April 17, 2014, 12:40:47 AM

Author Topic: High MP "stress"  (Read 6973 times)

te4o

  • EOS M2
  • ****
  • Posts: 287
    • View Profile
High MP "stress"
« on: December 12, 2011, 02:11:27 PM »
Apart from ISO (as in all recent discussions) what else do you think will very high (36 or more)MP sensor put stress on?:
Lens technology ? - how many MP (FF) would augment lens aberrations to an unacceptable level?
Production precision ? - above how many MP would you need focus adjustments in extensi of all existing, and would a  highMP sensor be more prone to optical misalignment
Diffraction ? - where is the sweet spot and can it be pushed around with technology
DR ? How would highMPs affect DR and what is the current electronic workaround to augment this (and give us more detail in shadows)

Please let's not discuss ISO and MB per shot (storage) neither FPS limitations. There are enough other threads.
I'd like to ask you to help me to overcome the fear of 36 MP  :-\ or more seriously : Canon seels to steam up lens upgrades - does this mean the existing lines would be less usable for a MegaMP sensor?
« Last Edit: December 12, 2011, 06:56:06 PM by te4o »
5D3 (04/12), Carl Zeiss ZE 21, 35/1.4, 50MP, 100MP
Canon 135/2, Sigma 85/1.4
SONY RX100

canon rumors FORUM

High MP "stress"
« on: December 12, 2011, 02:11:27 PM »

Mt Spokane Photography

  • 1D X
  • *******
  • Posts: 7703
    • View Profile
Re: MegaMP "stress"
« Reply #1 on: December 12, 2011, 05:38:41 PM »
The Term Mega MP means 1 Billion Pixels??

A mere 50MP is not going to stress my equipment much, the other factors are a much larger concern

te4o

  • EOS M2
  • ****
  • Posts: 287
    • View Profile
Re: High MP "stress"
« Reply #2 on: December 12, 2011, 06:59:13 PM »
You're right, I mean high MP and not the exact 1 billion - sorry, I changed the word for the sake of precision whereas high is not really precise. So, at or above 36 MP.
5D3 (04/12), Carl Zeiss ZE 21, 35/1.4, 50MP, 100MP
Canon 135/2, Sigma 85/1.4
SONY RX100

dr croubie

  • 1D X
  • *******
  • Posts: 1397
  • Too many photos, too little time.
    • View Profile
Re: High MP "stress"
« Reply #3 on: December 12, 2011, 07:24:20 PM »
A 36 MP FF sensor equates to a 14.5MP APS-C sensor. Near enough to the 50D's 15MP sensor. The Diffraction Limited Aperture for the 50D is f/7.5, still perfectly acceptable for landscapes, with hyperfocal on wide lenses only at a few meters.

I don't think we're anywhere near lens-tech limits yet, look at the centres of the TS-E 24, 24/1.4L, 35/1.4L, MTF figures well above what the sensors can provide (sure, edges could be better, but they're getting there).

For focus alignment, that's more a function of the aperture and not the sensor, although more MP will make mis-focus more noticeable when you blow it up to billboard size at 150dpi and inspect it with a 10x loupe.

(i had more to say, but gotta run)

High-MP is nowhere near its limit yet, i've seen a Phase-One IQ180 80MP shot, cropped to the same size as a 21MP 5D2 frame, and the detail absolutely trounced the 5D2. Maybe we'll max-out at the 50mp FF mark (more for marketing than technical difficulties), but not for a few years yet...
Too much gear, too little space.
Gear Photos

JR

  • 1D X
  • *******
  • Posts: 1244
    • View Profile
Re: High MP "stress"
« Reply #4 on: December 12, 2011, 07:51:13 PM »
I think all the recent "L" lenses wont have a problem with a 30-36MP camera, assuming this is the direction Canon is taking.
1DX, 24mm f1.4L II, 35mm f1.4L, 50mm f1.2L, 85mm f1.2L II, 135mm f2L, 24-70mm f2.8L II, 70-200mm f2.8L IS II :  D800, D4, and a whole bunch of Nikon lenses

Lee Jay

  • 6D
  • *****
  • Posts: 500
    • View Profile
Re: High MP "stress"
« Reply #5 on: December 12, 2011, 08:24:51 PM »
Apart from ISO (as in all recent discussions) what else do you think will very high (36 or more)MP sensor put stress on?:
The processing pipeline, cache and card writing speeds.
Quote
Lens technology ? - how many MP (FF) would augment lens aberrations to an unacceptable level?
Production precision ? - above how many MP would you need focus adjustments in extensi of all existing, and would a  highMP sensor be more prone to optical misalignment
Diffraction ? - where is the sweet spot and can it be pushed around with technology
Pixels don't enhance any of those.

