Okay, so we've all heard people whine and moan about camera manufacturers cramming more and more megapixels into bodies. Sure there's been many technological improvements coming along at the same time, yet people continue to complain.
Assuming all other things equal, how much more noise will a higher resolution sensor have when downrezzed versus a sensor with all the same technology but at a lower native resolution?
I guess this relates to shooting sRAW and reduced resolution JPEGs as well.
I'm interested more in effective noise in web-sized images and actual prints, not in pixel-peeping at 100%... you know, stuff that actually matters in the end when photos go on display.
Ariel Bravy wrote:
Okay, so we've all heard people whine and moan about camera manufacturers cramming more and more megapixels into bodies. Sure there's been many technological improvements coming along at the same time, yet people continue to complain.
Assuming all other things equal, how much more noise will a higher resolution sensor have when downrezzed versus a sensor with all the same technology but at a lower native resolution?
I guess this relates to shooting sRAW and reduced resolution JPEGs as well.
I'm interested more in effective noise in web-sized images and actual prints, not in pixel-peeping at 100%... you know, stuff that actually matters in the end when photos go on display. ...Show more →
It depends on the implementation. For instance, the total light collected by the sensor may drop as the pixel count increases; Canon is attemptint to avoid or minimize this with their "gapless" microlenses.
The best answer is probably that it's decreasing all the time, and with other concurrent advances improving the per-pixel noise of even smaller pixel, there's not much reason to put the brakes on resolution increases. Look at the 50D, with lower per-pixel noise than the 40D.
The increased-resolution gripe originated in laments over past-generation point and shoot sensors, where increasing the pixel count WAS drastic for many reasons.
Pixel binning interests me alot. Not just as a way to keep noise down, but also as a way of extending the life of my existing lenses.
I say this because I suspect megapixel yields over the next couple of years are going to go through the roof, and higher resolution sensors may show up the short comings of my existing lenses. If this occurs one option I assume would be for me to pixel bin my 2010 5D Mk3's 40 megapixel sensor down to 20 megapixels, and then all my old lenses would look good again!
Am I being paranoid or is anyone else thinking like this?
Right Jeff, it's not a hard and fast rule anymore that more pixels = more noise due to improvements in technology. You're right.
Now, holding all other things equal, does a higher res sensor pixel binned or shot at reduced resolution have more or less noise than an identical sensor at a lower native resolution?
Ariel Bravy wrote:
Right Jeff, it's not a hard and fast rule anymore that more pixels = more noise due to improvements in technology. You're right.
Now, holding all other things equal, does a higher res sensor pixel binned or shot at reduced resolution have more or less noise than an identical sensor at a lower native resolution?
It depends on the implementation. For instance, the total light collected by the sensor may drop as the pixel count increases; Canon is attempting to avoid or minimize this with their "gapless" microlenses.
So it may have more. It can't have less. If that's what you're driving at, I agree with you. sRAW seems mostly good as a way to save space when shooting at high ISO, when higher resolution would be largely wasted on noise.
jvarszegi wrote:
For instance, the total light collected by the sensor may drop as the pixel count increases; Canon is attempting to avoid or minimize this with their "gapless" microlenses.
I gotcha. On the sensor, I presume that not 100% of the surface area is used for photon collection? As the pixel size decreases, the percentage of the surface area used for purposes other than light collection increases?
jvarszegi wrote:
So it may have more. It can't have less. If that's what you're driving at, I agree with you.
Okay, so let's say that microlenses have been put in place and that 100% of the light coming in is being collected by the sensor.
I take it a higher res sensor downrezzed would then have the same amount of noise as a lower res sensor?
Ariel Bravy wrote:
I gotcha. On the sensor, I presume that not 100% of the surface area is used for photon collection? As the pixel size decreases, the percentage of the surface area used for purposes other than light collection increases?
The 50D is now "gapless". Don't have the unused surface area anymore.
Ariel Bravy wrote:
I gotcha. On the sensor, I presume that not 100% of the surface area is used for photon collection? As the pixel size decreases, the percentage of the surface area used for purposes other than light collection increases?
That seems to be the case so far, but not necessarily the case in the best implementation. The 50D pixels seem to be advertised as "virtually" gapless, so we're not at perfection yet in terms of surface area coverage. In addition the shape of the microlens itself may lead to photon loss at the quantum level, so more total area devoted to microlens edges may result in slightly more photon loss. I don't know much about this area, but I'd guess any actual differences would mean nothing in daily use.
One thing I don't understand is that when the images of the microlens array arrangement were posted, showing each microlens directing its light to a smaller sensel (photon bucket), some were claiming that there was still room for improvement in dynamic range and sensitivity by using gapless sensels. I also don't know whether, if they came up with gapless sensels, they would need a microlens at all.
Okay, so let's say that microlenses have been put in place and that 100% of the light coming in is being collected by the sensor.
I take it a higher res sensor downrezzed would then have the same amount of noise as a lower res sensor?
gfiksel wrote:
Wow, David! This the best photography Zen I've heard! Buddha can rest in peace now
It would be hard to measure, but still present, no? The output would be a result of both the photons counted by the bucket and the residual noise of that one tiny circuit.
Would it help to expose to the right, or would that just cause a hot pixel?
jvarszegi wrote:
The increased-resolution gripe originated in laments over past-generation point and shoot sensors, where increasing the pixel count WAS drastic for many reasons.
I think you're being charitable to the gripers. I suspect the real reason for the exaggeration of the noise cost when increasing sensor density is that photographers, including well-known review sites like Luminous Landscape and DPReview, are comparing 100% shots between cameras with different numbers of pixels. They are inadvertently magnifying the denser sensor's noise by a greater amount, in the ratio of the densities, and one things for sure, the more you blow it up, the worse that noise is going to look.
This myth was brought to us by a bad method of comparison.
As an illustration of this phenomenon, here is the D700 which is really usable at 12800 ISO, compared to the 1Ds3, which as you can see, is useless at ISO 12800: http://cyberphotographer.com/d700v1ds3/12800a.jpg
This is ISO 12800 with the same lens at the same aperture and cameras set to the same shutter speed. To ensure that the two files are equally magnified, the 1Ds3 file has been downrezzed to 12 megapixels.
I did a little test a few days ago, and sRaw did seem to exhibit slightly more noise than downsized full rez. That means that the pixels are really being binned, instead of funnelled, which is a shame. There should really be 4 sRaw modes: 1/2 pixels funnelled, 1/2 pixels binned, 3/4 pixels funnelled, 3/4 pixels binned. That way the user could choose between speedy or higher quality sRaw modes, at both half and a quarter the total number of pixels.
In brief: don't pixel bin for lower noise (yet, probably), because downrezzing in post seems to be better.
DavidP wrote:
Well, it wouldn't be measurable by the human eye, as there would be nothing to compare to.
I guess my point was that TO THE HUMAN EYE, reducing image size DOES improve signal/noise ratio.
Absolutely agree, and of course that's why a camera with double or quadruple the number of pixels of another will yield cleaner looking prints if both are printed at the same size and have otherwise similar technology.
