Friday, October 8, 2010

Is the D3100 an Improvement over the D3000 at High ISO?

In a word – holy sh*t!

Not only is the D3100 a *massive* improvement over the D3000 in image quality at high ISO, it is in fact in a virtual tie with the D90 at high ISO. And since it adds 2 megapixels and sits at the bottom of the lineup, this is really big news! This camera leaves nothing on the table in my opinion. And note – I am strictly talking about high ISO image quality here.

So how did I divine this, since I do not own any of the camera in question? The answer, as always, is that I fashioned a set of crops from the images at imaging-resource. I favor their low light images in the high ISO series of the mannequin … I always take crops that encompass the left eye and some hair, so we can see how the high contrast eye will fare next to the low contrast hair, which is much more difficult to get right.

I equalized the sizes up to the D3100. The difference in resolution at every level was minor, in fact trivial for D90 up to D3100.

So … the crops. Click through because it is obviously much larger than the 400px version shown here …

d3000_d3100_d90_crops_IR[1] My Analysis

  • The D3000 images are mediocre at best and excrutiatingly noisy at worst. I hope the camera is better than that in practical use.
  • The D3100 matches the D90 almost exactly at every ISO. Wow.
  • 3200 ISO is usable now from the entry level D3100. That’s a miracle.
  • 14mp is a trivial improvement over 12mp, a bit better on top of the 10mp of the D3000. But it’s all psychological, let’s face it.

Nikon has a real winner here. Small, light and packing a fantastic sensor. I really am very impressed. This would be a definite buy if it supported my AF lenses and AI lens. Who knows …. it still might be.

Edit: ReD BaRoN added a comment asking me to compare the D5000 and T2i as well. I agree that more cameras could be compared in this way to try to guess a bit which ones will be satisfying in low light and which might not be …

First off, I did not include the D5000 because it has the same sensor as the D90 and the jpegs are awfully similar. I don’t really consider the images coming out of the D5000, D90, D300 and D300s to have any differences worth contemplating.

But I *did* include the T2i. And I added the NEX-5, since people are pretty excited about that camera nowadays. And I droppe din the Pentax Kx since many think it has the best jpegs out there today. So let’s have a look, shall we?

d3000_d3100_d90_et_al_crops_IR[1]

My further analysis (and please click through to see the crops at 100%):

  • First, let me correct a slight oversight. The D90 crop at 3200 ISO has visible color blotching around the eye. Not a lot, but enough to tip me towards the D3100 … wow, Nikon really scored with that cam.
  • Now, the Canon T2i. Nice (very) detail around the eye. Looks like the cam is really good with high contrast details. Pretty good detail in the hair as well. This might be because Canon has a lot of pixels and so this cam was downsized quite a bit to match the D3100. That said, I dislike the white balance … all the others are close to each other, and this one is way off. It is also a bit dark, hiding some important details like eyebrow hairs. (Yes, I know these are normalized to 100% crops from the D3100, and this is like looking at 40” posters.) Anyway, the 3200 crop looks like pretty much a tie with the two Nikons. So get whichever one you want … the only difference I see is the white balance.
  • The Sony NEX-5 certainly holds its own at 800 ISO. Not bad at all. 1600 seems to be on par with the upsized D90, but the D90 and the other cams hold better detail in the hair. It’s not massive and it is absolutely invisible at web sizes, but it is visible at these sizes.
  • And finally the Pentax Kx … the cam everyone thinks is the second coming … and if you look at the eye detail – especially that upper eyelash – you will find it amazing. There is also no chroma noise anywhere to be found around the hair and eye … amazing. Yet there is a weakness, and it is exactly the kind of weakness I really dislike … the hair is excessively smoothed. The bands of flat loking non-detail are wider than they are on any of the others. The hair looks wet … almost helmet-hair. And that’s even on the 800 ISO shot (quite subtle, but there.) This is really disappointing to me … I would much rather have coraser looking hair with the impression of individual strands that very nice fine hair with patches of helmet. I am frankly surprised.

