Reading mode: Light Dark

D800 trick for low light?

Go to last post
B

Bluefin

New member
Messages
9
Reaction score
1
Location
CA, US
Love DPReview for many years but this is my first post!

I'm sure I read someplace awhile ago that you can get better low light/high ISO performance out of the D800 by sacrificing file size. I think it was part of the media blitz re: D4s/DF/D800 low light capabilities.

Not sure if they meant going into the shooting menu and switching from RAW to TIFF and then changing the IMAGE SIZE from large to medium or small or going to the next item in the shooting menu IMAGE AREA and changing to DX or changing the image area.

Actually, I'm not sure if what I think I read is possible in the first place! I would have a lot more confidence in someone who knows the answer vs testing on my own.

If it is, would someone let me know how to do it and why it works?

Thanks.........Mark
 
Solution
S
Hi,

I think what you mean is: if you reduce resolution from 36 MP to 12 MP (file size from 7360 x 4912 = 36 MP to 4300 x 2872 = 12 MP) you will reduce visible noise.
hewhosculpts said:
Bluefin said:
Love DPReview for many years but this is my first post!

I'm sure I read someplace awhile ago that you can get better low light/high ISO performance out of the D800 by sacrificing file size. I think it was part of the media blitz re: D4s/DF/D800 low light capabilities.

Not sure if they meant going into the shooting menu and switching from RAW to TIFF and then changing the IMAGE SIZE from large to medium or small or going to the next item in the shooting menu IMAGE AREA and changing to DX or changing the image area.

Actually, I'm not sure if what I think I read is possible in the first place! I would have a lot more confidence in someone who knows the answer vs testing on my own.

If it is, would someone let me know how to do it and why it works?

Thanks.........Mark
Click to expand...
Mark, you've been given the answer and now I'll let you in on another trick. Shoot raw and reduce the image size in post processing because in general post processing can expend more computational effort to do a better job.
Click to expand...
Thanks everyone....I think I am now "good to go!"
 
Oy.

The D800 does faithfully record shot noise like any other camera. It does also exhibit some read noise, like any other camera, though somewhat less so. In that sense - and only that sense - you've gotten some answers as they might apply to some cameras.

Not to this one.

The biggest problem with the D800 in low light is thermal noise. The Exmor sensor, and especially the D800 sensor, has an enormous amount of active circuitry on the sensor. All of the amplification and A-D are done on the sensor. And not just that, but it is done for each column separately. The reason is that the Exmor depends upon a slow read method. The only way to get reasonable quality and reasonable throughput is to multiply out the processing elements.

This is a problem.

As you will plainly see if you examine a dark frame (at, say, ISO 12800, 1/40), there is prodigious self-noise that manifests as a bluish-magenta cast across the entire frame with local hotspots. Trying to use NR or color-balancing to remove it is hopeless. The effect is much worse after using live view.

There is one way to fix this, and that is to do a dark-frame subtraction. Using a dark frame taken with the same exposure, and captured with the same effective gain settings, the same white balance, and no other noise reduction applied, you can subtract this from the original image and restore the image. The blacks will return to black. The noise will cancel out to a very close approximation. You can do NR in addition to this, as others have suggested, but only a light treatment is needed, and mainly only the chroma noise reduction. This combined technique is the only way to truly do low light with this camera.

You might note that the recent medium format cameras that use the same sensor design (PhaseOne IQ-250, for example) do a dark frame subtraction automatically on every exposure. It is mandatory. Got it everyone? On every exposure, not just low light.

I've done album covers with this technique at ISO25600, in a room painted black. The results are impressive by the standards of any other camera. But you can't do it without the appropriate treatment.
 
Sagittarius said:
Hi,

I think what you mean is: if you reduce resolution from 36 MP to 12 MP (file size from 7360 x 4912 = 36 MP to 4300 x 2872 = 12 MP) you will reduce visible noise.
Click to expand...
This is not a complete answer. No fault of yours Sagittarius. Someone jumped the gun on this.
 
Luke Kaven said:
Oy.

The D800 does faithfully record shot noise like any other camera. It does also exhibit some read noise, like any other camera, though somewhat less so. In that sense - and only that sense - you've gotten some answers as they might apply to some cameras.

Not to this one.

The biggest problem with the D800 in low light is thermal noise. The Exmor sensor, and especially the D800 sensor, has an enormous amount of active circuitry on the sensor. All of the amplification and A-D are done on the sensor. And not just that, but it is done for each column separately. The reason is that the Exmor depends upon a slow read method. The only way to get reasonable quality and reasonable throughput is to multiply out the processing elements.

This is a problem.

As you will plainly see if you examine a dark frame (at, say, ISO 12800, 1/40), there is prodigious self-noise that manifests as a bluish-magenta cast across the entire frame with local hotspots. Trying to use NR or color-balancing to remove it is hopeless. The effect is much worse after using live view.

