Have they given up on noise?

[peter_nelson1]

My old D100 has finally bitten the dust so I need a new DSLR. My other DSLR

is a 20D so I have both Nikon and Canon lenses.

 

The biggest thing I'd like to improve is noise. I mostly do figure, fashion,

dance,and portrait photography. I shoot available light even though I have a

full studio because I like the atmospherics of natural light filtered through

cloth, or in old buildings with old, dirty windows. An extra stop or so of

ISO would buy me lots longer shooting time during the day.

 

The 20D is legendary for its low noise. The 30D is about the same and,

according to most of the tests I've seen, the 40D isn't even quite as good!

In fact, according to DPReview the full-format 1D's are only a teensy bit

better than the 20D. Meanwhile the Nikon D200 is significantly worse, and

according to recent reports, the D300 is marginally worse than the 40D.

 

So what's up with this? Have they decided that the current noise levels of

DSLR's are "good enough" and they can concentrate on other things like screen

size, Live View, or have they just hit a technical wall? Why is the 40D not

quite as good as the 20D/30D? I know it has more pixels (=smaller

photosites) but I assumed the technology should be better by now.

[Matt Laur]

Don't confuse noise as an absolute across a given 100-pixel chunk of the image, and noise as a factor in the finished piece at a workable display size. And, higher resolution allows software to make a better distinction between per-pixel noise and multi-pixel objects in the image.

[ellis_vener_photography]

having shot fairly extensively with all of the cameras you mention, I feel obliged to say that you are sadly misinformed about the D300.

[peter_nelson1]

<i>having shot fairly extensively with all of the cameras you mention, I feel obliged to say that you are sadly misinformed about the D300.</i><p>

 

Many people on the web have posted side-by side images of identical subjects with D300 and competing cameras and none of them showed the D300 being lower noise than any of the Canons. Most of the images I've seen of the D300 showed that without its built-in noise-reduction it lokks very noisy at ISO 800 and up, and with it enabled it looks very soft. It also seems to use excessive supression of chrominance noise, creating a blotchy luminance noise pattern and reduced saturation.<p>

 

I'm glad you've had a better personal experience with it, but plenty of other people have reported the opposite in <b>their</b> personal experience, so I'm inclined to look at actual image comparisons people have posted for the tie-breaker. Can you point us to some objective tests that suggest that the D300 is better than (or even as good as) an old 20D?

[peter_nelson1]

<i>Don't confuse noise as an absolute across a given 100-pixel chunk of the image, and noise as a factor in the finished piece at a workable display size. And, higher resolution allows software to make a better distinction between per-pixel noise and multi-pixel objects in the image.</i><p>

 

I'm not sure what your point is.<p>

 

<b>My point</b> is that as good as the 20D is, I'd like my next camera to be lower noise. For the kind of shooting and printing I do I find that the 20D good up to ISO 400-600-ish but if I go higher, say to ISO 800, it's too noisy.<p>

 

A typical late-afternoon interior exposure is 1/8th-1/15th of a second at f2.8-4 (tripod mounted of course). When it gets any darker than that I have to stop shooting, so if I can go to ISO 800-1600 I get extra shooting time. And even before it gets that dark, I'd like to be able to put the model or her clothes in motion which means I'd like to use a higher shutter speed.<p>

 

I shoot dancers in better (available) light, but I also have to use a high shutter speed, which means I have to shoot at a high ISO (typically 1600), and that produces a noisy image.

[JDMvW]

Noise, schmoise.

 

Those of us, sonny, (says the old codger) who grew up with GAF 500 and 1000 film, find the noise in most modern digital cameras to be of no concern at all.

 

Here's what we put up with for low light back when<div></div>

[Matt Laur]

My point is that you're comparing the noise that appears in, say, a 100 x 100 pixel patch of an image while looking at an image that in the D300 is 4288 x 2848, and in the 20D is 3504 × 2336. When you have good light, noise isn't a consideration, and you have the extra resolution. When you have poorer light, and either ask the camera or post-production software to reduce noise, the higher resolution image of the D300 lends itself to such cleanup very nicely. Many noise artifacts are gone entirely when you down-sample a D300's image to the size the 20D is producing, and the 20D is still going to have whatever noise it has. Too many comparisons don't take that fundamental difference into account. When you remove 20% of the information from the image, that's a lot of opportunity for noise reduction processes to work very well while still preserving great detail. And when you don't have noise to worry about, you've still got 20% more information with that newer sensor.

