Does f/2=f/2?

[Medic Laviña]

... others mistakenly use the F numbers when they try to compare FOV + DOF between cameras with different sensor sizes.

--
Medic
-----------------------------------------------------

• The camera is mightier than the pen.

 

[Haider]

I'm sorry, but you don't have the right understanding of how noise
happens in sensors.

There are many different types of noise on sensors. If your photosite is correctly recording a max value, it's recording a max value. There is no room for noise to be recorded. Do we agree on that? Equally when a phoosite is not recording a max value that truly exists it is could be recording a 'noise' value.

Let's compare two theoretical sensor cells. Each is from a 12MP

Can we please make it just one pixel (if it foveon) or just one photosite for bayer.

sensor, one is from an APS-sized sensor (like the Nikon D300) and one
is from a 35mm sensor (like the Nikon D3). For the purpose of
focusing on just the size difference, let's assume that all else in
the sensor design is identical.

When you say sensor size I'm assuming you mean the size of the photosites?

For most of the things that
contribute to noise, for the purposes of this comparison, we can
assume that a given sensor cell will have X amount of noise in a
given circumstance. The value X has nothing to do with the size of
the cell so both the D300 sized cell and the D3 size cell have the
same absolute amount of noise in them.

NO SOME PHOTOSITES RECORD NOISE VALUES WHILST OTHERS RECORD CORRECT (OR TRUE) VALUES. The bayer interpolation will use the noise values and the good values to calculate pixels which will then have wrong RGB values.

Now expose the two cells to an image with the same focal length lens,
the same aperture setting and the same shutter speed.

Can we stick to aperture (hole size) and shutter speed please.

The larger cell collects more photons because it has a larger area.

Depends on the volume of light entered whilst the shutter was open. Can we say we used an extremely fast shutter speed and only 2 photons came in. Now say the smaller photosite has a maximum capacity of 2 photons whilst the 35mm sensor with it's bigger photosite has a capacity of 8 photons. The smaller sensor photosite has maxed out it has collected 2 photons and there is no room for noise to increase this value any further. This is how full the bucket is (signal) and noise are related. S'n'R value you see in sensor spec sheets is averaged value across the dynamic range, it is not true for every single input value or any single value. It is an average value. Whilst the 35mm sensor has recorded a 25% value there is still room for noise to bump this value up. Say noise bumps it up to say 50%. No when process this value my cameras says (I've hit the limit of DR here) it's 2 photons which is correct, your cameras say it 4 photons - incorrect. That's a noise value for you and good value for me. Of course as by product my photo will have a lower DR than yours that's the trade-off. UNFORTUNATELY the associated electronics on smaller sensors produce more 'noise' or interference when the sensor is not recording a true 100% value. That's why people say smaller sensors are noisier BUT THEY FORGET a smaller sensor/photosite needs less light to hit high values thus reducing the room available for noise to bump up the values. Noise is inversly proportional to the decibel of the signal input i.e. the bigger the signal input (more decibels or photons) the lower the noise content will be of the value recorded by the sensors.

The number
of photos is proportional to the signal that this cell produces. So,
let's say the amount of photons collected by the smaller D300 sized
cell is N. Since the D3 sized sensor cells are 2.3x as large the
area as the D300 sized sensor cells, that would mean that the D3
sized cell collects around 2.3N the number of photons.

So, you can see that the absolute value of the noise is approximately
the same in the two examples, but the signal is 2.3 times higher for
the larger sensor cell. If you then want to look at the
signal-to-noise ratio (which is what determines how visible the noise
is), you'd see that the larger sensor cell as a 2.3 times higher
signal-to-noise ratio.

The signal-to-noise ratio has absolutely nothing to do with how
"full" the bucket is.

It is take a shot in very bright light than take a shot in low light. Any DSLR user will tell as light (signal) gets less, noise get worse.

BTW, in the above example (assuming identical
sensors that just differ in size), the two sensors buckets are
equally full.

NO 35mm sensors have bigger photosites than smaller sensors at a given MP. That's how they get bigger DR, that's the whole point and room to associated electronics further apart to reduce electrical interference.