Quote
DR ? How would highMPs affect DR and what is the current electronic workaround to augment this (and give us more detail in shadows)

Should increase it a bit, assuming read noise is kept down.

Please let's not discuss ISO and MB per shot (storage) neither FPS limitations. There are enough other threads.
I'd like to ask you to help me to overcome the fear of 36 MP  :-\ or more seriously : Canon seels to steam up lens upgrades - does this mean the existing lines would be less usable for a MegaMP sensor?
[/quote]

RayS2121

  • Guest
Re: High MP "stress"
« Reply #6 on: December 12, 2011, 08:54:47 PM »
As stated by the last poster, I am sick of the high MP vs ISO threads... how many is enough? have we not rehashed this enough? :)

canon rumors FORUM

Re: High MP "stress"
« Reply #6 on: December 12, 2011, 08:54:47 PM »

jrista

  • 1D X
  • *******
  • Posts: 3204
  • POTATO
    • View Profile
    • Nature Photography
Re: High MP "stress"
« Reply #7 on: December 12, 2011, 09:06:58 PM »
A 36 MP FF sensor equates to a 14.5MP APS-C sensor. Near enough to the 50D's 15MP sensor. The Diffraction Limited Aperture for the 50D is f/7.5, still perfectly acceptable for landscapes, with hyperfocal on wide lenses only at a few meters.


The diffraction limit is not the point at which you get "worse" quality, its simply the point at which diffraction affects the virtual image enough that it begins to affect IQ at all. Plain and simple, if you doubled the 50D's APS-C sensor from 15mp to 30mp, assuming all other characteristics were exactly the same (including noise), the 30mp would trounce the 15mp in IQ. Even realistically, where noise increases a bit with higher resolution, a 30mp sensor will still produce a better image than the 15mp sensor. The 30mp will RECORD the effect of diffraction from the lens sooner than the 15mp, but at no point will the IQ of the 30mp ever be worse than the 15mp. At worst, it can only be "as bad".

Here are some examples. Below are two images, the first taken with the Canon 5D (original) with a diffraction-limited aperture (DLA) of f/13; the second taken with a Canon 1Ds III with a DLA of f/10. I've placed red lines at nyquist rate for both sensors, the point at which recording more detail is sketchy at best, and beyond which more detail is impossible (and confusing...details below the nyquist rate are effectively random samples of spatial frequencies.) Both samples are shot at f/8. Its obvious to see that the 1DsIII produces a much better image:




Now, here are two samples at f/16 of the same two cameras. Were past the DLA now for both cameras as well, however were significantly past it for the 1Ds III, which has a much smaller pixel. Despite that, the higher-density sensor produces a much clearer image with far more definition than the lower-density sensor.




Its become a bit of a myth that capturing a photo beyond the DLA of a higher density sensor means you get worse quality than you might have gotten at the same aperture with a lower density sensor. That is simply not true. The DLA is the point at which diffraction begins to affect IQ. Until you actually try to image down to or beyond the actual wavelength of light (and we'll not get into superresolution right now), there is no point at which you can get worse quality with more resolution than you could with lower resolution for any given aperture.
My Photography
Current Gear: Canon 7D | Canon EF 600mm f/4 L IS II | EF 100-400mm f/4.5-5.6 L IS | EF 16-35mm f/2.8 L | EF 100mm f/2.8 Macro | 50mm f/1.4
New Gear List: Canon 5D III/7D II | Canon EF 300mm f/2.8 L II

jrista

  • 1D X
  • *******
  • Posts: 3204
  • POTATO
    • View Profile
    • Nature Photography
Re: High MP "stress"
« Reply #8 on: December 12, 2011, 09:07:29 PM »
To answer the OP, I don't think, with FF sensors at least, we will even reach the limits of current manufacturing technology for quite some time. From a density perspective, a FF sensor would have to have 46.7mp 4.3 microns in size to have the same density as the 7D with its 18mp APS-C sensor. Assuming we use the same fabrication technique to create a hypothetical 46.7mp FF sensor, its safe to assume that it would have roughly the same characteristics, including the noise. There are obvious differences, for one with it being larger, it affects FoV differently...but I think thats an out of scope factor here. All things being equal, we could assume that this 46.7mp FF sensor functions the same as its 18mp APS-C counterpart.