 

And, of course, splitting hairs over those noise issues doesn't take into account the extremely nimble AF system on the D300. Faster focusing with better tracking, in the real world, may have more to do with producing a clean, compelling shot of those dancers. All depends on your circumstances. None of this adds up to the manufactures "giving up" on noise. There are many other factors that contribute to the final output's overall quality.

[danielleetaylor]

The D3 and 1D mkIII seem to be the low noise champs right now. Either would appear to out perform a 20D.

 

A 5D is the same as a 20D per pixel, but the noise appears to be less in print because you're enlarging less.

 

You don't mention your lenses but if you don't already have f/2.8 zooms or even faster primes, that could also help your situation. (Maybe...if you don't need the DoF from stopping down.)

 

I don't think anyone has decided DSLRs are "good enough" on noise, both the D3 and 1D mkIII pushed the envelope. But I do think manufacturers are starting to hit the limits of what can be done.

[StuartMoxham]

The best way to compare noise is by making prints. Looking at 100% crops won't tell you much, monitors are backlit and the noise will often look very different in a print. I find that there is far to much fuss made about the noise from DSLRs we've never had it so good and even if ISO 1600 gets a bit noisy there is always software like neat image to help out.

[StuartMoxham]

Why not go for a 5D full frame and better noise performance.

[johan_de_groote]

Well, there is a trade-off between noise, photo-site size (pixels) and sensor size. Larger pixels have less noise, but to get the same resolution you need a larger sensor. And large sensors cost more because you can fit less sensors on a silicon wafer and the chances on defects are larger. So it is a delicate balance act between it all and the price they can ask for it at the end. Proces changes and control play a role as well.

 

So they could make low-noise large high res sensors, but would anyone buy them at the price they will have to ask?

[mag_miksch]

I vote with Ellis, the D300 is amazing at higher ISO; beside this: Its a near perfect camera, with just one very stupid issue, the front wheel and the on-off trigger are much to near together and are to similar to feel, so its very easy to turn off the camera instead of setting something. Regards Martin

[ellis_vener_photography]

Peter,

 

I have no idea how other people process their photos. Ican onl yspeak from my experience. At ISo 800 and 1600 the D300 in my experience outperforms every other recent and current APS-C class D-SLR, up through the Canon EOS 30D and Nikon D200. I haven't shot with an EOS 40D so I'm not qualified to comment on that camera.

 

If you want low noise at high ISOs and that is a real killer priority for you, go straight to the D3.

[danielleetaylor]

I have to admit that I'm going off online samples here, but the D300 doesn't look any better than the 40D at high ISO. I think it's actually a bit too aggressive in controlling noise and the result is lost detail relative to the 40D.

 

Neither is that much better than a 20D. If you're going to try and shoot in lower light by upgrading your body, you would be much better off with a D3, 1D mkIII, or 5D. You're not going to see any real difference in ability to shoot in low light by just going to a 40D or D300 from a 20D.

 

Focus speed on any modern body is a function of lens design. A human would be unable to discern the AF speed difference between two AF modules. That's why there are no actual tests of body AF speed on the Internet, no numbers or graphs on sites filled with numbers and graphs like dpreview.com. A human can't use a stopwatch and figure it out. You would have to connect an external computer to the DSLR body and somehow detect both AF start and AF confirmation accurately in order to get any kind of speed measurement. When the lenses are different a human can easily discern the difference.

 

Choosing a lens with a sonic motor is a no brainer, but even among those lenses there can be a huge variance in AF speed. The size and weight of the elements which have to be moved, plus the distance they have to be moved, plus the max aperture of the lens versus the light level are all huge variables.

 

Peter - are you shooting at f/2.8-f/4 because of lens limitations or because of desired DoF? If the former, have you considered adding some primes? That could buy you 1-3 stops right there.

