takowasa wrote:
As I've been saying, the greater number of smaller pixels results in more electronic noise, which becomes significant at high ISOs,
So for those who advocate that smaller pixels = more noise (ceteris paribus) ... there are scenarios when this holds true as a byproduct of the multiple sources of noise contributing to the sum total. Is that correct?
takowasa wrote:
As I've been saying, the greater number of smaller pixels results in more electronic noise, which becomes significant at high ISOs, and that's a big part of it. However, as I recall, there was some sort of issue with the Sony A7 and A7R sensors ("purple haze"?) that was also an issue, and unique, so far as I'm aware, to those two sensors. I also seem to remember that the issue was fixed, or there was some sort of remedy, but I don't really know
It's not really unique to the Sony sensors, it's just that it presented itself at lower ISO and a bit more severely for the first gen Sony FF sensors. You can fix it relatively effectively with dark frame subtraction but that's not always practical. The newer Sony variants (A7S, D810, D750) seem to have improved the issue significantly (except for the A7II.)
takowasa wrote:
As I've been saying, the greater number of smaller pixels results in more electronic noise, which becomes significant at high ISOs,
RustyBug wrote:
So for those who advocate that smaller pixels = more noise (ceteris paribus) ... there are scenarios when this holds true as a byproduct of the multiple sources of noise contributing to the sum total. Is that correct?
Yes. As a general rule, more smaller pixels have more total electronic noise than fewer larger pixels. Since the electronic noise is significant in the portions of the photo made with very little light, this becomes an issue when pushing deep shadows at base ISO or for high ISO photography.
Conversely, depending on balance between the additional detail that a greater number of smaller pixels resolve and the noise differential, the application of noise filtering can offset, and even reverse, the noise advantage of fewer larger pixels even in these extreme situations.
For example, it would not be unreasonable to expect that the 5Ds photos were intrinsically more noisy than 5D3 photos at ISO 1600 and above, but by applying noise filtering to the 5Ds photos (and 5D3 photos, if we choose), we may find that the 5Ds photos come out cleaner, while still resolving more detail, all the way to ISO 6400 or even higher.
gdanmitchell wrote:
If all else is equal, but photo site "size" is smaller, the image is more susceptible to noise at the photosite leve (S/N ratio is smaller)l, though the noise may be less visible to viewers of a print at a given size since its "grain size" (to use a film way of thinking) is smaller.
The "grain size" is indeed smaller, but the amplitude is higher -- if the viewer can is standing close enough with sufficient visual acuity to see the actual "grain". I believe the examples in these photos that I've linked a few times in this thread tell the story rather nicely:
As I said before, I would not call the greater number of smaller pixels in the middle example "more noisy"; rather, I'd call the fewer number of larger pixels "more blurry". However, if the light gets low enough, the situation can most certainly reverse.
RustyBug wrote:
All this talk about noise ... got me to look at some comps on multiple sites. My intent was to evaluate a possible "high ISO" (Canon) mate to the forthcoming 5Ds or 5Ds.
The Canon 1DX seemed to have some of the better noise than almost all other cameras I compared it to ... both Canon & other brands. The only cameras that seemed a touch better (to my eye) was the D4s and the A7s ... obliterating the likes of the A7R.
Coming from the former conversation that suggests that pixel size did NOT have any influence on the noise, rather based on the aggregate light and QE ... what rationale exists for such a significant difference between the 1DX vs. the A7R @ high ISO?
Given the 1DX being older than the A7R and the amount of dialogue about Canon being behind Sony and dismissal of build components, etc. ... ... a bit of a surprise here if pixel size is not somehow involved.
I understand if folks have tired of this ... but there seems to be more to the picture. Hopefully the 5DS will be an improvement over the current 5DIII ... given the 1DX @ 2011 technology, that could bode hope beyond what we may hear otherwise. The intriguing comp though is the 1DX vs. the 5D III @ why the 5D III ...Show more →
As indicated before (here), the midtones between high MP and lower MP sensors of the same technology will have nearly identical noise since the midtones are dominated by shot noise, which is a function of QE. Where the difference occurs is in the shadows at high ISO, where the higher MP sensor's pixel count results in a higher aggregate read noise, which translates into lower High ISO DR, which is a proxy for shadow noise/read noise. To see this properly you need to evaluate both the midtones and the shadows at various ISOs between the sensors; the dpreview samples you posted are all shadows, where the pixel count difference correlation to High ISO performance manifests.
Here is a High ISO comparison I did 8 months ago that included both the A7s and A7r, along with the Nikon Df and A7. I purposely used a scene containing both midtones and shadows, to demonstrate the how the elements of low vs high MP count for an equal size (and near-equal technology) sensor manifests in terms of noise.
