snapsy wrote:
On an ideal sensor an ISO 100 shot pushed 4 stops should have the noise characteristics identical to an ISO 1600 shot. The D800 is pretty close in that regard due to the linear increase in its electronic read noise across ISOs, whereas more traditional sensors like the 5DM2/5DM3/D3 have non-linear increases in noise, making them perform better at Higher ISOs vs a Lower ISO image that's entirely pushed in post. Every sensor has an ISO level at which any increases in ISO produce no benefit vs underexposing at a lower ISO and pushing in post. This level is much lower on the D800 but defining it precisely depends on whether one is interested in the purely technical aspect of analog vs PP/digital gain or instead interested in the practical applications. The lower the ISO you can use and push with acceptable results, the more dynamic range you'll have available for your images.
Even with an ideal sensor there are potential pitfalls in pushing. One is posterization, which can occur because the encoding of shadows in the digital realm uses less precision than the midtones and highlights. In practice however it works out that the level of noise in the shadows (even as low as it is on the D800) is about balanced with the precision necessary to represent its tonality, so even after pushing there isn't much posterization. This was my experience with the D7000 but I need to repeat the experiments with the D800 to be sure nothing has changed.
Another pitfall is color/hue shifts, which generally occur because the color translations performed by many raw processors (such as ACR/LR) order their exposure adjustments before or after color application within their raw processing pipelines. I haven't seen much evidence of this on the D800 but I'm at the early stages of analyzing it, plus I don't do much skintone work so any shifts aren't as bothersome to me anyway.
I'm still in the early throws of figuring out the best base ISO to use for the D800, both in terms of any potential posterization but also because there is still some non-linearility in the D800's ISO gain. Right now I'm using ISO 100 because it yields the most latitude (DR) but it may turn out that ISO 200 or 400 have slight noise advantages (or perhaps even one of the intermediate ISOs in between)....Show more →
If you have a properly exposed lower ISO file for the highlights and pull the shadows, don't the highlight areas stay at the noise level of the original ISO? I'm thinking skies in particular, expose for good cloud detail and not worry about much else.
Hrow wrote:
If you have a properly exposed lower ISO file for the highlights and pull the shadows, don't the highlight areas stay at the noise level of the original ISO? I'm thinking skies in particular, expose for good cloud detail and not worry about much else.
The noise that dominates an image in the midtones and highlights is photon shot noise, which is a function of exposure only, ie how much light you allow to reach the sensor via aperture+shutter speed - ISO is not part of exposure (never has been really) and thus has no effect on shot noise.
If a High ISO image is metered in the same fashion as the low ISO image in terms of highlights, the midtones and highlights of both images will be from the same scene LVs (Light Values), meaning the same number of photons are captured and so with the same amount of photon shot noise. Naturally this will cause blocking of the shadows for the High ISO image and thus loss of DR.
If you instead meter the High ISO image for shadow detail, you will lose the higher LVs due to clipping, resulting in the same loss of DR vs the low ISO image but for a different subset of LVs. So the LVs represented in the midtones and highlights of the High ISO image will be different than those of the low ISO scene but the midtones and highlights which are represented will have the same noise as the low ISO image.
So in the end High ISO results in a loss of DR but noise is dependent only on exposure.
Thanks for the explanation. I'll read it again in the morning to allow the Crazy Glue to set on the parts of my head that exploded and were put back together.
Here's another ISOless "high iso" example, hand-held. This metered for ISO 6400 but was shot at ISO 100 and pushed in ACR 6.7RC/CS5. Default ACR NR (chrominance=25, luminance=0). Final image is resampled to 12MP bicubic sharper, selective USM micro-contast applied.
If anyone wants to play with the raw file you can download at the following link. I've also included the ACR side-car file so you can see the full settings. Link is good for 5 days. Raw File
Mike Yamin wrote:
That's crazy stuff snapsy! Do you just meter at the ISO you want to push to or gives you the shutter speed you need and then drop to ISO 100?
I normally don't even bother metering - I just set my desired aperture and shutter speed and adjust as necessary in post. I metered in this case just to demonstrate what the effective ISO was for this image.
snapsy wrote:
I normally don't even bother metering - I just set my desired aperture and shutter speed and adjust as necessary in post. I metered in this case just to demonstrate what the effective ISO was for this image.
Thanks for the quick response. This is really interesting information and could change the way I shoot if I decide to pick up a D800.
snapsy wrote:
Ditched the tripod today to try my hand at, well, handheld. Before leaving the house I tested/calibrated how low I can shoot handheld with the 50 f/1.8G and get pixel-sharp photo...came in around 1/40, which is better than I expected because I'm about average when it comes to hand'holdability and stability.
These are all shot at base ISO, pushed in post when necessary (ISOless). The flower shots are indoors at the Bellagio Hotel in Vegas, with sunlight coming through the glass ceiling. Tried to keep the shutter at or above 1/50. All shot raw, processed with ACR 6.7RC, including ACR sharpening + CS5 micro-contrast sharpening.
.../image links/... ...Show more →
I guess someone forgot to tell you that you need 1/1000s speeds and extremely expensive lenses to get sharp captures with the D800.
Fortunately they forgot to tell me too. Doing just fine with the new 85 and 50mm 1.8AFS lenses. I do wish they had made the 85 stabilized though. 1/60s without arm- or hand-support is hit/miss as for image usability.
theSuede wrote:
I guess someone forgot to tell you that you need 1/1000s speeds and extremely expensive lenses to get sharp captures with the D800.
Fortunately they forgot to tell me too. Doing just fine with the new 85 and 50mm 1.8AFS lenses. I do wish they had made the 85 stabilized though. 1/60s without arm- or hand-support is hit/miss as for image usability.
hehe, I must admit I was worried about shutter speeds before I got hold of the D800. I would love to see Nikon release a stabilized 50mm. I just acquired the Sony OSS 50mm for my NEX and it's a low-light monster. Usable f/1.8 at 1/13 shutter speeds on a crop body. In DOF-limited situations my NEX+50mm OSS will outperform any full-frame body in low-light hand-held usage for static shooting.
I do have an acceleration meter at the lab, fitted to a brass/chrome M42 adapter. But as in photographic tests, no - not really. I actually find the D7000 more sensitive to slap-effects. But that's just a feeling, and not a very concrete one at that.
Using self-timer with the camera placed on a soft cushion (with a 100mm+ lens mounted) is a very revealing test.
The 5D3 was very nice in that aspect btw. It feels / sounds almost like a D5100, but faster and snappier of course. Quiet, but still mechanically very affirmative. I do wish they had put some more damping on the D800 mirror end-stops, it's still more "loud" than what I would have preferred.
50mm is right on the edge where VR starts to make less difference to me, I'm usually target-movement limited at faster than 1/50s unless I get a lucky shot. But a stabilized 85mm lens could give me 2/3Ev longer shutter speeds (1/80 to 1/50s), something I notice with the 70-200 (which of course is a better stabilized lens just by the length and weight). Getting both the F1.8 and stabilization would probably be the most economical "real performance" upgrade for me.
