So you've shot the D5000 and 3100 then compared the results after equalizing the resolution difference?
There's not much to normalize for only a 2MP delta. I did shoot a bunch of High ISO D3100 RAWs at various levels of exposure and noise reduction. Here is the link:
brass tacks, if those first high-ISO images were legitimatificated (a.k.a., 'truthy'), i'm not interested in the scientific quantification of the differences between this sensor and that - they look to be in D700 territory based on my use of that camera, which is more important to me than measurement minutiae.
i should hope you would agree, sir and/or madam. what a lovely chapeau you're wearing, by the way.
Amazing how technology continuous to improve with new sensors. These images were impossible to obtain just a few years ago with DX sensors. The images look great for the high ISO used.
Another thing that has improved considerably is the rendition of JPEG files. We do not see the artifacts we used to see in the past. I see nothing wrong with JPEG if the right WB has been set prior to shooting.
I agree with others, the images are impressive for the little camera.
snapsy wrote:
There's not much to normalize for only a 2MP delta. I did shoot a bunch of High ISO D3100 RAWs at various levels of exposure and noise reduction. Here is the link:
I had a D90 and a D700 (2 actually) and I agree that the D3100 high-ISO IQ is about equal to the D90 and about 1EV to 1-1/2EV behind the D700.
On a side note, I have NOT seen the ridiculous banding that your D3100 is showing in the shadows. I've shot ISO12800 with strong shadows and still not seen the banding. Any chance the banding is a result of eletronic interference (continuous-AF=on maybe?) and not a normal flaw?
If that's a personal copy, I'd send it to Nikon. It's performing like a pre-release copy.
The one problem with the D3100 that I have noticed is how incredibly awful the in-camera NR algorithm is. It's as though Nikon intentionally made it awful to differentiate it from other classes of cameras. ISO3200 - ISO12800 with NR=ON results in artistic watercolor paintings, not photos. :-P
Oh, and the reason people post 1000px images isn't to make the camera look better, they are simply abiding by the forum rules/suggestions so that everyone can enjoy their work.
Forum FAQ
There are no restrictions on image dimension, but please keep files around 700px (1000px maximum), so that members using monitors with lower resolution can also enjoy viewing your photos.
AngryCorgi wrote:
I had a D90 and a D700 (2 actually) and I agree that the D3100 high-ISO IQ is about equal to the D90 and about 1EV to 1-1/2EV behind the D700.
On a side note, I have NOT seen the ridiculous banding that your D3100 is showing in the shadows. I've shot ISO12800 with strong shadows and still not seen the banding. Any chance the banding is a result of eletronic interference (continuous-AF=on maybe?) and not a normal flaw?
If that's a personal copy, I'd send it to Nikon. It's performing like a pre-release copy.
The one problem with the D3100 that I have noticed is how incredibly awful the in-camera NR algorithm is. It's as though Nikon intentionally made it awful to differentiate it from other classes of cameras. ISO3200 - ISO12800 with NR=ON results in artistic watercolor paintings, not photos. :-P...Show more →
Those shots are from a floor sample at Best Buy. It had an electronic anti-theft device attached to the base - maybe that gives off EMF and induced the banding (although I didn't see banding in the underexposed ISO 100 shots I posted to that same thread). To compare can you please take a ISO 3200 black frame on your D3100 (lens cap on, f/8.0, shutter = 1/4000) and post the raw to youshare.com?
Also worth noting is that you have to circumvent NX2 to achieve RAW -> JPG conversions without any NR. Even with NR=OFF, the NX2 conversions apply some slight luminance NR and chromatic NR. I'll post some comparisons tonight between an NX2 RAW and a RPP RAW. Nikon is tripping over its own feet in its software (computer and in-camera) for this body.
I just have not had any luck in keeping the EXIF intact on the RPP conversions, though.
snapsy wrote:
Those shots are from a floor sample at Best Buy. It had an electronic anti-theft device attached to the base - maybe that gives off EMF and induced the banding (although I didn't see banding in the underexposed ISO 100 shots I posted to that same thread). To compare can you please take a ISO 3200 black frame on your D3100 (lens cap on, f/8.0, shutter = 1/4000) and post the raw to youshare.com?
AngryCorgi wrote:
Also worth noting is that you have to circumvent NX2 to achieve RAW -> JPG conversions without any NR. Even with NR=OFF, the NX2 conversions apply some slight luminance NR and chromatic NR. I'll post some comparisons tonight between an NX2 RAW and a RPP RAW. Nikon is tripping over its own feet in its software (computer and in-camera) for this body.
I just have not had any luck in keeping the EXIF intact on the RPP conversions, though.
Interesting observation. Are you sure it's NX2 still applying NR instead of maybe RPP just doing a poorer job demosaicing vs. NX2?
snapsy wrote:
Interesting observation. Are you sure it's NX2 still applying NR instead of maybe RPP just doing a poorer job demosaicing vs. NX2?
