p.4 #1 · A1 EVF resolution -- theoretical vs actual
speedmaster20d wrote:
as for your last question are you asking about scaling factor between 0.9 and 0.7 magnification?
I assume the way the EVF achieves magnification 0.7 is by truncating the OLED display from 1536x2048 to 1200x1600. That is what is need to change from 0.9 to 0.7. Any other way I can devise requires some actuation of the EVF optics and/or the OLED display, which seems too unlikely to be believed. How can both magnification 0.7 be achieved and display of 1536x2048? I can't come up with a way I believe. Maybe I am not understanding something or am missing a simple explanation. As I stated above, unless there is an explanation for this, I loose confidence in the other conclusions. I hope you or someone has a good explanation.
p.4 #2 · A1 EVF resolution -- theoretical vs actual
dclark wrote:
The 240/sec mode is less compromised than the 120/sec mode if only one axis has its resolution lowered.
I don't see how it is possible to not reduce the resolution of both axes when the magnification is lowered to 0.7 since I assume the OLED display is truncated.
That's why I say I am baffled by that result. I don't believe any explanation I can devise.
Because the sensor is skipping lines in only one axis, so you're only losing resolution in one axis. The following article depicts this for the D800's Live View implementation.
p.4 #3 · A1 EVF resolution -- theoretical vs actual
dclark wrote:
If you look at the FM thread cited you will see that I complained that the Help Guide uses a lot of ill defined terms, rather than specific numbers and specifications. The numbers that have (*) are not specifically used in the Help Guide. For example, "Standard", "High", and "Higher". It's possible to find in the Help Guide or the A1 Specifications that "Standard" is 60/sec and that "Higher" is 240/sec, but I cannot find anyplace in either document that "High" is 120/sec. Maybe it's there someplace, but I could not find it. There are other non-Sony sources for the data in the table with (*). I tagged those items since so many sources are putting out unreliable or demonstrably incorrect information. Regardless, I believe the data in the table are all correct. ...Show more →
I guess I'm not clear who the author of the table is!
p.4 #4 · A1 EVF resolution -- theoretical vs actual
dclark wrote:
I cannot discern the difference between High and Standard resolution by looking at the display. They both look good to me. That is why I used the Siemens Star pattern.
I don't see how to get both High resolution display, which requires the 1536x2048 of the entire OLED display, and magnification of 0.7 which is achieved by truncating the OLED to 1200x1600. Maybe someone can come up with an explanation. I have not come up with one.
Yeah, I don't see the difference looking at the EVF display alone... I need to point it at something first. In most cases, whether far away (cityscapes) or closer (I used my monitor with this forum displayed), I can see the difference as there is enough detail to distinguish between the two modes.
I can also confirm I see the same behavior with the Siemens star, so in 240hz there is only horizontal (left & right) aliasing pattern...
What is interesting is that High quality in 0.9x mag. behaves the same way as in 0.7x so no aliasing patterns visible. Imho this suggests either High option relates more to PPI (Pixel Per Inch / or Dot Per Inch in priting terms) rather than the bare resolution or it controls also line skipping as @snapsy@ suggests...
p.4 #5 · A1 EVF resolution -- theoretical vs actual
snapsy wrote:
Because the sensor is skipping lines in only one axis, so you're only losing resolution in one axis. The following article depicts this for the D800's Live View implementation.
Yes, it is clearly possible to down res in one or two directions. That's not the problem I see. The problem is that I believe magnification 0.7 is achieved by truncating the OLED from 1536x2048 to 1200x1600, a factor of 0.78 along each axis . Note that 0.78x0.90=0.70. Consequently, if there is only 1200x1600 available in the OLED display, how to display 1200x2048? If both axes are down res'ed why only one axis of aliasing? I can't make sense of it.
