D7000 at high ISO(again)

[arthurking83]

well the other day, I had the opportunity to have a quick play with a D7000, and you reckon I was going to say no!? (Thanks Helmut)

Of course I wanted to see how wel the camera handled high ISO, and from what I know of the D300, it's leagues above that(and hence the D90 and D300s too).
All images were raw mode processed the proper way via Nikon software.. but I did take the time with one version of an image to use LR3(considering that it has a very good noise reduction facility and USED to be miles ahead of CaptureNX2's ability to reduce noise. Well... that was until I got smart and really tried hard to get CNX2 to produce some proper NR editing.(if you are keen to know the guts of how to use NR in CaptureNX properly, I'm more than willing to provide some gory and boring details. But I will refrain from boring all other viewers with the tedium of such banal commentary.


NOTE: I had to go back into CNX2 to turn NR off, as Helmut had his D7000 set to NR on at high ISO(this switch is easily switched off in CNX on NEF images, so really makes no difference if you have it set to on or off. With NR turned on in camera there is a "develop" edit step in CNX where the NR used in camera is set to automatically be enabled. This can be subsequently turned off, and hence the extra applied NR is then not applied to the NEF. The camera does apply some noise reduction anyhow, but this is the intricate workings of the camera itself and cannot be configured by the user anyhow.. that's just how Nikon cameras work.(ie. if you don't like it, then bad luck!)

#1 NO NR at all.

you can see some colour(chroma) noise even in the full frame, viewed at 100% pixel level it may look disappointing with lots of red channel noise(but remember this is ISO12800!)

#2 NR applied in LR3, which I immediately liked for it's super easy to use ability(despite my lack of respect for LR3, and it's annoying quirks)

One thing I really tried hard to do was to get the image looking as close as I could to what ViewNX would produce. Using LR's idea of Nikon's Picture Control settings wasn't even close for starters, and it needed a few other tweaks. One thing I I finally gave into was the 'over saturation' of that gold colour on the lettering. it took me too long to get the images looking close to what I expect from Nikon.. so for the sake of comparative effort, I'm surprised I persisted for so long with this single image in LR3(I give up easily with PP). But as I already said, LR's noise reduction is brilliant for a non dedicated NR reducing software(like Neat Image or Monkey Ninja Noise Dfiner.

That was until I started playing with NR in CNX with a bit more dedication.

You simply can not do this in CNX using the Develop tool's NR reduction facility. Hence if you use CNX and want noise free images, remember to turn NR in camera to off, and do NR correctly via the edit step process. If you have NR in camera turned on, and you want more detailed NR edit steps too, you will find that it will be frustrating, because you end up going back to the Develop tool and turning it off anyhow!(so do your self a favour). Unfortunately there is no NR on off switch available in ViewNX.. so a loud complaint to Nikon to add one may be appropriate.

all images are the same image processed differently, with different software.
With image #1 all I did was to reset the Picture Control set in camera from Neutral to Standard via VNX(which BTW produced more noise) and then open the image in CNX to switch off NR set in camera. #3 was not processed in any way using CNX other than two NR edit steps. I think the only other 'processing' made to the two Nikon versions was setting Picture Control to Standard, and also WB to 5700K(also done in LR3 on that version. No brightening, no darkening, no curves nada! Just the two simple edit steps in CNX2.

#4 Again in CNX2 and only NR edit steps. ISO25600 this time and slightly over exposed.


and for pixel peeping confirmation that there is both great detail still left intact and almost zero chroma noise in the image...
#5

the one or two percent 'hot pixels' is more than acceptable unless 100% detail retention is of absolute importance.

Now remember that the last two images are at the highest ISO setting on a Dx camera and the level of detail is (what I believe to be) amazing.. in coming from a D300. From my limited use and images I have from a D700, I still think that the D700 is better at ISO6400 in terms of noise quality, have no images made at ISO12800 and above from the D700, so I can't comment on those settings. But the D7000 is very usable.

If you are a Nikon software devotee, I can try to upload the NR edit steps I made for your perusal via the software itself. This would be easier than to describe the boring bits in detail and for you to subsequently find that it doesn't work for your camera, or on any particular exposure.
I haven't yet tried the edit steps on any of my D300 files as I don't have ISO12800 or 25600 available on that camera

[I @ M]

You made me sit down and experiment between heaps of other chores.

I hadn't considered noise reduction, especially the C NX version, on very many of the D700 files so I went and had a play with a couple from the same day as your shots. I managed some very credible efforts from C NX after experimenting with the varying opacity layers and it certainly does compare with Neat Image (but still a little slower) for final output.