It only has to do with the absolute value of the
signal compared to the absolute value of the noise. Holding
everything else constant, raising the signal will raise the signal to
noise ratio and reduce the visible effect of noise.

You mean Absolute value of the signal compared to the maximum value of the signal that the snsor can handle.

 

[nickleback]

There are many different types of noise on sensors.

Correct! And that's about all you wrote that is correct.

If your photosite
is correctly recording a max value, it's recording a max value.

aka saturated.

There is no room for noise to be recorded.

There is also not much information. It's blown. We don't know how far it is blown, so we can't compute the noise.

Let's compare two theoretical sensor cells. Each is from a 12MP

Can we please make it just one pixel (if it foveon) or just one
photosite for bayer.

That's exactly what the other poster is doing. He's comparing one from a 12mp 35mm sensor, and another from a 12mp APS-C sensor.

sensor, one is from an APS-sized sensor (like the Nikon D300) and one
is from a 35mm sensor (like the Nikon D3). For the purpose of
focusing on just the size difference, let's assume that all else in
the sensor design is identical.

When you say sensor size I'm assuming you mean the size of the
photosites?

Since the assumption was number of pixels is the same, and the sensor size is not, then it can be either the sensor size or the pixel size. Your choice.

For most of the things that
contribute to noise, for the purposes of this comparison, we can
assume that a given sensor cell will have X amount of noise in a
given circumstance.

NO SOME PHOTOSITES RECORD NOISE VALUES WHILST OTHERS RECORD CORRECT
(OR TRUE) VALUES.

All photosites record noise. BTW, your caps lock key appears to be stuck.

The bayer interpolation will use the noise values
and the good values to calculate pixels which will then have wrong
RGB values.

Bayer interpolation doesn't matter for this discussion. Every raw pixel has noise. Bayer interpolation is done afterwards.

Now expose the two cells to an image with the same focal length lens,
the same aperture setting and the same shutter speed.

Can we stick to aperture (hole size) and shutter speed please.

I believe that's what the other poster said. Same focal length and same aperture setting is the same apparent aperture (i.e. hole size). Same shutter speed means same shutter speed.

How do you see it as different?

The larger cell collects more photons because it has a larger area.

Depends on the volume of light entered whilst the shutter was open.

Larger cells collect more photons. The only time it isn't true is when you saturate the cells, at which point you don't have an image, you have a white blob.

Can we say we used an extremely fast shutter speed and only 2 photons
came in. Now say the smaller photosite has a maximum capacity of 2
photons whilst the 35mm sensor with it's bigger photosite has a
capacity of 8 photons. The smaller sensor photosite has maxed out it
has collected 2 photons and there is no room for noise to increase
this value any further.

False assumptions.

That would be 2 photons per unit area, and since the 35mm pixels are more than double the size of the APS-C pixels, each pixel on the 35mm sensor will see more than double the photons.

The size of the 35mm vs APS-C pixels is overstated (but actually to benefit the 35mm sensor, if you knew the math). If it is as stated, then the base ISO of the 35mm sensor would be 1/2 the base ISO of the APS-C sensor.

The example of two photons is absurdly low and masks the effect and magnitude of signal and noise.

This is how full the bucket is (signal) and
noise are related.

A full bucket stil has noise, you just don't know how much because it is saturated.

S'n'R value you see in sensor spec sheets is
averaged value across the dynamic range, it is not true for every
single input value or any single value. It is an average value.

Sure. Hey wait, that's two thing you have correct!

UNFORTUNATELY the associated electronics on smaller
sensors produce more 'noise' or interference when the sensor is not
recording a true 100% value.

When you are recording a true 100% value, you have a shot of a polar bear in a snowstorm. i.e. white nothing. Not a very interesting image.

That's why people say smaller sensors
are noisier BUT THEY FORGET a smaller sensor/photosite needs less
light to hit high values thus reducing the room available for noise
to bump up the values.

And you forget that noise will be a bigger part of the small pixel image at anything short of blown.

It is take a shot in very bright light than take a shot in low light.
Any DSLR user will tell as light (signal) gets less, noise get worse.

