p.2 #2 · Sony A7R IV Photographic Dynamic Range (PDR)
Reading some of this thread reminds me of my relationship with my wife. One her friends just got back from a month vacation in Europe, another just purchased a three million dollar home with 50% down and another was given a large diamond anniversary ring.
So, she wants to know why I haven't done all of these things for her.
p.2 #3 · Sony A7R IV Photographic Dynamic Range (PDR)
snapsy wrote:
The A7RIV's conversion gain change vs the A7rIII may not be an intentional decision made by Sony to rebalance/target low ISO performance but instead just the natural consequence of the A7rIV's smaller pixels, which means a lower FWC-per-pixel, which translates to a naturally lower LCG/HCG switching point, with consideration for potential differences in per-pixel read noise of the new sensor's design as well.
The lower full well capacity of the smaller pixels of the A7R4 has already been taken into account by the design capacitance of the sense node at base ISO. The sense node capacitance of the A7R4 at base ISO is already smaller than the capacitance of the sense node of the A7R3. The ratio is the ratio of the FWC. There is no reason to doubly compensate by also changing the point at which the capacitance is further lowered to boost the high ISO conversion gain. The PDR of the A7R4 was closely tracking the A7R3 in the ISO range of 100-250 and would have continued to do so to ISO 500.
p.2 #4 · Sony A7R IV Photographic Dynamic Range (PDR)
dclark wrote:
The lower full well capacity of the smaller pixels of the A7R4 has already been taken into account by the design capacitance of the sense node at base ISO. The sense node capacitance of the A7R4 at base ISO is already smaller than the capacitance of the sense node of the A7R3. The ratio is the ratio of the FWC. There is no reason to doubly compensate by also changing the point at which the capacitance is further lowered to boost the high ISO conversion gain. The PDR of the A7R4 was closely tracking the A7R3 in the ISO range of 100-250 and would have continued to do so to ISO 500....Show more →
That's certainly possible but makes presumptions about the altered pixel design on the A7RIV. If such adjustments were possible the A7S might not have a dual-gain conversion point all the way up at ISO 2000.
p.2 #5 · Sony A7R IV Photographic Dynamic Range (PDR)
Interesting to note how the compulsory raw NR kicks in 1EV earlier than the A7rIII, coinciding with the A7rIV's 1EV earlier dual-gain switch, even though the net effect on EDR of that earlier switch is only 1/3EV at the Ultra High ISO where the compulsory NR kicks in. This tells me the firmware switch for that NR may be tied/scaled directly to the dual-gain point rather than to a subjective measure of where Sony thinks the noise in the raw image would become objectionable.
p.2 #6 · Sony A7R IV Photographic Dynamic Range (PDR)
chiron wrote:
Nice summary of the differences and comparisons to alternatives.
I have trouble visualizing what a quarter or a third of a stop of dynamic range actually would look like in an image. Would one be able to see it? In what way would it matter?
In principle it might matter in situations where you deliberately underexpose so as to not burn out a prominent highlight, and then want to selectively raise the exposure or boost the shadows in shadow areas. In that case, the tiny bit less dynamic range could mean that that, after they have been boosted, there is just a tiny bit more noise in the deep shadows.
In practice a quarter of a stop will be very hard to detect (and is within the margin of error of exposure) and a third of a stop probably too.
While the discussion in this thread is interesting (especially why the Aptina dual gain point has been moved) and while the *in principle* differences are interesting (fractionally better ISO 400 in exchange for fractionally worse high ISO), in practice the differences in both directions are so minuscule that these won't matter in practice. In particular they should not play any role in decisions about bodies:a very small ergonomic difference, or price difference or AF different would be far more important.