I think that the C NX noise reduction is well suited to the type of noise rendered from the CMOS sensors and Neat Image is better at dealing with CCD sensor noise because after TONS of experimenting with the D200 files in C NX I could never get them as good as Neat Image.

[arthurking83]

As I said, each camera will need different levels and methods of NR, and the D7000 was supremely easy to produce chroma noise free images.. down from like 100% levels to close to zero(or may 1%).
The trick to getting good quality is changing the opacity in the blending mode, firstly from the [all] setting to [luminance and chrominance] leaving chroma levels at 100% and only lowering the luminance level to zero and working your way up as required.
The luminance level affects image sharpness, so the more luminance you have set in percentage terms the more image detail you lose as well.
I reckon for lower ISO noise levels you could probably get away with lowering the chroma levels as well.. all in all, this method makes noise reduction via CNX a viable proposition.

You can also change to the RGB blending mode as well and varying each colour level according to what colour noise is present. But you need to be more careful of adjusting noise levels on a channel level, as too much imbalance causes a colour cast on the image.

Note too tho, even tho CA was pretty ineffectual in that image(maybe 0.1pixel or so), and so the CA removal tool in CNX wasnt' really required, I went back into the image this morning to have another play with it and set the CA tool to 100% and it removed a slight green tint on the blacks. so the CA removal tool acts as a pseudo NR filter as well(maybe not in every case.. but at least in this case :th3)

I'm attaching the noise edit steps I've made in CNX2 which I saved as a batch job.
All you need do is to extract the zip file into it's .set format and use the load adjustments option in the batch area in CNX and point it to the noise reduction.set file. Then you can easily see what the edits are doing for yourselves, and tweak as you please.

These steps don't work well in my D300 files at high ISOs.. considering how well they've worked on the D7000 files, so I'm about to mess about with a few more variables to see how far I can clean up the D300's high ISO 3200 and 6400 NEFs.
The problem so far has been that in removing the chroma noise(which is the ugliest looking noise) it leaves behind a harsh looking white speckle, which tries to look like grain, but look more like the hard bokeh of a kit lens.. even with the 105VR!

neat Image is great and I love it to bits.. but I'd prefer clean NEF files as a priority.. and only really use it on tiff files because there used to be no other choice.
I'm going to persevere with NR in CNX to get as clean files as possible and then maybe add a lot less using NI if need be.

[I @ M]

The same image reproduced with 3 different noise reduction settings.
Your CNX preset, my CNX experiment (similar to yours) and then Neat Image.
All 3 have had the black and white control points applied to the same place and then brightness levels adjust to "somewhere near" the same as they varied quite a bit between versions, Neat image was the lightest, your preset was the darkest and my CNX version was in the middle.
The shot isn't close enough to pixel peep but I was after a comparison of more the full sized image.

AK83 preset.


My CNX experiment.


Neat Image.

[reaction]

your photos are well exposed, but where D7000 supposedly shines is in base noise
if you could compare 1EV underexposed photos in a dark room (simulating a party or dinner) pushed back in p/p you 'should' notice it has less noise than older offerings.

[I @ M]

Why would anyone want to shoot underexposed photos in a dark room?
Why would you want to rely on p/p to rescue photos when they can be taken without extensive or indeed minimal p/p?
Why would anyone want to shoot a party or dinner at the base ISO when the ability is there to have perfectly acceptable images at higher ISO levels.
Seeing as technology improves all the time and 2 years is a long time in camera models it is only to be expected that image quality at both base level and high ISO levels are going to improve.

[arthurking83]

Yeah!... apparently the D7000 has a lot of shadow recovery headroom at low ISO levels.
Forgot to test this out whilst I had the opportunity to play with the D7000, as we were busy helping out the owner with camera settings that beep! (seriously!!)

(for me) If there is any reason to switch from a D300 to a D7000, it would be for that reason. But I still remain unconvinced.

I can see perfectly valid reasons to shoot underexposed and so forth when dynamic range may be too great for the sensor to cope with, and heavy reliance on shadow recovery is then required too.

My biggest problem with the D7000 is that it was very uncomfortable for me to hold. Grip is small, compared to the D300.. and even the D70s! After even a short while of holding the camera, my hand would cramp up I reckon and I'd probably have to rely on the shoulder strap more... which I also hate(I don't use mine)

Without the benefit of access to either camera, I think that in an overall sense, and up to at least ISO6400, that the D700s noise handling ability is better(at least in NEF mode). How each camera handles and processes jpgs is usually a win for the newer camera, and the D7000 is now 2 generations ahead of the D700, so that we'd be within our rights to demand that a D7000's jpgs at just about any exposure level would turn out to be the equal or better of a D700.
I'd totally forgotten that the D700 is ISO25600 capable too, I only thought it did ISO12800.