True with any digital camera, regardless of sensor or pixel size.

BTW, in the above example (assuming identical
sensors that just differ in size), the two sensors buckets are
equally full.

NO 35mm sensors have bigger photosites than smaller sensors at a
given MP.

Bingo! Eureka! You got it. Or at least half of it. Now you just have to get it that total amount of light falling on a larger area is... larger.

--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed

 

[marf]

well F/2 is F/2 but the bigger sensor is a bit more sensitive because more light particles are collected from the bigger area.

But my opionion is, the bigger sensor is not linear more sensitive to increased area, the bigger sensor is more sensitive but not so much:

i wrote an article why 35mm is not 2 stops less noisier than the 4/3" sensor.
http://forums.dpreview.com/forums/read.asp?forum=1022&message=25338715
--
regards
Martin

-----------------------------
Typing errors are intended to provide a basis for global amusement.

 

[jfriend00]

Can we please make it just one pixel (if it foveon) or just one
photosite for bayer.

Can you please re-read my last posting. That's EXACTLY what I did. I compared one sensor cell in a D300 APS-sized sensor to one sensor cell in a D3 35mm-sized sensor. Since I've already done exactly that, I'd ask you to please tell me what you don't understand in what I already wrote rather than ask me to do something I've already done.
--
John
Popular: http://jfriend.smugmug.com/popular
Portfolio: http://jfriend.smugmug.com/portfolio

 

[Tristan Cope]

The people mentioned in the original post seem to think that changing
the sensor size will alter the intensity of light falling on the
sensor. That IS utter rot - like imagining that the sun becomes more
intense the more square inches of skin you expose to it.

Assuming the same f-stop is used (e.g. f2), the intensity of light (i.e. photon per unit area) will not change - as you say.

But: If you have two sensors with different areas (let say one 35mm and one 4/3) the total amount of light gathered on the larger sensor will be greater. Simple equation - light gathered = light intensity per unit area x sensor area.

If both sensors have the same number of photosites (pixels), then the larger sensor will gather more photons per photosite than the smaller sensor, because the photosites are larger. Its the same equation - light intensity x photosite area. This assumes the same fill factor for the photosites.

So, yes f2=f2=f2. But at f2, a larger sensor will gather more light per pixel than a smaller sensor with the same number of pixels. It will therefore have less noise. if you doubt this in any way whatsoever, have a look at compact camera with 10MP compared to a 10MP dSLR.

--
I think I'll change my signature

 

[Tristan Cope]

Your explanation is completely wrong.

--
I think I'll change my signature

 

[Haider]

Sorry about the caps, I'm not being rude and it is educational

If your photosite
is correctly recording a max value, it's recording a max value.

aka saturated.

There is no room for noise to be recorded.

There is also not much information. It's blown. We don't know how
far it is blown, so we can't compute the noise.

True it's blown but there is no noise because the photosite is CORRECTLY recording it's max value.

Since the assumption was number of pixels is the same, and the sensor
size is not, then it can be either the sensor size or the pixel size.
Your choice.

Instead of pixel size could we say photosite volume. As pixels are formed by bayer interpolation.

All photosites record noise.

Agreed but only when the photosite is not saturated

Bayer interpolation doesn't matter for this discussion.

Agreed

How do you see it as different?

I'm just keeping it simple hole size and shutter speed - pinhole camera my 1st camera.

The larger cell collects more photons because it has a larger area.

You agree it depends on the volume of light entered whilst the shutter was open?

Larger cells collect more photons.

They have a bigger dynamic range i.e. can collect more photons if they are provided so they won't saturate as quickly.

The only time it isn't true is when you saturate the cells,

No if you only provide two photons they cannot collect more light than what has been provided.

at which point you don't have an image, you have a white blob.

Agree but if you are taking a photo of white blob thats a perfect image.

False assumptions. That would be 2 photons per unit area,

Only two photons were provided to the sensors no more no less. Unit area is irrrelevant.

and since the 35mm pixels are more than double the size of the APS-C pixels, > each pixel on the 35mm sensor will see more than double the photons.

