Pages: 1 [2]   Go Down

Author Topic: What does it take to make you switch to smaller sensor digital camera?  (Read 13577 times)

SZRitter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 384
Re: What does it take to make you switch to smaller sensor digital camera?
« Reply #20 on: April 29, 2016, 10:11:50 am »

I'll throw this in here, although it may be easily debunked and I don't have data to back it up, so feel free to prove me wrong. And if it was already mentioned, I apologise.

But to me, larger negatives/sensors seem to have an easier time creating more pleasing out of focus areas. My guess is, that this is do to the smaller reproduction factor, and thus the actual bokeh effect is diminished. And yes, I know that it is also partly controlled by the shape of the aperture, but it seems like smaller sensors seem to have a harder time getting lenses with smooth bokeh.
Logged

TomFrerichs

  • Full Member
  • ***
  • Offline Offline
  • Posts: 108
Re: What does it take to make you switch to smaller sensor digital camera?
« Reply #21 on: April 29, 2016, 02:05:27 pm »

Aside from the depth of field issues already discussed, the FF (24 * 36 mm, 367 mm^2) has over twice the surface area of APS-C (15.6* 23.5 mm, 366 mm^2) and thus, for a given aperture, the FF collects more than twice the number of photo-electrons than the APS-C. Since signal:noise varies as the square root of the number of photoelectrons collected, the FF will have about 1.4x the S:N as the APS-C. With APS-H as with Canon (18.6*27.9 mm), the difference is somewhat less.

This difference is confirmed by DXO for the latest generation of Nikons (6 dB = 1 f/stop)

Bill

I'm not disputing the DxO curves, nor am I saying that this assertion is in error, but I would like some clarification.

I'm assuming we're talking about shot noise here; the camera contribution is not counted. 

In that case, isn't our "signal" the measurement from one photosite, and not the aggregation from all photosites?  What I'm trying to say is that visible noise appears to me to be localized to pixels.

I understand with a larger sample, the effects of variation within the sample become smaller.  Flip a coin ten times and you may end up with ten heads.  Flip it a hundred times, and if you got 100 heads I'd worry about the coin, giving Rob C his coin back and asking for a different one.  For this reason, a FF sensor would show a lower aggregate S/N ratio.

So, with equal photosites, the chances of getting too many or two few photons at a _single_ photosite are the same, regardless of how many total photosites you have. Therefore, given equal photosites--leaving out the biggest variable: camera performance--a FF and cropped-frame sensor would seem to have the same shot noise performance on the pixel level.

Now, where am I wrong?
Logged

Dave Ellis

  • Full Member
  • ***
  • Offline Offline
  • Posts: 103
Re: What does it take to make you switch to smaller sensor digital camera?
« Reply #22 on: April 30, 2016, 02:16:04 am »

I'm not disputing the DxO curves, nor am I saying that this assertion is in error, but I would like some clarification.

I'm assuming we're talking about shot noise here; the camera contribution is not counted. 

In that case, isn't our "signal" the measurement from one photosite, and not the aggregation from all photosites?  What I'm trying to say is that visible noise appears to me to be localized to pixels.

I understand with a larger sample, the effects of variation within the sample become smaller.  Flip a coin ten times and you may end up with ten heads.  Flip it a hundred times, and if you got 100 heads I'd worry about the coin, giving Rob C his coin back and asking for a different one.  For this reason, a FF sensor would show a lower aggregate S/N ratio.

So, with equal photosites, the chances of getting too many or two few photons at a _single_ photosite are the same, regardless of how many total photosites you have. Therefore, given equal photosites--leaving out the biggest variable: camera performance--a FF and cropped-frame sensor would seem to have the same shot noise performance on the pixel level.

Now, where am I wrong?

Hi Tom

I can see where you are coming from. I think Bill's analysis probaby assumes the same number of pixels on each sensor. Pixel size rather than sensor size is surely what matters with shot noise S/N.

Dave
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600

I think Bill's analysis probaby assumes the same number of pixels on each sensor. Pixel size rather than sensor size is surely what matters with shot noise S/N.

Dave
Yes and no.  Yes, pixel size is what counts when you measure per-pixel DR (when comparing sensors of equal technological level). But when it comes to looking at the whole image, say in prints of the same size, and judging tonal range, shadow handling, fineness of tonal gradations and so on, a sensor having more pixels of the same size will have an advantage, due to the "dithering" effect of the lower degree of enlargement needed: printing at higher PPI.  This is akin to the advantages in tonal gradations and such that are often seen with a larger film format compared to using the same emulsion in a smaller format.