RustyBug wrote:
All this talk about noise ... got me to look at some comps on multiple sites. My intent was to evaluate a possible "high ISO" (Canon) mate to the forthcoming 5Ds or 5Ds.
The Canon 1DX seemed to have some of the better noise than almost all other cameras I compared it to ... both Canon & other brands. The only cameras that seemed a touch better (to my eye) was the D4s and the A7s ... obliterating the likes of the A7R.
Coming from the former conversation that suggests that pixel size did NOT have any influence on the noise, rather based on the aggregate light and QE ... what rationale exists for such a significant difference between the 1DX vs. the A7R @ high ISO?
Given the 1DX being older than the A7R and the amount of dialogue about Canon being behind Sony and dismissal of build components, etc. ... ... a bit of a surprise here if pixel size is not somehow involved.
I understand if folks have tired of this ... but there seems to be more to the picture. Hopefully the 5DS will be an improvement over the current 5DIII ... given the 1DX @ 2011 technology, that could bode hope beyond what we may hear otherwise. The intriguing comp though is the 1DX vs. the 5D III @ why the 5D III ...Show more →
snapsy wrote:
As indicated before, the midtones between high MP and lower MP sensors of the same technology will have nearly identical noise since the midtones are dominated by shot noise, which is a function of QE. Where the difference occurs is in the shadows at high ISO, where the higher MP sensor's pixel count results in a higher aggregate read noise, which translates into lower High ISO DR, which is a proxy for shadow noise/read noise. To see this properly you need to evaluate both the midtones and the shadows at various ISOs between the sensors; the dpreview samples you posted are all shadows, where the pixel count difference correlation to High ISO performance manifests....Show more →
Spot-on. Then we have to consider what happens when we apply noise filtering, the effectiveness of which, of course, will vary greatly depending on how much additional detail was captured. That is, for a low light hand held shot with a bit of motion blur and/or camera shake, the sensor with fewer larger pixels may come out ahead. But for a nice detailed capture, the photo made with more smaller pixels can certainly be the clear winner.
Here is a High ISO comparison I did 8 months ago that included both the A7s and A7r, along with the Nikon Df and A7. I purposely used a scene containing both midtones and shadows, to demonstrate the how the elements of low vs high MP count for an equal size (and near-equal technology) sensor manifests in terms of noise.
and the A7s has a new Aptina sensor that has a rather interesting new design. If you compared the Df, D610, and D810, I think that would illustrate the principles a bit better.
Btw Rusty, the dpreview samples you posted don't appear to have had the 'print' tab selected, so you were also comparing them at their native pixel sizes rather than normalized to sensor area. To see how much of a difference that makes, here are 100% crops of a D3s vs D800 High ISO comparison I did in 2012 at ISO 12,800, first at their native pixel sizes and then normalized to 12MP. Notice again how the midtones are almost identical and the shadow favors the D3s.
takowasa wrote:
The A7 and A7R sensors fare poorly compared to their Nikon counterparts, for some reason:
There are a few theories for this, one being that the Sony implementations run the sensor hotter due to the full-time LV of their MILC design, not to mention in a smaller body with less heat dissipation. I've never experimented to see if that's the case. Other theories might relate to the power planes/rails differences of the PCBs built around the sensors though that seems less likely.
takowasa wrote:
The A7 and A7R sensors fare poorly compared to their Nikon counterparts, for some reason.
snapsy wrote:
There are a few theories for this, one being that the Sony implementations run the sensor hotter due to the full-time LV of their MILC design, not to mention in a smaller body with less heat dissipation.
That's the explanation I've heard.
I've never experimented to see if that's the case. Other theories might relate to the power planes/rails differences of the PCBs built around the sensors though that seems less likely.
I find it odd that it is not talked about as much as Canon's base ISO DR deficiency -- they seem like flip sides of the same coin, each equally problematic, just Canon's on the low ISO end and Sony's on the high ISO end.
takowasa wrote: takowasa wroteThe A7 and A7R sensors fare poorly compared to their Nikon counterparts, for some reason.snapsy wroteThere are a few theories for this, one being that the Sony implementations run the sensor hotter due to the full-time LV of their MILC design, not to mention in a smaller body with less heat dissipation.That's the explanation I've heard.
Earlier quoteI've never experimented to see if that's the case. Other theories might relate to the power planes/rails differences of the PCBs built around the sensors though that seems less likely.I find it odd that it is not talked about as much as Canon's base ISO DR deficiency -- they seem like flip sides of the same coin, each equally problematic, just Canon's on the low ISO end and Sony's on the high ISO end.