Disagree as only two photons were provided. The 35mm sensor cannot create more photons. You agree with this? This is the bit of your argument I don't understand...How does the 35mm collect more than the 2 photons provided?

The size of the 35mm vs APS-C pixels is overstated (but actually to
benefit the 35mm sensor, if you knew the math). If it is as stated,
then the base ISO of the 35mm sensor would be 1/2 the base ISO of the
APS-C sensor.

I was using theoretical values but actually 35mm is x4 the size of 4/3rds.

The example of two photons is absurdly low and masks the effect and
magnitude of signal and noise.

Ok increase it to a 1000 photons of light.

A full bucket stil has noise, you just don't know how much because it
is saturated.

You're confusing what is provided to the sensor and what it records i.e. signal Vs. signal+noise.

S'n'R value you see in sensor spec sheets is averaged value across the dynamic > > range, it is not true for every single input value or any single value. It is an > > average value.

Sure. Hey wait, that's two thing you have correct!

I'm on a roll now, no stopping me now do I get a Phd?;-)

And you forget that noise will be a bigger part of the small pixel
image at anything short of blown.

Let stick to photosites to get to pixels we have do that bayer interpolation.

It is take a shot in very bright light than take a shot in low light.
Any DSLR user will tell as light (signal) gets less, noise get worse.

True with any digital camera, regardless of sensor or pixel size.

3 things right & the boys on a home run!;-)

Now you just have to get it that total amount of light falling on a larger area
is... larger.

You're jumping to an assumption here that a larger volume of light is being provided to the larger area. If you provide the same amount of light (due to the same aperture and shutter speed) the concentration of light in the smaller area is higher. As you increase the concentration of light you reduce the noise.Think of it as 1/2 pint of beer whether it's in a pint glass or 1/2 pint glass it's still 1/2 pint of beer. The concentration of beer is higher in the 1/2 pint glass and the concentration is lower in the pint glass (half the glass is full of air). Agreed?

Await your reply in earnest and the beers are on you and no buying 1/2 pints and putting in pint glasses

 

[Ron Parr]

You seem very confused about some basic concepts here, so rather than trying to dissect all of the misunderstandings, I'm going to state a few general things in hopes that this helps clear things up a bit.

• We can view a pixel's job as estimating the rate at which it is getting hit with photons. (Even though the pixel is actually estimating the total photon incidence and not rate, the difference between these two is just a scaling factor determined by the exposure time.)

• We can view noise as the difference between the true rate of photon incidence and the reported rate. Typically this is reported as the standard deviation in the output.

• The pixels in your sensor can be thought of as converting photons to electrons. Some of the electrons stored at the pixel will come from photons and some will come from random noise in the chip. Even if there were no random noise in the chip, there would still be noise in what is recorded by your sensor, due to randomness in the flow of photons hitting your sensor. This type of noise is called photon shot noise.

• The photon shot noise described above has a Poisson distribution. The standard deviation for a Poisson distribution with mean n is sqrt.

• When a pixel saturates, it is unable to detect any more photons. Once a pixel has saturated, the error in the estimate of the rate of photon incidence can be arbitrarily large since we have no way of estimating what the true rate was other than the fact that it is above a certain level. For this reason, one can make talk meaningfully about noise up to the point of saturation, but not at or beyond it.

• Consider two pixels of different sizes which are exposed to light at the same intensity and have the same efficiency. Suppose pixel A has 2X the area of pixel B. At any time before saturation, pixel A will record 2X as many photons as pixel B. Suppose pixel B records n photons. If we consider the photon shot noise only, then pixel B will have an expected signal to noise ratio of n/sqrt = sqrt. On the other had, pixel A will have signal 2n and an expected signal to noise ratio of 2n/sqrt(2n) = sqrt(2n) = ~ 1.4 sqrt. So, we see that pixel A has higher signal to noise ratio. When the pixels saturate, we don't know what the true signal should be, so we can't even estimate the noise.

--
Ron Parr
Digital Photography FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[highsee3]

well F/2 is F/2 but the bigger sensor is a bit more sensitive because
more light particles are collected from the bigger area.

