Benefits to Greater/Lesser MP

Discussion in 'Open Discussion' started by Cruzan80, Jun 3, 2014.

  1. Cruzan80

    Cruzan80 Mu-43 All-Pro

    Aug 23, 2012
    Denver, Co
    Sean Rastsmith
    So looking around at various camera manufacturers, it seems like several are going with 16MP as a sweet spot for overall size. I also see people talking longingly about their older m4/3 or even older cameras, with less MP. Looking around, I have heard various discussions about each. Wanting to codify it in one place, so people can understand the reasons. I am going to try and leave off sensor size arguements (except as it results to pixel density, aka. 16MP m4/3 has tighter pixels than 16MP 35mm, as they have to be packed tighter to fit on the sensor). As far as I know, there are benefits to more, but unsure of the benefits of less.

    More MP Benefit: smoother transitions between sharp areas of contrast, due to more pixels fitting in the space, thereby able to make the jump between pixels smaller. Think the original NES Mario vs current games. You are able to crop more and still retain details, since the overall limit may still be above your threshold of detail.

    More MP Drawbacks: Bigger file size. It needs somewhere to put all that information. Not needed for everyone, especially since fewer people are printing (and printing large at that).

    I know there is something relating to noise vs MP, but unsure of the relationship. I do have a technical and engineering background, so should be able to follow any of the science out there. Please add your comments and observations, preferably backed up with the why (if you know).
     
    • Like Like x 1
  2. Reflector

    Reflector Mu-43 Veteran

    406
    Aug 31, 2013
    You might want to add this on:
    More MP benefit: Sharp lenses produce sharp photos.
    More MP drawback: Not so good lenses are much more apparent in the "not so good" qualities.

    For reference, the 16MP m43 sensor has a tighter pixel pitch than a D800. By that, m43 lenses actually are insanely sharp for the tiny image circle they project. It could be said that the m43 sensor is tougher on adapted lenses than a D800 in that sense. 16MP is plenty at this level, going to 24MP on the current m43 sensors will start getting into the territory of the RX100 series in pixel pitch. The RX100s have BSI sensors which have a significant advantage... So bumping up the resolution at this point wouldn't be too useful except in satisfying a small user base at best.

    Noise performance is another story that is more sensor dependent than pixel density. Even within the same "generation" of sensors there are design decisions down at the sensor level that influence the final output in many more ways than just talking about pixel pitch and sensor size. The noise performance for the relatively small pixels on m43 sensors combined with Olympus' 5 axis IBIS was one of the factors for me buying the E-M5.

    All of that said, the current set of 16MP sensors right now are a nice set of compromises. Good IQ, highly post processable and enough resolution (I've been saying this forever to people: 4MP was enough for professionals using a D2H to make 24x36 prints. See: http://www.grafphoto.com/articles/printdogma.html ).

    Unfortunately, with the introduction of the D800E, Nikon restarted the MP wars again (Just when Canon was killing that stupid plague off years back). Now there's a weird fetishistic crowd that demands something 144MP on a 135 format AA-less sensor that shoots between ISO 5-200 (Minus pixel binning or anything nice) without understanding the implications of what they're pushing for...
     
    • Like Like x 2
  3. Cruzan80

    Cruzan80 Mu-43 All-Pro

    Aug 23, 2012
    Denver, Co
    Sean Rastsmith
    Thanks for the extras. I know that going from different generations can impact noise and things as well, just wondering why Nikon and Sony (for two examples off the top of my head) are offering both a high and low MP camera, both targeted at relatively higher-end markets. Is it sensor characteristics, and the MP just happen to go along with it, or does the MP play a part?
     
  4. Vivalo

    Vivalo Olympus Loser

    941
    Nov 16, 2010
    Finland
    More MP fills buffer/memory card faster and needs more processing power or time.
     
  5. Ulfric M Douglas

    Ulfric M Douglas Mu-43 Hall of Famer

    Mar 6, 2010
    Northumberland
    The final use is critical to how many megapixels is ideal, so the subject doesn't get any simpler. :frown:
    I like 10mpx, I like 12mpx, I also like 16mpx but my older computer isn't so quick with the 16mpx RAWs.
    I find 5mpx (E-!) too little if sharpness or cropping are factors, but for A4 people prints it seems quite alright.

    I was happy to see a Sony A7S announced with just 12mpx, whereas other customers are jumping on the 36mpx A7R ... horses for courses.
     
  6. agentlossing

    agentlossing Mu-43 Hall of Famer

    Jun 26, 2013
    Andrew Lossing
    Here's another approach to more MP. Nokia uses high-MP sensors in their smartphone cameras, which are still very small sensors compared to any dedicated camera. However, the finished file size you get when using them is scaled down, with a very smart algorithm that eliminates most of the noise and other imperfections you would normally get from cramming so many pixels into a small sensor. My Nokia 1520 has a 19MP sensor, but outputs (I think) 5MP images, which are quite simply extraordinary for phone images, and often look as good or better than images from a dedicated camera.

    I'm sure Nokia's technology is proprietary, but it would be awesome if today's dedicated cameras used the same approach, downsizing image size but smart-filtering only the bad stuff out. Imagine a M4/3 sensor with say 40MP (remember, Nokia's 1020 has a 40MP sensor, so it's not beyond the capabilities of the hardware if a two year old phone can do it), that downsampled to a 20MP image with most of the undesirable data removed by an algorithm.
     