The new bottom line:

  • The D3100 and T2i are tied on tier 1
  • The D90 is only a hair behind so it remains on tier 1, but in 3rd place
  • The Kx is on tier 2 because I *hate* clumping of the hair at high ISO, so all that work for wonderful high contrast details kind of feels wasted
  • The NEX 5 is definitely tier 3
  • The D3000 should pick up its marbles and go home

Interestingly … for those willing to shoot RAW, my guess is that the D40x through D3000 still make a fine and very cheap travel camera. But if you want a party cam, please opt for the more modern sensors …

Oh … and one more thing … I am a huge fan of the Canon and Nikon hegemony … and the reason is that I like their systems (huge), I like their used markets (huge) and I like the idea that I can rent a lens or body anywhere in the world if I need to for whatever reason.

Sunday, October 3, 2010

F300EXR – Review Part 22 – Fuji Color vs D700

A couple of images from yesterday … shot during ASTIA-gate. A good study in Fuji color …

D700, processed with the ACR vivid picture control setting for Nikon … closest I could come to the overall tone. But the colors are still not quite warm enough.

DSC_6927_d700_vivid[1]

F300EXR

DSCF0288_f300[1]

Both look terrific at this size … in fact, give Fuji credit … the colors look really nice straight from the cam. Now, here is a case where the image contains no shadow detail worth worrying about so ASTIA works great.

A couple more images.

D700

DSC_6920_d700[1]

F300EXR (you probably noticed by now that I was shooting 3:2 yesterday)

DSCF0281_F300[1]

The Fuji again looks really great straight from the cam … but how about the details? Are they really clean?

Well …

F300vsD700_building[1]

The D700 slaughters the F300EXR in a fair fight here … in fact, the F300 has the advantage of 3600mm versus 300mm lenses and of course the depth of field advantage as well. I even shot it in M3:2 mode to avoid corrupting its edges with NR-itis …

Just to complete the picture, let’s revisit the F300 and F70 shoot out from today. Here is the same building, shot much earlier in the day by both cameras set to different film modes. So I cropped the F70’s270mm to match the framing of the F300’s 360mm and then cranked the saturation and contrast to 11 for the F70. They match pretty well now.

F300 in ASTIA

DSCF0301_f300_astia[1]

F70 cropped and curved

DSCF4905_f70_crop_curves[1]

The colors are cooler than the previous set because that late afternoon sun was burning the oranges into the paint … but these are a decent match. I obvoiusly would have had to spend a lot of time to match them exactly, but I chose to stop after a while and leave the F70 a tad cooler than the F300. I’ll do an ASTIA to ASTIA comparison one day …

Here are a pair of crops from the upper middle of each shot …

f300_f70_crop[1]

Yeah … both images were pretty much hammered by NR in the dark sections. A very good reason to keep the tone curves under control at capture time …

F300EXR – Review Part 21 – Film Modes and Saturation

I shot a lot of Kodachrome slide film back in the day … and one thing you learn to do with slide film is slightly underexpose to (a) protect your highlights and (b) increase color saturation.

The cool thing about digital is that it acts a lot like slide film. For the most part, digital capture has poor dynamic range, matching slide film in behavior and character. And, of course, underexposing deepens color and increases saturation. This can be very handy …

So I whipped up a matrix with six images of a red car sitting in front of my car in the parking lot just before breakfast this morning. I shot the 6 film modes (this was the moment when I realized that my entire morning of shooting had to be written off as a valid comparison between the F70 and the F300 … *sigh*

The two exposures here are 0EV and –1.33EV. Now, that will seem like a lot of negative compensation, but this was a wicked bright day outside and I found that this much compensation made for some really nice looking images. The following crops are in fact not crops. They are complete images reduced to 512px on the long side, for a total width of 1024px for the images side by side and a bit more for the text.

All that to suggest that you click through on this image, as it needs to be seen big to see the subtle and not so subtle tone changes through the six variations …

film_modes_f300[1]

As always … if you want to process your images, the 0EV + PROVIA image top right is the way to go. You can as as much or as little extra saturation and contrast as you like. But try and make the middle left image look like the top right image and, well, that’s not gonna happen.

