Re: Nikon D80/200 - Canon 30d

newsmb@xxxxxxxxx <newsmb@xxxxxxxxx> wrote:
On Jun 10, 2:52 pm, Wolfgang Weisselberg <ozcvgt...@xxxxxxxxxxxxxx>
wrote:

Ughh.
Serves me right for feeding the trolls.
I'd argue *I* am feeding a troll.
You'd lose. Again.

Of course! You, being a troll, cannot ever be wrong.

This will be my absolute, hope-to-die last post on this topic and I
will let people decide for themselves who the troll is. OK?

Fine with me.

The typical troll tactic is to continuously escalate the debate into
something other than what is at issue through inflammatory straw man
argumentation that prompts a less than amicable response. You have
repeatedly attributed arguments to me which I have not made.

I have? Where?

And from that, in your world, follows that even a full frame CMOS
sensor with large pixel sizes, well made, will have more noise
than an el cheapo Point&Shoot camera with a tiny teeny-weeny CCD
chip and far too many megapixels squeezed upon it.

No, from that I treat with a certain degree of scepticism any claims
from a manufaturer that its CMOS sensor has less noise than a CCD
sensor of a similar size and pixel density. It's not the same as
saying it doesn't happen or it can't be done.

I can live with that. However, in that case I'll take your links
as "I have some doubts", not as "proof that it cannot be", the
latter being what I understood you to indicate with these links.
That, however, may well have been a misunderstanding on my part.

Furthermore, that is a slippery slope tactic on your part - another
variation of the straw man fallacy.

Reductio ad absurdum is the phrase you were looking for. I take
(what I understand to be) your position and show that the logical
conclusion of that position is obviously wrong.

Obviously, both conclusions are completely wrong, and irrelevant
to boot --- we are not talking about CMOS and CCD sensors as such,
we are talking about a *specific* CMOS sensor and a *specific*
CCD sensor. I would be delighted if you were able to grasp that
some day.

Let's not rewrite history. You are arguing that the Canon CMOS sensor
itself is less noisy,

I am arguing that the 20D and 30D sensor is less noisy than
the D200 sensor --- not "the Canon CMOS sensor". There is no
single "Canon CMOS sensor".

and I am arguing that the differences between
the images of the two cameras is (1) entirely attributable to stronger
NR by the Canon and (2) insignificant in real-world photography.

Re (1): I understand that Canon RAW files have no NR beyond the
scope of a single pixel (i.e. no blurring or averaging,
though dark frame substraction is applied if available)
According to
http://www.astrosurf.com/buil/d70v10d/eval.htm
Nikon, at least on the D70, does run a median filter,
i.e. a multi-pixel averaging, over the RAW file.

That being said, I doubt that Canon does a stronger
NR (eating resolution), especially when we come to RAW.


Secondly, have another look at
http://www.dpreview.com/reviews/canoneos30d/page20.asp
. Ignore the text, just look at the images (not the
black or grey stripes), especially at
http://a.img-dpreview.com/reviews/CanonEOS30D/Samples/ISO/d200_iso0100_crop.jpg
versus
http://a.img-dpreview.com/reviews/CanonEOS30D/Samples/ISO/d200_iso0800_crop.jpg
(Nikon D200 at 100 and 800 ISO). You'll see that at
800 ISO the Nikon _does_ smoothing (look at the hair,
and the right part of the crown).

Now compare
http://a.img-dpreview.com/reviews/CanonEOS30D/Samples/ISO/30d_iso0100_crop.jpg
versus
http://a.img-dpreview.com/reviews/CanonEOS30D/Samples/ISO/30d_iso0800_crop.jpg
versus
http://a.img-dpreview.com/reviews/CanonEOS30D/Samples/ISO/30d_iso1600_crop.jpg
(Canon 30D at ISO 100, 800 and 1600). I see _very_ light
NR at 800, and comparable (to D200 @ ISO800) NR at 1600.

Again, I _doubt_ that Canon's 30D uses a stronger
multi-pixel (blurring/detail hurting) NR than the D200,
I am inclined to believe the opposite is true: the D200
uses NR earlier, and stronger. (Of course, if you stay in
the ISO 100-400 range, both cameras should be excellent.)

Comparing
http://a.img-dpreview.com/reviews/CanonEOS30D/Samples/ISO/d200_iso3200_crop.jpg
and http://a.img-dpreview.com/reviews/CanonEOS30D/Samples/ISO/30d_iso3200_crop.jpg
my eyes tell me that the D200 image is splotchy and
resolves less (compare the hair, for example) than
the 30D image at ISO 3200.