So, some people are saying that because 35mm's surface area is 4x larger than 4/3, it's 4x more sensitive with the same f stop? IE 35mm ISO 400 is "roughly equivalent" to 4/3 ISO 100?

Therefore, 6x4.5 is 3.6x larger than 35mm and would be "roughly equivalent" to ISO 1250?

4/3 (225 sq 3.8x smaller than 35mm)
APS-C (329 sq 2.6x smaller than 35mm)
35mm (864 sq)
6x4.5= (2352 or 2.7x larger 35mm)
6x7= (3752 sq or 4.3x larger than 35mm)
4x5= (12,954 sq or 15x larger than 35mm)

I'm gonna round to make things easier, but I guess if you use the idea above (if you 2x the sensor size, you gain a stop) you would get the following

4/3 = ISO 100
APS C = ISO 160
35mm = ISO 400
6x45 = ISO 1250
6x7 = ISO 1600
4x5 = ISO 6400

Wow! who knew my Mamiya was such a light eating moster with the 80mm f/1.9

I agree the that the gains are not linear to the sensor size area, IHMO the are very dependent on

1. sensor technology
2. output size

if you never print larger than 8x10, you have no need for 6x7 and would see no gain what so ever over 4/3, just a much larger, more expensive camera system.

at the rate we are going, DSLRs 35mm and up are becoming overkill for 90% of photographers. The 40D and D300 proves this point. They offer a quality and resolution thought to be only attainable in FF 35mm a couple of years ago.

--
http://www.highsee3.smugmug.com

'A camera maker that simply copies others' idea has no right to call itself an original
maker in the first place.' -Mr. Maitani, creator of the OM photographic system.

 

[Haider]

Antonio see my reply in the other posts as well for a fuller explanation.

Pedagydusz wrote:

The photosites are small buckets. You put them out to > collect the rain for a short, well determined, time interval.
Let call it shutter speed

The smaller the bucket, the less water it collects.

Yes but that is dynamic range not noise, you are confusing the two. The smaller bucket will get saturated quicker i.e. have lower dynamic range. When a bucket is saturated there is no noise. As you head toward saturation the noise decreases.

I'll give you an example. One thing photosites are three dimensional, that why digital sensors require telecentric lens. If you have rain-fall of 5cm3 per second a 10cm3 bucket will get half full where as 20cm3 bucket will be 1/4 full in 1 sec. The level of concentration of water in the 10cm3 bucket is 50% where as in the 20cm3 bucket the concentration is 25%. As you increase the concentration of water (in terms of a sensor, you increase the concentration of light) noise gets less. We need to be clear on this now matter what sensor you have as you increase the concentration of light to it, it noise will go down.

Now I'll throw the spanner into the works it all comes down to how noisey the sensors are. Smaller sensors are noiser to start of with therefore an APS sensor at 50% concentration may record more noise than a 35mm sensor at 25% concentration. It's nothing do with the amount of light the 35mm has captured, we only provided it with 5cm3 of light same as the APS sensor. You tell me without increasing the aperture size and/or increasing the exposure time where is the 35mm sensor going to get the extra light. Like I said a 1/2 pint beer is 1/2 pint of beer whether you put in a pint glass or 1/2 pint glass. 35mm sensors have better noise characteristics that allow them produce less noisier photos. That is the point I'm getting at and have been on this thread. This is down to the fact that you are squeezing same electronics into a smaller (APS or 4/3rds) space which results in less isolation for each of the components in the circuit which results in greater interference or noise.

 

[nickleback]

I'd reply point by point, but it's pointless. Just one thing...

If you provide the same amount of
light (due to the same aperture and shutter speed) the concentration
of light in the smaller area is higher.

The concentration of light is the same, because the aperture is the same. If you shine this same concentration of light on a larger surface the larger surface will see more light. That's the whole definition of aperture. Understand this, and you'll understand the rest.

Think of it as 1/2 pint
of beer whether it's in a pint glass or 1/2 pint glass it's still 1/2
pint of beer.

As much as I like beer, this is a flawed analogy.

But since you tried, here's a better beer analogy.

