FF vs APS-C: being serious...

IMO there is room for both high MP and lower MP in both APS-C and FF:

Sports: I would want APS-C for higher frame rates and the 1.5x crop. High MP for bright conditions, lower MP for dim condtions.

Landscape: High MP FF would be my choice (tripod of course).

Studio: High MP FF

Wildlife: Same as sports

Portrait: Same as Studio

--
fjbyrne

Carry your arguments (that smaller sensors have benefits over larger if you throw out the benefits of the larger sensor) and you should be looking at a P&S with a pinky-nail sized sensor but good glass.

Because, really, a 12mpx P&S at ISO 50 isn't really any worse than a D3 .... if shooting in perfect light, only want deep DoF, and you don't PP.

bah.

--
JOE FEDERER
Websites:
Misc personal stuff: http://www.joefederer.com
Minneapolis / St. Paul Wedding photography @ http://www.federerphotography.com

Zoomstein said:

1) The noise's out there and it is inevitable. Film or sensor - the
light's noisy. That's something we can't avoid or complain about.
2) The techniques of noise reduction are based on collecting data
over space. The larger the space and the lower the density - the less
precise your results are.
3) The lesser the pixel density, the less noise You're going to bite on.

1. The noise isn't "out there" it is IN the sensor. All light that hits the sensor is by definition, signal, all else is noise. It is the signal to noise RATIO that is important. It CAN be avoided, or at least, minimized by using a larger pixel. Assuming the same technology, the larger pixel gathers more light and therefore more signal than a smaller pixel and has a higher signal to noise ratio.

Click to expand...

2. I am not talking of noise REDUCTION once it has been recorded as a RAW file or JPEG, but noise avoidance and minimization. Again, bigger pixels will PRODUCE less noise than smaller ones, all things being equal.

3. This, I just didn't even understand!

More pixels for any given sensor will give you GREATER RESOLUTION, but they will also give you GREATER NOISE. The fewer the pixels for any given sensor, the LESSER the RESOLUTION and the LESSER the noise also. Translated into this "debate" it means I can have a FF sensor with the same pixel density as an APSC sensor and have the same noise, field of view from all lenses, but GREATER resolution. Or, I might as a manufacturer choose to have the same NUMBER of pixels on each sensor but have a FF sensor with LESS noise than the APSC sensor.

It appears that my point No 1 disappeared from the previous post, sorry.

1. Noise isn't "out there" it is in the sensor. All light hitting the sensor is by definition, signal, all else is noise. The bigger the pixel, the more signal that it can gather and the greater the signal to noise RATIO will be, all else being equal.

Sure, we can agree on this. FF does have disadvantages, (size, weight, cost) but poorer resolution and noise performance ain't one of them (claiming poetic license here)

Zoomstein said:

Not really. Let's see: noise's a natural thing. It is a part of
light. The bigger the surface of projection, the more noise there is.
The lower the pixel density is, less probable it is to manage noise
while there's actually less noise caught per pixel. Remember that our
pixels are still much-much-much larger than photons

Click to expand...

See also my previous posts.

Noise by definition, is neither a "natural thing" or a "part of light". All light that hits a sensor is SIGNAL, it provides INFORMATION which we are trying to record.

The process of amplification and recording this signal, as well as the very materials used in the technology and that we use the camera at room temperature, introduces unwanted NOISE. This noise tends to be at a fixed level PER PIXEL all things apart from pixel size being equal, and the greater the ratio of SIGNAL or light to this NOISE the less noisy the image will be. In other words, big pixels produce an image with less noise than small pixels.

Zoomstein said:

Come on, 't was a joke I don't consider it a "bummer" at all and
I'm a very happy about A900 coming around. For at least two reasons:

Click to expand...

So the title of your post was...satirical? And I must have missed your jokes. My bad.

Zoomstein said:

1) A900 is going to give Sony a-world the right range of products to
make it a serious player. Since it's "my system" I'm generally happy.

Click to expand...

as you should be.

Zoomstein said:

2) A900 will make more glass thrown at the market which will
certainly make prices go a bit down. And that's nothing to be sad
about too

Click to expand...

I don't know about the price of used glass going down...but new glass, yes, it should become more affordable as time goes on.

Zoomstein said:

I did my school math too The 1.5 ratio does not refer to the actual
surface and we both know that

Click to expand...

Good.

Zoomstein said:

If lower pixel density is such a goody, than why, for heavens sake,
does my 7D have lower noise tolerance and lower IQ than A700 that has
twice as much pixel density?

Click to expand...

