Seeking some clarification - DX vs FX

[bconolly]

Hi Everyone,

Just seeking some clarification on the ISO performance DX versus FX. By way of background, I shoot a lot of low light shots mainly of my little boy (soon to be joined by his little sister) with my D7000. I'm not in any major hurry to upgrade to FX but will head in that direction. One thing that may sway me though is the ISO performance. I'm very happy with the noise up to 1600 on the D7000, but above that the loss of detail is noticeable. My question is - if I were to go FX (say D600) my thinking is that I should pick up about 1 full stop of ISO performance so that 3200 on the D600 would be roughly equivalent to 1600 on the D7000. Is that roughly right or am I dreaming?

TIA!

Brenden

[ricktas]

Low light performance is not FX DX relative, it is about pixel density more than anything. Take a D3 with 12MP and a D800 with 36MP on the same sized sensor (FX), the D3 will outperform the D800 as each pixel site is larger, thus allowing more light to hit each pixel site. The smaller the pixel sites the less light particles hit it and thus in low light/high ISO performance, noise becomes more prevalent in a higher pixel density sensor. However, tech is improving and pixel density is not the ONLY factor.

You can shoot at ISO 1600 relatively cleanly with a D3, but at ISO 800 a D800 is starting to show considerable noise in low light

[bconolly]

Thanks Rick! So looking here: http://www.digicamdb.com/compare/nikon_d7000-vs-nikon_d600/ the D600 has about a 55% lower pixel density than the D7000 (which seems pretty significant?). Along with the 56% larger pixel size in the D600 that's looking promising.

[ricktas]

Thanks Rick! So looking here: http://www.digicamdb.com/compare/nikon_d7000-vs-nikon_d600/ the D600 has about a 55% lower pixel density than the D7000 (which seems pretty significant?). Along with the 56% larger pixel size in the D600 that's looking promising.

exactly.

[swifty]

This may be controversial but from my understanding, pixel density only matters when evaluating noise at pixel level. Yes, a larger pixel surface area can collect more photons but if there are 4 smaller sensors pixels occupying the same surface area as the one large one, does it collect any less given we have gapless micro lens over each pixel.

When evaluating the entire image, the two most important aspects are total sensor area and underlying sensor technology, regardless of pixel density. The identical technology part is a bit tricky since increasing pixel inevitably changes the tech in one way or another and manufacturers constantly tweak and improve their sensors on subsequent models even if the underlying designs are similar/same.

For identical/similar sensor technology, FX should perform a little better than 1 stop over DX purely due to the increased total light collecting area.
Its useful to look at some recent Sony sensors eg. the D7000 vs D800. They appear to be very similar Sony tech (D800 being a newer iteration), scaled from DX to FX and ISO performance are as predicted. D800 is better than a D7000 by more than a stop when looking at the whole image but is very similar at pixel level.
It is less useful to compare a D3 sensor and a D800 since a D3 uses a Nikon designed sensor.
But looking at a D3->D3s->D4, all have Nikon designed FX sensors although the manufacturers are unknown (likely same for D3 and D3s but different for the D4) D3-D3s yielded about a 1/2 stop improvement despite the same pixel count (tech improvement) whilst D3s->D4 yield similar performance (D3s slightly better) despite an increase in pixel density on the D4 from 12mp to 16mp.

From the samples I've come across, I disagree the D800 performs worse than a D3 (D700 in my case which has an identical sensor to the D3), despite having 3x the pixel count and hence higher pixel density. In fact I think a D800 performs better due to superior tech. But viewing both files at 100% on screen is not a fair comparison since you're at a higher magnification on the larger pixel count file when viewed on screen at the same percentage reproduction. I think this is probably where most ppl's opinions about higher pixel density sensors are formed.
But use the entire image and resize both to the same final size (pick any size whether to enlarge or reduce) and then that would be a fair comparison.

Like I said, a bit controversial and definitely not universally accepted but from what I've read I accept this explanation currently.

- - - Updated - - -

So it seems the D600 also uses a Sony sensor so I'd expect it to perform almost identically to the D800 in terms of high ISO despite having 24MP vs 36MP, and a bit more than 1 stop better than the D7000 when looking at the whole image and also a little better at pixel level.
Just my 2c

[bconolly]

Thanks swifty - awesome feedback!

[ricktas]

This may be controversial but from my understanding, pixel density only matters when evaluating noise at pixel level. Yes, a larger pixel surface area can collect more photons but if there are 4 smaller sensors pixels occupying the same surface area as the one large one, does it collect any less given we have gapless micro lens over each pixel.

When evaluating the entire image, the two most important aspects are total sensor area and underlying sensor technology, regardless of pixel density. The identical technology part is a bit tricky since increasing pixel inevitably changes the tech in one way or another and manufacturers constantly tweak and improve their sensors on subsequent models even if the underlying designs are similar/same.

