p.11 #1 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
skibum5 wrote:
Can't you just do that by taking two different shots now, one long, one shot and combine? Does that help much?
The key step in the patent above is that the two exposures are "at the same time" - not sequentially as the HDR or the blend two exposures method currently requires.
In the patent, the two exppsures are centered on the same notional instant t=0 - e.g. shadows exposed from -8 msec to +8 msec, - highlights from -2 msec to +2 msec ; This results in 16 msec for shadows, and 4 msec for highlights. The luminance values put in the file for the shadow pixels would be calculated by dividing the recordedl 16 msec values by 4. Any motion that might occur during the 16 msec (~1/60 sec) needed for the shadows would be centred on the image contributed by the highlights (captured in 1/250 sec). Useful in some situations, maybe not all. If you are at the limit of light sensitivity and need a long shutterspeed even for midtones, extending the exposure to improve the shadows may not be a great help. In the patent, IIRC, the user can set a slider to determine the threshold luminance used in the map to distinguish between "shadow pixels" (for the long exposure) and the rest for the normal exposure. The wording was also open (as it often is) to allow for more than just two exposure times for the pixels.
It may be OK in theory but in practice , who knows. It's also something that could be added on top of any other DR improvement strategies.
p.11 #2 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
I'm not sure why they haven't examined the idea of a photochromatic filter on the sensor or even a photochromatic version of the micro lens array. Could liberate 1-2 stops DR without any change to the sensor tech.
p.11 #3 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
AJSJones wrote:
The key step in the patent above is that the two exposures are "at the same time" - not sequentially as the HDR or the blend two exposures method currently requires.
In the patent, the two exppsures are centered on the same notional instant t=0 - e.g. shadows exposed from -8 msec to +8 msec, - highlights from -2 msec to +2 msec ; This results in 16 msec for shadows, and 4 msec for highlights. The luminance values put in the file for the shadow pixels would be calculated by dividing the recordedl 16 msec values by 4. Any motion that might occur during the 16 msec (~1/60 sec) needed for the shadows would be centred on the image contributed by the highlights (captured in 1/250 sec). Useful in some situations, maybe not all. If you are at the limit of light sensitivity and need a long shutterspeed even for midtones, extending the exposure to improve the shadows may not be a great help. In the patent, IIRC, the user can set a slider to determine the threshold luminance used in the map to distinguish between "shadow pixels" (for the long exposure) and the rest for the normal exposure. The wording was also open (as it often is) to allow for more than just two exposure times for the pixels.
It may be OK in theory but in practice , who knows. It's also something that could be added on top of any other DR improvement strategies....Show more →
yeah I know centered helps a bit, but only so much, well I guess you don't have the mirror flap down to and all so it helps a fair bit, but it's still only a partial solution and not really as generally as good as just having better actual direct read noise by any means it doesn't sound like
p.11 #4 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
Pixel Perfect wrote:
I'm not sure why they haven't examined the idea of a photochromatic filter on the sensor or even a photochromatic version of the micro lens array. Could liberate 1-2 stops DR without any change to the sensor tech.
I think a photochromic layer on the sensor would be great if they could make them respond fast enough to be useful. The other possibility is a neutral LCD that lined up with the sensor. Take a prexposure and then use the LCD image as a contrast mask (remember those?) to attenuate the highlights and expose the shadows more - then use the preexposure map to expand the contrast post facto.
p.11 #5 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
Pixel Perfect wrote:
I'm not sure why they haven't examined the idea of a photochromatic filter on the sensor or even a photochromatic version of the micro lens array. Could liberate 1-2 stops DR without any change to the sensor tech.
p.11 #6 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
skibum5 wrote:
yeah I know centered helps a bit, but only so much, well I guess you don't have the mirror flap down to and all so it helps a fair bit, but it's still only a partial solution and not really as generally as good as just having better actual direct read noise by any means it doesn't sound like
Indeed - if you boost the shadow well counts way above the read noise, they become cleaner. Add that eeffect on top of an additional reduction in noise by other methods, then you have detail in the coal stains on the black cat in the cellar
The dual Green process offers the following significant technical advantages:
1) Doubles the effective saturation level of the summed Green output video
2) Increases the noise of the final Green output only by a factor of square root of two
3) Combination of 1) and 2) increases the effective dynamic range of the green signal
and as a consequence, that of the matriced Luma signal
4) Increases the effective output green video bit depth
5) The half-pixel offset between the two separate green sampling lattices both
horizontally and vertically virtually eliminates the first order sideband spectra
associated with the sensor sampling process. This eliminates green aliasing.
6) Creates an effective FIR* filter within the readout process that aids the optimization
of the horizontal MTF and the progressive vertical MTF and associated aliasing
Topics 5) and 6) are dealt with in a companion white paper that discusses imager resolution.