[ellis_vener_photography]

While I think the D300 at least as good as the 40D maybe we have just reached the limits of how many photosites and transitors can be crammed into an APS-C (roughly 16x24mm) size sensor before the laws of physics can no longer be beaten into submission by today's limits on materials and clever image processing. maybe if a future CMOS and CCD imaging chip is supercooled, or the chiprocessors o nthe chCMOs board made more efficient in terms of heat dissipation the signal to noise ratio will improve to Peter's liking. It could happen.

[peter_nelson1]

<i>Many noise artifacts are gone entirely when you down-sample a D300's image to the size the 20D is producing</i><p>

The D300 is 12 MP; the 20D is 8 MP. That's less than a 2:1 downsampling so I don't think that's going to result in much noise reduction. I agree that if the D300 were something like 32 MP, so you could average 4 pixels down to 1 then your theory might be valid. Can you point to any formal tests that confirm that when you take your theory into account it's any better?

[peter_nelson1]

<i>Why not go for a 5D full frame and better noise performance.</i><p>

 

In DPReview's tests the 5D wasn't significantly better - see http://www.dpreview.com/reviews/canoneos5d/page21.asp.<p>

 

I mentioned that in my original comments - most of the full-frame cameras seem, at best, only marginally better. The D3 may prove to be an exception to this - I've seen conflicting reports and I'm waiting for a strict DPReview-style comparison.

[peter_nelson1]

<i>Peter - are you shooting at f/2.8-f/4 because of lens limitations or because of desired DoF? If the former, have you considered adding some primes? That could buy you 1-3 stops right there.</i><p>

 

I like the flexibility of a zoom for figure and dance work, and also that means I can clean my sensor before shoot and not introduce dust while changing lenses. I hate stopping in the middle of a model shoot to clean the sensor.<p>

 

But the 28-70 f/2.8 benefits from a little stopping-down - it's hard to make a zoom that is optimimum wide open. I wish Canon or Nikon would come out with a 28-70 f/2.

[rjacksonphoto]

Peter,

 

I tried to check your website to see what kind of stuff you shoot, but couldn't get the link to work.

 

I think you need to rent some cameras and try them out for yourself. My impression is that the D300 surpasses the smaller sensor Canons in low light performance, but, like you, I have not used it. If you are really questioning if noise lower than the 20D (I'm not convinced this is the low-noise gold standard) is possible, you should try the D3. All reports I've seen suggest that this is clearly the new benchmark and is unlikely to be surpassed by an APS-sized sensor.

 

If noise-free low-light performance were as critical to my work as it seems to be to yours, I wouldn't hesitate to snap up a D3.

[danielleetaylor]

Peter - again I'm admittedly going off online examples, but if you have to squeeze more low light performance out of the body as opposed to the lens, your choices would appear to be the D3 or the 1D mkIII. I think you would see an improvement in print with a 5D, even though the per pixel noise is the same, because you're not enlarging as much. But you would see a larger improvement with one of the other two bodies. The D3 sort of took the thunder from the 1D mkIII in the press and forums, but the Canon's high ISO performance is amazing as well.

 

I would agree with you that outside of those two cameras, and to a lesser extent the 5D, noise performance isn't much better than the benchmark 20D. Because of the multiple D300 mentions in this thread I went searching for more online examples, but they just confirmed that it sacrifices detail for smoothness at high ISO. For some subjects that might look better, but I would rather make those choices myself in front of a computer. (Not that it matters much either way. In print I think it's probably a wash between the 20D/30D/40D and the D300.) We're about at the limit of what can be done on APS sensors so long as engineers trade every incremental S/N increase for the opportunity to squeeze more pixels on the same space.

 

I know you've said you like zooms and don't like sensor dust, but if you really want to extend your shooting time when natural light fades, having a prime to switch to would buy you a lot of light. And if you're shooting wide open or near wide open any way, you're not going to see any sensor dust. I never notice it until about f/16. I don't know what your shooting situation is, but if you are able to move around and you're using a 28-70, are you sure you couldn't substitute foot work and a 30 or 50 f/1.4 for the zoom? That's a couple stops of light right there.