For identical/similar sensor technology, FX should perform a little better than 1 stop over DX purely due to the increased total light collecting area.
Its useful to look at some recent Sony sensors eg. the D7000 vs D800. They appear to be very similar Sony tech (D800 being a newer iteration), scaled from DX to FX and ISO performance are as predicted. D800 is better than a D7000 by more than a stop when looking at the whole image but is very similar at pixel level.
It is less useful to compare a D3 sensor and a D800 since a D3 uses a Nikon designed sensor.
But looking at a D3->D3s->D4, all have Nikon designed FX sensors although the manufacturers are unknown (likely same for D3 and D3s but different for the D4) D3-D3s yielded about a 1/2 stop improvement despite the same pixel count (tech improvement) whilst D3s->D4 yield similar performance (D3s slightly better) despite an increase in pixel density on the D4 from 12mp to 16mp.

From the samples I've come across, I disagree the D800 performs worse than a D3 (D700 in my case which has an identical sensor to the D3), despite having 3x the pixel count and hence higher pixel density. In fact I think a D800 performs better due to superior tech. But viewing both files at 100% on screen is not a fair comparison since you're at a higher magnification on the larger pixel count file when viewed on screen at the same percentage reproduction. I think this is probably where most ppl's opinions about higher pixel density sensors are formed.
But use the entire image and resize both to the same final size (pick any size whether to enlarge or reduce) and then that would be a fair comparison.

Like I said, a bit controversial and definitely not universally accepted but from what I've read I accept this explanation currently.

- - - Updated - - -

So it seems the D600 also uses a Sony sensor so I'd expect it to perform almost identically to the D800 in terms of high ISO despite having 24MP vs 36MP, and a bit more than 1 stop better than the D7000 when looking at the whole image and also a little better at pixel level.
Just my 2c

sort of..yes...

remember we are squeezing 36 million pixel sites on to a sensor, so each sensor site is damn small. So what we are talking about is almost single photons (or very small numbers of) hitting each pixel site. The smaller and tighter we pack these sites, the less photon hit each one. The science of it all gets quite in-depth, cause then you add the electrical interference from the circuits in the sensor, etc and then add in higher ISO's and it all gets down to physics.

[geoffsta]

I'd like to see arthurking's, I @ M's and Lance's take on this conversation..... From all I have seen about the D800, on here and through various reviews obtained mostly from this site. That the D800/800E are excellent in low light.

[ricktas]

I'd like to see arthurking's, I @ M's and Lance's take on this conversation..... From all I have seen about the D800, on here and through various reviews obtained mostly from this site. That the D800/800E are excellent in low light.

It is, until you really push it. At ISO's of 2400 and more in low light it tends to not be as competitive as current models with less pixels.

[swifty]

I'm more than willing to concede there's limits to how high the density can get so I have to retract my words and say that pixel density may be a factor. Just that anecdotally I haven't seen evidence to suggest we've hit that density so currently it appears, at least to me, that the 2 major factors are total sensor area and technology ie. Signal-to-Noise-Ratio (quantum efficiency of sensor which is the signal part relative to noise - read noise really since I don't think you can change shot noise).
Again, I'm not an engineer but do read some of the theories presented from those apparently in the know. And it appears to fit my real life observations so this is just my current accepted theory. More than happy to be updated or proven otherwise.

[Lance B]

Actually, the D800 just slightly outperforms the D3 at all ISO's (although it is so close as to probably call it equal) and both outperforms the D7000 by about a stop. See DXO Mark:
http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/%28appareil1%29/792|0/%28brand%29/Nikon/%28appareil2%29/680|0/%28brand2%29/Nikon/%28appareil3%29/438|0/%28brand3%29/Nikon

The D800 and D600 are line ball as far as noise is concerned and both outdo the D7000 by about a stop:
http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/%28appareil1%29/834|0/%28brand%29/Nikon/%28appareil2%29/792|0/%28brand2%29/Nikon/%28appareil3%29/680|0/%28brand3%29/Nikon/original.jpg

The other benefit of FX over DX is Dynamic Range which is alos a big benefit. The D800 and D600 both outperform the D7000 by about a stop of DR which is very handy for not blowing highlights or losing shadow detail.

Don't forget that these benefits will mostly only be seen when using RAW.

Whatever the case, you can further get better results by using a top shelf post process noise reduction program like Noiseware Professional which plugs into Photoshop.

[ricktas]

Actually, the D800 just slightly outperforms the D3 at all ISO's (although it is so close as to probably call it equal) and both outperforms the D7000 by about a stop. See DXO Mark:
http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/%28appareil1%29/792|0/%28brand%29/Nikon/%28appareil2%29/680|0/%28brand2%29/Nikon/%28appareil3%29/438|0/%28brand3%29/Nikon



Which I always find interesting cause in real world situations my D3 outperforms my D800 for low light work above about ISO 800.