* FIR stands for Finite Impulse Response being a class of digital filters used in digital signal processing (DSP
Each photosite has noise generating sources associated with the various separate physical
mechanisms taking place inside it:
Input Mechanism shot or photon noise from random nature of photoelectron
collection over a small interval of time
Reset Noise thermal noise generated when the signal readout portion in the
pixel (floating diffusion) is reset
Output Noise 1/f noise from an amplifier in the pixel
Output Amplifier Noise On-chip MOSFET amplifier to extract tiny signal from
the charge capacitor
The new image sensor employs a number of strategies to effectively reduce the last three noise sources which collectively define the noise floor of the imager.
The noise in the output analog signal is primarily reset thermal noise and the fixed pattern noise associated with the photosite. The reset noise is random and appears during the previous reset period, whereas the fixed pattern noise is primarily due to the variance among the threshold voltages of the amplifier MOS transistors. These noise components are read out separately from the signal readout (the latter still containing those noise components) and these two readouts are applied to a differential amplifier which successfully cancels the noise.
A low-noise narrow-band pre-amplifier then raises the resultant analog voltage to a level optimized for the application to the subsequent ADC process. Correlated double sampling -- which refers to a method whereby the output signal is sampled twice for each photosite - just after precharging capacitor and after the pixel charge packet is added, is employed. The difference between these two values does not include the noise component induced by the switch.
They are only claiming a 73 db DR, but teh base ISO is 800.. They are saying thatthey maintain a 12 stop DR from ISO 350 to 20,000.
They are also doing high speed signal readouts to get 60P in 1/120 second, then splitting the signal to 2 frames.
Any application to still image DR in a higher MP sensor?...Show more →
Thanks for the link.
The emphasis is on power consumption and readout speed to reduce problems with rolling shutter.
Then there is emphasis on how the dynamic range is utilized. This is only choices in how they encode the signal, nothing new in the signal quality of the raw capture.
The base characteristics of the sensor is normal. 40000 e- full well and 12 e- read noise. That is nothing new, it is about the standard for Canon sensors at lowest ISO.
They still use analog CDS as they have for many years. Nothing new here. The implications for still photography dynamic range is uninteresting.
When you say they maintain 12 Stops DR up to ISO 20000, that is completely false and a misinterpretation.
Where in the paper do they state the base ISO of 850 and the ISO range you mention? Do you confuse base ISO with unity gain ISO?
p.11 #8 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
skibum5 wrote:
Yeah the numbers actually do work out that way. So long as they give it a 6fps mirror box (and they certainly can since the D800 and D700 and 5D3 all have FF mirrors that can do at least 6fps for a not insane price).
So no they are not made up (beyond the possibility that the original rumor of 46+5 is made up).
What on earth does some have two digic or some have one have to do with my 45+5 vs 38+6
So if 46mp=5 fps and 38mp=6fps then why does the 5DIII at 22mp=6fps? And why is a 6D at 20mp=4.5fps?
Sorry, unless there is a more detailed technical explanation other than "the numbers work out that way" then these numbers are still made up.
p.11 #9 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
artd wrote:
So if 46mp=5 fps and 38mp=6fps then why does the 5DIII at 22mp=6fps? And why is a 6D at 20mp=4.5fps?
Sorry, unless there is a more detailed technical explanation other than "the numbers work out that way" then these numbers are still made up.
Because the 5D3 is single digic 5+. The 46MP cam would be dual digic 5+. The 1DX is 18MP and does 12-14fps. The 6D is maybe using only digic 5 and not 5+ plus (?) they they may simply be using a cheaper old 4.5fps mirror box most likely (since mirror boxes have real physical cost and they want as high a margin as they can on it).
p.11 #10 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
skibum5 wrote:
Because the 5D3 is single digic 5+. The 46MP cam would be dual digic 5+. The 1DX is 18MP and does 12-14fps. The 6D is maybe using only digic 5 and not 5+ plus (?) they they may simply be using a cheaper old 4.5fps mirror box most likely (since mirror boxes have real physical cost and they want as high a margin as they can on it).
So there ARE reasons.
Based on the math, a 46mpx camera could be about 5 fps. However, this would likely require the camera to be a 1Dx type body as otherwise I doubt there would be enough power and/or space for the dual chips. If that is the case, then canon would have to charge at least 6700 for the camera. Probably more.
p.11 #11 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
Schlotkins wrote:
Based on the math, a 46mpx camera could be about 5 fps. However, this would likely require the camera to be a 1Dx type body as otherwise I doubt there would be enough power and/or space for the dual chips. If that is the case, then canon would have to charge at least 6700 for the camera. Probably more.
