Dynamic Range Hilarity in the Real World

kain · 12-02-2015, 09:23 AM

As many of you know, the 645Z has a hilarious level of dynamic range. Back on Halloween, I was shooting a costume party, and my flash malfunctioned, and started only shooting on maximum power. Posted here is an example of the levels you can recover out of an overblown flashed portrait. Not necessarily art-book quality left over, but this seemed completely blown at first glance!

FWIW, my workflow is Darktable, using DNGs from the 645z, and using a custom profile generated from shooting and profiling a known IT/8 target using Argyll CMS.

Still half-blown out at the end, but I was able to recover a set of tones that 'fit the theme'

kain · 12-02-2015, 09:24 AM

Heh, forgot the shots!

Heie · 12-02-2015, 12:06 PM

I was expecting a recovery of blown highlights so epic that I'd be floored and reduced to fits of "hilarity."

I'm sorry, but I don't see a significant improvement. Perhaps it's your processing style, but I would not have posted the same under the auspices of "I have proof that the 645Z's DR is so amazing it's SILLY!!"

The same highlight recovery (although we have to be honest - no blown highlights were actually recovered, as your "After" photo clearly demonstrates) can be achieved using a lowly APS-C camera.

-Heie

gavincato · 12-02-2015, 03:09 PM

to be honest the 645z is pretty poor in highlight recovery versus the canons & nikons - it's strength is in shadows.

biz-engineer · 12-02-2015, 03:19 PM

Originally posted by gavincato

to be honest the 645z is pretty poor in highlight recovery versus the canons & nikons - it's strength is in shadows.

Sometime ago was a thread about dynamic range, full frame etc.... Some were mentioning the new sensor technology from MIT using auto-reset and a counter... But I'm not sure who understood what is the real deal with DR of digital sensor compared to the DR of film. Film is non-linear, so for instance it has ISO100 sensitivity in the middle of the light range, and a lower sensitivity near the highlights. That does not happen with a digital sensor, no matter how big the pixels are, because the sensitivity is constant across the light range. Digital sensor get their DR from the lower noise level in the dark zone, and underexposing followed by a software increase of exposure. Even the sensor from MIT with a folding approach does not offer what film can offer because multi-reset + counter lead to lower ISO...hence longer exposure time. So far, only film is able to provide the largest DR at a constant average ISO and constant exposure time.

**Ed Hurst** · 12-02-2015, 03:32 PM

I think there are two issues here. One is the 645Z's DR, which is exceptional (comparisons to film notwithstanding). The other is the way the sensor's mid-point is calibrated. I too have found that its ability to pull detail from shadows is superior to its ability to pull them from highlights. But that takes nothing away from the overall number of stops of DR, which is very high. So I think the answer is that the mid-point is set at a higher than usual level, meaning that 'correct' exposures according to the meter are in fact on the light side (maybe a good thing in terms of ETTR, but it means you are closer to the over-exposure limit than you think). One answer to this is to under-expose consistently (perhaps with exposure compensation), which distributes the ability to recover detail from highlights and shadows more evenly. Another is to be careful of over-exposure and make sure you don't blow the highlights (we can never be certain of such things of course). But none of this means that the DR isn't amazing - it is. It's just that the camera's metering tends to expose things a little towards the lighter end. Correct for that (or be aware of it) and you are still left with a camera that has superb DR.

itshimitis · 12-02-2015, 04:43 PM

Originally posted by gavincato

to be honest the 645z is pretty poor in highlight recovery versus the canons & nikons - it's strength is in shadows.

+1 which is why I tend to expose to the left rather than the right as conventional wisdom would recommend...

kain · 12-02-2015, 05:57 PM

Well, keep in mind that metering had nothing to do with this -- I was pre-calibrating my exposure to a given camera setting and flash power. My (manual) flash malfunctioned and decided to -only- fire at its maximum power, so this was essentially a trash shot to begin with. Apologies if I wasn't so clear about that. Also Heie, you are right, no 'blown' highlights were recovered. That is basically impossible without inpainting. What was recoverable though, is enough DR in the near-blown highlights to reproduce an appreciable amount of the photo structure.

In that regard, I've read that the K-5 and K-3 also have 14-bit encodings for sensor readings, but is the luminance response on those as good as the Z? I suppose I could test that, since I have the k-5iis and k-3 available.

mee · 12-02-2015, 09:11 PM

Curious as to why are you shooting at 1/50s and ISO 3200 if you knew your flash was malfunctioning and firing at max power?

You'd have gotten even more room to work with at ISO 100 and 1/160s as these digital sensors do better at pulling shadow detail than overblown highlights.

2351HD · 12-02-2015, 11:27 PM

Originally posted by itshimitis

+1 which is why I tend to expose to the left rather than the right as conventional wisdom would recommend...