We live in Bavaria and it is raining beer (or, at least it seems like it does, whenever I've been there). I put a (clean!) garbage can out to collect the beer-rain. You put out pint glass.

After the beer storm passes, we go out and see how much beer fell. I pull out my ruler, measure the depth of the beer in my garbage pail and say "woohoo! looks like we had an inch of beer". You pull out your ruler, pop it into your pint glass, and say "yep, we got an inch".

Guess who has more beer?

--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed

 

[highsee3]

If both sensors have the same number of photosites (pixels), then the
larger sensor will gather more photons per photosite than the smaller
sensor, because the photosites are larger. Its the same equation -

light intensity x photosite area. This assumes the same fill factor
for the photosites.

So, yes f2=f2=f2. But at f2, a larger sensor will gather more light
per pixel than a smaller sensor with the same number of pixels. It
will therefore have less noise. if you doubt this in any way
whatsoever, have a look at compact camera with 10MP compared to a
10MP dSLR.

My 8mp Nikon 8400 shows no noise @ ISO 50 (a very old sensor design, I'm sure a modern 2/3" sensor would do better). The RAW files are super clean and the lens is very sharp.

I would say it had no more noise than the 20D (which has a sensor 7x larger) @ ISO 100.

at ISO 400, yes not contest, the DSLR clearly could be pushed to ISO 1600 or more for a similar result, but shouldn't it be linear across all ISOs?

at base ISO I don't see a big difference between P&S and DSLR in terms a noise despite the huge difference in sensor size. The P&S does have a little less DR, but when shooting RAW and not letting the camera choose the exposure, it's still good.

I use the 8400 for it's wide angle and because it has so much more DOF, I don't need to stop it down at all, so it's not a big deal to keep it at ISO 50, 100. It doesn't even stop down to f/8.

I don't disagree that a larger sensor makes a big difference, I just can't figure out how much that difference is or when it occurs.

I've tried some simple formulas to figure it out, but it's never true across the board.

Fuji makes very clean P&S cameras.

Canon got lucky and made a very clean CMOS that reduces noise at the chip level despite the fact that CCDs were supposed to be better.

Kodak make wonderful, but noisy CCD sensors.

Foveon doesn't have great high ISO

Medium format has a huge sensor sensor size, but never has clean high ISO.

The 40D has excellent noise and isn't that far off from the 5D. I really see very little difference between the two even up to ISO 1600.

--
http://www.highsee3.smugmug.com

'A camera maker that simply copies others' idea has no right to call itself an original
maker in the first place.' -Mr. Maitani, creator of the OM photographic system.

 

[highsee3]

We live in Bavaria and it is raining beer (or, at least it seems like
it does, whenever I've been there). I put a (clean!) garbage can out
to collect the beer-rain. You put out pint glass.

After the beer storm passes, we go out and see how much beer fell. I
pull out my ruler, measure the depth of the beer in my garbage pail
and say "woohoo! looks like we had an inch of beer". You pull out
your ruler, pop it into your pint glass, and say "yep, we got an
inch".

Guess who has more beer?

so why does the medium format keg not have good high ISO? too much foam?

further more, do you really want to buy and carry a keg when a couple of pints is all you'll drink?

--
http://www.highsee3.smugmug.com

'A camera maker that simply copies others' idea has no right to call itself an original
maker in the first place.' -Mr. Maitani, creator of the OM photographic system.

 

[Ron Parr]

My 8mp Nikon 8400 shows no noise @ ISO 50 (a very old sensor design,
I'm sure a modern 2/3" sensor would do better). The RAW files are
super clean and the lens is very sharp.

I would say it had no more noise than the 20D (which has a sensor 7x
larger) @ ISO 100.