[arthurking83]

I like Swifty's response to this thread.

You can't really compare sensors of varying pixel levels directly with respect to noise at a pixel level only. The entire image should be taken into account, and with that, the output resolution of the final image.

Where the D800's lose out in noise comparisons against lower pixel count sensors, they usually make up for it in comparison where the image output is equalized.

A couple of other caveats with this noise comparison too tho, are related to colour reproduction(or accuracy of colour) and also the software used to render raw files.

Even tho the D800 may come out ahead in a noise comparison against lower pixel sensors where output is equalized, in some instances the lower quality noise it produces at the pixel level can lead to lower quality colour in the image, even in an equalized output comparison. The simple fact is that if noise at the pixel level is not very good, then the image won't render colour as well as the sensor with higher quality noise suppression at the pixel level.
Equalizing the output resolution is an important point in the comparison, but not the only manner in which to assess noise quality. Downscaling an image always masks the aberrations within the image, and noise is no exception.

Also on the topic of noise, I've noticed that various raw converter software also impact on the quality of the noise, even at the pixel level.

I have at my disposal 3 different raw conversion software, each with their own interpretation of how a raw image should be rendered.
They are CaptureNX2(Nikon), Lightroom 4.4 and Silky Pix Developer Studio 4.0.

While I am much more fluent with using CNX2, this doesn't impact in the demosiacing raw file.
My fluency with any of the software only impacts my workflow speed and ability to produce a final image.
So, with all noise reduction in software turned off, to create an untouched rendering of the raw file as can possibly be made, there is a clearly marked difference in the way that noise is rendered.
Using a D800 NEF shot at ISO6400, CNX2 a clear winner followed by Lighroom4 (an easy second) and Silky Pix (easily) last in this comparison.
Same raw file was used for each software, and this is not a test to see how well each software can reduce noise, but a simple test to see how each software renders a raw file in a basic, raw, untouched, unmodified state without any interference(as best as I could possibly set). And it must be noted that with Silky Pix, I can't even work out how to set NR to completely zeroed out settings, other than to say they look zeroed out going by the quality of the image, but the actual noise reduction values are greyed out and can't be set to any other value manually .. only via some preset edit. Even taking that into consideration, Silky Pix definitely renders much more (colour) noise in the raw file.

So there's not only the comparison between sensors quality of noise to be made here, but also the software used to make the comparison needs to be taken into account in this comparison as well.
The problem with comparisons is that while we can assume that using one software to compare different sensors should provide for some consistency, unless we know absolutely that the software developer has the same understanding of each sensors democaising requirements, the results may actually not reflect the true status.

I'm more inclined to believe that the manufacturer of the hardware has a fuller understanding of how the sensor will produce the raw data in the file, so if a comparison between raw file noise quality were to be as fair as it could be, I think the manufacturers software should be used to balance out the software equation.
If you use ACR to evaluate different sensors for their noise quality, you may not be evaluating the differences between the sensors noise only, but you may also be evaluating that Adobe has a better understanding of sensor A compared to sensor B.
I don't think that it's too far fetched to believe that the manufacturer has a full and complete understanding of their own sensors too tho.

I haven't had all that much experience with the D7000 other than a few quick plays with a couple of them, and then about a week with a D5100(same sensor) .. and all I can say is that ISO6400 and even ISO12800 are easily very usable values for real world usage. That is, if you don't pixel peep to the nth degree, or if you don't crop the bejeezuz out of every image!
I've captured a macro image with a D7000 shot at ISO12800 and you'd be hard pressed to see it in the image. From memory only a very slight(but carefully crafted) tweak of NR using both LR3(at the time) and then subsequently with some via CNX2 as well.

So, to the OP: I'd say if you can't get usable images shooting at ISO12800, you're probably doing something wrong either at the fundamental level(in camera), and/or at the secondary level(processing).
If I had a D7000, I'd have it set to AutoISO with ISO Hi1 as my upper ceiling.(note that this is how I have my D800E setup too tho).

I can post any images up if required, other than a crop of the Silky Pix- with zero noise reduction sample! ... but only because I can't figure out how to crop the image :p

[I @ M]

Brenden, you have been handed a heap of good advice and points to consider and I agree with a lot of it and I would like to summarise with my thoughts.

Your signature says "various el cheapo lenses" so I would advise you yo stick to the D7000 and upgrade your lens collection a little, learn and use good software to edit your (NEF) images and have them printed at a reasonable size by a good print lab.