In digital, one brute-force way to achieve "tonal" advantages from a fully exposed(*) larger sensor of equal pixel size is to downsize to the same pixel count as the smaller sensor being compared to, since that improves the "per pixel" DR and SNR, giving roughly the same results as with a sensor of the larger size but with the same pixel count as the smaller sensor.  (Why?: adding pixel values together increases the signal level in larger proportion than it raises the noise level, so improves any measure of per-pixel SNR.)

And then throw in the fact that that larger format sensors most often have both more and larger pixels.


(*) I say "fully exposed" because these advantage will be lost at the other extreme where both formats are seeking the same DOF and the same limitation on motion blur and so using the same exposure duration but with a high aperture ratio in the large format, and thus with the larger format using an "ISO" exposure index that is higher in proportion to sensor area, which limits both formats to the same photon counts, and the same photon shot noise levels.
Logged

TomFrerichs

  • Full Member
  • ***
  • Offline Offline
  • Posts: 108

Thank you all for your answers.

I have read many times “because a full frame sensor receives more light than a cropped-frame sensor, it will have better noise performance.” I’m not saying these are reliable sources, you understand. This assertion didn’t make sense to me. I was hoping to learn how that statement could be true.

The responses above convince me that while a full-frame camera may…probably does…have better shot noise performance than a cropped-frame camera the difference is due to other factors than simply “receiving more light.” The bare statement is, if not incorrect, at least a major over-simplification that can be easily misinterpreted.
Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com

I have read many times “because a full frame sensor receives more light than a cropped-frame sensor, it will have better noise performance.” I’m not saying these are reliable sources, you understand. This assertion didn’t make sense to me. I was hoping to learn how that statement could be true.
This would be true as long as the overall sensor resolution is the same for both cameras. f the larger sensor as the same resolution, each sensel is then physically larger so more light is going to strike it.

Of course there's a lot more to noise than just that, but if the two sensors are built with the same technology then the larger pixels should offer better noise.  This may be why the top of the line Canon and Nikon are full frame cameras but lower resolution ... this allows higher ISO's, suitable for the market these cameras are designed for (sports, journalism, some types of wildlife).
Logged

TomFrerichs

  • Full Member
  • ***
  • Offline Offline
  • Posts: 108

This would be true as long as the overall sensor resolution is the same for both cameras. f the larger sensor as the same resolution, each sensel is then physically larger so more light is going to strike it.

Of course there's a lot more to noise than just that, but if the two sensors are built with the same technology then the larger pixels should offer better noise.  This may be why the top of the line Canon and Nikon are full frame cameras but lower resolution ... this allows higher ISO's, suitable for the market these cameras are designed for (sports, journalism, some types of wildlife).

My point is made again.  This qualification (same resolution->larger pixels) is an addition to the "larger sensor has less noise because it gets more light" claim. The "larger pixel" idea is one of those other factors not covered by the claim.

Thank you.
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387

This would be true as long as the overall sensor resolution is the same for both cameras. f the larger sensor as the same resolution, each sensel is then physically larger so more light is going to strike it.

Of course there's a lot more to noise than just that, but if the two sensors are built with the same technology then the larger pixels should offer better noise.  This may be why the top of the line Canon and Nikon are full frame cameras but lower resolution ... this allows higher ISO's, suitable for the market these cameras are designed for (sports, journalism, some types of wildlife).

The larger pixels will have improved signal:noise ratio when one is evaluating them on a per pixel basis. However, this is the wrong way to evaluate performance. When the images are printed at equal linear size, the small pixel camera image will be downsized and acquire better SNR due to pixel binning. See Emil Martinec Big pixels vs small pixels for a rigorous explanation. These considerations are the basis of screen vs print DR on Dxomark. See the DXO documentation. Also see their explanation on how how more pixels offset noise.

One reason that top of the line full frame cameras such as the Nikon D5 do not have the highest resolution is that when shooting sports or journalism under difficult conditions is that D810 resolution can not be achieved under such conditions and the overhead of large file sizes is only a disadvantage. D5 resolution is sufficient for Sports illustrated and news paper usage. For these applications a high frame rate and good low light performance trump resolution.