So, no, I don't think that it's a case of Canon simply
blurring the image more or things like that.

Re (2): In almost all cases, yes. However, once you have to
push images several stops --- and I gave an example
--- noise can very obvious, and significant.

I
pointed to the electronics pages as grounds for my scepticism over
your claim.

You did, thoough I understood them as "no, and here is why".

So your links are very nice, but not relevant.

Yes. Links that explain the technology and the relative strengths and
weakness of each type of sensor are "irrelevant".

When it comes to comparing a given, specific sensor of one type
to a specific sensor of another type, yes, then general relative
strenghts and weaknesses are irrelevant. If you are to compare,
um, Xena with Dr. Watson, I know who'd be tying whom into knots,
even if women are supposedly weaker and all that.

If we were to construct a new sensor, that would be different.

I also pointed to Ken Rockwell's site, which had pretty convincing
tests that demonstrated that differences in noise between the two
cameras is negligible.

Your point was:
| If you run the D200 image (especially the RAW file) through Noise
| Ninja or NeatImage you can get results at least as good as what the
| 30D produces.
(Message-ID: <1180574646.014178.324...@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>)

That implies one of two things:
a) (Noise Ninja or NeatImage) are not improving the results and
the D200 images are already as good as the 30D images.
b) The images of the D200 are inferior in noise to the 30D images,
by a margin not larger than what (Noise Ninja or NeatImage)
can correct.
True?

More or less. And admittedly this is a qualitative judgement on my
part based on what I have been able to do with Noise Ninja and ISO1600
shots from my D200.

Here is a photo I took last week.

http://gallery.plcom.net/public/DSC_1450.jpg?width=800

I can see some JPEG artifacts (e.g. around the (c)-text and
the left side of the face), but no noise as noise.

Let me counter with
http://weissel.smugmug.com/photos/160107906/800x800.jpg
which is a 20D shot at effectively ISO 32.000(!).


You on the other hand have pointed to Canon's marketing page

After all, they ought to know their products better than
_you_ do.

And they wouldn't be trying to sell anyone anything, would they?

Of course noone expects them to point out all the drawbacks
while hiding the good sides, but they do provide a couple of
interesting factlets on how they "cheat" the CMOS noise.

and to dpreview -- which basically just regurgitates Canon's marketing
hyperbole,

Unproven claim.

They don't exactly treat it with a critical eye.

And they handle Nikon and the rest of the world completely
critical?

and grossly overstates the > differences in the images.

I don't need them to understate or overstate the differences,
I have my own eyes.

Well, then we have to agree to disagree here. You see big differences
and I see very similar 100% crops that have no bearing on real-world
photgraphy.

That probably depends entirely on what you define as "real word
photography". On 4x6", as one extreme example, you certainly
won't see noise, on large prints (20x30" and above), especially
images with very high detail, inviting you to look at it closely
.... there differences may become quite important.

The only other "proof" you have is an obscure astronomy site,

Clarkvision is an "obscure astronomy site"? HUH! Try auditioning
as a clown, some day. Especially "astronomy" should give gales
of laughter.

whose results don't mean what you think they do.

Well, if _you_ know so much, why don't you go and show us
what they mean?

Photographically, nothing.

Really!
Ignoring for a moment that you, too, care about noise (using
NoiseNinja, which uses e.g. 100% crops of supposedly identical
colour (or raw pixels, in Bibble) like that to understand the noise
it combats), astrophotography was a valid part of photography,
last *I* looked.

But is it not a measurement of the "statistical" read noise of the
resulting images at various ISOs and not of the sensor itself?

Huh?
Do you say that not just the electron-collecting sensor but
also the read-out circuits and amplifiers and A/D converters
are measured?

In that case, yes, of course, as they are the ones producing the
RAW data ... someone has to do the electron counting, after all.



I have not yet found a single shred of proof for
| If you run the D200 image (especially the RAW file) through Noise
| Ninja or NeatImage you can get results at least as good as what the
| 30D produces.

I don't care if you believe me.

It's not a question of belief, this is not religion.

I just found it positively rich the way you say it's irrelevant, then
expend so much time and energy gazing at 100% crops of ISO3200 test
images looking for noise.

You are missing a zero there.

