tried an experiment, I have just taken a scene's measure in high light and shadow using the spot measure and locking to iso100 at f5,6. The highest speed in the highlights gives me 1/1250 and 1/15 in the shadow, that's a bit less than 6 stops difference. Now if I go slower than 1/250, the highlights are blown out. At the same time, at 1/250 I lose tons of details in the shadow...
I don't have another camera to compare by taking the exact same shot with same lens, settings and scene so I'm not sure if I'm wrong. I'm not an expert so I'm not confident enough either to send the camera to repair or not... What do you guys think?
Could it be a hardware problem? Or am I completely wrong somewhere?
Thanks!
Ps English isn't my mother language, I hope you understand Smiley Happy

---

Are you trying to do this in camera without post processing?

That should be no issue to bring up the shadows if you shoot RAW and do post processing.

If you are doing are trying to do it in camera be sure Auto Lighting Optimizer is on, and possible Highlight Tone Priority also.

[Dukai]

Thank you both for your comments and suggestions.

Are you sure it's 7,25 and not 6,25 stops?
I am usually shooting raw+jpeg but when I can refrain heavy retouching and spending time on the computer I try to do so. That's why when possible I prefer to have the best possible result straight from the camera. I was moving the spot meter around the scene (tried to find the darkest and brightest points in dark/bright areas to find them, as you said it might still evaluate a certain range of 10° or so, and take a medium evaluation, so exposition for the brightest could be brighter and darkest darker eventually, but I don't think it would make a big difference especially I had plain areas to take the measure).
With post processing I'm certainly able to bring back the shadows... But in this case, even exposing in the middl at around 1/250 just before blowing out any of the highlights, the non processed jpeg has a lot of totally dark areas and isn't looking too good... Plus when shooting at 100iso as here it's ok, but with higher iso, bringing the shadows up also brings lots of noise. The solution left is exposure bracketing/hdr I guess. I just feel that my 600d was giving me better jpeg in similar situations.
I have a question regarding the histogram tho. When you said you were able to capture the whole 7 stops and seeing no clipping to the histogram, you mean the histogram from the hdr jpeg? Is there a way to display the "raw histogram" on the camera? I mean the part that can be recovered from the raw file? Cause even tho I can recover shadows in post-processing, the histogram I see on the camera corresponds to the jpeg image and is clipping the right or left side in any case in the present situation (even exposing in the mid tones.
I will take new pictures illustrating this and send the 16bits tiff tomorrow (the pictures I took previously for test were taken in my room and it's a total mess, haha :). Tho if I get your explanation nothing seems wrong with my camera^^

Anyway thanks again for your help!

[BurnUnit]

This is an interesting bit of experimentation you're doing here. But since the camera's spot metering area is a bit wider than ideal, would it maybe be better to include a piece of pure white and pure black poster board in the scene you are shooting? Maybe place it near enough to the camera so that the 10* metering area is safely within the white and the black sample areas?

[TCampbell]

Let me see if I can walk you through this by example.

Here's a straight-out-of-the-camera image of an old covered bridge (forgive the horse dung... I crop that out later.)

You might be thinking this image lacks dynamic range... the interior of the bridge is clearly black, right?

Well not so fast... take a look at the histogram for this image:

A bit about histograms... the extreme left edge represents pixels which are completely black.   The extreme right edge represents pixels that are completely white.  Everything else is some level of tonality somewhere in the middle.  The hight of the graph at any point represents the number of pixels with that level of tonality.  

So this histogram indicates that nothing is completely white (the graph is has no height at all at the extreme right edge.)   It also has data very near the dark side... but it's not actually slammed into the left wall... the graph returns to to the zero level before it hits the left edge which means no pixels are completely black.

I'd need a video to demonstrate this... but as I hover my mouse around (I'm using Lightroom) at the very black pixels in the middle of hte interior of the covered bridge ... finding the darkest pixels I can find... I am struggling to find anything that reads less than an 8-9% light level.  So sure they're "dark"... but they aren't actually clipped pixels.