For those of you who do not process … any of these would be fine. And there are three variations missing from each line directly between the left and right versions …

F300EXR – Review Part 20 – Birds and Planes

I’ve been trying for two days to successfully shoot a bake off between the F300 and the D700 (yesterday) and the F70 (today) … I have finally achieved some sort of lighting parity in today’s tests, which you will not see in this part of the review :-)

Bait and switch … I like it. Well, technically it is not bait and switch … I just wanted to set the tone for the pain I have been feeling with the film modes … the icon for the ASTIA film mode is “S” … which to me screams STANDARD, which is what Fuji sometimes calls PROVIA. I;m getting pretty sick of the incompetence in Fuji’s software and user interface design people … I shot the first hundred or so images today with the two camera’s in different film modes. Scrap it all …

Anyway … so here are some images that came out of ASTIA-gate … this part will concentrate on birds and planes … I went a little nuts with it today.

A little private jet blasting overhead on its way to the airport. Shot in M4:3 at +2/3EV … but ASTIA made sure you can’t see the bottom anyway :-)

DSCF0322_jet[1]

Now, I cheated a bit here … the background pole and bird have been softened using the blur brush in CS5. This is the perfect use for something like that … by the way, this is a pretty heavy crop from an L4:3 image … yes, I have been trying it for a while … all I can say is that distant objects have their edges ravaged really badly … but this bird was close enough to escape that ignominy …

DSCF0343_flying_gull[1]

There were a large number of shit-hawks, er, gulls near the restaurant where I had breakfast this morning. The puddles attracted a bunch of them for bathing and drinking (yuck) … the sun made it a challenge to capture them, especially with ASTIA on board.

DSCF0385_gull[1] The F300EXR is capable of some pretty magnificent detail on close subjects … I hope that you’ve been clicking through …

DSCF0394_gull[1]

My only gripe with these EXR cams is that sharpening is a tad high … there are sometimes halos already in the image …

DSCF0403_gull[1]

I was standing right in the middle of the gulls … and some of them were downright annoyed …

DSCF0404_gull[1]

Backlit birds were much harder to process … Topaz Adjust gives me a hand here …

 DSCF0412_gull_crisp[1]

I was fooling around with L4:3 and happened to have focused on the light standard there just as this bird took off … although not perfectly sharp, it’s pretty neat …

DSCF0428_gull[1] Here’s what I meant about shooting distant subjects with L4:3 … and here in soft light (sun behind clouds for a moment) … the results are underwhelming …

DSCF0430_L43[1] Here’s a 100% crop from yesterday … the D700 … and yes, this is not all that impressive … those of you who really hate dSLRs can use this as your latest reason :-)

DSC_6914_d700_100crop[1] And the F300’s 100% crop from yesterday …

DSCF0258_f300_100crop[1]

And one more from the F300 …

DSCF0262_f300_100_crop[1]

So … getting close is the key with most small sensors. Keep the details reasonably large and the camera will reward you with a great image. In this case, I am very pleased with the Gull images … but of course I was right on top of the little buggers … and they were ticked …

Saturday, October 2, 2010

F300EXR – Review Part 19 – Smearing test part 1

Smearing is something the new Fujis do quite aggressively. The EXR cams do it with abandon when shot at L size, but it is much rare at M size … still, you can stimulate it by reducing contrast until the camera thinks it is looking at noise. Then it will smooth away all the detail for you …

How can you reduce contrast?

  1. Shoot things from further away … the smaller the details, the less they look like details and the more they look like grain (noise)
  2. Shoot things with uniform colors … contrast requires a change in tone or color
  3. Shoot in low light or soft light … contrast increases as light increases (ever try to read something in near dark?)
  4. Underexpose … if you deliberately or accidentally allow less light to the sensor, the camera will see less contrast
  5. Increase ISO …. the noise injected into the image tends to reduce the contrast of the actual data

There are probably other ways to reduce contrast, but these are the obvious ones. So, the corollaries are probably worth noting. I.e. how do you increase contrast and thus the crispness of your images?