The link you provided is interesting but it relates to down sampling a Bayer array. The downsampling for the EVF display is simpler since it is for downsampling an RGB array to a lower res RGB array. The EVF display is a partial res RGB image or a subsampling of the full res RGB image. The subsampling is 8640/2048=135/32 for High Display Quality and 8640/1600=27/5 for Standard Display Quality. Since these are not integers the sampling is a mix of integer steps that average those ratios. For example 5, 6, 5, 6, 5 to average 27/5 to downsample a full res image to Standard Display Quality. The 135/32 downsampling is a series of 32 steps consisting of 25 4's and 7 5's. (BTW, those are not the only solutions but the most uniform solutions. For example 1,1,1,1,23 also averages 27/5).
Regardless, the issue that concerns me is not how to downsample, it is that I see no way to display 1200x2048 at 0.7 magnification.
p.4 #6 · A1 EVF resolution -- theoretical vs actual
Holger wrote:
Why is the image moving, but the settings aren't? If it is the phone than everything should move at the same time. Your test was a very good idea, but not well executed.
Otherwise, the EVF behaves in accordance to Sony's description, the only thing not clearly disclosed is the aperture dependent behavior at Standard as shown by dclark.
Out of interest: Do you shoot BIF at 60Hz or 120Hz? Would you see the difference between 5.76 and 9.44 Million dots when panning and shooting BIF?
the image is not moving.
It does appear to me you don't have an A1 so I don't understand what is your business trying to justify or defend(?) a product you don't own and may never actually own by making assumptions and speculation about how the actual owner of this camera will use it and what they should expect. Sony themselves are admitting there is an unexpected issue and looking into it and a random guy on the internet who doesn't even have the camera is trying to prove otherwise...hum doesn’t that sound bizarre what do you gain?
As I said when and if you get an A1 test it for yourself and check if you see the resolution drop or not. That should clear it
p.4 #7 · A1 EVF resolution -- theoretical vs actual
dclark wrote:
I assume the way the EVF achieves magnification 0.7 is by truncating the OLED display from 1536x2048 to 1200x1600. That is what is need to change from 0.9 to 0.7. Any other way I can devise requires some actuation of the EVF optics and/or the OLED display, which seems too unlikely to be believed. How can both magnification 0.7 be achieved and display of 1536x2048? I can't come up with a way I believe. Maybe I am not understanding something or am missing a simple explanation. As I stated above, unless there is an explanation for this, I loose confidence in the other conclusions. I hope you or someone has a good explanation. ...Show more →
Hi Dave,
The resolution is not chaining when you go from 0.9 to 0.7 the camera is just putting a black frame around the image the black frame is 270 pixels wide. the image in the center is zoomed back and rescaled to fid in the smaller window
p.4 #8 · A1 EVF resolution -- theoretical vs actual
j4nu wrote:
Yeah, I don't see the difference looking at the EVF display alone... I need to point it at something first. In most cases, whether far away (cityscapes) or closer (I used my monitor with this forum displayed), I can see the difference as there is enough detail to distinguish between the two modes.
I can also confirm I see the same behavior with the Siemens star, so in 240hz there is only horizontal (left & right) aliasing pattern...
What is interesting is that High quality in 0.9x mag. behaves the same way as in 0.7x so no aliasing patterns visible. Imho this suggests either High option relates more to PPI (Pixel Per Inch / or Dot Per Inch in priting terms) rather than the bare resolution or it controls also line skipping as @snapsy@@ suggests......Show more →
Concerning your three points:
Your eyes are better than mine.
I am pleased you checked the 240/sec pattern and see the same thing I saw. I was concerned that I was making a mistake. Unfortunately, I still see no way to explain the result.
You will see by my other comments that I still don't see any explanation that works. I hope someone comes up with something.
p.4 #9 · A1 EVF resolution -- theoretical vs actual
speedmaster20d wrote:
Hi Dave,
The resolution is not chaining when you go from 0.9 to 0.7 the camera is just putting a black frame around the image the black frame is 270 pixels wide. the image in the center is zoomed back and rescaled to fid in the smaller window
best
Where did you find this information?
270 pixels? Seems like it should be different on top/bottom and on the sides for a 4:3 aspect ratio display.