Not sure what you mean there. With better shadow recover it would be more important to expose to the right in order to preserve all highlight data and just let the shadows fall where they will, then recover them.

Unless I am missing something, the logic seems flawed.

itshimitis · 12-03-2015, 02:23 AM

Originally posted by 2351HD

Not sure what you mean there. With better shadow recover it would be more important to expose to the right in order to preserve all highlight data and just let the shadows fall where they will, then recover them.

Unless I am missing something, the logic seems flawed.

If you're preserving highlights then your histogram will be towards the left. With a canon I would preserve the shadows and expose to the right. As a result the histogram will bias to the right.

2351HD · 12-03-2015, 02:50 AM

Nope, not seeing your logic here.

You have a 645Z right? So you should be exposing to the right ETTR to preserve the highlights, then recover the shadows.

ETTR means that you preserve the highlights and if it's a high DR scene, you will have a lot of bunching of the shadows on the left because that's where it's running out of range and clipping. EG you turn the exposure compensation UP until just before the highlights clip.

With canon you are ETTL (to the left) to try and save the shadows because apparently there is more detail to recover in the canon highlights. EG you turn the exposure comp down until just before the shadows clip.

If you are preserving shadows you are exposing to the left, not the right.

Even with Canon, the very best Canon shooters I know still ETTR because it's better to let the shadows fall where they may, i.e. Inside a church cathedral.

itshimitis · 12-03-2015, 04:38 AM

Originally posted by 2351HD

Nope, not seeing your logic here.

You have a 645Z right? So you should be exposing to the right ETTR to preserve the highlights, then recover the shadows.

ETTR means that you preserve the highlights and if it's a high DR scene, you will have a lot of bunching of the shadows on the left because that's where it's running out of range and clipping. EG you turn the exposure compensation UP until just before the highlights clip.

With canon you are ETTL (to the left) to try and save the shadows because apparently there is more detail to recover in the canon highlights. EG you turn the exposure comp down until just before the shadows clip.

If you are preserving shadows you are exposing to the left, not the right.

Even with Canon, the very best Canon shooters I know still ETTR because it's better to let the shadows fall where they may, i.e. Inside a church cathedral.

My understanding of ETTR is to do with the direction of the histogram. When you expose FOR the highlights (to retain detail in the highlights) then the histogram is leaning to the left. When you expose for the shadows the histogram is leaning to the right.

https://en.wikipedia.org/wiki/Exposing_to_the_right

I wouldn't be using the 5D mk III and exposing so I had to recover detail from shadows, as that was what made me jump from Canon in the first place. Lots of noise banding.

Edited to add: I think that we are agreeing vociferously just seeing the ETTR from different directions. I've always seen the ETTR in books and magazines talking about the direction of the histogram. What you expose for is very different.

Kolor-Pikker · 12-03-2015, 04:52 AM

The Z actually has extremely poor highlight recovery, from my testing about only -10 highlights in Lightroom before it starts to produce false colors. A Canon 5D3 would've worked wonders here, especially if processed in Capture One.

You really do not want the histogram anywhere near the right edge, IMO Pentax should adjust the scale and add several stops to the left to make the histo more representative of the total DR, as it is you're only looking at the brightest half.

Right half of this chart is what you see when previewing your image, left half is how much you have left to recover in post. This is assuming you don't have shadow recovery enabled in D-range settings.

For reference, here's a really old photo that I've kept just to test highlight recovery on the 5D2, it's over exposed by about 2.5 stops.

texandrews · 12-03-2015, 09:55 AM

Originally posted by Ed Hurst

... So I think the answer is that the mid-point is set at a higher than usual level, meaning that 'correct' exposures according to the meter are in fact on the light side (maybe a good thing in terms of ETTR, but it means you are closer to the over-exposure limit than you think)....

I wholeheartedly agree with this assessment. I also believe the midpoint is set high---I think at least a stop.

12-02-2015, 09:23 AM	#1
kain Senior Member Join Date: Dec 2012 Photos: Gallery Posts: 162	Dynamic Range Hilarity in the Real World As many of you know, the 645Z has a hilarious level of dynamic range. Back on Halloween, I was shooting a costume party, and my flash malfunctioned, and started only shooting on maximum power. Posted here is an example of the levels you can recover out of an overblown flashed portrait. Not necessarily art-book quality left over, but this seemed completely blown at first glance! FWIW, my workflow is Darktable, using DNGs from the 645z, and using a custom profile generated from shooting and profiling a known IT/8 target using Argyll CMS. Still half-blown out at the end, but I was able to recover a set of tones that 'fit the theme'

