D800E -Truly Amazing Low-Light Performance

[Ray]

Just a couple of years ago I was considering upgrading from the Canon 50D to a 7D or 60D, mainly because of the video capability which none the DSLRs I owned at that time had.

However, when I saw the DR specs of the D7000 on the DXOMark website, I got very interested in that camera despite the fact that, at that time, I owned only one Nikkor lens, the 14-24/2.8 ultra-wide-angle zoom.

After some discussions on LL about the significance of shot noise at such low signal levels (some posters on LL seemed to think that any DR claims beyond about 12 stops for a small sensor would be an irrelevancy because shot noise would unavoidably mask any benefits), I took the plunge and bought a D7000 with Nikkor 24-120/F4 zoom.

After comparing the DR of my new D7000 with that of my Canon cameras, such as the 50D and the full-frame 5D, it became very apparent that the claimed 2-stop DR advantage of the D7000 was very real and only slightly obscured by the proportional increase in shot noise. I was very pleased with the camera's performance in general and would often marvel at how miraculously an almost totally dark image, underexposed at ISO 100 instead of properly exposed at ISO 6400, would spring to life with just a click on the 'auto' button in ACR.

The main attraction of the D800E for me was its incorporation into a larger sensor of all the qualities of the D7000. If fact, at equal image size, the D800E has about 1/2 a stop greater DR than the D7000, due to its larger sensor. However, I haven't taken the trouble to compare the DR of both cameras. I'm too busy enjoying photographing the real world.

It was while photographing the real world recently, in Thailand, I made the mistake of taking a shot at 1/60th, F8 and ISO 100 in manual mode, in near darkness at the end of the day, not realising the D800E's built-in flash was not still popped up. The review image was totally black. Not even a hint of a highlight. I didn't delete it due to laziness and the fact I was in a hurry to get back to the hotel before it got completely dark.

After downloading the day's images onto my Dell Notebook, and deleting a few mistakes and duds, I was about to delete the totally black preview in Bridge, of the F8 shot at 1/60th without flash, when curiosity got the better of me, and I decided to see just how noisy and awful this image really was.

When the image opened in ACR, the histogram showed just a single, thin, vertical line on the far left. There was absolutely no other information. Not even a little hump or two on the horizontal axis. Refer attached image.

However, after some extreme adjustments in ACR, such as +4 exposure, +100 Brightness, 100 Fill Light, +100 Darks in the Tone Curve etc, the image looked almost presentable, apart from an over all green cast. A single click on 'AutoColor' in Photoshop fixed that.

I'm just flabbergasted that so much detail actually exists in a shot I was about to delete. The purpose of the shot was to attempt to give a perspective on the actual size of this "Night Butterfly" (as it is locally known). I'd taken a few shots close up, with 24-120 zoom at 120mm, using the built-in flash, and as I was leaving the scene, still amazed at the enormous size of this butterfly, I decided to take a shot from a greater distance, hoping it would give a better impression of its great size.

Of course, "Night Butterfly" sounds more romantic than bloody Moth. But a check on the internet reveals it is in fact a moth, but the largest moth in the world, that goes by the name of Attacus Atlas. It's total wing surface area can be in excess of 62 sq inches (or 400 sq cm), and wing span in excess of 10". It's larger than my notebook on which I've processed the attached images. 

I've also shown a close-up of the moth taken when the flash was popped up. I thought it would immediately fly away, but it just twitched. I began to think it was maybe supremely confident because it had two snakes attached to its wings. It also had simulated patches of bamboo leaf on its wings as extra camouflage protection. Refer image 04, which I think is also rather abstract. If you think that detail is moire, don't hesitate to say so. 

[Dustbak]

That is pretty amazing. The absence of horrible noise in the OOF areas is remarkable!!!

[Ray]

That is pretty amazing. The absence of horrible noise in the OOF areas is remarkable!!!

Forgot to mention, I applied no noise reduction or masking in ACR. Sharpening was at a default level of 25.

[francois]

I'm impressed to say the least!

[kaelaria]

That's insane!

[bjanes]

However, after some extreme adjustments in ACR, such as +4 exposure, +100 Brightness, 100 Fill Light, +100 Darks in the Tone Curve etc, the image looked almost presentable, apart from an over all green cast. A single click on 'AutoColor' in Photoshop fixed that.

I'm just flabbergasted that so much detail actually exists in a shot I was about to delete.

Ray,

Thanks for posting your findings. I also have the 800e but have not experienced such an underexposed shot and would not have thought that such results were possible. The take home point for me is that one should not carry ETTR too far--with such shadow performance, one can afford a bit of headroom for the highlights with the D800.