Its well known that the 7D is a bit noisier than one might expect at ISO100 when viewed at 100%. Scaled down on screen to 300dpi print size, however, and it looks fantastic, and provides more fine detail than older cameras, so the noise doesn't seem to be a deal-breaking problem. According to DxO labs, the 7D has the highest DR of any Canon APS-C sensor. Its rumored that Canon newest L-series lenses can currently resolve about 45mp of FF resolution, which puts the 7D right at the limit...its capturing all the detail it can from current lenses. Lens aberrations are an optical phenomena and projected in the virtual image. Assuming you print at a consistent size, say A3+ 13x19", the 7D should not have any affect on optical aberrations at all, while concurrently producing clearer, sharper details at the same print size as a lower resolution sensor. Scaling for preview on screen, say via the web, generally requires down-scaling images to a small fraction of their original size, and there should be no reason for concern of any kind regarding optical aberrations, noise, or any other possibly undesirable artifact.

Back to our hypothetical 46.7mp FF sensor...all the benefits of the 7D for realistic scenarios, like printing A3+ or scaling for the web, should be as good or better now. One could also choose to print twice as large, say 24x36 or even 30x40, and expect the same amount of clarity and detail, without any additional enhancements to optical aberrations beyond what the larger print size might mean. This hypothetical 46.7mp sensor is still not significantly outresolving the best glass on the market today, either. Simply put:

A 46.7mp FF sensor shouldn't "stress" anything from an IQ standpoint beyond any level we have today (although it might indeed put stress on frame rate.)

That would mean a 36mp FF sensor also shouldn't stress anything, and should actually do very well considering it has about 10.7mp of headroom until it hits the limits of anything as of today. Given that the 5D Mark II has some of the best DR of any Canon sensor, and its only marginally better than the 7D, one could assume that a 36mp sensor would have even better DR, possibly by quite a margin, over the 7D. This is nothing to say of the kinds of design and manufacturing improvements that have apparently been realized in the 1D X's 18mp FF sensor, and in many new Sony sensors of a variety of sizes up to FF.

This is all assuming that other technology, such as optics and sensor fabrication, don't also improve right along with increases in MP. Were moving along with optics, getting there but not really close to the quality of a hypothetical diffraction-limited "perfect" lens. The effect an aperture has on diffraction is an aspect of the lens, and it affects the maximum spatial resolution of the virtual image the lens projects. Thats an abstract concept, independent of other factors, and is the same regardless of what camera you may use that lens on. (NOTE: Don't read that to mean that the sensor does not affect IQ of a whole camera system...it most definitely does. The sensor just doesn't have anything to do with how an aperture affects IQ in a virtual image.) Assuming Canon improves lenses right along with MP, and goes beyond the current 45mp cap. So long as the sensor is not significantly outresolving the lens, we can continue to see benefit from more megapixels. I say significantly, because there is something to be gained by outresolving the lens a little bit, as it can help improve moire of fine repetitive spatial frequencies as you approach the nyquist rate of a sensor.