"zoomed back" implies that the EVF optics are actuated. Is that what you are saying? I discounted the possibility but if the optics and/or OLED can be moved that opens up a lot of possibilities that can explain just about anything. Maybe I should not so quickly discount that possibility. I searched for some information on this but could find nothing. Not even speculation let alone a credible source.
There is no need to move anything for most of the cases. Only the High Display Quality at 0.7 magnification, and the 240/sec mode. Everything else works very well by just truncating (digital zooming) the OLED, which is pretty trivial.
p.4 #10 · A1 EVF resolution -- theoretical vs actual
dclark wrote:
Yes, it is clearly possible to down res in one or two directions. That's not the problem I see. The problem is that I believe magnification 0.7 is achieved by truncating the OLED from 1536x2048 to 1200x1600, a factor of 0.78 along each axis . Note that 0.78x0.90=0.70. Consequently, if there is only 1200x1600 available in the OLED display, how to display 1200x2048? If both axes are down res'ed why only one axis of aliasing? I can't make sense of it.
The link you provided is interesting but it relates to down sampling a Bayer array. The downsampling for the EVF display is simpler since it is for downsampling an RGB array to a lower res RGB array. The EVF display is a partial res RGB image or a subsampling of the full res RGB image. The subsampling is 8640/2048=135/32 for High Display Quality and 8640/1600=27/5 for Standard Display Quality. Since these are not integers the sampling is a mix of integer steps that average those ratios. For example 5, 6, 5, 6, 5 to average 27/5 to downsample a full res image to Standard Display Quality. The 135/32 downsampling is a series of 32 steps consisting of 25 4's and 7 5's. (BTW, those are not the only solutions but the most uniform solutions. For example 1,1,1,1,23 also averages 27/5).
Regardless, the issue that concerns me is not how to downsample, it is that I see no way to display 1200x2048 at 0.7 magnification. ...Show more →
From my view the only reason for the EVF to have a lower resolution and/or frame rate for various AF operations is because the sensor feed being delivered to the EVF changes, ie lower resolution from line skipping or binning, both of which are required to accommodate a higher sampling rate into the AF system. This means a subsampled feed of the raw sensor data.
If I'm understanding you correctly, you believe there might be other reasons for the EVF resolution to drop in these AF scenarios. If that is correct, what do you believe those reasons to be? For example, why would it be limited to a downsample of the post-demosaiced RGB data For the EVF?
p.4 #12 · A1 EVF resolution -- theoretical vs actual
dclark wrote:
Where did you find this information?
270 pixels? Seems like it should be different on top/bottom and on the sides for a 4:3 aspect ratio display.
"zoomed back" implies that the EVF optics are actuated. Is that what you are saying? I discounted the possibility but if the optics and/or OLED can be moved that opens up a lot of possibilities that can explain just about anything. Maybe I should not so quickly discount that possibility. I searched for some information on this but could find nothing. Not even speculation let alone a credible source.
There is no need to move anything for most of the cases. Only the High Display Quality at 0.7 magnification, and the 240/sec mode. Everything else works very well by just truncating (digital zooming) the OLED, which is pretty trivial.
270 pixel is the horizontal border you can calculate the vertical border the same manner
the optics is fixed. what I meant was the image is scaled digitally (the same way you resample an image in photoshop) to fit into a smaller window with a black frame around it when you set the magnification to 0.7x.
p.4 #13 · A1 EVF resolution -- theoretical vs actual
QuietOC wrote:
Gordon Laing covers the viewfinder extensively in his review.
I accidently posted this in the other thread. He shows the resolution differences pretty well.
yes it is quite evident yet actually for shooting birds I don't notice it as much, I am so focused on the action and getting the shot that this is the last thing I would notice. When I notice it more is when I shoot portrait which is a lot slower and more relaxed shooting... The moire and jagged artifacts are suddenly visible and distracting on lady's long hair at portrait range, especially blond or light color/shiny hair. looks like 1980's crunchy TV's. the worst part is that it shows you how nice and smooth the image ought to be as soon as you are not focusing! kind of like showing a kid the candy then taking it away from them
p.4 #14 · A1 EVF resolution -- theoretical vs actual
snapsy wrote:
From my view the only reason for the EVF to have a lower resolution and/or frame rate for various AF operations is because the sensor feed being delivered to the EVF changes, ie lower resolution from line skipping or binning, both of which are required to accommodate a higher sampling rate into the AF system. This means a subsampled feed of the raw sensor data.