I see that you are using PV2010 with an older version of ACR. It would be interesting to see how PV2012 with its improved tone rendering would handle this image.

Regards,

Bill

[SunnyUK]

That's impressive! I have previously extracted what I thought was an astounding amount of detail from a very dark shot from my D800, but nothing like this.

[Ray]

Ray,

Thanks for posting your findings. I also have the 800e but have not experienced such an underexposed shot and would not have thought that such results were possible. The take home point for me is that one should not carry ETTR too far--with such shadow performance, one can afford a bit of headroom for the highlights with the D800.

I see that you are using PV2010 with an older version of ACR. It would be interesting to see how PV2012 with its improved tone rendering would handle this image.

Regards,

Bill

Hi Bill,
I'm travelling at present and don't have access to the latest version of Photoshop, and because of the calibration limitations on laptops, I don't spend too much time attempting to get the most from an image.

However, as a result of my surprise with this extremely underexposed shot, I checked the DXO test results again to confirm that the D800E does have close to 1/2 a stop more DR than the D7000, at base ISO and equal size, and also noticed that the graph for the D800E DR is not completely straight from ISO 100 down. Between ISO 100 and ISO 200, the drop in DR is only about 1/2 a stop. This implies it would be better to use ISO 200 if one intends to underexpose significantly, rather than raise ISO significantly. The DR should be close to 1/2 a stop better at ISO 200. Would you agree?

[EricV]

... noticed that the graph for the D800E DR is not completely straight from ISO 100 down. Between ISO 100 and ISO 200, the drop in DR is only about 1/2 a stop. This implies it would be better to use ISO 200 if one intends to underexpose significantly, rather than raise ISO significantly. The DR should be close to 1/2 a stop better at ISO 200. Would you agree?

That sounds about right.  As I recall from looking at the DxO curves, raising ISO from 100 up to around 400 improves signal/noise, but after that there is no benefit to increasing ISO (apart from not scaring you with very dark preview images and histograms).

[Ray]

That sounds about right.  As I recall from looking at the DxO curves, raising ISO from 100 up to around 400 improves signal/noise, but after that there is no benefit to increasing ISO (apart from not scaring you with very dark preview images and histograms).

Looking again at those DXO curves, I don't see any advantage that would be noticeable, in moving from ISO 200 to 400. SNR at 18% falls by 2.8dB, which is close enough to 1 stop, or 3dB. And DR falls by 0.98 EV, which is close enough to 1 stop.

[jeremypayne]

I was very pleased with the camera's performance in general and would often marvel at how miraculously an almost totally dark image, underexposed at ISO 100 instead of properly exposed at ISO 6400, would spring to life with just a click on the 'auto' button in ACR.

Given the nature of the Exmor sensors ... and your interest in such things ... I'm surprised you are surprised.

Shooting in that manner and shooting at ISO 6400 should be almost indistinguishable on such an "iso-less" camera.

From the looks of the result, however, this would appear to be effectively past 6400 on the scale ...

[bjanes]

However, as a result of my surprise with this extremely underexposed shot, I checked the DXO test results again to confirm that the D800E does have close to 1/2 a stop more DR than the D7000, at base ISO and equal size, and also noticed that the graph for the D800E DR is not completely straight from ISO 100 down. Between ISO 100 and ISO 200, the drop in DR is only about 1/2 a stop. This implies it would be better to use ISO 200 if one intends to underexpose significantly, rather than raise ISO significantly. The DR should be close to 1/2 a stop better at ISO 200. Would you agree?

Ray,

I think you are correct. Unlike the D7000, the D800e is not "ISO less" as Marianne Oelund notes in this post. With the D7000 one may leave the ISO at base and increase exposure in the raw converter when one is underexposing. With the D800e, one should use at least 200 and there may be some minimal improvement beyond that.

It is interesting that DXO rates the DR of the D800e at 14.33 at base ISO, which is interesting since a 14 bit linear file can encode only 14 stops of DR, at least in theory.

Regards,

Bill

[ErikKaffehr]

Bill,

It's only 13.23 in screen mode. On my computer the graphs don't update reliably, but clicking a couple of times between print mode and screen mode I get 13.23 in screen mode.

Best regards
Erik

Ray,

I think you are correct. Unlike the D7000, the D800e is not "ISO less" as Marianne Oelund notes in this post. With the D7000 one may leave the ISO at base and increase exposure in the raw converter when one is underexposing. With the D800e, one should use at least 200 and there may be some minimal improvement beyond that.

It is interesting that DXO rates the DR of the D800e at 14.33 at base ISO, which is interesting since a 14 bit linear file can encode only 14 stops of DR, at least in theory.