I will stick my neck out and guarantee that you won't see noise in a 12" x 18" print of a properly exposed and processed D7000 NEF file at 3200 iso. :)

[Lance B]

Arthur is correct. You have to compare images from a D800, D3, D600 and D7000 at the same viewing size not at the pixel level to gauge an accurate indication of what the camera is like for noise. We view images on our screens and print them to certain sizes and they are the size we should check for noise not at the pixel level as it is an unfair comparison and not how we view our images. However, even at pixel level, the D800 and D600 are extremely good at high ISO and would still be more than a match for the D7000. I have had both the D7000 (now sold) and own a D800E (did have aD800 but sold it) and the noise from the D800E is still better than the D7000, and at pixel level it is still a tad better than the D7000, but we are splitting hairs at the pixel level.

I shoot birds mainly, and many times I have to shoot in lowish light with long lenses and in order to freeze motion of the birds I need high shutter speeds and this requires the regular use of ISO6400. This is to ahow you what you should be able to achieve very similar noise results as these below with a D600 if you use similar ppst processing.

D800 + 500mm f4 VR, 1/320sec, f5.6, ISO6400.

http://www.pbase.com/lance_b/image/143716479/original.jpg

Crop of above to DX size:

http://www.pbase.com/lance_b/image/143716480/original.jpg

D800 + 500mm f4 VR, 1/250sec, f5, ISO6400.

http://www.pbase.com/lance_b/image/143238579/original.jpg

Crop to DX:

http://www.pbase.com/lance_b/image/143238630/original.jpg

This is stupidly low light as attested to by the very low shutter speed for a 300mm lens handheld. This is less than 3EV light level!:
D800 + 300mm f2.8 VRII, 1/60sec, f3.2, ISO6400 cropped to square

http://www.pbase.com/lance_b/image/144063024/original.jpg

Another very low light level shot, equivalent to about 5EV
D800 + 300mm f2.8 VRII, 1/50sec, f5, ISO6400 cropped to DX

http://www.pbase.com/lance_b/image/145737024/original.jpg

Cropped to less than DX.
D800 + 300mm f2.8 + 1.4x TCII (420mm), 1/250sec, f5.6, ISO6400.

http://www.pbase.com/lance_b/image/145657975/original.jpg

[bconolly]

So, to the OP: I'd say if you can't get usable images shooting at ISO12800, you're probably doing something wrong either at the fundamental level(in camera), and/or at the secondary level(processing).
If I had a D7000, I'd have it set to AutoISO with ISO Hi1 as my upper ceiling.(note that this is how I have my D800E setup too tho).


Thanks Arthur - To me in a lot of ways noise is a taste issue, and for me, using Aperture (which is admittedly pretty crummy in terms of both its RAW converter and certainly noise reduction) ISO 1600 and below in the D7000 is excellent but above that it's not to my taste (i.e. too noisy for my liking). If I can pick up an extra stop, along with the dynamic range benefits etc that it may make sense to "upgrade" to FX. I also use auto ISO but only to 1600. I'm also chasing more than usable photos - these are keepsakes often shot in low light of first milestones (e.g. first steps etc) so I want them to be as "perfect" as possible.

- - - Updated - - -


Brenden, you have been handed a heap of good advice and points to consider and I agree with a lot of it and I would like to summarise with my thoughts.

Your signature says "various el cheapo lenses" so I would advise you yo stick to the D7000 and upgrade your lens collection a little, learn and use good software to edit your (NEF) images and have them printed at a reasonable size by a good print lab.

I will stick my neck out and guarantee that you won't see noise in a 12" x 18" print of a properly exposed and processed D7000 NEF file at 3200 iso. :)

Hi Andrew - agree 100% with this. Current work horses are the 35mm 1.8 Nikon (which I think is a reasonable lens), and also the Tamron 17-50mm 2.8 fine tuned to the nth degree in terms of focus etc. Regardless of change of body next lens will be the Tamron 24-70 VC, but based on the (as mentioned) excellent advice (as usual) I think the D600 would help to get more from that lens. Certainly the range will be more friendly for my purposes on FX. I also have the Nikon 50mm 1.8D but don't use it alot. Sharpness isn't the issue for me - it's been a fussy body re noise I think!

I do think it's high time to invest in some additional noise reduction software or perhaps move away from Aperture (but I've been hanging in for the illusive upgrade and it really suits my workflow) so that is absolutely something I will look more closely at.

Happy to take more advice and be pointed in a different direction - really appreciate all of the time spent and amazing feedback everyone!

Brenden

- - - Updated - - -

Hi Lance, I can't see these from my current Internet connection but will check them out tonight. Thank you for going to the trouble to shoot those shots though!