  1. Get close. The HS10 smears like mad from a distance … but fill the frame and things start to look nice.
  2. Shoot subjects with inherent contrast.
  3. Increase light. Add light (flash) if you have to. (Note: At some point too much light makes things look harsh, as with midday sunlight.)
  4. Expose to the right, but don’t overexpose because that tends to lower contrast in the highlights :-\
  5. Decrease ISO … keep noise at bay

I decided to run a test that would attempt to simulate a low contrast subject at distance and simultaneously show the effect of underexposure.

First, I tried the infamous Jamaican dollar test. Show from 15 feet in low light.

0 EV

DSCF0221_0ev[1] -1.33 EV

DSCF0226_1_33ev[1]

The difference is pretty obvious, although less so on my out of whack TN panel. Obviously, details like the dark area around the head become almost uniform to the eye … but the question is when they become uniform to Fuji’s noise reduction algorithms?

Here is a set of crops … the left side is a set of exposures of increasing underexposure and the right side is the same crops equalized in exposure by dialing exposure up in ACR6. Click through for full sized 100% crops.

bill_crops[1]

One side effect that I was not aware of is that the contrast increases in the adjusted crops. Pulling up an image from a severe underexposure results in a harsher image than you get by exposing properly. This could make a real difference in the quality of the iage, further strengthening the argument for getting the exposure right … learn your meter!

The main observation with these is that the low contrast detail, which includes the area surrounding Sir Bustamante’s head and the almost white flower petals, is compromised pretty badly by the time we reach the –1.33 EV underexposure. Edges of higher contrast detail – BANK OF and ONE D are good examples – are also looking a little compromised.

Obviously, these details are pretty small … but imagine them repeated across the frame as a field of like-colored foliage or lots of repeated edges. They would become noticeable as being rather degraded. I think that’s a bit of what we see every day in images posted to the Fuji forum.

The bill test leaves something to be desired though. Even the low contrast details are pretty large, so how about we try something a little different. The paint brush test seemed like a good idea. I happened to have three paint brushes lying around, so off I went.

Here is the 0EV image

DSCF0227_0ev_OOC[1] Click through to see the superb details in this one. But even here at 100 ISO we can see some compromise in the tips of the brush on the right. EVen though there are individual strands there, they are the same color and contrast is not very high so they are starting to bunch up. This is a classic issues with Fuji cameras and hair … or fields of foliage from a distance.

Here is what –2EV looks like

DSCF0233_2ev_OOC[1]

Here is what –2EV looks like when equalized to 0EV

DSCF0233_2ev_equalized[1] That’s surprisingly good. There is more latitude than you think in these little EXR cameras. But the areas where colors are similar are actually under attack here. And ISO as at base, so there is nowhere to go but down.

The crops again …

exposure_effect[1]

It is obvious even at the 400px wide blog version that things are deteriorating as the underexposure gets more severe. But it is hard to see the extent of the smearing from the crops presented this way. So here is a set of three crops … 0EV in the middle, –1EV on top and –2EV on the bottom. All equalized to 0EV … and what we see is that there is definitely smearing of the areas where contrast is low to start with. So apply this to any foliage shot where the light is low and you can start to imagine what will happen …

pull_ups[1]

Again, you must click through to the larger version to really see the effects of underexposure.

Bottom Line

A little underexposure is safe … and even warranted when you want to saturate colors a little more. But if you underexpose by 1 or more stops, you are going to see a lot of extra shadow noise (see the final image above) and a lot of extra noise reduction destruction …

I will repeat one of my primary suggestions for learning to nail exposures here: read John Shaw’s Field Guide to Nature Photography …

Crank it to 11

That’s a reference from “This is Spinal Tap” … a pretty funny movie with some pretty decent music in it.

MV5BMjE3MzExNzU1NV5BMl5BanBnXkFtZTcwMTIwMzMzMQ@@._V1._SY314_CR6,0,214,314_[1] Bill from the Fuji Talk Forum passed the following cartoon along to me in a comment on a recent post here and I thought it worthy of its own post …

spinal_tap_amps[1]