There is one way to fix this, and that is to do a dark-frame subtraction. Using a dark frame taken with the same exposure, and captured with the same effective gain settings, the same white balance, and no other noise reduction applied, you can subtract this from the original image and restore the image. The blacks will return to black. The noise will cancel out to a very close approximation. You can do NR in addition to this, as others have suggested, but only a light treatment is needed, and mainly only the chroma noise reduction. This combined technique is the only way to truly do low light with this camera.

You might note that the recent medium format cameras that use the same sensor design (PhaseOne IQ-250, for example) do a dark frame subtraction automatically on every exposure. It is mandatory. Got it everyone? On every exposure, not just low light.

I've done album covers with this technique at ISO25600, in a room painted black. The results are impressive by the standards of any other camera. But you can't do it without the appropriate treatment.
Click to expand...
Or he can wait for the next generation D900 with 65MP to print his 6x9 pictures.
 
You're quite accurate in what you say in your response. But you didn't address the OP's question. The question, simply stated, is why does noise get reduced if you downsample the photo.

Having supplied one of the previous answers to the OP I am a bit offended that you call all the previous responses "wildly incomplete". The OP asked why reducing resolution decreased the noise. The question was answered by many.
 
You are right about the thermal noise (aka amp glow) on the D800, 6400 and above. This problem with dark frame subtraction is that the "glow" is actually random red/blue speckles. Since they are random, they don't really subtract on a pixel by pixel basis. Thus, you are still left with much of the noise (though to a lesser degree, maybe 40% less). Subjectively, the photos will look a bit better.

Something I have done is to do the dark frame, then do a low pass on it to average out over several pixels. This gives a variable color cast in the dark image. Then subtract out this from the original .

Returning to the OPs original question, re-sampling to lower res will reduce the random shot noise - but won't fix this thermal glow.
 
michaeladawson said:
You're quite accurate in what you say in your response. But you didn't address the OP's question. The question, simply stated, is why does noise get reduced if you downsample the photo.

Having supplied one of the previous answers to the OP I am a bit offended that you call all the previous responses "wildly incomplete". The OP asked why reducing resolution decreased the noise. The question was answered by many.
Click to expand...
I apologize, Michael. I should not have used the word "wildly", and it was certainly not said to diminish the value of your reply. It is true that one kind of "noise" is high frequency noise that will get reduced upon downsampling, due to the fact that downsampling is a low pass filter, which eliminates high frequencies wholesale.

I should also say that when the word "noise" is used by itself, it does not imply only one kind of noise -- e.g., high frequency noise. For anyone who wants to achieve quality results in low light with the D800, the other sources of noise need to be taken into account.
 
breivogel said:
You are right about the thermal noise (aka amp glow) on the D800, 6400 and above. This problem with dark frame subtraction is that the "glow" is actually random red/blue speckles. Since they are random, they don't really subtract on a pixel by pixel basis. Thus, you are still left with much of the noise (though to a lesser degree, maybe 40% less). Subjectively, the photos will look a bit better.

Something I have done is to do the dark frame, then do a low pass on it to average out over several pixels. This gives a variable color cast in the dark image. Then subtract out this from the original .

Returning to the OPs original question, re-sampling to lower res will reduce the random shot noise - but won't fix this thermal glow.
Click to expand...
Interesting. I've been using a small amount of gaussian blur to achieve those results. I'm curious what you use for a LP filter in this case?
 
Luke Kaven said:
Oy.

The D800 does faithfully record shot noise like any other camera. It does also exhibit some read noise, like any other camera, though somewhat less so. In that sense - and only that sense - you've gotten some answers as they might apply to some cameras.

Not to this one.

The biggest problem with the D800 in low light is thermal noise. The Exmor sensor, and especially the D800 sensor, has an enormous amount of active circuitry on the sensor. All of the amplification and A-D are done on the sensor. And not just that, but it is done for each column separately. The reason is that the Exmor depends upon a slow read method. The only way to get reasonable quality and reasonable throughput is to multiply out the processing elements.

This is a problem.

As you will plainly see if you examine a dark frame (at, say, ISO 12800, 1/40), there is prodigious self-noise that manifests as a bluish-magenta cast across the entire frame with local hotspots. Trying to use NR or color-balancing to remove it is hopeless. The effect is much worse after using live view.

There is one way to fix this, and that is to do a dark-frame subtraction. Using a dark frame taken with the same exposure, and captured with the same effective gain settings, the same white balance, and no other noise reduction applied, you can subtract this from the original image and restore the image. The blacks will return to black. The noise will cancel out to a very close approximation. You can do NR in addition to this, as others have suggested, but only a light treatment is needed, and mainly only the chroma noise reduction. This combined technique is the only way to truly do low light with this camera.