Regards,

Bill

[bjanes]

Bill,

It's only 13.23 in screen mode. On my computer the graphs don't update reliably, but clicking a couple of times between print mode and screen mode I get 13.23 in screen mode.

Best regards
Erik

Erik,

Thanks for pointing that out, since I didn't think about screen vs print. I get 13.24 stops for screen from the site. It is nice to know that the laws of mathematics are not violated.

Bill

[BJL]

Unlike the D7000, the D800e is not "ISO less" as Marianne Oelund notes in this post. With the D7000 one may leave the ISO at base and increase exposure in the raw converter when one is underexposing. With the D800e, one should use at least 200 and there may be some minimal improvement beyond that.

It is interesting that DXO rates the DR of the D800e at 14.33 at base ISO, which is interesting since a 14 bit linear file can encode only 14 stops of DR, at least in theory.

Is there a reliable list of maximum "relevant" ISO settings for various cameras, beyond which results are no better than underexposing at a lower ISO speed setting and compensating later? Like seemingly 200 or 400 for the D800(E), and base ISO speed for the D7000.

In particular, I would like to know this for the EM5, because it could help me refine my use of the auto-ISO mode when there is not enough time to optimize settings before taking a photo.

P. S. wouldn't the rounding error in 14-bit be at most half a bit (and probably less in RMS terms), so that data rounded to 14 bits by a perfect 14-bit ADC could have a SNR of up to 2^15, which might be reported as 15 stops, or 15 bits?

[moonpeep]

Very interesting. I'm wondering then what the ideal iso would be for something like shooting the night sky. Is the thought here that 200-400 or so would capture all the detail needed such that you could modify the image in post to achieve the results of say, 6400 iso? (if not better it)

So in theory... if you have proper post processing software... no reason to go above iso 400?

[Ray]

Given the nature of the Exmor sensors ... and your interest in such things ... I'm surprised you are surprised.

Shooting in that manner and shooting at ISO 6400 should be almost indistinguishable on such an "iso-less" camera.

From the looks of the result, however, this would appear to be effectively past 6400 on the scale ...

Jeremy,

Have you been spending too much time on the Coffee Corner? 

I did use the past tense when I wrote, "I was very pleased with the camera's performance in general and would often marvel at how miraculously an almost totally dark image, underexposed at ISO 100 instead of properly exposed at ISO 6400, would spring to life with just a click on the 'auto' button in ACR."

When I bought the D7000, I was coming from a number of years experience with Canon DSLRs which had unrivalled high-ISO performance until the Nikon D3 hit the market, but unspectacular DR at base ISO. I have one or two shots from my 5D taken at night when the attached flash inadvertently did not fire. Despite the histogram of such shots looking a lot better than the D800E histogram pictured above, the noise and banding in the underexposed 5D shots was so bad it became a feature in itself. It created the impression of a photograph of a tapestry with a rather coarse weave, which I found quite interesting.

My renewed surprise at the low-light performance of the D800E was due partly to; (1) never having taken or processed such a drastically underexposed shot before; (2) partly due to my forgetting that the D800E should have around 1/2 stop greater DR than the D7000 at same image size, (half a stop of DR is at the threshhold which is clearly noticeable) and; (3) my previous version of Photoshop which I'd used to process D7000 images was CS3. I skipped CS4. I'm now using CS5 which I suspect might be doing a noticeably better conversion job with these underexposed NEF files.

Hope I've clarified this issue for you. 

[mac_paolo]

So in theory... if you have proper post processing software... no reason to go above iso 400?

Yes and no.
I have presets for every and each ISO value for my main camera on Lightroom.
If I import an ISO 3200 shot it will receive specific denoise values, very different from the ones that will be applied to an ISO 200/400 shot.
If you shoot at fixed low ISO you get the most of the quality and the most of the time in front of Lightroom

[Ray]

Very interesting. I'm wondering then what the ideal iso would be for something like shooting the night sky. Is the thought here that 200-400 or so would capture all the detail needed such that you could modify the image in post to achieve the results of say, 6400 iso? (if not better it)

So in theory... if you have proper post processing software... no reason to go above iso 400?

As a result of starting this thread I've learned something myself about the performance of the D800E which I hadn't previously given much thought to. Looking again at the DR and SNR figures at various ISOs, on the DXOMark site, it seems to me that an exposure at base ISO that is 6 stops underexposed should have about one full stop worse DR than the same exposure at ISO 6400, with the D800E.

The figures I'm using to deduce this, are 14.33EV of DR at base ISO, and 9.23EV 6 stops down at ISO 6400. However, 14.33-9.23=5.1EV. What appears to be happening is one initially gains a 1/2 stop advantage by underexposing at ISO 200 instead of 100, then with each additional increase in ISO one gains an incremental DR advantage of around 0.1EV. These small amounts, in addition to that initial 1/2 stop advantage, add up to a total of a 0.9EV DR advantage using ISO 6400 instead of ISO 100.