Bill
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600

This would be true as long as the overall sensor resolution is the same for both cameras. If the larger sensor as the same resolution, each sensel is then physically larger so more light is going to strike it.
Yes, but with the qualification that this also requires using the larger sensor with a larger aperture size; more precisely with a lens of larger effective aperture diameter, and thus with stronger out-of-focus effects – which can be good or bad depending on what is wanted from the image.  The only ways to get more light to each of an equal number of of pixels, or more generally to deliver more light in total from the scene to the sensor are to (a) use a larger effective aperture diameter (= focal length divided by aperture ratio), or (b) a longer exposure time.  And option (a) increases out-of-focus blurring, as has been debated a few thousand times in this forum over the years.

So many comparisons between format sizes about light gathering and thus photon shot noise are potentially misleading because of the assumption that cameras in different formats will be used at the same aperture ratio [f-stop] and with lenses giving the same field of view, which means that the larger sensor is used with both a greater focal length and a larger [effective] aperture diameter.  Such comparisons can be reasonable for some purposes, but not for example when one is choosing the aperture to get sufficient depth of field, which leads to the use of a higher f-stop in the larger format. In fact it leads to approximately the same effective aperture diameter, so that the lens is gathering light from the scene at the same rate.  (I say "approximately" because things deviate a bit from this pattern at close range, as with macro photography.)


By the way: as a landscape, architecture and nature macro enthusiast, I almost always use lower f-stops with my Four Thirds format gear than I would if still using my 35mm film cameras.  Do any other multi-format photographers in this forum care to comment on how their f-stop choices vary [or do not vary!] depending on the format in use?


P. S. I am ignoring differences between lenses in transmission efficiency (as in T-stops vs f-stops) because this is not particularly related to format size, as far as I know.
« Last Edit: May 05, 2016, 11:24:13 pm by BJL »
Logged

happypuppy

  • Newbie
  • *
  • Offline Offline
  • Posts: 7
What does it take to make you switch to smaller sensor digital camera?
« Reply #29 on: May 04, 2016, 06:28:29 pm »

It's sort of the same argument that has been going on forever. Serious work was with 8x10. Then we had the the era of the medium formats etc.

What won was the 35mm due to size and convince it's really no different today.

I don't like to have to carry massive bodies and lens. I like to hike and ride my bike so the need is critical

Right now it's the Sony NEX 6000 a zoom and a legacy zoom plus prime.

The bicycle camera is a Nikon V1 due to size and a couple lens. 

I find I shoot a LOT more with these two than I ever did with my old D90 and its lens

Sure there are tradeoffs. But the they do the job and I get shots I never would have been able to haul my old cameras to shoot. 


Sent from my iPhone using Tapatalk
Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com

When the images are printed at equal linear size, the small pixel camera image will be downsized and acquire better SNR due to pixel binning.
Guess I’m a little confused, if the two sensors in question have the same resolution, why would the small pixel camera benefit from “down sizing” any more than the other, and in fact to down size wouldn’t that mean only when printing smaller prints (depending on resolution).

The question I was trying to answer was simply why the larger pixel sensor gets “more light”.  I mentioned there is a lot more to it than that when considering all factors, but it seems pretty straightforward, it’s physically bigger with a larger surface area, so all other things being equal more light is striking it’s surface. As to whether that’s actually  beneficial and how beneficial depends on many factors, and as technology improves those differences have diminished quite a bit.
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387

Guess I’m a little confused, if the two sensors in question have the same resolution, why would the small pixel camera benefit from “down sizing” any more than the other, and in fact to down size wouldn’t that mean only when printing smaller prints (depending on resolution).

The question I was trying to answer was simply why the larger pixel sensor gets “more light”.  I mentioned there is a lot more to it than that when considering all factors, but it seems pretty straightforward, it’s physically bigger with a larger surface area, so all other things being equal more light is striking it’s surface. As to whether that’s actually  beneficial and how beneficial depends on many factors, and as technology improves those differences have diminished quite a bit.

If the resolutions of the sensor are equal, there would be no need for downsizing as you correctly state. If you have a crop frame camera and a full frame camera with the same resolution, the full frame pixels will be larger and collect more light. Except for the deep shadows, the signal to noise is largely determined by shot noise and varies as the square root of the number of photo-electrons collected and will be better with the larger pixels, other things being equal. I was thinking more along the comparison of full frame sensors having different resolution, such as is the case with the Nikon D5 and D810 and the new Canon 50 MP sensor and the older 24 MP or similar sensors. In these cases, the smaller pixel sensor will have more resolution but not necessarily worse DR when the image is downsized. This relationship may not apply for very high ISOs. Sorry for the confusion.