How about new manufacturing techniques, new technologies, oh, and the fact that the CMOS in the a700 has ON-CHIP noise reduction, which can be a blessing or the devil, depending on who you ask, right Barry?

Zoomstein said:

As I explained above, "cropping down" is nowhere close "actually
shooting that way". For once because the center pixels are going to
be "polluted" by the side light penetrating the chamber. So "the best
of both worlds" is not really going to pass a reality check.

Click to expand...

Cropping a large imaging sensor down to the pixels within the APS sized region will produce exactly the same results as chopping off those parts of the sensor in the camera manufacturing process and calling the camera APS-C. Of course, cropping an APS-C 12 mp image down to 6mp is not anything like a 6mp APS-C camera...

Zoomstein said:

Not really. Let's see: noise's a natural thing. It is a part of
light. The bigger the surface of projection, the more noise there is.
The lower the pixel density is, less probable it is to manage noise
while there's actually less noise caught per pixel. Remember that our
pixels are still much-much-much larger than photons

Click to expand...

See the other post just above this. Noise that we see in our photos has nothing to do with the light that is entering the front element of the lens. The noise is a characteristic of the image sensor and data processing path. Ok, there is a small amount of radical photons floating around creating "noise" in our "light", but when we get to the point in technology where that noise is what we complain about...then I guess photography equipment discussion will be pretty darn dry. Maybe we'll be forced to talk about photography instead?

Zoomstein said:

I'm not actually saying FF is useless or bad. I'm just saying that
for most of us it is as useful as the 51MP Hasselbad. Or some other
ultraheavy and very expensive gear. That's all.

Click to expand...

I respectfully disagree. The portion of the population who can afford and will enjoy using a FF D-SLR is enormous compared with the few thousand (maybe 10-20k at the very most) US residents who will ever OWN a 51mp Hassy. If you don't think Nikon is selling a TON of full frame DSLRs this year, you're either blind, or...blind?
--
-Matt
Rent Alpha Mount Lenses! - http://www.alphalensrental.com
http://www.ouatphotography.com

the A900 may be the exception to that!

Without getting into speculation about particular brands or models, it is of course possible to get some terrible images from medium format (film, and by extrapolation, digital) cameras. ANY 35mm film SLR will provide an incomparably better image in every respect to the single element, uncoated plastic meniscus lens of the Holga, for instance.

This is why a mainly theoretical discussion such as this contains such a lot of, "generally speaking"s, "all else being equal"s and so on.

For the time being, all of the 35mm FF DSLR cameras, all four of them, can and do produce higher overall IQ than any APS camera. The A900 may or may not do that.

Zoomstein said:

The CONs of FF:

1) It costs more. Not only now, it'll always cost more, because it
costs more to make.

Click to expand...

Yes it costs more to make a full frame. It also costs more to make a 1.5x sensor then, say, a 1.7x sensor. So why are 1.7x sensors not made by Sony, Pentax, Nikon or Canon? Actually, the question should be: would you buy a 1.7x crop camera because it is cheaper than your A700? Most people wouldn't do it. In the future, 1.5x crop cameras may be looked at the same way than 1.7x cameras are today: i.e. nobody will bother to make them.

Zoomstein said:

So actually the pixel density on an FF sensor is lower. Which means
that interpolation (i.e. the process that approximates RGB value for
every pixel from combining several channels measurement) is less
precise and hence is the actual colour, detail and sharpness.

Click to expand...

Actually color accuracy is dependent on a lot of factors, one of which is the strength of the signal received by the sensor and another factor is the dynamic range. Smaller pixels receive a weaker signal, with less dynamic range and more noise. That means the camera has less space to work with to represent all of the colors. Color accuracy is therefore diminished, not enhanced, when you have greater pixel density.

Zoomstein said:

Another thing I actually dislike about FF is that it takes the
tele-range back to the stone age, making it impossible to get good
shots with "portable" glass. And given that a 100-450mm zoom costs
much more than a 70-300, that's a big problem. Wide angle, on the
other hand, does not cost that much even if digital crop is taken
into consideration and is generally less useful. So "the blessing of
digital crop" is actually an important issue.

Click to expand...

APS-C is really involuntary cropping. If you want to crop your image to simulate APS-C it is easy enough to do. But you cannot process the image to simulate a wider angle. There is no such thing as reverse cropping.

Zoomstein said:

Last, but not least - the lame belief that FF is "more
professional"... I think it's some sort of a psychological block. Why
35mm is "professional"?

Click to expand...