p.2 #7 · Sony A7R IV Photographic Dynamic Range (PDR)
dclark wrote:
The lower full well capacity of the smaller pixels of the A7R4 has already been taken into account by the design capacitance of the sense node at base ISO. The sense node capacitance of the A7R4 at base ISO is already smaller than the capacitance of the sense node of the A7R3. The ratio is the ratio of the FWC. There is no reason to doubly compensate by also changing the point at which the capacitance is further lowered to boost the high ISO conversion gain. The PDR of the A7R4 was closely tracking the A7R3 in the ISO range of 100-250 and would have continued to do so to ISO 500....Show more →
At low enough ISO it is not required to use an extra capacitor connected to the pixel: the pixel itself is enough to store the charge (the parasitic capacitance of the node). This is the trick used by the DR-Pix tech. Smaller pixels receiving less light could allow for using a lower ISO to do the conversion gain switch in case they still can retain a similar capacitance in the node. If that is not the case and they can't hold as much amount of charge, the switch point should remain unmodified. I'd suspect the second, but just my two cents, I'm not a sensor engineer
p.2 #8 · Sony A7R IV Photographic Dynamic Range (PDR)
Steve Spencer wrote:
A third of a stop DR will matter a lot more in some images than others. In those images in which you are clipping highlights and you would still like to pull up shadows a bit more, being able to pull up those shadows a third of a stop more without more noise is something you may well notice. You could just live with more noise in the shadows, but that will limit how big you can present the image. It isn't a huge difference, and even a whole stop difference isn't a night and day difference, but if you work with enough files over a range of high DR situations you can recognize the difference and what it means for your shooting....Show more →
The A7R3 "advantage" of 1/3 stop is in "shadow improvement". And this difference is constant from ISO 640 to 12800. The A7R4 has higher dynamic range at high ISO if properly used (cap at ISO 320).
We shouldn't forget the big picture: if you keep the ISO at 640 (A7R3) or 320 (A7R4) and never go further, the A7R4 has around 0.75 EV higher dynamic range (or 0.66 EV if you use the PDR measure). With this additional headroom, if you manage to expose an additional +1/3 EV (despite the low light) you'll have just the same shadows performance of the A7R3, with still less clipping on these marvellous church stained glass windows. To be exact +0.36 EV (the extra 0.03 to compensate the shadows improvement lost by using ISO 320 instead of 640 in the A7R4, from the data posted a few days ago). At low signal levels, shadows are anyway just garbage. And we are splitting hairs here, nobody meters a scene so accurately (well, maybe except myself, which I'm a zebras maniac )
p.2 #9 · Sony A7R IV Photographic Dynamic Range (PDR)
Fred Miranda wrote:
With a high performance lens, I thought that higher sensor resolution = less moire.
With a soft lens, it should not matter as much as moire is masked just like when using a low pass filter.
I agree. Diffraction is the AA filter of high resolution sensors... for the good enough lenses. This is the A7R3 near the centre from a Sony 12-24 I tested (despite I was not testing the moire!):
The diffraction has not a clear "frontier" or an exact f-stop in which it appears. In fact, each color diffracts differently. But thanks to physics we know that if we set a certain "acceptable" f-stop for the green, the red color diffracts just the same -0.35 EV before, and blue diffracts the same +0.46 EV later. The same formula used to compare wavelengths can be used to compare megapixels. We can calculate how the change in sensor resolution has additionally moved these values in average. The A7R4 is limited by diffraction around 0.5 f-stop before:
log2(42/61) = -0.54 EV
There is a calculator here. I think that with many lenses, in the Ar74 we'll notice that F8 is clearly worse compared to F5.6 (much worse than we felt in the A7R3). But I'm not so sure that lenses sharper at F6.3 could now need F5.6. If the A7R5 will have 100MP (we hope!) it will move another -0.71 EV the frontier, so same the diffraction damage at F8 (A7R3) would be noticed at F5. I prefer looking it this other way: the image will have even more resolution (slightly in the worst case) but will be much less prone to moire and aliasing.