[reaction]

Why would anyone want to shoot underexposed photos in a dark room?
Why would you want to rely on p/p to rescue photos when they can be taken without extensive or indeed minimal p/p?
Why would anyone want to shoot a party or dinner at the base ISO when the ability is there to have perfectly acceptable images at higher ISO levels.
Seeing as technology improves all the time and 2 years is a long time in camera models it is only to be expected that image quality at both base level and high ISO levels are going to improve.

If you read about what the difference btwn the D7000 sensor and other sensors you will know why that test is important.

If you want to test brightly lit well exposed photos, that's your choice. I'd wager a D7000 beach scene at ISO 25600 would look equally good as a $100 P&S set to 'beach mode'

Now you may never shoot underexposed photos, but I'm sure many others here sometimes either mistakenly or due to lens+ISO combos choose to take underexposed photos rather than to blur or miss a shot. Not everyone owns a flash or knows how to set their camera settings optimally. Should they give up before they even started?

You can also say why would anyone want to take photos of the barrel of another lens, but the thread about testing the shadow recovery of this new tech, not whether a photo deserves a prize!

P/p is needed for this test because a noisy black photo and non-noisy black photo are hard to distinguish.

[arthurking83]

No one has ever said that newbies should give up because they can't understand the concept of setting their camera up optimally.
But the intent in Andrew's reply is of a sound nature... get your exposure right and noise becomes less of a problem.
So that, if newbies can't get it right, they should both ask about, and experiment with, the various settings available to them to effect an appropriate exposure.

I can't be sure as to which of either the D700 or D7000 will have better recovery in the shadows, but at the moment, I dare say that the D700(no typo!) may be better at shadow recovery.
D7000 may have a slight edge in actual noise reducing ability at high ISOs(6400 and above) where noise is heavily chroma based(remembering there are different types of noise to deal with)

.....

If you want to test brightly lit well exposed photos, that's your choice. I'd wager a D7000 beach scene at ISO 25600 would look equally good as a $100 P&S set to 'beach mode'

........

my tests shots were random and interfered with(as we were more interested with helping the D7000 owner with settings.. I was really only fooling around. Blacks and noise can be a big issue in a lot of situations even in very good light.
The images I shot there of the lens was with a very high quality Sigma 150macro lens, and the purpose of the test was to see how well detail is maintained .. as seen in the texture of the lens's surface.
From experience, I'd never shoot that on my D300 at anything above about ISO800.. maybe ISO1600, but with the D700 ISO6400 is a cinch and still maintain good detail levels.

I think that people need to get this idea that high ISO and noise reduction is or should be limited only to low light photography. Anyone that thinks in that manner, is basically limiting their scope for capturing images with greater variety.
Even in a low light environment, it's a trivial matter to create the illusion that the conditions were actually brighter than the reality. Low light bright light... I don't think it's of any consequence to how noise is dealt with.

BTW Andrew's question as to why you would underexpose a dark scene makes perfectly good sense.. why would you? Especially at higher ISOs. If there is going to be underexposed elements in the scene due to dynamic range.. well that wasn't what I had in mind at the time when I was playing around with this D7000. I'll look into that the next opportunity I have to play with it.
My part of the thread is not about shadow recovery. All the test images I captured with the D7000 in the limited time I had available, were set to slightly overexpose with blowing too much in the highlights. I darkened all images appropriately to minimise the impact of noise in the image.

ATM, from what I can see and have seen(with the software I have available to me).. the D7000 looks to be at least 2stops better than the D300 in an overall sense, where overall sense means not just noise levels but quality of bokeh and detail retention and all that stuff too(that no one else really seems to worry about). Noisy D300 images recovered to display less noise always seem to kill the quality of the OOF areas too, where it tends to go harsh even if it was smooth to begin with. Much less so with the D7000.

[reaction]

I duno, when we're talking about testing high ISO, and noise, then what we should be doing is trying out things at high ISO that show a different result with various cameras.

"Why would you want to " shouldn't come into it all, and if you want to talk about 'correct technique' or 'correct settings' all the time then testing is not for you. After all, "Why would you want to shoot at ISO25600 at all?"

When someone tests a lens' bokeh you don't criticise their framing and composition, or whether their subject was posed correctly. What's wrong then with using p/p to enhance noise in a noise test to ease comparison?