Considering that RPP is resulting in a finer noise grain and NX2 results in splotchier noise and slightly softer details, I think if anything its the opposite...also, I'm not referring to Capture NX2, rather View NX2 that is bundled with the camera. It doesn't let you change the NR settings invoked by your camera settings and it puts similar NR to what the camera JPG NR=off option does.
AngryCorgi wrote:
Considering that RPP is resulting in a finer noise grain and NX2 results in splotchier noise and slightly softer details, I think if anything its the opposite...also, I'm not referring to Capture NX2, rather View NX2 that is bundled with the camera. It doesn't let you change the NR settings invoked by your camera settings and it puts similar NR to what the camera JPG NR=off option does.
Ok, thanks for the clarification. I used Capture NX2 for my conversions, which does let you completely disable NR. It also lets you control the amount of luminance vs. chrominance NR.
The pics posted as examples are not a good comparison to the D700. When you run Noise reduction then its not a sample anymore. Samples would be ISO 1600 and above with no post at all. That's a sample.
No comparison, in terms of noise ,is going to be a good comparison unless it's done under controlled conditions with the cameras in question shooting side by side with similar processing. Everything else represents user samples that we have to make our own suppositions from.
There's no problem with user samples, but everyone is going to have different opinions.
Oh no, I can't believe I'm going to link to this... sigh:
At least it's controlled, looks like JPG noise reduction was on for both cameras. D3 vs D3100 Jpg normal with NR. D3 unpressed to match D3100 resolution (similar to making a 30x40" print from each.)
Up to 1600 they look similar, the NR @ 1600 looks more aggressive in the D3100, you loose a bit of high frequency detail in the fur. Same thing at 3200. @ 6400, the D3100 and D3 look very similar, they have a different look indeed. The D3 seems to have slightly more actuance but the chroma and lum noise in the D3100 are lower (either because of better NR alog or lower noise to begin with) @12800, the D3 looks better, but not by a whole stop.
Unfortunately the metering is off (the D3 tends to expose a bit lower than optimal at higher ISO's I think). The amount of relative "overexposure" in the D3100 files gives it at least 1/2 to 2/3 Ev advantage over the darker exposed D3 shots.
As soon as you get > ISO1600, 2/3Ev of over-exposure means exactly the same noise-reduction as lowering ISO by 2/3 stops.
Considering the difference in in-camera NR, the D3 should be set on "NR-High" and D3100 on "NR-Low" - that would mean about the same NR in the cameras.
But totally disregarding this nitpicking (which anyway results in a +1Ev unfair comparison advantage to the D3100) I still find the new D3100 (and the D7000) totally amazing. We're not allowed to publish D7000 testsamples yet, so the shootout will have to wait... :-)
@ Suede, I had assumed in that KR test the cameras were in manual mode and just the shutter and ISO were changed manually while the differences in tone and brightness were because of sensor differences. You are most likely correct though, the lighter white fur on the D3100 monkey could be because of differing exposure. The EXIF is stripped
I agree, the samples from the D3100 and D7000 are pretty amazing as far as their sensor mojo goes.
I completely agree with what Jammy stated above. The NR does seem more aggressive in the D3100. However, after doing a few side by side tests against my D700 at home, the D3100 comes very close. The slight edge goes to the D700, but if you are printing smaller than 11 x 14 inch prints, it would be hard to tell the difference. I just finished running some prints taken at 3200 ISO (20 x 30" equivalent) with color NR + minor sharpening and they looked amazing! The detail is excellent!
Which is close to what I've suspected. D3100's High-ISO performance seems a hair better than D5000/D300s. Low-ISO performance seems a hair worse than D5000/D300s. I don't think D3100's IQ is comparable to D700.
baumgarten0712 wrote:
I completely agree with what Jammy stated above. The NR does seem more aggressive in the D3100. However, after doing a few side by side tests against my D700 at home, the D3100 comes very close. The slight edge goes to the D700, but if you are printing smaller than 11 x 14 inch prints, it would be hard to tell the difference. I just finished running some prints taken at 3200 ISO (20 x 30" equivalent) with color NR + minor sharpening and they looked amazing! The detail is excellent!
Here is the DxO mark just released:
http://94.23.12.156/index.php/en/Camera-Sensor/All-tested-sensors/Nikon/D3100#measurement
w.r.t. ISO performance score:
D700 2303
D90 977
D3100 919
D5000 868
D700 is more than 1 stop ahead of D3100, not "very close". That's much more than the difference of D90 & D70.
D3100 matches the ISO performance of D90 at slightly high resolution. That's very good, but your assessment is WAY OFF! Not sure what you & OP were smoking that day.