12-03-2015, 04:38 AM	#13
itshimitis Pentaxian Join Date: May 2011 Location: All over the place Posts: 3,535	Originally posted by 2351HD Nope, not seeing your logic here. You have a 645Z right? So you should be exposing to the right ETTR to preserve the highlights, then recover the shadows. ETTR means that you preserve the highlights and if it's a high DR scene, you will have a lot of bunching of the shadows on the left because that's where it's running out of range and clipping. EG you turn the exposure compensation UP until just before the highlights clip. With canon you are ETTL (to the left) to try and save the shadows because apparently there is more detail to recover in the canon highlights. EG you turn the exposure comp down until just before the shadows clip. If you are preserving shadows you are exposing to the left, not the right. Even with Canon, the very best Canon shooters I know still ETTR because it's better to let the shadows fall where they may, i.e. Inside a church cathedral. My understanding of ETTR is to do with the direction of the histogram. When you expose FOR the highlights (to retain detail in the highlights) then the histogram is leaning to the left. When you expose for the shadows the histogram is leaning to the right. https://en.wikipedia.org/wiki/Exposing_to_the_right I wouldn't be using the 5D mk III and exposing so I had to recover detail from shadows, as that was what made me jump from Canon in the first place. Lots of noise banding. Edited to add: I think that we are agreeing vociferously just seeing the ETTR from different directions. I've always seen the ETTR in books and magazines talking about the direction of the histogram. What you expose for is very different. Last edited by itshimitis; 12-03-2015 at 05:10 AM.

12-03-2015, 04:52 AM	#14
Kolor-Pikker Veteran Member Join Date: Dec 2014 Posts: 344	The Z actually has extremely poor highlight recovery, from my testing about only -10 highlights in Lightroom before it starts to produce false colors. A Canon 5D3 would've worked wonders here, especially if processed in Capture One. You really do not want the histogram anywhere near the right edge, IMO Pentax should adjust the scale and add several stops to the left to make the histo more representative of the total DR, as it is you're only looking at the brightest half. Right half of this chart is what you see when previewing your image, left half is how much you have left to recover in post. This is assuming you don't have shadow recovery enabled in D-range settings. For reference, here's a really old photo that I've kept just to test highlight recovery on the 5D2, it's over exposed by about 2.5 stops. Last edited by Kolor-Pikker; 12-03-2015 at 05:57 AM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Pentax 645Z - Real World Dynamic Range	thapamd	Pentax Medium Format	18	09-11-2014 10:59 AM
K-3 in camera HDR (Dynamic Range)	Iam1of2	Pentax K-3 & K-3 II	3	01-12-2014 12:29 PM
Bringing out the shadow details in a high dynamic range image	atupdate	Post-Processing Articles	3	01-17-2011 02:11 PM
The Pentax K-7: The era of in-camera High Dynamic Range Imaging has arrived!	HermanLee	Pentax News and Rumors	2	05-20-2009 10:40 PM

12-02-2015, 12:06 PM	#3
Heie Veteran Member Join Date: Nov 2011 Photos: Gallery \| Albums Posts: 968	I was expecting a recovery of blown highlights so epic that I'd be floored and reduced to fits of "hilarity." I'm sorry, but I don't see a significant improvement. Perhaps it's your processing style, but I would not have posted the same under the auspices of "I have proof that the 645Z's DR is so amazing it's *SILLY!!" The same highlight recovery (although we have to be honest - no blown* highlights were actually recovered, as your "After" photo clearly demonstrates) can be achieved using a lowly APS-C camera. -Heie

12-02-2015, 03:09 PM	#4
gavincato Senior Member Join Date: Apr 2015 Location: NSW, Australia Posts: 247	to be honest the 645z is pretty poor in highlight recovery versus the canons & nikons - it's strength is in shadows.

12-02-2015, 03:19 PM	#5
biz-engineer Pentaxian Join Date: Feb 2015 Photos: Gallery Posts: 12,231	Originally posted by gavincato to be honest the 645z is pretty poor in highlight recovery versus the canons & nikons - it's strength is in shadows. Sometime ago was a thread about dynamic range, full frame etc.... Some were mentioning the new sensor technology from MIT using auto-reset and a counter... But I'm not sure who understood what is the real deal with DR of digital sensor compared to the DR of film. Film is non-linear, so for instance it has ISO100 sensitivity in the middle of the light range, and a lower sensitivity near the highlights. That does not happen with a digital sensor, no matter how big the pixels are, because the sensitivity is constant across the light range. Digital sensor get their DR from the lower noise level in the dark zone, and underexposing followed by a software increase of exposure. Even the sensor from MIT with a folding approach does not offer what film can offer because multi-reset + counter lead to lower ISO...hence longer exposure time. So far, only film is able to provide the largest DR at a constant average ISO and constant exposure time.