p.2 #10 · Sony A7R IV Photographic Dynamic Range (PDR)
cgarcia wrote:
At low enough ISO it is not required to use an extra capacitor connected to the pixel: the pixel itself is enough to store the charge (the parasitic capacitance of the node). This is the trick used by the DR-Pix tech. Smaller pixels receiving less light could allow for using a lower ISO to do the conversion gain switch in case they still can retain a similar capacitance in the node. If that is not the case and they can't hold as much amount of charge, the switch point should remain unmodified. I'd suspect the second, but just my two cents, I'm not a sensor engineer ...Show more →
The problem is not whether the photodiode can hold the full well capacity (it can) it is what capacitance at the sense node is needed so that the voltage at the sense node is approximately 1 volt or a bit more when the FWC is transferred to the sense node. The parasitic capacitance of the sense node is small so some additional capacitance is needed at the sense node to accommodate the FWC charge and not exceed the voltage. That is the sense node capacitance at base ISO. If the sensor has a smaller pixel the FWC is lower so less capacitance needs to be added. The A7R4 has a smaller FWC than the A7R3 so it has a smaller added capacitance.
A dual conversion gain sensor adds another transistor (the dual conversion gain, DCG, transistor) to the pixel that can isolate the added capacitance from the sense node. If the transistor is conductive the capacitance is added to the sense node and it can handle the FWC. If the transistor is open the capacitance at the sense node is much smaller and is usually determined by the parasitic capacitance of the sense node. That can be controlled by the design of the node and attached electronics. The sense node capacitance determines the maximum charge it can handle without exceeding ~1V in the high ISO conversion gain mode. The A7R3 switches to HCG (opens the DCG transistor) at ISO 640. That means the max charge in the pixel should be 6.4X smaller than FWC. To generate 1V at the sense node the capacitance needs to be ~6.4X smaller than the sum of the sense node capacitance and the added capacitance.
For the A7R4 the FWC is smaller, so the added capacitance at the sense node is smaller. The A7R4 switches to high conversion gain at ISO 320, which means the pixel max charge is 3.2X smaller than the FWC and the sense node capacitance is 3.2X smaller than the sum of the sense node capacitance and the added capacitance. If we assume that the FWC of the A7R4 is ~(42.4/61) times the FWC of the A7R3, we can compute the ratio of the sense node capacitance of the A7R3 and the A7R4. I will not show my algebra but I get that the sense node capacitance of the A7R4 is ~1.7X the sense node capacitance of the A7R3. That means that in order to move the gain change from ISO 640 to ISO 320, the sensor design had to be modified to significantly increase the sense node capacitance in a smaller pixel sensor. This does not seem to be consistent with it just being a by product of the smaller pixel. In fact it is the opposite of what may be expected from a smaller pixel. This seems to be a deliberate act.
Someone may argue that my assumption that the ratio of the FWC is 42.4/61 is suspect. You can work through the problem of what would the FWC ratio needs to be for the sense node capacitance to remain unchanged. I will leave that as an exercise for any interested persons. Then the question is, if the FWC of the A7R4 is that much less than the FWC of an A7R3, is it possible to see similar PDR performance?
This is a long note. I hope I did not make too many errors.
p.2 #11 · Sony A7R IV Photographic Dynamic Range (PDR)
Fred Miranda wrote:
With a high performance lens, I thought that higher sensor resolution = less moire.
With a soft lens, it should not matter as much as moire is masked just like when using a low pass filter.
Just a point of clarification. Higher resolution sensor does not equal less moire unless the sampling is exceeding the lens performance (which is likely in the corners, for example).
What happens with a high resolution lens and sensor is the moire will be pushed down into smaller portions of the image where it may become irrelevant. For example, instead of seeing it in bricks that are 100 feet away, it will be in bricks 200 feet away, or instead of seeing it in cordura fabric, it will be in silk.
If the lens is still exceeding the sensor, there will be the possibility for moire, but the size of the moire pattern will become smaller and less significant both on likely image size (proportion) and in terms of the level of detail you have to look at in a normal scene to see it.
