High ISO noise CCD's vs CMOS

Roger N. Clark (change username to rnclark) wrote:

The point about the exposure times and f/stop being different
has nothing to do with noise, because the light levels between
the two different scenes are probably different. It is all
a matter of the total amount of photons received in each
exposure. (A laboratory with fixed lighting would make
for better tests.)

Hello,
I don't disagree with your points below (and indeed the photos were not
even taken on the same day, so the light levels must have been
different), however, the noise in the D200 image is much worse than I
see in my photos, therefore I suspect it was underexposed. Hence my
remark about the different exposures. The point here being that these
samples aren't representative of the difference between the two
cameras, at least in my experience with both of them (and various raw
converters which I used to test them).

My conclusion in the end was that the difference is in high ISOs and in
shadows. Probably, from a brief look at your analysis of the D50 (lack
of time, drowning in work), due to higher read noise. By the way, do
you have any idea which factors affect the read noise (ie why would the
D50 have higher read noise than the 1D)? The amplifiers (seems unlikely
to me)? Presumably, this is random (eg thermal), otherwise they'd just
model it and remove it.

I've though of taking photographs of printed targets with various kinds
of noise on them just to see what kinds of noise reduction are done on
the raw data. [The Nikons certainly do this, see
http://astrosurf.com/buil/d70v10d/eval.htm
and I can't see how the Canons get their noise so low, but haven't
confirmed anything.]

Do you have any idea if anybody has done this? I could not find
anything on the web.

Cheers.



Noise in DSLR camera images is greatly affected by raw converter
software, so the only true way to understand the noise is a proper
noise analysis on raw data that has not gone through a raw
converter.

Examples:
Procedures for Evaluating Digital Camera
Sensor Noise, Dynamic Range, and Full Well Capacities;
Canon 1D Mark II Analysis
http://www.clarkvision.com/imagedetail/evaluation-1d2

The Nikon D50 Digital Camera:
Sensor Noise, Dynamic Range, and Full Well Analysis
http://www.clarkvision.com/imagedeta...tion-nikon-d50

The factors that will influence the noise between different cameras
is directly related to quantum efficiency, fill factor,
pixel size, and at the low intensity end, read noise
(and for long exposures, thermal noise).

CCD quantum efficiencies tend to be slightly higher than
CMOS sensors, so the advantage there is the Nikon (by perhaps
10%).

Fill factors are essentially 100% by the use of micro-lenses
over the detectors (CCD or CMOS), so no advantage to either.

Read noise: Canon's CMOS has 4 electrons on good cameras like the
20D (and by extension 30D; same sensor). CCDs are typically
7 to 15 electrons (the D50 above is ~ 7.5 electrons.
So the CMOS has an advantage of ~2x at the very lowest signals,
not the main things you see in the steves-digicams.com images
which are much brighter.

That leaves the major factor in noise: the pixel size.
The important factor is delivering photons, and to do that,
you need aperture. See:

The f/ratio Myth and Digital Cameras
http://www.clarkvision.com/photoinfo/f-ratio_myth

The D200 has 6.1 micron pixels versus the 30D at 6.4 micron
pixels, so a small difference (actually area is the important
factor: 37.2 versus 40.1 square microns, again not much
difference).

There should be a slight advantage to the 30D but it should be
small. (I would choose the camera based on other factors.)

I will be evaluating a 200 in the next couple of months.

Other sensor data are located in Tables 1-3 at:
The Signal-to-Noise of Digital Camera images
and Comparison to Film
http://www.clarkvision.com/imagedeta...ignal.to.noise

Roger

Roger

ian wrote:
"mswlogo" wrote in message
ups.com...

As far as build quality goes the 20D and 30D are more than adequate
unless you are demanding weather sealing. EOS 5D owners jump in here?

True. Although in spite of my whining, I have yet to even experience my
first sensor-dust speck on my 5D, much less any trouble with other seals.
Would I prefer 1D type sealing? Absolutely. Is it life or death? No.
Should Canon provide it anyway...on a $3K DSLR? No question. -Gaskets are
super-cheap. Nikon proves that.

As for higher ISO the alternative is spending more on lenses with a
larger maximum aperture. That usually means brighter viewfinder and
better low light focusing. I know the eos 20D can make use of F2.8
or wider lenses. An extra set of autofocus sensors come into play. 2
vertical ones. My one gripe with canon equipment is the tendency of
flash system to underexpose. Nikon is said to be superior in this
regard.

I've had excellent flash performance with my 5D and the 580EX. Improvement
over my 10D w/ 550EX is significant. Still not perfect, but it's an
improvement. With full frame, I've had to get back in the habit of flipping
down the built-in wide screen on the flash...but it works well.

--
Images (Plus Snaps & Grabs) by Mark² at:
www.pbase.com/markuson

mswlogo wrote:
Ok, I'm in the market for my first DSLR (upgrade from 35 mm SLR and
Nikon 5700).

I've been looking at reviews on the Sony A100, Nikon D80/D200 and
Canon 30D.

These are both 1600 ISO (see reviews for more detailed information
about conditions etc).

http://www.steves-digicams.com/2006_...s/IMG_8337.JPG

http://www.steves-digicams.com/2006_...s/DSC_3490.JPG

What a HUGE difference !!!