[bconolly]

Arthur is correct. You have to compare images from a D800, D3, D600 and D7000 at the same viewing size not at the pixel level to gauge an accurate indication of what the camera is like for noise. We view images on our screens and print them to certain sizes and they are the size we should check for noise not at the pixel level as it is an unfair comparison and not how we view our images. However, even at pixel level, the D800 and D600 are extremely good at high ISO and would still be more than a match for the D7000. I have had both the D7000 (now sold) and own a D800E (did have aD800 but sold it) and the noise from the D800E is still better than the D7000, and at pixel level it is still a tad better than the D7000, but we are splitting hairs at the pixel level.

I shoot birds mainly, and many times I have to shoot in lowish light with long lenses and in order to freeze motion of the birds I need high shutter speeds and this requires the regular use of ISO6400. This is to ahow you what you should be able to achieve very similar noise results as these below with a D600 if you use similar ppst processing.



Well these are great shots Lance and the noise is perfectly acceptable to my eye. Did you process these in Lightroom and use its NR functionality or are these straight conversions of the RAW?

[arthurking83]

Thanks Arthur - To me in a lot of ways noise is a taste issue, and for me, using Aperture (which is admittedly pretty crummy in terms of both its RAW converter and certainly noise reduction) ISO 1600 and below in the D7000 is excellent but above that it's not to my taste (i.e. too noisy for my liking). .....


Unfortunately I don't have access to Aperture to compare how it renders 'noisy' raw files natively. But I'm surprised that you see too much noise above ISO1600(or more accurately at ISO3200).

Something just twigged as I was reading your reply, and that was also on the issue of sharpening.
You may be getting the situation whereby if you over sharpen a high ISO image, you're magnifying the graininess of the noise in the high ISO image.
(If so) be very careful with your sharpening .. obviously starting off with zero, and adding any sharpening in small amounts to see how it affects IQ.


I did end up finding an image that may help you to see how insignificant noise can be from the D7000.
The D7000 was a new acquisition from by a friend, and the image shot was about the third one I took, for my purpose just to see how well the D7000 works.
the actual image itself is not quite at the macro level, just focused very closely, and the image I'm posting is a 100% pixel view.

101295

Hopefully full exif is still intact, image shot in 2010, ISO value is Hi1(or 12800), or as I've also seen it 1Ev over 6400.
Note that in blacks you always get some of the worst noise quality rendering, and this image is no exception, you may be able to just make out a bit of red and green noise blotches in the black areas.
There is admittedly also some softening of the detailed areas but (I think) still a pretty high quality rendering of such areas, and it should be noted that focus was actually on the black textured finish on the lens barrel.
Also of note in this image(as it's a shot from 2010), the noise reduction was more of a test run on using it more effectively in CaptureNX2. There are multiple ways to use NR in CNX2, so the NR routine in this particular image is not as refined as it could be.

See on the whole tho, the image at 1920x1080(my screens resolution) gives not indication that any noise is present in the image. It's only at the 100% pixel view where you can see the traces of red/green noise artifacts.
From this, I reckon I'd feel confident in printing an image shot at ISO12800, with some well crafted noise reduction!!) at about 30" .. possibly more.

If you feel that noise from the D7000 is limiting in some way, I think that two areas that you should reassess firstly(which shouldn't cost any money!) are:
1/ your processing technique.
2/ your software.

These are not criticisms against you or your software(as I know nothing of either of them) .. but it's more of a comment that you may not fully realise that more could be done with what you already have.

When I first began using CaptureNX2, noise was one of those things you just put up with in your images, because(quite frankly) NR in CNX2 wasn't something you'd look forward to doing.
But as I trialled more techniques in using NR, I found that it's actually a lot more powerful than the obvious and basic Noise Reduction tick box!

Alternatively, if you can't eke out any more quality from Aperture, have a look at ViewNX2(free) and have a go at the trial of CaptureNX2 if you can be bothered.

And it should be reiterated time and again, when using higher than normally used ISO values, make sure exposure is more to the right, and don't allow it to sink deep into shadow.
It may be a technical aspect here, but the way you need to view the issue is that ISO is basically the noise in the term signal to noise ratio, and where exposure value is the signal.
At any given ISO value the level of noise ratio is basically the same value irrespective of exposure.
But the more signal you give your image(ie. exposure) the lower is the value of noise as a ratio compared to the amount of signal.