Blow up the images so that the pixels are the same size, and if the image has the pattern to produce moire, it will still be there (as long as the lens can out resolve the sensor).
p.2 #12 · Sony A7R IV Photographic Dynamic Range (PDR)
cgarcia wrote:
At low enough ISO it is not required to use an extra capacitor connected to the pixel: the pixel itself is enough to store the charge (the parasitic capacitance of the node). This is the trick used by the DR-Pix tech. Smaller pixels receiving less light could allow for using a lower ISO to do the conversion gain switch in case they still can retain a similar capacitance in the node. If that is not the case and they can't hold as much amount of charge, the switch point should remain unmodified. I'd suspect the second, but just my two cents, I'm not a sensor engineer ...Show more →
p.2 #13 · Sony A7R IV Photographic Dynamic Range (PDR)
mjm6 wrote:
Just a point of clarification. Higher resolution sensor does not equal less moire unless the sampling is exceeding the lens performance (which is likely in the corners, for example).
What happens with a high resolution lens and sensor is the moire will be pushed down into smaller portions of the image where it may become irrelevant. For example, instead of seeing it in bricks that are 100 feet away, it will be in bricks 200 feet away, or instead of seeing it in cordura fabric, it will be in silk.
If the lens is still exceeding the sensor, there will be the possibility for moire, but the size of the moire pattern will become smaller and less significant both on likely image size (proportion) and in terms of the level of detail you have to look at in a normal scene to see it.
Blow up the images so that the pixels are the same size, and if the image has the pattern to produce moire, it will still be there (as long as the lens can out resolve the sensor)....Show more →
Sure.
But if you are comparing whole images at fixed print sizes or screen sizes, the image with higher capture MP will usually have less visible moiré, and never more visible moiré.
p.2 #14 · Sony A7R IV Photographic Dynamic Range (PDR)
DavidBM wrote:
Sure.
But if you are comparing whole images at fixed print sizes or screen sizes, the image with higher capture MP will usually have less visible moiré, and never more visible moiré.
It is entirely dependent on the reproduction size of the image and for many people, higher MP means larger prints.
With 10 MP, would you do a 20x30 image from 35mm? Now, people are doing that regularly because they have the MP to support 360ppi on a print of that size.
Once you do that (run up the image to 360ppi or something similar), you will find the same amount of moire, but it will be at a different scale in the image because the image is reproduced larger.
This is what I meant about the moire being driven down into the details of the image with high MP.
If you take a 24MP image and a 62MP image and print them both at the same size, the moire in the 62MP image will be smaller scale and may become mostly invisible (or completely invisible) at that reproduction ratio, but it isn't less in the source file. The potential for moire remains exactly the same as long as the lens renders more than the sampling frequency of the sensor.
The caveat here is that the wells don't change in terms of cross-talk or other noise factors that will change the actual ability of the sensor to accurately collect the color information. That could very well happen as the sensors get more densly packed.
p.2 #15 · Sony A7R IV Photographic Dynamic Range (PDR)
mjm6 wrote:
It is entirely dependent on the reproduction size of the image and for many people, higher MP means larger prints.
With 10 MP, would you do a 20x30 image from 35mm? Now, people are doing that regularly because they have the MP to support 360ppi on a print of that size.
Once you do that (run up the image to 360ppi or something similar), you will find the same amount of moire, but it will be at a different scale in the image because the image is reproduced larger.
This is what I meant about the moire being driven down into the details of the image with high MP.
If you take a 24MP image and a 62MP image and print them both at the same size, the moire in the 62MP image will be smaller scale and may become mostly invisible (or completely invisible) at that reproduction ratio, but it isn't less in the source file. The potential for moire remains exactly the same as long as the lens renders more than the sampling frequency of the sensor.