In addition, one of the advantages of CMOS is
that you can apply NR circuitry onto the sensor itself, so that even
the RAW files have NR applied to them. In fact there's nothing to stop
the camera manufacturer from applying NR to RAW files from CCDs for
that matter. Which is one of the reasons why noise measurements are
largely meaningless.

If you had actually read my posts, you would know that Nikon
does exactly that ... and still comes off worse. Now, what
would that tell you?

Oh, you mean the dpreview test where the Nikon images are underexposed
by about half a stop relative to the Canon's?

If Nikon cameras underexpose, it would have something to do with
their metering --- after all, the camera should be the one to
know how to meter for itself.

On the other hand, Canon's ISO is 1/3rd of a stop low, so the
difference is (or would be) 1/6th of a stop, which doesn't matter
in photography.

That in itself would
skew the results, wouldn't it?

Only if you compare naked sensor to naked sensor, instead of
camera to camera.

Not to single out dpreview, but that's
the problem with these comparisons in general. There are enough subtle
differences in the way the cameras handle exposure, contrast and
colour rendition -- as well as a different approach towards noise
reduction -- that it is rarely a simple matter of "better" or "worse".

True. But testing cameras means that *the camera* gets to
decide how to expose and handle contrast and so on. If the
camera can't hack it, too bad for the camera.

And zooming in necessarily means that you don't judge how the camera
renders the entire image as a whole, which in the end is the only
thing that really matters (at least to me). It's the classic case of
not seeing the forest for the trees.

You are aware that dpreview offers these type of shots as well.

I think all the stuff out
there is great. Canon, Nikon, Pentax, Olympus, Panasonic, Sony and the
others. People should just find one they like and not worry about any
of this crap we're discussing.

Of course.

And I maintain that noise IS largely meaningless.

I disagree.

Of course, any decent DSLR will not have 'real' noise problems
in every day photography, even with ISO 400 and 800.

But the run-of-the-mill (ultra)compact camera with their tiny
sensor and too many megapixels crowded on it ... that camera either
gets very noisy, especially past ISO 200, or uses an ungodly NR,
rendering skin like plastics.

In the film world

.... grainyness is part of the image. It also has a different
"taste" compared to digital noise. That may be an aquired taste
(of our society in general), though, which may change over the
years and decades.

photographers still use Kodak's Tri-X decades after much sharper,
finer-grained film came along. Why? Because black and white
photographers are infinitely more concerned with contrast, tonal range
and half-tone separation than they are with grain and resolution. The
grain was part of the image and no one worried about it. And now
digital comes along and people go berzerko over noise, which is much
less obtrusive than grain ever was. Go figure.

Film grain does happen in mid-ranges, not near pure black or
pure white. Digital grain is in the shadows. It is (perceived
to be) much uglier than the "natural" grain of film.

-Wolfgang
.

Relevant Pages

  • Re: More on Canon Rebel XT noise at high ISO - 2 main new data points
    ... > No it's YOUR fault.The firmware in the camera can't adapt, ... And, when exposure is very bad, I don't understand ... >>As to noise reduction, I don't currently own Neat Image, ... > is free and will batch process up to ten images as a time, ...
    (rec.photo.digital.slr-systems)
  • Re: The Joy of Pixel Density
    ... critics here have no viable working model in their heads of what a sensor ... That is the only way to compare pixel density, ... that higher pixel density lowers DR or increases noise. ... Who said I was trying to tell people what camera to buy? ...
    (rec.photo.digital)
  • Re: Any thoughts /news on Foveon sensors?
    ... >>the design of the sensor and the fact that the pixel count hasn't changed. ... > because it lacks an anti-aliasing filter, ... >>anti-alias filter of a Bayer sensor camera. ... The aliased images from the Sigma camera just look unnatural to ...
    (rec.photo.digital)
  • Re: Any thoughts /news on Foveon sensors?
    ... >the design of the sensor and the fact that the pixel count hasn't changed. ... The X3 sensor doesn't suffer from colour moire on B&W images, ... >anti-alias filter of a Bayer sensor camera. ...
    (rec.photo.digital)
  • Re: Canon 5D stomps Sigma SD14 into the dust
    ... equipped 5D's images. ... ISO 1600 (from what I've seen, things currently turn nasty much before ... There were lots of ISO 800 images, and there was quite a bit of noise; ... personally wouldn't pay so much for this camera. ...
    (rec.photo.digital.slr-systems)