To visualise this concept easier we'll try to give each component a numerical value:

Lets give ISO a value between 1 and 10, where 1 is base ISO(ISO100) and Hi 1(or ISO12800) is 10
We'll give exposure some random values of 1 for almost black(underexposed) to 100 for almost pure white(over exposed).
(obviously you don't want to go fully black nor fully over exposed tho!! ;))

So if we use base ISO(1) and we expose almost too, or just past, the point of over exposure(90) our signal to noise ratio will be something like (eg. 90:1) that's a pretty high signal to noise ratio, which means you basically will not see the noise at all.
But if you now boost noise up to Hi1(10) and vary the exposure, with under exposure(eg. 10).... you basically get 1:1 signal to noise ratio! ie. (as a parallel)for every pixel of signal you'll see a pixel of noise.
But if you push the exposure again up to the uppermost limit(say 90) your signal to noise ratio is 9x better than before .. ie. 90:10 or 9:1 .. again using a similar doublespeak, you may see 9 pixels of detail and colour for every pixel of noise.
Of course the parallels aren't how it actually works in real life, but it gives you an idea of how important exposure(signal) is to ISO(noise)... as a ratio.

I'd be curious to see an image of how badly you beleive noise affects your images, at what almost any D7000 owner(ex or current .. or temporary like me!) seem to think is fine.
This could be simply a matter of a higher expectation on your part relative to our expectations.

Now don't get me wrong here, I'm not trying to dissuade you from getting a new camera. LOL! far from it. I'm usually the one offering encouraging to get a new camera, and if you need some form of confirmation to get the missus off yer back .. I'm more than happy to write up a 10,000 word dissertation on the apparent benefits of upgrading your camera to this model or that :p
But I also like to assist with folks having some of the info I've gained over the last few years from others and from personal experience.

if you have (say) $1500 for a new camera, and are thinking of new lenses anyhow, at this stage, your camera is (most believe) more than adequate for almost any task and a couple of well targetted lens acquisitions may give you better overall results.
But if a full frame camera is what you really want, then there's no way around it, a full frame camera is what you should spend your money on!

BTW, so far .. the Tammy 24-70 is more than I expected it to be in many ways(mainly performance) .. but not as much of a leap as I really wanted it to be, if compared to say a 17-50/2.8 on a good APS-C camera.
Funny thing is tho, that I've always had the Tammy 28-75/2.8 as well, so this is probably just me. 28-75 has been pretty much OK on the D800, but obviously old hat when pixels were being peeped at.
The VC(on the D800) has been a godsend in some situations.
One thing I've never had issues with is the T17-50/2.8. Mine is screw driven, not internally focus driven, but has been a great lens on Dx.

[Lance B]

Well these are great shots Lance and the noise is perfectly acceptable to my eye. Did you process these in Lightroom and use its NR functionality or are these straight conversions of the RAW?

Thank you for your nice comment.

Shot RAW, I processed them in Capture One Pro 7 and converted to 16bit TIFF. I then open in Photoshop CS6 and if necessary, I use Noisware Professional plug-in to reduce noise. I also do any other Photoshop specific tweaks and then reduce for web and sharpen.

- - - Updated - - -

Arther has given some excellent advice and I also agree with Arthur that your D7000 should still give exemplary high ISO performance with the right post processing etc:

D7000 + 500mm f4 VR, 1/160sec, f5.6, ISO3200.

http://www.pbase.com/lance_b/image/141556734/original.jpg

[bconolly]

Hi Everyone, just in the process of moving house over the next couple of days. As soon as I'm a bit organised I'll put some more info including some examples up. However the sharpening mentioned Arthur could indeed be an issue (and one I'd not considered). More ASAP and thanks again!

BC

[bconolly]

Hi again everyone - back into the world of the interwebs today!

I've had a look at the default settings in Aperture for RAW fine tuning for the D7000 and Sharpening is set at .8 (scale from 0 to 1.0) so it is at the higher end. I can't remember if I've played with this or not so it may partly be to blame. In looking for samples I must say up to ISO 1600 is great and as I've capped the Auto ISO at that level for some time now finding samples was a bit difficult. I have attached one though for review. I think this seems a bit noisy but that said lighting was shocking (but not entirely unusual for when I'm able to get shots). Am I being too picky?

Brenden

[I @ M]

Brendan, after looking at the exif data I would suggest a few things.

The shutter speed, 1/50 is too slow to stop the movement in the subject.
The subject is underexposed by about 1/2 to 2/3 stop.
Having the sharpening set that high in the camera is not going to help matters as it will accentuate any noise that is occurring (mostly due to the underexposure) and I feel would be better handled by having no sharpening set in camera and applying it selectively in PP.
The focus is not spot on to help with obtaining a crisp shot.

My suggestions towards a "cure" for that particular image.

Spot meter on the subjects face. Matrix is doing a good job of the whole scene but ignoring the main point. If the background is a little over exposed it won't matter as the subject will be fine and after all that is what the image is all about.
Add 1/3 positive exposure compensation.
Your 35mm F/1.8 lens will give you great results wide open. Use it at F/1.8 as that immediately doubles your shutter speed at 1600 iso. untill you factor in any + exposure compensation.
Set the focus mode to AFC and practice locking the focus on the subjects eyes while they are moving as in the above scenario.