It is more "professional" because professionals demand better image quality. A professional camera is defined as a camera that is better than most cameras. Full Frame cameras fit that description.

Zoomstein said:

Why is it "the best format"?

Click to expand...

35mm is not the best format for every purpose. There are other formats that excel. For example, James Bond would not want to carry a 35mm camera to take pictures. He would carry the smallest camera he can find. OTOH, Ansel Adams would not use a 35mm camera either because the prints he made demanded a bigger negative.

Zoomstein said:

Isn't it because
back some 90 years ago it turned out to be (after some trials and
failures) the best format for reasonably portable cameras? So if
APS-C is becoming something of the same sort for DSLRs why is it
wrong?

Click to expand...

APS-C is "wrong" because full frame is better.

Zoomstein said:

If APS-C has given us a chance to take longer shots, have a
larger variety of focal lengths in our backpack with less investment
and granted us SSS and high-ISO noise reduction - why is it all to be
tossed away in favor of a presumably better (only because it's
larger) format?

Click to expand...

Sony is not "tossing away" SSS. The full frame Sony will have SSS.

Zoomstein said:

Robert Lynch said:

If you are going to explain why, please get it right. "Larger sensor
means better IQ" is for the same pixel count. That is why everyone
drools over the output from the FF 12MP cameras on the market. Those
big, fat light collecting pixels produce lower noise, higher IQ
images.

Click to expand...

Lower noise? That's not exactly true actually. Let me try to explain:

1) The noise's out there and it is inevitable. Film or sensor - the
light's noisy. That's something we can't avoid or complain about.

Click to expand...

If you are talking about the light that originates from the subject you are trying to photograph, it makes no sense. Light is composed of all sorts of focal lengths. Do you think light coming from something should be pure? If you are talking about light that reaches the film or sensor, then there is of course noise in the recording medium. This noise does not come from "the light." So, light is not noisy. You do know that different films have different noise levels, I hope. So do different sensors. Different sensors have different noise levels because some sensors have larger individual pixels that can collect more light to give a signal a higher signal to noise ratio. So it is true that noise cannot be avoided, it can be reduced by choosing a larger sensor and keeping the individual pixels larger to minimize noise.

Zoomstein said:

2) The techniques of noise reduction are based on collecting data
over space. The larger the space and the lower the density - the less
precise your results are.

Click to expand...

Noise reduction does not depend on data collection. Noise reduction is dependant on algorithm. It depends on how well the software can distinguish noise from signal. Noise reduction can never be perfect. Hence a part of the signal is also removed. The more noise reduction that is applied, the more signal is removed. Hence it is best to have a sensor that can collect a low noise signal, so that less noise reduction is required. The result is a clean, detailed image.

Zoomstein said:

3) The lesser the pixel density, the less noise You're going to bite on.

Click to expand...

That is a strange expression. I am not sure what you mean. The lesser the pixel density, then the larger the individual pixel, so the signal can be stronger, which reduces the effect noise will have on the image.

Zoomstein said:

So what happens here is that the amount of noise will decrease as
well as our ability to detect and annihilate it.

Click to expand...

If noise is decreased, there is less need to eliminate it.

Zoomstein said:

The question is: are
the two effects proportional? I'm afraid they're not. So basically
the FF sensor camera will have "to think more" to achieve the same
results APS-C camera has now. And the moment algorithms better, APS-C
size will be again better and quicker...

Click to expand...

Here you do make some sense. You are saying that the larger files of a full frame will require longer processing time for the in camera software to identify and eliminate the noise. However, if the full frame has lower pixel density, then noise will be lower.

I'm actually surprised as to how aggressive people get here at times.... Especially if someone questions the solidity of daytime dreaming about FF)))

--

------------------------------------------------------------------------------------
Omne solum forti patria est.

I'm quite surprised as to how aggressive people get here. Got some rotten eggs tossed at me (together with "You should've bought Olympus" and "Get yourself a P&S") )) It's ok, no hard feelings. I have also learned a thing or two I think (especially from those who were more informative and realistic) for which I thank you all.

We'll have to wait and see what A900 looks like and how FF is going to change the Sony world. I personally and emotionally wish for the things to go wel and beyond what some of the posters are dreaming about. Rock-solid SSS, extra-sharp, low-noise FF DSLR that shoots at 15 fps and can take up to four memory cards (2 cf + 2 msd) storing raws on one side and JPGs on the other. And lots of other extremely important and useful features, qualities, etc, etc, etc. Believing in this dream, however, is pretty hard for me. At least at the moment. Not because I'm a dead-end sceptical pessimist, but because of the way things work. I'd like to be proven wrong on this one and I shall be the first to applaud when I'm proven wrong.