The caveat here is that the wells don't change in terms of cross-talk or other noise factors that will change the actual ability of the sensor to accurately collect the color information. That could very well happen as the sensors get more densly packed. ...Show more →
That’s what I said! Viewing the whole image at a fixed print or screen size (from the same distance) there will be less moirč visible in the higher capture res file. Of course if you compare at 1:1 this may not be so, because 1:1 with a higher res file is effectively greater magnification. But what matters is the final image - and for an intended final image viewing size choosing a higher res sensor will give you less moirč than a lower res one, with no extra perceived softness such as you might get by using a moirč tool or a low pass filter.
p.2 #16 · Sony A7R IV Photographic Dynamic Range (PDR)
DavidBM wrote:
That’s what I said! Viewing the whole image at a fixed print or screen size (from the same distance) there will be less moirč visible in the higher capture res file. Of course if you compare at 1:1 this may not be so, because 1:1 with a higher res file is effectively greater magnification. But what matters is the final image - and for an intended final image viewing size choosing a higher res sensor will give you less moirč than a lower res one, with no extra perceived softness such as you might get by using a moirč tool or a low pass filter. ...Show more →
That's what I said...
However, you are dismissing the actual moire potential in the file with the practical moire in a final reproduced image (digital or on paper). They are different things, and that was my point.
p.2 #17 · Sony A7R IV Photographic Dynamic Range (PDR)
mjm6 wrote:
That's what I said...
However, you are dismissing the actual moire potential in the file with the practical moire in a final reproduced image (digital or on paper). They are different things, and that was my point.
We really (I think) don’t disagree! Of course there’s pixel level moirč as it's at the effective frequency of the pixel level where the moirč occurs. I don’t know what “dismissing the moirč potential” means. If it means denying the moirč at the pixel level I’m certainly not - that’s why I was careful to specify fixed print or screen sizes. If it means realising that for photographic purposes there are significant improvements in the moirč of finished images holding print or screen size fixed to be had by using higher resolution sensors then I am.
p.2 #18 · Sony A7R IV Photographic Dynamic Range (PDR)
mjm6 wrote:
Just a point of clarification. Higher resolution sensor does not equal less moire unless the sampling is exceeding the lens performance (which is likely in the corners, for example).
What happens with a high resolution lens and sensor is the moire will be pushed down into smaller portions of the image where it may become irrelevant. For example, instead of seeing it in bricks that are 100 feet away, it will be in bricks 200 feet away, or instead of seeing it in cordura fabric, it will be in silk.
If the lens is still exceeding the sensor, there will be the possibility for moire, but the size of the moire pattern will become smaller and less significant both on likely image size (proportion) and in terms of the level of detail you have to look at in a normal scene to see it.
Blow up the images so that the pixels are the same size, and if the image has the pattern to produce moire, it will still be there (as long as the lens can out resolve the sensor)....Show more →
No... because there is more than lens aberrations.
Yes, usually the lenses are aberration limited before becoming diffraction limited, but even a perfect lens with zero aberrations will be diffraction limited.
So the sensor doesn't neccesarily needs to outresolve the lens to get rid of aliasing: at worst, at a given aperture, it only needs to have pixels small enough to cause the lens to become diffraction limited if it was not already limited by its aberrations (as likely was if not somewhat stopped down).
Totally removing the aliasing by the means of diffraction requires really small pixels (see recent Jim's article). But diffraction is increasingly filtering out the aliasing each time the pixels get smaller. And it is not merely getting scaled down. That's what I pointed out a few posts above. A sensor with twice the pixels gets the same aliasing "safety" a full f-stop before. If we felt safe shooting a fabric at F11 in the A7R3, the A7R4 will get us just the same feeling at F9.5 (not just the same artifacts smaller, but less artifacts compared to the A7R3 also at F9.5).
p.2 #19 · Sony A7R IV Photographic Dynamic Range (PDR)
This detailed discussion is very interesting although it vividly points out some holes in my understanding
At the end of the day, the charts tell me to essentially expect similar performance with the 4 that has an approximate 50% increase in resolution. I'm good with that
p.2 #20 · Sony A7R IV Photographic Dynamic Range (PDR)
schlotz wrote:
This detailed discussion is very interesting although it vividly points out some holes in my understanding
At the end of the day, the charts tell me to essentially expect similar performance with the 4 that has an approximate 50% increase in resolution. I'm good with that