I feel that with your camera, that lens and some more prudent focussing, appropriate exposure and realistic shutter speeds the image would have been entirely printable at 20 inches on the longest size from a 3200 iso file.

[bconolly]

Thanks Andrew - absolutely get the shutter speed, metering and focus point issue. Most of which I've addressed in later shots (will attempt to find an example). Here's another not so great one at ISO 3200, 1/250th, f2.8, Matrix Metering. As you can see I've missed timed the shutter and hit his gob in terms of focal point rather than the eyes. This one to me does indeed seem quite noisy though, but I can sorta live with it at ISO 3200 (but would like less!). If this looked more like the ISO 1600 shot if using a D600 I'd be sold...

- - - Updated - - -

Here's another couple with spot metering etc just as a contrast. One at ISO 3200 and one at ISO 2200. 2200 looks great to me!101413101414

[arthurking83]

Of those last three images, has there been any processing done to them, even if this processing was not made by yourself. That is, has the raw converter attempted any form of shadow recovery .. as some tend to do by default with their contrast curve.

In some respects the two shots at ISO3200 aren't as bad as they ought to be(considering that they're ISO3200 images!!) if you look at some of the detail in the images.
My suspicions are that the top two of the recent three are basically similar images a few moments apart.
And the (first image)ISO3200 image has been recovered in the shadow areas, either by yourself, or via your raw converter(in this instance Aperture).

Either that or the second image was much brighter as an actual exposure and hence recovered back a little bit with some negative exposure compensation via the raw converter.

to get the best possible quality of image from an exposure shot at higher ISO values, the image must be exposed much more to the right(basically ETTR) .. and even over exposed, as far as you dare push the exposure.
This over exposed image will subsequently require a fair amount of exposure compensation to recover the highlights to a point as much as you can provide.
You will know how well your software can recover image detail prior to making the exposure .. so if for example you know you can recover up to 1Ev of highlight detail, then it's likely that you can shoot 1Ev over your normal exposure point and then apply -1Ev compensation on the image later on the PC.

So back to the discussion of the first image. For an image shot at ISO3200, if you look at the quality of noise in the brighter areas of the image, you should see that the noise quality in those areas is actually not too bad.
But where the image falls into a bit of a heap(in terms of noise quality) is in the shadow areas .. as already said by many .. this is how noise usually affects IQ.
So this should be the key to understanding that noisy images is generally tied to lower exposures. Coupled with recovery in those areas, it makes for an almost disastrous case of IQ .. and add to that some sharpening and well it just compounds the problem.

So the conclusion is that:
If you choose to go with a D600(and a damned fine camera it seems to be) .. you may well be wasting your money attempting to chase better quality image rendering in the form of lower noise, as in these instances the camera is not the limiting factor. If the issue was (say) ISO12800 image quality, then my guess would be that the D600 would indeed provide better ISO based noise quality than the D7000 .. simply due to the improved technology level of the hardware.
It may not be by much, but I think there will be a slight improvement.
But had you shot those images(supplied here) with the D600 in the same manner, I believe that you would have seen exactly the same results ... namely ok quality in the well exposed sections, lot so of noise in the shadowy areas that have been recovered.

And FWIW: to highlight the difference in exposure value between the two similar images of 1(ISO3200) and 2(ISO2200), you have over 3 stops of Ev difference in terms of the actual exposure. So not only was #1 a much higher ISO value, but the recovery job was more than 3Ev more difficult in the shadow areas in that image. Nearly 3Ev of shutter speed and almost one stop of ISO difference.

Lastly even tho it won't have any real impact on the outcome of these images, as already said in previous replies, your raw converter software will also impact on the noise quality of the affected images.
I'd suggest downloading a trial version of LR5 which, being the latest, probably has some very good noise reducing qualities(for a non specialist NR software).

[bconolly]

Thanks Arthur. Looking more closely at the original RAW converter options I've definitely messed with the sharpening at some point. Dropping that back to .5 from .8 and bumping the noise reduction from nothing to .2 makes a considerable improvement in the appearance of the image. I'll post one tomorrow as a comparison. Your point about the shadow recovery was interesting and took a bit of playing around to see what was going on. By default Aperture doesn't do anything in the recovery space on import. If however you use, which I have early on in the piece whilst learning the tool, the auto fix option it does mess quit extensively with shadow recovery in particular. Whilst I'd not used it on the above samples I tested turning it off and on and it certainly makes a difference (negatively) to the noise in the darker parts of the image.