This means that in theory I should be able to grab the exposure adjustment and slide it up... and start to see detail emerge in the beams inside that bridge.  And if I do that... that is precisely what I see.  The wood deetail is there... it's just extremely dark.  So the camera does indeed have the dynamic range to handle the shot (this is shot with my 5D IV.)

So here's that VERY SAME image... but this time I went into the adjustments and drove up the "black" levels a little... and drove up the "shadows" adjustment very heavily (I've changed nothing else)... now have a look at the interior of the bridge.

And there's the detail in the wood beams on the inside of the bridge.

That's possible by just learning to use adjustment software while watching the histogram to while performing adjustments.  You can do this by just adjusting the blacks, whites, shadows, and highlights... or you can use the curves tool... or you can use the levels tool.  There are lots of ways to do it.  

But it's also possible to take photos when the true dynamic range exceeds what the camera can handle.... by using "HDR" techniques (HDR = High Dyanimic Range). 

In the old days, camera's didn't have built-in digital HDR features... so you shot "bracketed" images.  "Bracketed" just means you shoot the exposure that you think is the correct "middle" exposure for the shot... then take a shot which is deliberately over-exposed as well as a shot which is deliberately under-exposed.    Typically you'd do this by 2 stops for digital HDR work. Also you'd only change the shutter speed (because changing the aperture will change the depth of field and that affects focus and you want all three images to have identical focus).  

So if I take a shot at 1/250th... then 2 stops down (the darker image) will be shot at 1/1000th and the 2-stops brighter image will be shot at 1/60th.  I then use softare to combine these.

The HDR process takes the darkest parts of the over-exposed image... the brightest parts of the under-exposed image, and the middle parts of the middle image.  

It then performs a process called "tone mapping".  I don't want to just copy each pixel "as is".  I want to fine a scale of brightness by comparing that pixel across the images to determine how much to skew the value of light (tonality) for each source image when mapping it into the final output image (hence the name "tone mapping").

So here are my source images:

Here's an image several stops under-exposed...

For the above image, we really only care about the brightest bits.... we want details from the sky.

Here's my middle exposure...

This is the middle exposure.  The software will ignore the darkest bits (the interior of the bridge) and also ignores the brightest bits (anything bright in the sky, etc.) and will just grab data with a middle level of tonality.

And here is an image several stops over-exposed...

For this image, the computer is only interested in the darkest bits... in this case it will use the data from the interior of the bridge where you see the wood beams.

I combine these using HDR software (Lightroom, Photoshop, and many many programs have a feature built in to do this.  But the popular commercial product that specialized ... software designed *just* for the purpose of doing HDR ... is a program called "Photomatix Pro" by HDRsoft.

Here's a merged result:

Sure I've cropped, boosted color, and messed with lots of stuff... but the point is I've "compressed" the tonal range so that nothing is extremely dark (so dark the eye thinks it is black even when it isn't really black) and done the same for the lights.  

You have LOTS of tone-mapping choices when creating HDR images and many do not look realistic.  I've tried to make this one a little more realistic.  I want the interior "dark" (because it should look dark even though you clearly saw we have the data to brighten that up quite a lot if we want) ... I'm just trying to get some detail on the wood beam construction inside the bridge.  

I could have done this using just one image... but I did it both ways to show you.

The adjustment method (bringing up shadows and bringing down highlights) can only be used IF the histogram shows the data isn't clipped.  (slammed against the left wall or right wall of the histogram)   If the data is clipped then you'll have to resort to HDR (bracketing).

BTW, you can shoot as many frames as you want for HDR.  It doesn't have to be 3 frames.  They can be 2 stops apart... they could be just 1 stop apart... they could be as much as 3 stops apart.  But I find that 2 stops apart seems to work quite well.