My thoughts exactly.
So, let's say this 46mp camera costs $7000.
Would that be in any way competitive with the $3000/36mp Nikon D800?
I would think not.
p.11 #13 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
skibum5 wrote:
Because the 5D3 is single digic 5+. The 46MP cam would be dual digic 5+. The 1DX is 18MP and does 12-14fps. The 6D is maybe using only digic 5 and not 5+ plus (?) they they may simply be using a cheaper old 4.5fps mirror box most likely (since mirror boxes have real physical cost and they want as high a margin as they can on it).
So there ARE reasons.
The 1DX does 12fps. The 14fps is jpg only. So if 18mp=12fps then 38mp=6fps?
But more importantly we come back to this point: there is more to fps than just the number of megapixels on the sensor. I would much rather they use a less expensive mirror box, for instance, so that I don't have to pay more for something that's not relevant to my needs. My guess is that most of those who are on the high megapixel bandwagon are not going to care about 1-2 extra fps.
p.11 #14 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
jorkata wrote:
My thoughts exactly.
So, let's say this 46mp camera costs $7000.
Would that be in any way competitive with the $3000/36mp Nikon D800?
I would think not.
Huh? Not enough power and space? My small form factor small battery 7D had dual Digic chips and could take 2500 exposures with the mirror flipping on one battery charge if I turned all image display off.
p.11 #15 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
The Digic5+ chips are HUGE.
Here's the 1DX PCB.
One of the three small chips in the bottom-left corner is a Digic4.
You can see the difference in size for yourself.
Also, you can be sure that this time around only the 1-series will have dual Digic5+ chips.
If there is a 7DII, for example, it won't have dual Digic5+ chips for sure.
p.11 #16 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
artd wrote:
So if 46mp=5 fps and 38mp=6fps then why does the 5DIII at 22mp=6fps? And why is a 6D at 20mp=4.5fps?
Sorry, unless there is a more detailed technical explanation other than "the numbers work out that way" then these numbers are still made up.
Theres also a limit on how quickly they can read data off the sensor, more read channels allow faster harvesting of the sensor data but introduce other issues. Its not all about the digic raw processing power.
p.11 #17 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
The whole idea of getting much improved dynamic range by doing two simultaneous exposures centered on the same point in time seems a little messy, but may have potential. In thinking about good ways to do that, my mind diverted to a little different concept to achieve the same effect. How about keeping the large photo sites for the lower light, and interlacing another full set of smaller photo sites among the larger ones, with the smaller ones handling the brighter part of the scene. Some pattern similar to the Bayer pattern could be worked out, and you may want to think of it as something like Dual Bayer for now. You could then still use the same exposure system, and still control exposure with a focal plane shutter, and have no time skew issues or different motion blur results between the bright and dim light exposures like the different duration exposure method described in the previous posts. Obviously it would be a little complex to build, and you would loose some of the sensitivity of the low light sensor wells because they would be smaller, and you would not keep the high dynamic range all the way down to very low light levels, but those are both limitations that a high resolution "landscape and studio" camera could work well with. And obviously you double the data handling requirements to digitize data and move it around, as well as to process it in camera.
You may even be able to put both the large and small sensor wells under the same micro lens, and if so there is no need to work on a "new" pattern of layout. The same Bayer concept would be used, and you end up with two independent but simultaneous images to merge. The output could feed a RAW file that keeps the images independent, but of course you would support in camera merge to JPEG for saving and in camera image viewing. You may even want to support in camera merge to some form of RAW style image, and if you did, all the image handling after the camera processor and the post processing software support could remain almost identical to what is in use now. By properly spreading the wider range of information over say 16 bits, with the right Gamma curve support, it just may all work.
I am not trying to design such a system on this Forum, but still, it is worth thinking about. And I have no tie in or experience to any real image sensor design, but am just thinking about it from an Engineers point of view.
p.11 #18 · 46.1 MP Canon EOS-3D X To Be Announced Before PhotoPlus ?
jorkata wrote:
The Digic5+ chips are HUGE.
Here's the 1DX PCB.
One of the three small chips in the bottom-left corner is a Digic4.
You can see the difference in size for yourself.
Also, you can be sure that this time around only the 1-series will have dual Digic5+ chips.
If there is a 7DII, for example, it won't have dual Digic5+ chips for sure.
The Canon 7D PCB would have space for 2 Digic5+ if they didn't make it U-shaped. The 1DX PCB isn't mcuh higher. The assembly would be a few mm thicker, that's it.
I am not talking about a sub $2000 crop camera, I am talking about a high megapixel camera that will certainly cost more than the 5DIII. But it doesn't have to be 1D form factor.