Thanks again for the in depth help - very much appreciated. Now, you mentioned something about a 10,000 word dissertation to help with the missus and D600 purchase? :lol:

[arthurking83]

LOL! I'm all worded out for the next week or thereabouts .. and I'll need to let the dark matter settle for a while, whilst I conjure up some incomprehensible arguments :D

Of course I don't think I can have access to Aperture, so your judgement as to what it does under the hood to a raw file is much better than mine, but I'm yet to find a raw converter that doesn't have some embedded tweaks it attempts to pretty up an image .. otherwise a raw image will look very dull.
In saying that tho, I haven't tried them all .. just quite a lot(more so out of curiosity than anything else).

BUT!! while on the topic of raw converters, I remember a long while back(v2) I had a thought to try RawTherapee, and another member rejigged my dark matter into partial operation again, and I had a go at actually installing it on my PC(Mac version available too from memory).
Quite impressive results for initial rendering on a couple of raw files, that look quite good in my normal workflow(of basically Nikon specific software).
I reckon RawTherapee may have even produced a slightly more impressive rendering on one of the raw files compared to Nikon's software .. except, once again, it sets all these parameters on initial loading of the raw file trying to make it look more impressive. But as you found with high ISO noise, those results aren't always the better option.
A great 'selling point' about RawTherapee is that it's free!
Whilst it's not really a true replacement for how I prefer to work with raw files(ie. ViewNX2 and CNX2), at least RT's image rendering is good-excellent .. once the image is 'evened out' with respect to how it renders in Nikon's software.


And I think I've already mentioned it, but the D600 is definitely a camera to aspire towards(if not for the D7000 like size, I'd probably have got one too) .. and I reckon coupled with a Tamron 24-70/2.8VC .. you'll get many photos to be proud of.

[ROA44]

I think you deserve a week off after all of this, if not just to collate it all into a downloadable format for easier reading, for everyone who will probably find it a task to try to read it all at once, online to try to take it all in. But certainly one of the very informative posts that have appeared recently. well done to all involved.

But if like me you still waiting to get a reasonable monitor it doesn't make much difference. Not to worry they're getting closer. :)

[I @ M]

Thanks again for the in depth help - very much appreciated. Now, you mentioned something about a 10,000 word dissertation to help with the missus and D600 purchase? :lol:

I will perform a feat of magic here and save Arthur's keyboard and grey matter from further abuse as well as supplying you with well under 10,000 words that should guarantee wholehearted support from your better half towards the purchase of a D600. -------



D600 ( and luverly lens ) versus sports car and mistress. :D

[bconolly]

LOL! I'm all worded out for the next week or thereabouts .. and I'll need to let the dark matter settle for a while, whilst I conjure up some incomprehensible arguments

Well I certainly don't blame you! The detail back from your good self Arthur and Andrew, supported by Lance, Rick and everyone else who's posted has been nothing short of amazing. It's why I recommend this site to folks I meet, even those who aren't in Oz - best community bar none I've come across (and I'm involved in a few).


I think you deserve a week off after all of this, if not just to collate it all into a downloadable format for easier reading, for everyone who will probably find it a task to try to read it all at once, online to try to take it all in. But certainly one of the very informative posts that have appeared recently. well done to all involved.

Hear hear! at least a week (although that's selfish cause I've got plenty of homework to do before I can re-annoy you all for advice) ;)


D600 ( and luverly lens ) versus sports car and mistress.

That argument Andrew could well be worth a try, or maybe just some nice flowers? :confused013 :D

I'll hope to have some re-worked Aperture images up later tonight all going to plan.

Brenden

[bconolly]

So, based on some one on one editing feedback from Andrew (thanks so much!) and all the other good advice here I've reworked two of the images and I think the results are a heap better. Note that this is still using Aperture which will I'm ok with I wouldn't call myself an expert in by any means. BTW I don't mean to impune Andrew's editing capability as a reflection of mine - his were MUCH improved over the samples I'm posting here.

What does everyone else think? Are these easier on the eye noise wise?

[arthurking83]

.....

What does everyone else think? Are these easier on the eye noise wise?

Definitely!
the noisy grey area of the child's face are much better quality with your edited version .. a little but speckled, but no colour noise to observe(at these resolutions).

There is some graininess of the blurred background(#1), but not hard on the eyes. Whilst this isn't really an issue, it can be edited with the use of gaussian blur.

I think the key here, as you've shown is that experimentation .. and experience is what can make the difference.
We've all been there, and in reality still there ... I don't know anyone that knows everything about all aspects of photography.. so there's always something new to discover.

But I think the more important point to note about the last comment is that it's important not to jump to conclusions.
Always ask if unsure.

(y'know what would work really nice between now and that D600!? .... a Simga 35/1.4 ;))

[I @ M]

Looking much better. :th3:

It is all a learning curve, get the exposure "correct" to start with and the processing to remove any untoward noise will happen a lot easier.

[bconolly]

(y'know what would work really nice between now and that D600!? .... a Simga 35/1.4 ;))

Oh yeah Arthur - that's a great idea!