You might note that the recent medium format cameras that use the same sensor design (PhaseOne IQ-250, for example) do a dark frame subtraction automatically on every exposure. It is mandatory. Got it everyone? On every exposure, not just low light.

I've done album covers with this technique at ISO25600, in a room painted black. The results are impressive by the standards of any other camera. But you can't do it without the appropriate treatment.
Click to expand...
Luke, So is the thermal noise why FX sensors that are less dense, ie: 16 MP vs 36, have less noise? And if so, why is the D4(s) only slightly better as per DXO?
 
Bluefin said:
Luke, So is the thermal noise why FX sensors that are less dense, ie: 16 MP vs 36, have less noise? And if so, why is the D4(s) only slightly better as per DXO?
Click to expand...
The main consideration in answering that question has to do with sensor architecture.

The D600/610 and the D800 both use Exmor sensors that perform all amplification and A-D conversion in massively parallel fashion. Each column of the sensor has a converter. The sensor relies on a slow readout (the D3x had a sub mHz readout) for lowest read noise and highest dynamic range. It runs hot. It has a low frame rate. But it produces very very clean results at base ISO.

In line with your hypothesis, the D600/610 would have less thermal noise due to having less active circuitry. These cameras are in a nice sweet spot. They produce very clean files with enough detail for good hair/skin/foliage rendering, even at web sizes.

DxOmark does not really isolate this kind of noise very effectively, and their numbers do not tell a complete story. I would not be surprised if they were not giving this issue some thought right now.

The D4/D4s/Df sensor is a Nikon design with a somewhat different architecture. Instead of a few thousand channels, like the Exmor, it uses 24 channels of on-chip processing. (The D3/D3s used off-chip processing with a set of 6 Analog Devices AD9978 chips.) In addition, the sensor uses readout circuits that are shared between adjacent columns and multiplexed, which cuts the amount of active circuitry dramatically (but also utilizes it more). Also, in a nice move, the sensor has deep wells with full-well capacity of 120k e- at ISO 100. The pixel count is kept low in order to ensure fast readout with good results. And it delivers in that regard with few compromises, and very low thermal noise.
 
Luke-

OK.......got the point of Dark Frame Subtraction.

I just searched the Nikon D800/D800E manual and the results yields zero.

How does a person actually do this with the camera?

I'd love to try it out ASAP.

Thanks..........
 
Lucky Sky said:
Luke-

OK.......got the point of Dark Frame Subtraction.

I just searched the Nikon D800/D800E manual and the results yields zero.

How does a person actually do this with the camera?

I'd love to try it out ASAP.

Thanks..........
Click to expand...
Long Exposure Noise Reduction (pg. 277 D800 USA PDF User Guide) will subtract a dark from for exposures longer than one second. So, if you turn on LENR and shoot a 2 sec. exposure it will take the shot and then take another shot with the shutter closed. The second exposure is subtracted from the first. The reason for this is that sensors get hot with long exposures. You tend to get more hot pixels the longer the exposure and with some cameras (D2x, and I think D700) you can get a glow from a nearby amp. The dark frame subtraction removes these problems.
 
Robin-

Thanks for the heads up on Long Exposure Noise Reduction.........will read up and give it a try.
 
Lucky Sky said:
Luke-

OK.......got the point of Dark Frame Subtraction.

I just searched the Nikon D800/D800E manual and the results yields zero.

How does a person actually do this with the camera?

I'd love to try it out ASAP.

Thanks..........
Click to expand...
I always thought that LENR should be a full-time option with this camera. Some of the newer cameras that use this kind of sensor (e.g., the medium format version in the PhaseOne IQ-250) do a dark-frame subtraction on every exposure.

Once you get to about ISO6400 and 1/40th, you will start to see the noise pollution. It's a matter of degree. The noise is always there. It's worse if you've been using live view. The blacks won't be black anymore, but a kind of blueish-magenta. There really should have been an in-camera treatment for this. It would be so easy for Nikon to do it.

What you'd need to do is take a dark-frame exposure at the same shutter speed and gain setting. Capture it with the same white balance setting, and no noise reduction. Next you can do one of two things. One, load both frames in photoshop and use the "Apply image..." option and choose "subtract" from the drop-down menu in the dialog box. Adjust percentage as needed. Two, load both frames as separate layers, with the original image on the bottom, and the black frame above it. Change the blending mode to "subtract", and play with the opacity until you get a good balance. The blacks should just return to black.

In some cases, just a bit of gaussian blur on the dark frame will average out the effects a bit. I created a library of a few dark frames that I can reuse. It is a bit of a pain, which is why I would not recommend using this camera in low light in a production setting with lots of images. It just takes too much time. The D4/Df is more obliging in low light with no questions asked.

It is helpful if you take a few moments with just the black frame and give it a good look to see what the thermal noise looks like by itself. Try adding a couple of stops of digital gain just to see the effects more clearly.
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top