However, SNR at 18% Grey diminishes more evenly as one reduces exposure. At base ISO it's 46dB. At ISO 6400 it's 28.1dB. 46-28.1=17.9dB. On the basis that 3dB is equivalent to a difference of one F stop in exposure, a drop in 6 stops of exposure should result in a drop of 18dB of SNR at 18%, which is approximately what one gets whether underexposing at base ISO or correctly exposing at ISO 6400.

These are the sorts of things one should easily be able to test for oneself using some practical real-world examples. I'll take a few underexposed shots tonight to see how they compare at ISO 100 and 6400.

[Alistair]

Just a couple of years ago I was considering upgrading from the Canon 50D to a 7D or 60D, mainly because of the video capability which none the DSLRs I owned at that time had.

However, when I saw the DR specs of the D7000 on the DXOMark website, I got very interested in that camera despite the fact that, at that time, I owned only one Nikkor lens, the 14-24/2.8 ultra-wide-angle zoom.

After some discussions on LL about the significance of shot noise at such low signal levels (some posters on LL seemed to think that any DR claims beyond about 12 stops for a small sensor would be an irrelevancy because shot noise would unavoidably mask any benefits), I took the plunge and bought a D7000 with Nikkor 24-120/F4 zoom.

After comparing the DR of my new D7000 with that of my Canon cameras, such as the 50D and the full-frame 5D, it became very apparent that the claimed 2-stop DR advantage of the D7000 was very real and only slightly obscured by the proportional increase in shot noise. I was very pleased with the camera's performance in general and would often marvel at how miraculously an almost totally dark image, underexposed at ISO 100 instead of properly exposed at ISO 6400, would spring to life with just a click on the 'auto' button in ACR.

The main attraction of the D800E for me was its incorporation into a larger sensor of all the qualities of the D7000. If fact, at equal image size, the D800E has about 1/2 a stop greater DR than the D7000, due to its larger sensor. However, I haven't taken the trouble to compare the DR of both cameras. I'm too busy enjoying photographing the real world.

It was while photographing the real world recently, in Thailand, I made the mistake of taking a shot at 1/60th, F8 and ISO 100 in manual mode, in near darkness at the end of the day, not realising the D800E's built-in flash was not still popped up. The review image was totally black. Not even a hint of a highlight. I didn't delete it due to laziness and the fact I was in a hurry to get back to the hotel before it got completely dark.

After downloading the day's images onto my Dell Notebook, and deleting a few mistakes and duds, I was about to delete the totally black preview in Bridge, of the F8 shot at 1/60th without flash, when curiosity got the better of me, and I decided to see just how noisy and awful this image really was.

When the image opened in ACR, the histogram showed just a single, thin, vertical line on the far left. There was absolutely no other information. Not even a little hump or two on the horizontal axis. Refer attached image.

However, after some extreme adjustments in ACR, such as +4 exposure, +100 Brightness, 100 Fill Light, +100 Darks in the Tone Curve etc, the image looked almost presentable, apart from an over all green cast. A single click on 'AutoColor' in Photoshop fixed that.

I'm just flabbergasted that so much detail actually exists in a shot I was about to delete. The purpose of the shot was to attempt to give a perspective on the actual size of this "Night Butterfly" (as it is locally known). I'd taken a few shots close up, with 24-120 zoom at 120mm, using the built-in flash, and as I was leaving the scene, still amazed at the enormous size of this butterfly, I decided to take a shot from a greater distance, hoping it would give a better impression of its great size.

Of course, "Night Butterfly" sounds more romantic than bloody Moth. But a check on the internet reveals it is in fact a moth, but the largest moth in the world, that goes by the name of Attacus Atlas. It's total wing surface area can be in excess of 62 sq inches (or 400 sq cm), and wing span in excess of 10". It's larger than my notebook on which I've processed the attached images. 

I've also shown a close-up of the moth taken when the flash was popped up. I thought it would immediately fly away, but it just twitched. I began to think it was maybe supremely confident because it had two snakes attached to its wings. It also had simulated patches of bamboo leaf on its wings as extra camouflage protection. Refer image 04, which I think is also rather abstract. If you think that detail is moire, don't hesitate to say so. 

It is a good device the 800e but not sure that it is any more amazing than other cameras. I am amazed at how clean the low iso files are from my d800e look on-screen, I am amazed at the resolution that my Nex7 produced with a good Leica tele and on a tripod on-screen at 100%, and I am amazed that I cannot tell the prints made on my 8 year old 1dsII with 70-200/II apart from the other two.