Be aware that differences in in-camera sharpening, and in noise reduction
algorithms may account for a large part of the difference you are seeing.

By the way: I recently had the chance to compare my Nikon 8400 with a
top-of-the-range Canon 5D. The viewfinder on the Canon was awfully dark,
even though the image quality was (obviously) better. People tell me that
was one of the better cameras, as well! It would seem to me that buying
these cheap lenses with a maximum aperture of f/5.6 may be a mistake, and
you should factor in the price of better (brighter) lenses,

David

David J Taylor wrote:
mswlogo wrote:
Ok, I'm in the market for my first DSLR (upgrade from 35 mm SLR and
Nikon 5700).

I've been looking at reviews on the Sony A100, Nikon D80/D200 and
Canon 30D.

These are both 1600 ISO (see reviews for more detailed information
about conditions etc).

http://www.steves-digicams.com/2006_...s/IMG_8337.JPG

http://www.steves-digicams.com/2006_...s/DSC_3490.JPG

What a HUGE difference !!!

Be aware that differences in in-camera sharpening, and in noise
reduction algorithms may account for a large part of the difference
you are seeing.
By the way: I recently had the chance to compare my Nikon 8400 with a
top-of-the-range Canon 5D. The viewfinder on the Canon was awfully
dark, even though the image quality was (obviously) better.

Ah...but that is entirely dependant upon what lens you had attached to the
5D. Your words below seem to indicate some awareness of this, but I wonder
if you realize how dramatic a difference it is...

A 2.8 lens lets in FOUR TIMES as much light at a 5.6 lens, so if you had an
f5.6 lens attached (for example), it would show you a viewfinder image that
exhibits only 1/4th the light of another lens. Heck...if you had the 50mm
1.4 lens atached, that would be a whopping 16x brighter than a 5.6!
Every stop doubles or halves the difference in light collected by lenses one
stop different. A 2.8 is 2 stops faster than a 5.6, meaning it doubles the
light (f4) and then doubles THAT at 2.8, for four times the light. A 1.4
lens doubles at f4 (2x the light), doubles that at 2.8 (now 4x), doubles
again at f2 (now 8x) and again at f1.4 (16x!!).

All that to say... You can't judge any viewfinder on a DSLR until you know
what lens is on it.

-Mark


People
tell me that was one of the better cameras, as well! It would seem
to me that buying these cheap lenses with a maximum aperture of f/5.6
may be a mistake, and you should factor in the price of better
(brighter) lenses,

Yes... So...which lens was on the 5D?


--
Images (Plus Snaps & Grabs) by Mark² at:
www.pbase.com/markuson

Mark² wrote:
[]
Ah...but that is entirely dependant upon what lens you had attached
to the 5D. Your words below seem to indicate some awareness of this,
but I wonder if you realize how dramatic a difference it is...

Yes, I used to own a Nikon F3 with some f/1,4 and f/2.8 lenses.

A 2.8 lens lets in FOUR TIMES as much light at a 5.6 lens, so if you
had an f5.6 lens attached (for example), it would show you a
viewfinder image that exhibits only 1/4th the light of another lens.
Heck...if you had the 50mm 1.4 lens atached, that would be a whopping
16x brighter than a 5.6!
Every stop doubles or halves the difference in light collected by
lenses one stop different. A 2.8 is 2 stops faster than a 5.6,
meaning it doubles the light (f4) and then doubles THAT at 2.8, for
four times the light. A 1.4 lens doubles at f4 (2x the light),
doubles that at 2.8 (now 4x), doubles again at f2 (now 8x) and again
at f1.4 (16x!!).
All that to say... You can't judge any viewfinder on a DSLR until
you know what lens is on it.

-Mark


People
tell me that was one of the better cameras, as well! It would seem
to me that buying these cheap lenses with a maximum aperture of f/5.6
may be a mistake, and you should factor in the price of better
(brighter) lenses,

Yes... So...which lens was on the 5D?


One of the f/5.6-ish zooms. The owner had both wide-angle and telephoto
zooms. I only handled the camera for a new moments, but as I have been
using my Nikon 8400 just before, the brightness difference was shattering.

If the 5D is supposed to be a "bright" camera, I hate to think what the
others are like. Certainly nothing as pleasant as my SLR experiences of
10-15 years ago. And, of course, no split-image focussing, micro-prism
etc.!

Just goes to affirm what I always suggest - try the equipment for yourself
before purchase.

David

People
tell me that was one of the better cameras, as well! It would seem
to me that buying these cheap lenses with a maximum aperture of f/5.6
may be a mistake, and you should factor in the price of better
(brighter) lenses,

Yes... So...which lens was on the 5D?


One of the f/5.6-ish zooms. The owner had both wide-angle and telephoto
zooms. I only handled the camera for a new moments, but as I have been
using my Nikon 8400 just before, the brightness difference was shattering.

If the 5D is supposed to be a "bright" camera, I hate to think what the
others are like. Certainly nothing as pleasant as my SLR experiences of
10-15 years ago. And, of course, no split-image focussing, micro-prism
etc.!