Until than (some say we've got a couple of months to wait) - no more FF threads for me

--

------------------------------------------------------------------------------------
Omne solum forti patria est.

There are two types of noise for a ccd/cmos sensor, shot noise and readout noise. Shot noise is device-independent, it is an inherent result of the quantized nature of light itself. In hand waving terms the light arriving at the sensor arrives in packets. There is a distribution in the number of packets, so your device might receive a mean of 100 packets/second with a standard deviation of 10 packets/second; the distribution is constant with increasing intensity so this source of noise is unimportant at high intensities and more important at low intensities. This is shot noise, and cannot be avoided.

Readout noise is sensor dependent and tends to decrease with increased pixel size.

Noise reduction can occur on the chip in a cmos sensor, giving such sensors superior high iso performance.

Look here for more info on shot noise:
http://en.wikipedia.org/wiki/Shot_noise

This is a very enlightening comment. I was actually only talking about shot noise and not readout noise. For once, shot noise can be coped with to a certain degree. Most people here argued that larger pixels mean better shots at high ISO (while that's not exactly true as high ISO does not change the readout noise factor, while increasing greatly shot noise).

I actually doubt that a 24MP FF sensor is going to have lower readout noise than a 12MP APS-C. These are my doubts that are not exactly in line with the common belief that says: "bigger sensor - less noise". We'll actually have to see FF A900 to be able to judge if it copes better (or worse) with r/o noise. Technically both are possibilities.

--

------------------------------------------------------------------------------------
Omne solum forti patria est.

I think you are still not grasping this fully. Increasing iso increases the gain on the sensor, producing higher readout noise.

Greater pixel size increases the area available to detect photons, and therefore the pixel can detect more signal in lower light than a smaller sensor. The pixel pitch of the 24MP sensor is 5.94μm (H) x 5.94μm (V) whereas the A700 has a pixel pitch of 5.4μm. The light gathering area of each pixel in the 24MP sensor is therefore 5.94^2/5.4^2=1.21, 21% greater than that of the A700. As long as the electronics are not noisier (and if anything they should be better) then we should be looking at a modest improvement in noise performance at the per pixel level. Since there are more pixels, the image level noise ought to be even better.

I agree completely though that on paper is one thing, the proof of the pudding is in the eating. And we are all very hungry...

Zoomstein said:

This is a very enlightening comment. I was actually only talking
about shot noise and not readout noise. For once, shot noise can be
coped with to a certain degree. Most people here argued that larger
pixels mean better shots at high ISO (while that's not exactly true
as high ISO does not change the readout noise factor, while
increasing greatly shot noise).

I actually doubt that a 24MP FF sensor is going to have lower readout
noise than a 12MP APS-C. These are my doubts that are not exactly in
line with the common belief that says: "bigger sensor - less noise".
We'll actually have to see FF A900 to be able to judge if it copes
better (or worse) with r/o noise. Technically both are possibilities.

--

------------------------------------------------------------------------------------
Omne solum forti patria est.

Click to expand...

--
Stuart / the Two Truths
http://www.flickr.com/photos/two_truths/
http://two-truths.deviantart.com/gallery/

mike_2008 said:

I think you are still not grasping this fully. Increasing iso
increases the gain on the sensor, producing higher readout noise.

Click to expand...

As far as I remember, increasing ISO changes the threshold for a signal considered "valid" on the sensor level. This results in more shot noise getting thru. (Note that this is an extremely simplistic explanation).

One thing makes me wonder at this point: if readout noise is our "greater enemy", than we're on a dead-end road with CMOS and CCD sensors. I.e. there needs to be a radical change in the way sensors work or else we'll get stuck at a certain MP count that would be sort of a compromise between density, sensor size and readout noise quantities.

--

------------------------------------------------------------------------------------
Omne solum forti patria est.

You're right, that's why the recent press release from Sony about mass-produced back-illuminated cmos sensors is exciting. It could offer 1-2 stops improved SNR. ISO 25600 anyone?

Zoomstein said:

One thing makes me wonder at this point: if readout noise is our
"greater enemy", than we're on a dead-end road with CMOS and CCD
sensors. I.e. there needs to be a radical change in the way sensors
work or else we'll get stuck at a certain MP count that would be sort
of a compromise between density, sensor size and readout noise
quantities.

--

------------------------------------------------------------------------------------
Omne solum forti patria est.

Click to expand...