I'm not so sure I'd agree with these statements. Here are Phil's sample shots:

http://a.img-dpreview.com/reviews/NikonCP8400/Samples/Compared/Studio/cp8400_iso050.JPG

http://a.img-dpreview.com/reviews/CanonEOS20D/Samples/Compared/Studio/20d_p2_iso0100.JPG

I'd say that both look pretty good, but the 20D shot is definitely better. Compared to the 20D shot the 8400 has a fair bit of noise, which can be seen in the artificial texture that everything seems to have - almost as if sandpaper were rubbed over everything. Perhaps the best way to see is to look at the dark glass bottles. In the 20D shot, the glass looks like smooth glass, while in the 8400 shot it seems like you could feel some texture on it if you rubbed your hand over it.

at ISO 400, yes not contest, the DSLR clearly could be pushed to ISO
1600 or more for a similar result, but shouldn't it be linear across
all ISOs?

I'm not sure what kind of relationship you expect, but the behavior where the small sensor performance seems to deteriorate more rapidly ought to be consistent with your expectations.

As you indicated above with your comments on the 8400, you seem to consider a certain level of SNR to be acceptable or visually noiseless. Suppose the 8400 is right at this level, but the 20D is well above it at the base ISO. The 20D could have a way to go before the noise becomes unacceptable or noticeable to you, but the 8400 might be right on the boundary to start with.

Fuji makes very clean P&S cameras.

Some have argued that Fuji just uses more powerful NR in camera. I'd like to see the raw data and decide for myself.

Canon got lucky and made a very clean CMOS that reduces noise at the
chip level despite the fact that CCDs were supposed to be better.

I'm sure there are some Canon engineers (and many engineers and scientists throughout the world who worked on CMOS sensors before Canon made the D30), who would bristle at your description of their hard work as luck.

--
Ron Parr
Digital Photography FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Haider]

I'll give you an example. One thing before start photosites are three dimensional, that why digital sensors require telecentric lens. If you have rain-fall of 5cm3 per second a 10cm3 bucket will get half full in 1s where as 20cm3 bucket will be 1/4 full in 1s. The level of concentration of water in the 10cm3 bucket is 50% where as in the 20cm3 bucket the concentration is 25%. As you increase the concentration of water (in terms of a sensor, it would light) noise gets less. We need to be clear on this now matter what sensor you have as you increase the concentration of light to it, noise will go down. The more light you feed it the less noise you will get fullstop.

Now I'll throw the spanner into the works it all comes down to how noisey the sensors are to start of with. Smaller sensors are noiser therefore an APS sensor at 50% concentration may record more noise than a 35mm sensor at 25% concentration. It's nothing do with the amount of light the 35mm has captured, we only provided it with 5cm3 of light same as the APS sensor. You tell me without increasing the aperture size and/or increasing the exposure time where is the 35mm sensor going to get the extra light from. Ther is only 5cm3 of light. Like I said a 1/2 pint beer is 1/2 pint of beer whether you put in a pint glass or 1/2 pint glass. 35mm sensors have better noise characteristics that allow them produce less noisier photos. That is the point I'm getting at and have been on this thread. This is down to the fact that you are squeezing same electronics into a smaller (APS or 4/3rds) space which results in less isolation for each of the components in the circuit which results in greater interference or noise.

 

[Haider]

Medium format backs use CCDs vs CMOS/NMOS sensors which give better ISO100 but not as good high ISO. You can't have it all

 

[nickleback]

so why does the medium format keg not have good high ISO? too much foam?

Most medium format sensors don't have microlenses.

--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed

 

[nickleback]

I'll give you an example. One thing before start photosites are three
dimensional

So are garbage pails and pint glasses.

that why digital sensors require telecentric lens.

Near-telecentric. Most 35mm system lenses fit the description.

If
you have rain-fall of 5cm3 per second a 10cm3 bucket will get half
full in 1s where as 20cm3 bucket will be 1/4 full in 1s.

That's only if you consider the buckets to be different heights, but the same area is exposed to the rain. Larger sensors have larger area. Heights are typically similar.

Just do your own experiment. Next time it rains, stick a big can and a small can out in the rain. Height doesn't matter as long as neither one overflows. Imagine the heights are the same if you like.

Now imagine the noise is a couple of extra drops.

Which one has more noise? (answer: neither, absolute noise is the same)

Which one has more signal to noise? (answer: the bigger one, it collected a lot more rain volume)

--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed

 

[Haider]

Give me 8MP and better IQ. I only print at 8R and 8MP is more than enough.