Also... VERY IMPORTANT... shoot RAW when you want this level of adjustment latitude.  Don't shoot JPEG.    If you shoot JPEG then your camera's 14-bit data is compressed to 8-bit data and similar (but non-identical) pixels will be normalized to make it easier to compress the file size and this causes a massive loss in data that wont let you recover your details.

JPEG is great for final output.

RAW is great for images you want to be able to adjust.

[BurnUnit]

Nicely done! Thanks for taking the time to put this together. There's another recent thread going here that explains how your monitor can "lie" to you and how your eyes can "lie" to you as well. But the histogram will lead you to the truth of whats really in the image, waiting to be uncovered.

Also, nice job on cropping out the "road apples".

[Dukai]

Hi!
Thanks a lot for all those explanations and the time spent for answering my questions.
I understand this, but in my example, the thing is the histogram is clipped. So my question was if it was normal that it does clip with a dynamic range of 7 stops. In other words, is it normal that I need to use hdr techniques (either bracketing and tone mapping in post or automatic hdr within the camera) to capture details in shadows and highlights when a scenes darkest shadow and brightest highlight is only far apart fom each other of 7 stops of light? Here, even exposing correctly to the right will left me with black pixels.
Thanks again

[TTMartin]

---

@Dukai wrote:
Hi!
Thanks a lot for all those explanations and the time spent for answering my questions.
I understand this, but in my example, the thing is the histogram is clipped. So my question was if it was normal that it does clip with a dynamic range of 7 stops. In other words, is it normal that I need to use hdr techniques (either bracketing and tone mapping in post or automatic hdr within the camera) to capture details in shadows and highlights when a scenes darkest shadow and brightest highlight is only far apart fom each other of 7 stops of light? Here, even exposing correctly to the right will left me with black pixels.
Thanks again

---

You want to expose to the right. Turn on your highlight alert 'blinkies'. You'll only want specular highlights to blink.

The 80D has some processing 'baked' into the RAW file that your 600D did not, so it behaves differently in post processing. You want some true black black pixels, and these are baked into the 80D RAW file. So don't worry that Lightroom shows some black clipping, you want that. Don't assume since you got noise in your shadows with your 600D you will with your 80D.

Using Lightroom with your 80D RAW file leave the black and white sliders set to zero, and pull the highlight slider to -100 and the shadow slider to +100 and then adjust the exposure as needed. You will not see noise in the shadows like you might have with your 600D. 

[TCampbell]

We'd need to see your data.  Your camera certainly has considerably more than 7 stops of dynamic range.  You'd be very hard pressed to find any digital camera that doesn't have at least 10 (most current cameras are in the 12-14 stops range).  

Even the Canon Digital Rebel (300D) had 10.8 stops.

It would be helpful to understand why your perception is that you got less than 7, when testing on that camera report that it does so much more.

Keep in mind that when you bracket exposures, that's just the exposure for the middle-tones of the image... there will be parts of the image that are both darker and brighter.    Also keep in mind that if you manually meter the darkest and brightest points in a scene via spot metering, the math to find the middle exposure isn't to take the average (exposures are exponential... each stop "doubles" or "halves" the amount of light.  If we keep ISO and f-stop constant but check only shutter speeds and my "dark" metering point reads that I should shoot at 1/60th and my "bright" metering point reads that I should shoot 1/1000th... don't use the average (don't use 1000 + 60 = 1060 ... ÷ 2 = 530.)   The order of stops is 1/60, 1/125, 1/250, 1/500, 1/1000.   So the middle exposure is really 1/250th (if you used something closer to 1/500th then you woould underexpose by a full stop).

We're flying blind with respect to knowing why you are having an issue (in part because we haven't seen any sample images.) But I'm sure we can help you work out what's going on if you're willing to stick it out.  

While I cannot guarantee this is not a camera defect (certainly the camera model is fine ... whether you have an individual unit with a problem is another quesstion) -- it's unlikely that there is a sensor defect.