Just goes to affirm what I always suggest - try the equipment for yourself
before purchase.

David

The Nikon 8400 has a video based view finder, so the brightness of the
viewfinder is based on the video gain, not the lens. So it is apples
and oranges. One of the problems with having auto focus and variable
ISO in digital cameras, and that the higher ISO results are so good, is
the slow lens (f3.5 - f5.6) is making its come back. With film SLRs we
were sold 50 f2 as the entry level lens, now we have a f3.5- f5.6 zoom.
With this we get a dim viewfinder. The less expensive Nikon and Canon
cameras also have mirror based viefinders dimming the view even more.
It is actually a good thing that the D70 and D50 Nikon can't take
manual focus lenses as it would be difficult to focus with these lenses
with the dim view finders.
When I bought my D200, the first DSLR I owned, I bought a 24 f2 mf
Nikon lens for it, mainly because I was used to 35 f2s on my film SLRs.
Not a big brightness difference between my D200 and my Canon F1n with a
35 f2 lens on it. There is a visible frame size difference, but the
brightness is about equal. The D200 screen though makes it tough to
accurately focus a wide angle, may get a Katz screen for it.
Back to high ISO, the D200 is only about a stop different from a Canon
D30 IF you nail the exposure, under exposure is a killer with that
camera in high ISO. Unfortunatly working in high ISO necessary environs
often is where you are shooting quickly rather than figuring out each
shot, if you are shooting in that evironment alot get a Canon. Other
wise which ever body is ergonomicly good for you.

Tom

mswlogo wrote:

But I think it is true that the Canon is at least 1 stop better, and I
keep reading excuses.

There isn't enough noise with any of these cameras that it should even
be a factor in your decision. I could make a picture with the Canon that
has far more noise than that Nikon "example"; does that mean it's a noisy
camera? No, of course not.

If you don't expose properly you get noise. If you do expose properly,
it's just not a factor. I shoot with low light and high ISO with a Nikon
D2x quite a lot, and noise just doesn't enter into it. I haven't seen
enough noise to even tempt me into getting any noise reduction software.
But, naturally, if you underexpose, then there's noise. So don't do that.

Most "noise comparison" shots you see are total nonsense. Comparing noise
from one camera to another is very nearly impossible to do in a way that
means anything at all.

But I just can't get paste this ISO thing

Get past it. It's not worth worrying about.

--
Jeremy |

Jeremy Nixon wrote:

If you don't expose properly you get noise. If you do expose properly,
it's just not a factor. I shoot with low light and high ISO with a Nikon
D2x quite a lot, and noise just doesn't enter into it. I haven't seen
enough noise to even tempt me into getting any noise reduction software.
But, naturally, if you underexpose, then there's noise. So don't do that.


The point, however, is that the D200 at high ISO reacts badly to
underexposure (ie worse than the 20D). If it gets enough light, there
is no visible noise. But of course a photograph with a wide dynamic
range will include shadows. Still, with postprocessing, I found that
the difference can be minimised, so it's a non-issue in reality (for
me).

Of course, someone else's definition of reality might be different from
mine.

"Jeremy Nixon" wrote in message
...
mswlogo wrote:

But I think it is true that the Canon is at least 1 stop better, and I
keep reading excuses.

There isn't enough noise with any of these cameras that it should even
be a factor in your decision. I could make a picture with the Canon that
has far more noise than that Nikon "example"; does that mean it's a noisy
camera? No, of course not.

If you don't expose properly you get noise. If you do expose properly,
it's just not a factor. I shoot with low light and high ISO with a Nikon
D2x quite a lot, and noise just doesn't enter into it. I haven't seen
enough noise to even tempt me into getting any noise reduction software.
But, naturally, if you underexpose, then there's noise. So don't do that.

Most "noise comparison" shots you see are total nonsense. Comparing noise
from one camera to another is very nearly impossible to do in a way that
means anything at all.

But I just can't get paste this ISO thing

Get past it. It's not worth worrying about.

--
Jeremy |

Of course it's not worth worrying about, and any thinking photographer
understands that. That is anyone who actually USES THE EQUIPMENT IN
QUESTION!!!
Think about it though -- if we didn't have these esoteric "noise" threads
periodically, it would get dull around here and all the "pixel peepers"
wouldn't have anything to discuss, let alone getting out to actually
photograph anything.
Bob

mswlogo skrev:
Ok, I'm in the market for my first DSLR (upgrade from 35 mm SLR and
Nikon 5700).

I've been looking at reviews on the Sony A100, Nikon D80/D200 and Canon
30D.

These are both 1600 ISO (see reviews for more detailed information
about conditions etc).

http://www.steves-digicams.com/2006_...s/IMG_8337.JPG

http://www.steves-digicams.com/2006_...s/DSC_3490.JPG

What a HUGE difference !!!

This comparison is ridiculous - the second is underexposed at leat one
step - more likely two steps. In Denmark we could call such a comparison
fraud without any riscs.

--
Med venlig hilsen, Ole Larsen.
New Images And Design, aug. 2006
http://Olelarsen.eu/
http://home.tiscali.dk/muggler