Taking the concept of outresolving the lens farther. The argument against more MP (excluding the anectodal statements that indicate that outresolving the lens immediately means bad IQ) is that once you do outresolve the lens, more MP doesn't offer any benefit. Generally speaking, this is true, but there are a few ways it can be beneficial. From a physics standpoint, there is a hard wall to resolution (again, ignoring superresolution): imaging at or beyond the wavelength of light. Its light were imaging, and the smallest spatial resolution you can resolve in an image would be the wavelength of light (which medians at about 550nm, but ranges from around 340nm to 780nm across the whole color spectrum.) Physical brick walls are a LONG way off, however, way beyond the scale of resolution we can realistically talk about off into the relatively near future. Assuming we take sensors to twice the maximum of current canon lenses...90mp. Were WAY outresolving the lens now...so whats the benefit? For one, you can pretty much eliminate color moire, which is a problem that exhibits with bayer type sensors (where there are alternating rows of RGRG and GBGB pixels.) Second, you could utilize the pixels on the sensor more effectively. Normal bayer demosaicing interpolates the intersections of every RGBG 2x2 pixel quad, which reuses overlapping sets of pixels to produce many RGB output pixels. This effectively reduces the color resolution of the final image, although it maintains full luminance resolution, not to mention that you start out with half the pixels for red and blue as you do for green, so your already a little color anemic. With a sensor double the resolution the lens can project, you can now use one full bayer quad for each single RGB output pixel. Eliminate interpolation entirely, and use a full constitute of color information for each and every output pixel. This has the added side effect that you are now averaging four bayer pixels, and all the noise they contain, into a single RGB pixel, so noise should be greatly improved. (As it turns out, Canon already does this in their cameras from the last couple years. They call it mRAW and sRAW, or medium raw and small raw. Both image formats use double or quadruple the bayer pixel information per output pixel to produce a much cleaner, less noisy result...albeit at half or quarter the native sensor resolution.) Outresolving a diffraction-limited lens may not be a horrid thing if we can make better use of the sensor pixels. Image output resolution would still contain the same amount of resolution that the lens projected, so you have to start questioning what it actually means to outresolve the lens. Could we use a 180mp sensor, and use 16 bayer pixels per RGB pixel to improve quality even more? To normalize noise even more? The 1 micron or less pixels of high resolution point and shoot cameras seems to be a problem at very narrow physical apertures (a millimeter or two). I am not sure if thats a physics problem, or a "cheapskate" problem, wherein the technilogy in P&S cameras are simply not up to snuff.

The only realistic stressers I can see with higher resolution sensors is image readout rate and file size. Readout rate may be something that may already addressed, as Canon prototyped a 120mp APS-H sensor earlier in 2011 that they proclaimed had a good readout rate. They did not state any explicit framerate, however its probably safe to assume 2-4fps, which is a pretty normal rate. The rate at which images can be saved to memory is also a factor, however soon 300mb/s CF\SD cards should be finding their way into photographer hands, which will help address that problem. That leaves file size, which itself is also something that can be addressed with faster image processors and more advanced compression algorithms, as well as cheaper space.
« Last Edit: December 12, 2011, 09:13:07 PM by jrista »
My Photography
Current Gear: Canon 7D | Canon EF 600mm f/4 L IS II | EF 100-400mm f/4.5-5.6 L IS | EF 16-35mm f/2.8 L | EF 100mm f/2.8 Macro | 50mm f/1.4
New Gear List: Canon 5D III/7D II | Canon EF 300mm f/2.8 L II

wickidwombat

  • 1D X
  • *******
  • Posts: 4034
    • View Profile
Re: High MP "stress"
« Reply #9 on: December 12, 2011, 09:21:22 PM »
that interesting so are you saying that mraw on the 5D2 works like the EXR sensor of the fuji X10?
APS-H Fanboy

dr croubie

  • 1D X
  • *******
  • Posts: 1397
  • Too many photos, too little time.
    • View Profile
Re: High MP "stress"
« Reply #10 on: December 12, 2011, 09:39:19 PM »
Its become a bit of a myth that capturing a photo beyond the DLA of a higher density sensor means you get worse quality than you might have gotten at the same aperture with a lower density sensor. That is simply not true. The DLA is the point at which diffraction begins to affect IQ. Until you actually try to image down to or beyond the actual wavelength of light (and we'll not get into superresolution right now), there is no point at which you can get worse quality with more resolution than you could with lower resolution for any given aperture.

Don't get me wrong, I use my 7D (dla=6.9) to take landscapes at f/8-11 or so, and macros up to f/14, and I'm definitely someone who'd rather a high-mp than high-iso (but can't we just have both a la d3s/d3x?). My general point was that if a DLA of 7 can take decent shots up to f/11, a 40-50MP FF can take decent landscapes up to f/8 or higher, and you're right, will get a lot more detail than a 21MP 5D2 with DLA of f/10 set at f/8.

Still, I know there's a lot of people who don't understand that, thanks for the good explanation. In short, diffraction of FF sensors is something we don't even need to think about until we get above 50MP (at the rate things are going, probably not within the next 7 years).
Too much gear, too little space.
Gear Photos

jrista

  • 1D X
  • *******
  • Posts: 3204
  • POTATO
    • View Profile
    • Nature Photography
Re: High MP "stress"
« Reply #11 on: December 12, 2011, 09:42:36 PM »
that interesting so are you saying that mraw on the 5D2 works like the EXR sensor of the fuji X10?