12-02-2015, 03:32 PM	#6
Ed Hurst Site Supporter Join Date: May 2010 Location: Sydney Posts: 1,657	I think there are two issues here. One is the 645Z's DR, which is exceptional (comparisons to film notwithstanding). The other is the way the sensor's mid-point is calibrated. I too have found that its ability to pull detail from shadows is superior to its ability to pull them from highlights. But that takes nothing away from the overall number of stops of DR, which is very high. So I think the answer is that the mid-point is set at a higher than usual level, meaning that 'correct' exposures according to the meter are in fact on the light side (maybe a good thing in terms of ETTR, but it means you are closer to the over-exposure limit than you think). One answer to this is to under-expose consistently (perhaps with exposure compensation), which distributes the ability to recover detail from highlights and shadows more evenly. Another is to be careful of over-exposure and make sure you don't blow the highlights (we can never be certain of such things of course). But none of this means that the DR isn't amazing - it is. It's just that the camera's metering tends to expose things a little towards the lighter end. Correct for that (or be aware of it) and you are still left with a camera that has superb DR.

12-02-2015, 04:43 PM	#7
itshimitis Pentaxian Join Date: May 2011 Location: All over the place Posts: 3,535	Originally posted by gavincato to be honest the 645z is pretty poor in highlight recovery versus the canons & nikons - it's strength is in shadows. +1 which is why I tend to expose to the left rather than the right as conventional wisdom would recommend...

12-02-2015, 05:57 PM	#8
kain Senior Member Join Date: Dec 2012 Photos: Gallery Posts: 162 Original Poster	Well, keep in mind that metering had nothing to do with this -- I was pre-calibrating my exposure to a given camera setting and flash power. My (manual) flash malfunctioned and decided to -only- fire at its maximum power, so this was essentially a trash shot to begin with. Apologies if I wasn't so clear about that. Also Heie, you are right, no 'blown' highlights were recovered. That is basically impossible without inpainting. What was recoverable though, is enough DR in the near-blown highlights to reproduce an appreciable amount of the photo structure. In that regard, I've read that the K-5 and K-3 also have 14-bit encodings for sensor readings, but is the luminance response on those as good as the Z? I suppose I could test that, since I have the k-5iis and k-3 available.

12-02-2015, 09:11 PM	#9
mee Veteran Member Join Date: May 2010 Posts: 7,403	Curious as to why are you shooting at 1/50s and ISO 3200 if you knew your flash was malfunctioning and firing at max power? You'd have gotten even more room to work with at ISO 100 and 1/160s as these digital sensors do better at pulling shadow detail than overblown highlights.

12-02-2015, 11:27 PM	#10
2351HD Veteran Member Join Date: Apr 2013 Location: Brisbane, Australia Photos: Albums Posts: 750	Originally posted by itshimitis +1 which is why I tend to expose to the left rather than the right as conventional wisdom would recommend... Not sure what you mean there. With better shadow recover it would be more important to expose to the right in order to preserve all highlight data and just let the shadows fall where they will, then recover them. Unless I am missing something, the logic seems flawed.

12-03-2015, 02:23 AM	#11
itshimitis Pentaxian Join Date: May 2011 Location: All over the place Posts: 3,535	Originally posted by 2351HD Not sure what you mean there. With better shadow recover it would be more important to expose to the right in order to preserve all highlight data and just let the shadows fall where they will, then recover them. Unless I am missing something, the logic seems flawed. If you're preserving highlights then your histogram will be towards the left. With a canon I would preserve the shadows and expose to the right. As a result the histogram will bias to the right.

12-03-2015, 02:50 AM	#12
2351HD Veteran Member Join Date: Apr 2013 Location: Brisbane, Australia Photos: Albums Posts: 750	Nope, not seeing your logic here. You have a 645Z right? So you should be exposing to the right ETTR to preserve the highlights, then recover the shadows. ETTR means that you preserve the highlights and if it's a high DR scene, you will have a lot of bunching of the shadows on the left because that's where it's running out of range and clipping. EG you turn the exposure compensation UP until just before the highlights clip. With canon you are ETTL (to the left) to try and save the shadows because apparently there is more detail to recover in the canon highlights. EG you turn the exposure comp down until just before the shadows clip. If you are preserving shadows you are exposing to the left, not the right. Even with Canon, the very best Canon shooters I know still ETTR because it's better to let the shadows fall where they may, i.e. Inside a church cathedral.

12-03-2015, 09:55 AM	#15
texandrews Moderator Join Date: Apr 2014 Location: Baltimore Photos: Albums Posts: 3,390	Originally posted by Ed Hurst ... So I think the answer is that the mid-point is set at a higher than usual level, meaning that 'correct' exposures according to the meter are in fact on the light side (maybe a good thing in terms of ETTR, but it means you are closer to the over-exposure limit than you think).... I wholeheartedly agree with this assessment. I also believe the midpoint is set high---I think at least a stop.