Bill
Logged

Hywel

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 294
    • http://www.restrainedelegance.com

I think digital photo tech is approaching maturity.

Good-enough-for-most-purposes images can be captured by everything from iPhones to Hasselblads and Phase MF.

So increasingly your choice of camera systems is going to be driven by your own artistic vision, and whatever requirements your clients, customers or final purposes for the images impose on you.

You can probably shoot billboard-sized images with an iPhone - someone has probably done so. (But I bet they had studio, lighting, art directors and so forth backing them up, like the film Tangerine, shot on iPhones which recently triumphed at Sundance. They put all their money into other areas of the production). 

You can shoot website photos with a Hasselblad. I know, because that's my day job. Could I get away with shooting everything on my iPhone instead? Yes, but the results wouldn't be pleasing to me, and not to my more discerning customers either.

If you shoot billboards REGULARLY, you'll doubtless find it a lot easier with a 5DIII, Phase or Hasseblad than with an iPhone or bottom end m43 compact.

If you shoot sports regularly, you'll find top end dSLR's from CaNikon offer performance above and beyond the capabilities of other systems, for the moment.

So it is all about how easy it is to get the results YOU want in YOUR workflow with each size or style of camera.


I have a Panasonic GH4 system, and while I love its flexibility, light weight, ergonomics and battery life, it cannot be denied that the detail and tonal richness present in the shots I take with it in the mountains is sub-par compared with my Sony A7Rii. So while I have taken lots of pics I love with the Panasonic, I no longer take it into the mountains. I just like the Sony's renditions better. Another photographer might find the absolute opposite, or find the extra weight of the Sony system or its crappy ergonomics or poor battery life a compromise to far.

For me the compromise too far in the mountains is the Hasselblad- too heavy, too fragile-feeling. Of course, there are many excellent photographers who use their Hasselblads for landscape and mountain work- and many who use m43 or APS-C systems, too. It's personal choice to improve the number of "keepers" you get each trip.

In the studio with flash lighting the tonal richness and detail of the A7Rii is a bit behind the Hasselblad to my eyes. The detail gets close but loses out in terms of artistic preference, and the colours are definitely easier to get the way I want them on the Hasselblad. Conversely, the Sony in available light with 85 mm f/1.4 GM lens absolutely kicks the Hasselblad's ass in terms of sharpness, camera shake reduction, and hence detail. And the bokeh is luminous.

For another photographer with different subjects and different workflow and different artistic preferences, they might prefer the A7Rii in the studio. Some will prefer Fuji's rendition, some Panasonic's. Some are so attached to Canon skin-tone rendition that they'll absolutely accept the marginally smaller dynamic range and plump for the 5Ds or 5DIII. Or they might have a need for lenses which only the big boys have in their ranges.

Some people may even like Nikon's skin tones ;) (I'm sure they're great in skilled hands, but I dread people turning up with Nikons for people photography tutorials- to me they seem to render everything glowing plastic orange and purple.)

So it is all about which compromises matter to you, and finding the system or systems which best suit your tastes, your personal vision and your workflow. It is possible to take extremely good photos on pretty much every modern imaging system, so pick the one which works best for you.

These days there will be factors which will matter MUCH more to you and your workflow than just the size of the sensor. Hire kit to figure out what they are, and buy accordingly.


Cheers, Hywel


 
« Last Edit: May 08, 2016, 06:43:07 am by Hywel »
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13983
    • http://www.flickr.com/photos/bernardlanguillier/sets/

Many interesting comments here.

It wasn't that long ago that the 11mp Canon 1Ds was assessed as being equal to 645 film, and now we look at the latest 20mp sports cameras from Canikon as being important compromises in terms of image quality in favor of speed.

Cheers,
Bernard

EinstStein

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 501

We all agree that with the same number of pixels, the larger sensor (hence target pixel) has the better signal to noise ratio. The question is if both sensors have the same pixel size, does larger sense (hence more pixel) still have better signal to noise ratio.
It is obvious the signal to noise ratio per pixel would be the same, but, does more pixels somehow can improve the overall signal to ratio with cross pixel correlation. taking average, for example.

I mean, suppose the larger sensor is 4X larger than a smaller sensor. Assume with the same pixel size, the larger sensor has 4X pixels. In the simplest form, suppose we take the average of every 4 pixels off the larger sensor, then compare the signal to noise ration,would the average of every 4 pixels beats the single pixel in the smaller sensor?