Well the differences between the A7r and D800/D810 are near zero per DxO, within the margin of their measurement error. The A7 trails further behind its Nikon-implemented D600/D610 equivalents though - part of that may be due to the on-chip PDAF sensors, or perhaps a different analog gain circuit.
snapsy wrote:
Well the differences between the A7r and D800/D810 are near zero per DxO, within the margin of their measurement error. The A7 trails further behind its Nikon-implemented D600/D610 equivalents though - part of that may be due to the on-chip PDAF sensors, or perhaps a different analog gain circuit.
I dunno. Going by the crops in the link I gave earlier:
those differences don't seem "near zero" to me, and are even clearly visible all the way down to ISO 1600. Perhaps I'm overly critical? Or is it all chroma noise and DxOMark is only measuring luminance noise?
takowasa wrote: snapsy wroteWell the differences between the A7r and D800/D810 are near zero per DxO, within the margin of their measurement error. The A7 trails further behind its Nikon-implemented D600/D610 equivalents though - part of that may be due to the on-chip PDAF sensors, or perhaps a different analog gain circuit.I dunno. Going by the crops in the link I gave earlier:
those differences don't seem "near zero" to me, and are even clearly visible all the way down to ISO 1600. Perhaps I'm overly critical? Or is it all chroma noise and DxOMark is only measuring luminance noise?
dpreview doesn't match absolute LV and exposure for their studio scene setup. They've started to do that with newer cameras and have redone some previous cameras as well but haven't disclosed which cameras exactly, and to my knowledge they're still not using the same lenses across bodies so there are t-stop differences as well. As such their comps can't be used for definitive noise comparisons, esp since the sensor differences are so small to begin with.
takowasa wrote: snapsy wroteWell the differences between the A7r and D800/D810 are near zero per DxO, within the margin of their measurement error. The A7 trails further behind its Nikon-implemented D600/D610 equivalents though - part of that may be due to the on-chip PDAF sensors, or perhaps a different analog gain circuit.takowasa wroteI dunno. Going by the crops in the link I gave earlier:
those differences don't seem "near zero" to me, and are even clearly visible all the way down to ISO 1600. Perhaps I'm overly critical? Or is it all chroma noise and DxOMark is only measuring luminance noise?snapsy wrotedpreview doesn't match absolute LV and exposure for their studio scene setup. They've started to do that with newer cameras and have redone some previous cameras as well but haven't disclosed which cameras exactly, and to my knowledge they're still not using the same lenses across bodies so there are t-stop differences as well. As such their comps can't be used for definitive noise comparisons, esp since the sensor differences are so small to begin with.I hear what you're saying, but those differences are more than 1/3 EV illumination errors, so DPR's studio errors would have had to have been huge.
Sensorgen's figures are based on DxOMark data, and at ISO 6400, the D810 and D750 are all but identical. However, the D750 seems to be half a stop cleaner in the shadows with DPR's test:
Weren't both those cameras tested after DPR started taking care to make sure the exposures were identical?
Before matching absolute lighting dpreview would use exposure wedges to match exposure so they can definitely be off by more than 1/3EV. And the D750 is a newer sensor design or at least a new integration of the sensor, with significantly reduced Exmor-type amp glow, a qualitative element of noise not captured in DxO's tests.
snapsy wrote:
Before matching absolute lighting dpreview would use exposure wedges to match exposure so they can definitely be off by more than 1/3EV. And the D750 is a newer sensor design or at least a new integration of the sensor, with significantly reduced Exmor-type amp glow, a qualitative element of noise not captured in DxO's tests.
I deleted my earlier post because I made a stupid error, which was even more stupid given that it was the subject of conversation. The D750 and D810 have the same electronic noise per pixel, but the D810 has more pixels, so more electronic noise, which is why it comes out more noisy in the DPR studio test.
Seriously, I could not have been more stupid. It was as if we were discussing gravity and talking about having to take wind resistance into account for projectile motion, and then I wondered why the hammer hit the ground before the feather.
takowasa wrote:
I deleted my earlier post because I made a stupid error, which was even more stupid given that it was the subject of conversation. The D750 and D810 have the same electronic noise per pixel, but the D810 has more pixels, so more electronic noise, which is why it comes out more noisy in the DPR studio test.
Seriously, I could not have been more stupid. It was as if we were discussing gravity and talking about having to take wind resistance into account for projectile motion, and then I wondered why the hammer hit the ground before the feather.
Glad to know it happens to even the best of us on any given day.