For as much as I understand EXR (or possibly their prior generation of sensors...I think EXR just rearranges the pattern of red, green, and blue pixels, which without considerably greater resolution, doesn't really do the same thing as mRAW), they address different concerns. Fuji sensors include an additional type of pixel, smaller luminance-only pixels as well as full-sized color pixels. The goal is to increase DR by capturing more light. If I had to compare this with other sensor tech, I would say gapless microlenses solve the same problem...they aim to capture into a photodiode as many photons as possible to maximize DR. Back when they first designed it, Fuji's extra photodiode was a novel idea, however I think gapless microlensing is more effective these days.

In contrast, mRAW aims to make more effective use of the available pixels. Its a digital thing, rather than a physical thing. Normal bayer interpolation makes redundant use of the RGBG pixels in a bayer array. Its not particularly obvious, but the pixels in a bayer sensor are not directly converted into RGB pixels. Rather, the intersections between each RGBG 2x2 pixel quad in a bayer sensor are interpolated to produce a single RGB pixel. Since intersections occur after every bayer pixel, there is some overlapping usage. Its effectively the same as doubling the size of an image in photoshop with nearest neighbor filtering. The actual mechanics of mRAW processing are fairly complex, but for all intents and purposes, it doesn't do any of that overlapping. Every RGBG quad is combined to form one and only one RGB output pixel...there is effectively no overlap (in reality, there is a little bit of "overlap", since mRAW is a variant of YCC, but the final result is still superior to natively resolution bayer interpolatio.)
« Last Edit: December 12, 2011, 09:45:20 PM by jrista »
My Photography
Current Gear: Canon 7D | Canon EF 600mm f/4 L IS II | EF 100-400mm f/4.5-5.6 L IS | EF 16-35mm f/2.8 L | EF 100mm f/2.8 Macro | 50mm f/1.4
New Gear List: Canon 5D III/7D II | Canon EF 300mm f/2.8 L II

wickidwombat

  • 1D X
  • *******
  • Posts: 4034
    • View Profile
Re: High MP "stress"
« Reply #12 on: December 12, 2011, 09:55:48 PM »
so does mraw give lower levels of high iso noise? additional DR then, what are the direct benefits of shooting MRAWor sRAW over regular RAW?
APS-H Fanboy

canon rumors FORUM

Re: High MP "stress"
« Reply #12 on: December 12, 2011, 09:55:48 PM »

jrista

  • 1D X
  • *******
  • Posts: 3204
  • POTATO
    • View Profile
    • Nature Photography
Re: High MP "stress"
« Reply #13 on: December 12, 2011, 10:00:53 PM »
so does mraw give lower levels of high iso noise? additional DR then, what are the direct benefits of shooting MRAWor sRAW over regular RAW?

The benefits include lower noise, better edge definition, greater clarity, etc.
The drawbacks include lower resolution (50% native for mRAW, 25% native for sRAW), and the fact that its not a "true" RAW image. Its actually a YCC image with 4:2:2 pulldown, storing full detail luminance data and interleaved Cyan/Yellow+Green/Magenta color layers.

I'll see if I can get some comparison shots to demonstrate. I was going to start making some diagrams of bayer pixels, but I think real photos will help demonstrate the concept. I'll try to get something by tomorrow night, so I have time to hunt for a variety of subjects that can demonstrate the effects for different purposes.
My Photography
Current Gear: Canon 7D | Canon EF 600mm f/4 L IS II | EF 100-400mm f/4.5-5.6 L IS | EF 16-35mm f/2.8 L | EF 100mm f/2.8 Macro | 50mm f/1.4
New Gear List: Canon 5D III/7D II | Canon EF 300mm f/2.8 L II

wickidwombat

  • 1D X
  • *******
  • Posts: 4034
    • View Profile
Re: High MP "stress"
« Reply #14 on: December 12, 2011, 10:06:00 PM »
thanks all very interesting info
so for example shooting an event in low light with a 5D mk2 and going to mRAW will give me equivalent resolution to say my 1Dmk3 however i could bump the iso up to 6400 and get lower noise images than if i was shooting at 6400 in full RAW? i usually wont shoot above 3200 if i can help it but if mRAW opens up 6400 as "usable" that would be pretty nice even with a loss in resolution.
APS-H Fanboy

canon rumors FORUM

Re: High MP "stress"
« Reply #14 on: December 12, 2011, 10:06:00 PM »