If I'm understanding you correctly, you believe there might be other reasons for the EVF resolution to drop in these AF scenarios. If that is correct, what do you believe those reasons to be? For example, why would it be limited to a downsample of the post-demosaiced RGB data For the EVF?...Show more →
that is a possibility but I don't think the sensor readout is being changed because if you actually output the image over HDMI to a 4K TV it shows no difference (the rear LCD also shows no evidence of interlaced reading or binning when focusing). If you bin at sensor level you won't see much difference on such a comparatively low resolution finder due to the huge down-sampling ( from 50 mega pixel to 3 mega pixel)
My speculation is that the SoC cannot simultaneously compute AF tracking which is quite sophisticated in this camera and run EVF pipeline at full bandwidth. it probably bins at the data link from SoC to EVF which results in such crunchy image. but only Sony knows what's happening and at the very least they should loud and clearly disclose this to the customer.
p.4 #15 · A1 EVF resolution -- theoretical vs actual
speedmaster20d wrote:
that is a possibility but I don't think the sensor readout is being changed because if you actually output the image over HDMI to a 4K TV it shows no difference (the rear LCD also shows no evidence of interlaced reading or binning when focusing). If you bin at sensor level you won't see much difference on such a comparatively low resolution finder due to the huge down-sampling ( from 50 mega pixel to 3 mega pixel)
My speculation is that the SoC cannot simultaneously compute AF tracking which is quite sophisticated in this camera and run EVF pipeline at full bandwidth. it probably bins at the data link from SoC to EVF which results in such crunchy image. but only Sony knows what's happening and at the very least they should loud and clearly disclose this to the customer.
p.4 #17 · A1 EVF resolution -- theoretical vs actual
speedmaster20d wrote:
yes it is quite evident yet actually for shooting birds I don't notice it as much, I am so focused on the action and getting the shot that this is the last thing I would notice. When I notice it more is when I shoot portrait which is a lot slower and more relaxed shooting... The moire and jagged artifacts are suddenly visible and distracting on lady's long hair at portrait range, especially blond or light color/shiny hair. looks like 1980's crunchy TV's. the worst part is that it shows you how nice and smooth the image ought to be as soon as you are not focusing! kind of like showing a kid the candy then taking it away from them ...Show more →
Exactly! I thought I was the only one seeing this that way ...
For action shots, I don't think it's really relevant as you will not be able to see all the details anyway because of the movement.
p.4 #19 · A1 EVF resolution -- theoretical vs actual
speedmaster20d wrote:
the worst part is that it shows you how nice and smooth the image ought to be as soon as you are not focusing! kind of like showing a kid the candy then taking it away from them
You could use 120Hz at all times in order to hide the sweets
p.4 #20 · A1 EVF resolution -- theoretical vs actual
speedmaster20d wrote:
270 pixel is the horizontal border you can calculate the vertical border the same manner
the optics is fixed. what I meant was the image is scaled digitally (the same way you resample an image in photoshop) to fit into a smaller window with a black frame around it when you set the magnification to 0.7x.
hope this helps
I don't see how you calculate 270. If the display changes from 1536x2048 to 1200x1600, the border with be (1536-1200)/2=168 pixels at the top and bottom, and (2048-1600)/2=224 pixels on the left and right.
I agree that the display is digitally zoomed not optically zoomed to change the magnification from 0.9 to 0.7. So I still don't see how to explain the results for the High Display Quality at 0.7 magnification, and the 240/sec mode which is also at 0.7 magnification. When the magnification is 0.7 the available OLED display is reduced to 1200x1600.
what do you gain? 