 

I think those are your choices.

[tonybeach]

"...again I'm admittedly going off online examples"

 

> Online examples are the worse way to evaluate a camera, the second worse being most of the DPR style tests. Dismissing on hands evaluations from seasoned photographers demonstrates a complete lack of understanding of how to appropriately evaluate equipment. I know for instance that one of the most commonly referenced online ISO comparisons used for the D300 have several out of focus shots. I will add my voice to those that say from hands on experience that the D300 is very good, a full stop better than the D200 at ISO 1600. Useful test results for the D300, 40D, S5, D3 and some others can be found here: http://www.diwa-labs.com/wip4/test_result.epl

[StuartMoxham]

Many wedding photographers used the 5D because of its good hi ISO performance. I would take the Dpreview noise tests with a grain of salt. You need to shoot a camera in bad light to see how it performs in bad light. Looks to me that most of the tests on Dpreview are studio shots so these don't realy give the whole picture about hi ISO performance. You can't really view a 100% crop from a 5D and compare it to a 100% crop from a 20D because the cameras produce different size images it is really best to make prints at the size you need and then decide which camera has the better noise performance.

[peter_nelson1]

<i>Dismissing on hands evaluations from seasoned photographers demonstrates a complete lack of understanding of how to appropriately evaluate equipment.</i><br><br>

 

When it comes to subjective qualities like "handling" or UI that might be true, but noise is an objective attribute that can be measured and compared. In any case, "Dismissing on hands evaluations from seasoned photographers" on this topic is the <b>only</b> option because they're entirely contradictory so they cancel themselves out as a useful metric. For every photog who likes the D300's noise characteristics I can find <b>at least</b> one other photog who thinks it's terrible. <br><br>

 

So all we're left with is the images people have posted on the web, and based on those my impression is that the in-camera noise reduction Nikon uses on the D300 is outrageously aggressive and creates soft, fuzzy images.<br><br>

 

I guess my original point was that up until the 20D it looked like every new generation of DSLR had lower noise than the previous one <b>despite</b> the fact that the pixel count was going up (=photosite size was going down)! But the 20D came out years ago yet that progress seemed to stop.

[tonybeach]

"For every photog who likes the D300's noise characteristics I can find at least one other photog who thinks it's terrible."

 

> Then check their credentials. You can take what Ellis says to the bank. What I see on the internet in general is those that are the most critical often have the least impressive credentials. I can tell you personally and straight up that anyone who says the D300 noise characteristics are "terrible" is either a troll or a dolt.

 

"So all we're left with is the images people have posted on the web, and based on those my impression is that the in-camera noise reduction Nikon uses on the D300 is outrageously aggressive and creates soft, fuzzy images."

 

> You are completely clueless about not only how NR is applied to D300 files but also how this is being handled by photographers you don't know personally or have any knowledge about their preferences or workflow. D300 NR is not outrageous unless you turn it up, and it can be turned completely off.

 

"I guess my original point was that up until the 20D it looked like every new generation of DSLR had lower noise than the previous one despite the fact that the pixel count was going up (=photosite size was going down)! But the 20D came out years ago yet that progress seemed to stop."

 

> Well, the D300 has 12 MP versus the 20D which has 8 MP, and that translates to more resolution and ultimately to bigger prints. The D300 is a close match pixel for pixel with the 5D (probably within a 1/3 of stop) when it comes to noise and you will enlarge those pixels less than 20D pixels because there are more of them (it's a matter of simply doing the math).

 

The bottom line is that the quality and amount of noise coming from a D300 will be better than a 20D. Is it worth upgrading? Considering that you are replacing a D100, the D300 is an enormous upgrade. If you want an enormous upgrade to the 20D, then the D3 is the breakthrough camera.

[Brad_]

>>> For every photog who likes the D300's noise characteristics I can find at least one

other photog who thinks it's terrible.

 

Peter, can you post a handful of links with respect to (serious) photogs who think D300

noise is terrible. Serious question as I'm looking to upgrade...

 

From anecdotal observation of on-line D300 pics, from what I've seen, high-ISO noise is way

better than my 20D (which I know well using it for 3-4 years).