And base noise does NOT mean noise at base ISO. There are different components and causes of noise, and there are different ways to test for each - none of these tests are things you would want to do in real life shots.

[Sar NOP]

if you could compare 1EV underexposed photos in a dark room (simulating a party or dinner) pushed back in p/p you 'should' notice it has less noise than older offerings.

Hi everyone,

I got my D7000 two weeks ago and now it has more than 3,000 actuations !
I mostly use it at its high ISO and am very impressed by its noise performances. My main subject : wildlife photography.

To show you how this little camera can handle severe underexposure at its very high ISO, I took a few shots of these 20 dollars notes :


Exifs : D7000, AI-S Micro-Nikkor 105mm f/4, f/8, @12,800 ISO. Subject lit by a single fluorescent tube at 2.5m high.


#1 : 1/15", image correctly exposed.




#2 : 1/50", image underexposed 1.83 EV.




Now, I reset the exposure of image #2 to the same level as the image #1 and show you the crop 100% of each shot :

Crop image #1



Crop image #2

[farmer_rob]

Thanks SarNOP - that clearly illustrates to me the value of exposing correctly to start, even at high ISO. (BTW, looking forward to seeing your pictures from the D7000 - how is it going against the D2x?)

[kiwi]

Nice tests. I think I'll stay with my d3. The noise in the blacks when pushed is what counts

[Sar NOP]

Thanks SarNOP - that clearly illustrates to me the value of exposing correctly to start, even at high ISO. (BTW, looking forward to seeing your pictures from the D7000 - how is it going against the D2x?)

Up to 1600 ISO, I can't see any difference in noise when compared the D7000 pictures to my NEF files from a D700.
Actually in image #2 above, the 1.83EV underexposure is equivalent to a 40,000 ISO correctly exposed !!!
From 1600 to 25600 ISO, when using Lightroom 3.3 to remove noise the D7000 NEF files will look the same as the NEF from a D700, especially when downsized (from 16mpix to 12mpix).

Compared to the D2X at 100 ISO, the D7000 is not as good as the D2 serie (the D2X is much sharper).
Have a look at the detail of this image :

D2X, 200-400 VR, 1/125", @400mm, 200 ISO.




Crop 100%




For most situations, IMHO the D7000 is the best value for money in all Nikon DSLR range for its overall performances. For same the price of the D3s, I prefer to buy a D7000 and a 200/2 VR !

For my wildlife work, the smoothness of its shutter/mirror system and its performances at high ISO are a real bonus.

D7000, 200-400 VR+TC-20EIII, 1/160", f/9, @700mm, 8 000 ISO, HH.
1800x1200px

[arthurking83]

Nice tests. I think I'll stay with my d3. The noise in the blacks when pushed is what counts

Apart from the narrower FOV, and hence apparent extra reach of the D7000 with the same lens, I can't see any reason to go from a D700/D3 to a D7000 either.
D3s is supposedly better again in noise terms, but as Sar said.. that's the price of a D7000 + 200VR.. and I'd have the 200VR over any body... any day

That under exposed image of the $20 note that Sar posted isn't even badly underexposed and in recovery you can clearly see that detail is lost in the fine writing(twentydollarstwentydollars). The 'purple fringing' in the high frequency pattern on the RHS could almost certainly be removed in post as well, but once you've lost detail, as in the writing here, well.... gimme a well exposed image to begin with any ol time. Hence the why would you want too idea coming into play. And As Andrew said.. expose correctly in the first place.

.....

"Why would you want to " shouldn't come into it all, and if you want to talk about 'correct technique' or 'correct settings' all the time then testing is not for you. After all, "Why would you want to shoot at ISO25600 at all?"

.....

Why would you want too is always relevant when producing tests.. the two are almost inseparable, otherwise what's the point of testing anyhow?

Once the test is complete, the final analysis is generally of a why would you want too or why would you not want too nature.
How that translates in this particular test is that with the D7000, why would not you want to shoot at high ISO(to get higher shutter speeds(even in bright midday sunlight!) ... or conversely why would you want to shoot underexposed(by too much) when the evidence is clear that shooting with a brighter exposure ensures better detail.