I am not sure my following argument is right, but this is what I think:
-- The overall noise can be categorized into several types, such as white noise, thermal noise, etc. Some of them have cross pixel correlation, some don't.
-- Thus the average of 4 pixels may average out the random noise while enforced the noise with localization correlation.
-- This implies the random noise would looks better with average effect, but noise with localization correlation will be less affected.
-- However, this means the noise with localization correction would also get the benefit in another subtle reason.
   This is because the "noise pattern" of the "clustering/clouding/grainy" effect, i.e., the size of the cluster/cloud/gain will be scaled down if the final display size is the same.

Based on the above argument, now I tend to believe the larger sensor will always has better signal noise ratio even if the size of each pixel is the same.

Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto

Hi,

645 film was perhaps not the ultimate summit of image making.

Personally, I felt that 6x7 Velvia was a good match for my 24 MP Sony A900 in things like sharpness. I had shot the same subject with Pentax 67 on Velvia in 2003 and on Sony SLT A99 in 2013. 30"x40" prints from these images were very similar.

An interesting note, LuLa did a comparison between 4"x5" film and the P45 back in 2006. But, for some reason they scanned that film at 2000 PPI. I guess that 2000 PPI was good enough. It is possible to scan 4"x5" at higher PPI, do more post processing and have even better results:
https://www.onlandscape.co.uk/2011/12/big-camera-comparison/

Personally, I have a P45+ that I use with the Hasselblad 555/ELD. It is a nice camera that I like using a lot. But, now I have a Sony A7rII. I feel it can do all the P45+ can and it has some features of it's own. So the Sony A7rII is now my travel camera and the P45+ has been relegated to nice to have stuff.

Best regards
Erik


Many interesting comments here.

It wasn't that long ago that the 11mp Canon 1Ds was assessed as being equal to 645 film, and now we look at the latest 20mp sports cameras from Canikon as being important compromises in terms of image quality in favor of speed.

Cheers,
Bernard
Logged
Erik Kaffehr
 

EinstStein

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 501

Typo, I meant the opposite.
I assume the cluster/cloud/grain size is physically the same, but with the larger sensor, the cloud size is subject to smaller scale up.
So strictly sensor, the larger sensor should be better even if it has the sea pixel size as the smaller sensor. -- I guess.

But it does not imply it is better in the system level -- after adding the lens considerations.
To achieve the same field angle, with the same diameter of the lens (hence the same size for portability), the smaller sensor would get brighter lens, so it can compensate the weakness of the smaller sensor.
Also, with he same lens diameter, assume they have the same resolution angle (angle of confusion), the smaller sensor will have smaller focal length, hence smaller resolution distance (circle of confusion), so would appear better resolution. 




Guess I’m a little confused, if the two sensors in question have the same resolution, why would the small pixel camera benefit from “down sizing” any more than the other, and in fact to down size wouldn’t that mean only when printing smaller prints (depending on resolution).

The question I was trying to answer was simply why the larger pixel sensor gets “more light”.  I mentioned there is a lot more to it than that when considering all factors, but it seems pretty straightforward, it’s physically bigger with a larger surface area, so all other things being equal more light is striking it’s surface. As to whether that’s actually  beneficial and how beneficial depends on many factors, and as technology improves those differences have diminished quite a bit.
Logged

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267

I'm quite pleased with what the highest resolution 135 FF cameras can do today in terms of technical image quality, especially if I design my own color profiles. It's certainly adequate for the type of photos I make.

Still I shoot with a Linhof Techno, Schneider Digitar lenses and a H4D-50 back, which by today's standards is a dinosaur system. In the right conditions it makes great image quality, and I use the system in that range. The lenses are also indeed quite sharp for those 50 megapixels.

What I'm after is the large format creative process though, without having to mess with film. Distortion-free lenses, plenty movements, tilt/swing, rise/fall & shift for all lenses, long or short. This camera gives me a unique feature set in that aspect.

If it was only about technical image quality I'd probably shoot a Canon 5Ds by now.

Unfortunately the tech cam genre is about to kill itself, newer sensors reduce movement capability, newer lenses add distortion so I don't really have anything to upgrade to. I take it year by year. Assuming the digital back electronics doesn't die from old age my camera system could work a lifetime, so what would stop me using it would be temptations from other formats. I don't know what the future holds, but so far temptations have been weak for my type of photography.
Logged
Pages: 1 [2]   Go Up