I've always maintained that it isn't solely our errors that define us (to err is human), but rather our ability to correct for and recover from them ... hopefully in a dignified and even humorous manner.
The D750 and D810 have the same electronic noise per pixel, but the D810 has more pixels, so more electronic noise, which is why it comes out more noisy in the DPR studio test.
I think this is the aspect that for many folks becomes translated from the multiple variable issue that it is to a reduction into a single variable one @ smaller pixels = more noise. The defensibility of it holds true ... but @ certain parameters, not necessarily universal for all parameters as a byproduct of the asymmetrical, multi-variable relationship of the other additional contributions. Thus while correlations are drawn, they are not fully universal 1:1 correlations.
As a general rule, more smaller pixels have more total electronic noise than fewer larger pixels. Since the electronic noise is significant in the portions of the photo made with very little light, this becomes an issue when pushing deep shadows at base ISO or for high ISO photography.
Likewise, the application of primary interest is reduced to the area where it is most exaggerated ... often used as the criteria for judgment ... @ times not giving credence to the other areas / applications where it has less differentiating impact.
snapsy wrote:
As indicated before (here), the midtones between high MP and lower MP sensors of the same technology will have nearly identical noise since the midtones are dominated by shot noise, which is a function of QE. Where the difference occurs is in the shadows at high ISO, where the higher MP sensor's pixel count results in a higher aggregate read noise, which translates into lower High ISO DR, which is a proxy for shadow noise/read noise.
Driving the (lay observation) correlation of smaller pixels = more noise.
snapsy wrote:
Btw Rusty, the dpreview samples you posted don't appear to have had the 'print' tab selected, so you were also comparing them at their native pixel sizes rather than normalized to sensor area. To see how much of a difference that makes, here are 100% crops of a D3s vs D800 High ISO comparison I did in 2012 at ISO 12,800, first at their native pixel sizes and then normalized to 12MP. Notice again how the midtones are almost identical and the shadow favors the D3s.
+1 @ the normalization.
The story repeats @ detectability to the low end. I also notice a color variation between the D3s and the D800 ... the D800 showing a less neutral (i.e. more blue) rendering in the camera body / lens.
snapsy wrote:
not to mention in a smaller body with less heat dissipation. I've never experimented to see if that's the case. Other theories might relate to the power planes/rails differences of the PCBs built around the sensors though that seems less likely.
This is akin to what I was referring, when I raised the "build" issue @ pro vs. consumer build comparisons.
takowasa wrote:
Canon sensors fare well at high ISOs (same quantum efficiency as everyone else, and the same, or even a tad lower, electronic noise per pixel). The "deficiencies" of Canon sensors compared to the competition are low pixel count and high electronic noise at low ISOs.
takowasa wrote: gdanmitchell wroteIf all else is equal, but photo site "size" is smaller, the image is more susceptible to noise at the photosite leve (S/N ratio is smaller)l, though the noise may be less visible to viewers of a print at a given size since its "grain size" (to use a film way of thinking) is smaller.The "grain size" is indeed smaller, but the amplitude is higher -- if the viewer can is standing close enough with sufficient visual acuity to see the actual "grain". I believe the examples in these photos that I've linked a few times in this thread tell the story rather nicely:.
You are basically saying exactly what I said: that "if all else is equal" the level of noise from smaller photo sites is higher ("amplitude is higher") but that "grain" size is smaller.
In reality, the end result is rarely noisier photographs. Instead, photographers who choose the higher MP cameras tend to do so because they value their higher resolution and will likely shoot at lower ISOs. Those who care more about higher ISOs may prefer lower MP sensors.
This is simply one of the trade-offs that camera designers (and then photographers) make.
RustyBug wrote: takowasa wroteCanon sensors fare well at high ISOs (same quantum efficiency as everyone else, and the same, or even a tad lower, electronic noise per pixel). The "deficiencies" of Canon sensors compared to the competition are low pixel count and high electronic noise at low ISOs.
Lower electronic noise @ high ISO vs. high electronic noise low ISO
Sounds like there is a crossover that occurs ... diff mfr's opting to lean one way vs. the other.
That's funny you say that because back in the D700 heyday the users didn't hesitate bragging about the virtues of their superior high ISO capabilities, they were proud owners.
That changed big time with the D800, DR became the cat's meow, the heck with high ISO.
Yes ... I even bought into the D700 "superior high ISO capabilities" and picked one up.
Kept it for a month and sold it.
It was nice to use with my F-mount glass and certain other things I did dig about it. But, (for me) it didn't live up to making THAT MUCH of a diff over my other bodies.