As for noise(in digital images) there are really only three types of noise that affect images in an adverse manner(not taking banding into account) chroma, grey and black, and they each have a different effect on the image, and each camera has different levels of each type.
From what I've seen so far, the D300 has too much luminance noise(grey) at high ISO(D300 can also take into account D300s, D90 and D5000(as I know of them). With that you lose sharpness detail.
Chroma noise is also higher for a given ISO setting, but is easily removed especially as NR software gets better so this is less of a problem. One thing with using CNX for removing chroma noise in D300 images is that the OOF sections(bokeh) take on a much harsher look, as the chroma affected pixels will usually be 'whited out'(when viewed at 100%). When viewing the image as a whole tho, this then translates into a harsh rendering, and some Gassian blur helps. but that means more PP. D7000 eliminates that need(even in CNX).

Haven't yet tried out comparative images in Neat Image(but I know that program does a better job than CNX does).

So far, as it stands, I can't see any reason (in my case) to go from a D300 to a D7000, but if this is any indication than a D400 wil make a compelling upgrade path. I really need a bigger camera for my hands. I could however see a very good reason to retire the D70s, and go with a D7000. For sure! A lot of money to spend on a 10yo tho, and I could use the D7000 on a tripod only. I have the D70s set to AutoISO for my sons use and combined with the 18-105VR lens he's able to get a decent keeper ratio. The D7000 will ensure a higher keeper ratio with it's better AutoISO ability.
I'd still prefer a D700 tho.

[Sar NOP]

When correctly exposed, the results are just amazing @25,600 ISO on a consumer DX sensor :




Crop

[farmer_rob]

...So far, as it stands, I can't see any reason (in my case) to go from a D300 to a D7000, but if this is any indication than a D400 wil make a compelling upgrade path. I really need a bigger camera for my hands. I could however see a very good reason to retire the D70s, and go with a D7000. For sure! A lot of money to spend on a 10yo tho, and I could use the D7000 on a tripod only. I have the D70s set to AutoISO for my sons use and combined with the 18-105VR lens he's able to get a decent keeper ratio. The D7000 will ensure a higher keeper ratio with it's better AutoISO ability.
I'd still prefer a D700 tho.

I don't think there is a compelling reason to go from a d300 to a d7000 - but that does not take away from the quality of the d7000. Sometime in the future, Nikon will be releasing a) a d300s replacement, b) a d700 replacement and c) a d3s replacement. This is not rocket science - however the timing will be opaque and rumour-driven. The d300s replacement will not compel a change from a d7000, but may be from a d300. The d700 replacement will compel some to change from either a d300/s or a d700 - if you want FX. The d3s replacement will compel change for some from all four cameras - d7000, d300/s, d700 and d3/s. Just look at the current line-up of cameras - currently 2 consumer (d3100/d5000), 3 pro-sumer DX (d90/d7000/d300s) 1 pro-sumer FX (d700) and 2 professional FX (d3s/d3x). There are 4 clear categories, and it is serving Nikon well - each product release gives a reason to move up a level for some photographers, and to a new camera at the same level every second generation (eg. d80 to d7000, but not d90 to d7000). I can't see Nikon moving away from this layering, but I can see a reduction of numbers with the d90 going. Maybe 2 consumer (d3100/d5500?), 2 pro-sumer DX (d7000/d400?), 1 prosumer FX (d800? or d8000?) and 2 professional (d4, d4x). (Although I can see an argument to not have a d300s replacement, but instead have the d7000 as the flagship DX and try to migrate d300 photographers to a FX d700 replacement - it depends on fitting between the Canon ranges.)

All of the cameras coming after the d7000 *will* have better ISO performance - tech doesn't go backwards. All of the newer cameras will have higher pixel counts. The top-end cameras will all have newer AF modules. The remaining features will be tweaks (including improved video).

So, examine the high ISOs above carefully - I think this is indicative of future directions.

[Sar NOP]

I don't think there is a compelling reason to go from a d300 to a d7000 - but that does not take away from the quality of the d7000.

For my photography style (action & wildlife), there are two good reasons to go from a D300 to a D7000 :

1- Frame rate : 2.5 fps @14-bits RAW on the D300 vs 6 fps @14-bits RAW on the D7000,

2- ISO range & performances : 200-6,400 on D300 vs 100-25,600 on D7000.

[kiwi]

How relevant is that fps for raw though when the buffer is so much smaller? I don't know how big the buffer is but that's one reason I'm happy with my d300s for sport anyhow though still won't go past iso 2000 on it

Theres no doubt the d7000 is a fine camera and all these examples prove it (thanks guys)

If I were in the market for a new general purpose body it's a no brainer I think right now

[Sar NOP]

On the D7000 when most in-camera settings are turned off, the buffer can take 10 NEF @14-bits at 6 fps.
I'm not sure about the D300...maybe around 13 NEF ?