  7. Cruzan80

    Cruzan80 Mu-43 All-Pro

    Aug 23, 2012
    Denver, Co
    Sean Rastsmith
    This is what I was referring to. What circumstances benefit from more or less MP? Not trying to troll, just see posts like this, and wonder what the benefits are for the 12, besides speed of processing/size of file. Is "not needed" a reason in this case?

    Sent from my LG-P769 using Mu-43 mobile app
     
  8. Vivalo

    Vivalo Olympus Loser

    941
    Nov 16, 2010
    Finland
    I think the physical size of the pixel has something to do with signal to noise ratio. Here are people who can answer to your question and probably will but their answer will be so long that they are still writing it.
     
  9. alex66

    alex66 Mu-43 Top Veteran

    715
    Jul 23, 2010
    In theory you could make a lower mp sensor that is more light sensitive but will contain less information, going back to film as an explanation might make it easier; If I took 3 films from Ilford Pan F, FP4 and HP5 and developed them in ID11(D76) for the standard time suggested by Ilford I would get ratings of 50, 125 and 400 ASA(ISO). Now my PanF is less sensitive to light and HP5 more so but I have to trade off information and the ability to get a grainless print for speed. The Silver grains are larger from the HP5 sort of like a 12mp sensor over a 24mp sensor, but just as sensors have improved over time film did the same, I could use Delta 400 with its tabular grain format and be able to enlarge a bit bigger as the grain was not so pronounced. So the theory is lower mp should be more sensitive but it mostly applies to the same generations of sensor tech, much like films just a pity the tonality of the Tabular grain films was never as nice to me as the classical films HP5, TriX etc.
     
  10. OzRay

    OzRay Mu-43 Legend

    Jan 29, 2010
    South Gippsland, Australia
    Ray, not Oz
    If you look at the Sony, and the Nikon 'sports' cameras, a lower number of pixels per sensor size (with today's technology) means vastly better high ISO performance. The cameras can also perform faster, because of reduced processing requirements and for those that use these cameras, they don't need the much higher resolution, as the final output is not usually used for very large prints (though are still fine for quite large prints). If Olympus put say a 5-6 MP sensor in the E-M1 (or any of the other bodies), it would have outstanding high ISO performance and you could still do an A3+ print that would look great. Swings and roundabouts with advancing technology.
     
  11. Reflector

    Reflector Mu-43 Veteran

    406
    Aug 31, 2013
    You can make a high pixel density sensor that has great high ISO performance by designing the sensor that way in the first place and utilizing pixel binning and some other tricks (Multi-ISO read, etc...) With the right software and hardware backend*, you can utilize some tricks like pixel binning and end up with a lower resolution image with a potentially nice benefit of having a little more DR.

    But does that mean it'll be as competitive as the low MP sensor designed as a high ISO monster in the first place? You'll be pushing it and I doubt you could ever get there against a dedicated design like that.

    Otherwise, taking the sensor designed for the crowd that wants to shoot at high resolution at base ISO and nothing else and trying to apply the same tricks to that? Good luck, even with resizing you'll end up with more noise. Remember that since most sensors aren't BSI, the microscopic wires do block some of your photons from reaching the photosites. The more pixels you have, the more of those you have. The low MP sensor with giant photosites? Less of those wires, way more imaging area. It might be a great sensor for shooting at base ISO all the time. It could have other characteristics that make it practically the best sensor for shooting at ISO 25 all day long. Does it mean it'll compete with a sensor of the same generation optimized for high ISO shooting in the darkness? No and nooooooooooooooooooooooooooooo.


    There's a lot of factors involved and that's why I get annoyed at the "equivalence debates" which stop holding true the moment the sensors aren't equal** (Which is practically always)

    *This is one of my complaints with Sony. They make great sensors for everyone else, I say that because the same sensor in a Sony body will generally have inferior performance than if Nikon or Pentax were to put it in their body. Why they can't is beyond me.
    **See my first point: The A7R has the same sensor as the D800E but the backend prevents it from reaching the same level of performance as the D800E. This is where the whole "but 2 stops..." etc stuff gets really silly fast.


    Nokia was the first to implement pixel binning as an integrated solution with the 808 Pureview, you can write your own code in Matlab or something and punch in a TIFF and pixel bin that. I know some people have experimented with that but they never really went far with it.

    Nikon's new sRAWs (As with Canon's current) aren't using pixel binning. They're cooked RAWs and you are better off with using your computer to post process.
     
    • Like Like x 1
  12. Fri13

    Fri13 Mu-43 Veteran

    353
    Jan 30, 2014
    That is same thing what Nikon already did with their first DSLR decade before Nokia. The down sampling isn't new idea or technology.

    Don't anymore remember the Nikon camera model first hand, might be the 1D or something but it was 2Mpix camera, with a actual 10Mpix sensor and they used 5 pixels downsampled to the final pixel to remove noise and increase sharpness.

    One of the benefit for high pixel amount by Olympus engineer is that circle of confusion isn't such a problem and can be dealt with software as the data to sharpen image is there. This Olympus presented with samples of processed and without processinf using small apertures like f/22.

    And it does work. compare 4/3 10Mpix sensor photo to 16Mpix m4/3 photo taken with a same lens from tripod distance same from sensor plane and the new sensor maintains details with small apertures much better and it isn't just because resolution.

    And with the resolutions we do have, I don't care any more at all do I use f/22 or not as the prints what I can print, I don't lose details because using apertures smaller than f/8.

    If the pixel density of the Nokia 808 but today's sensor technology and manufacturing would help m4/3 cameras, should we get it?
    With that density we would get around 60-70Mpix to next sensor, even if the camera would have a locked ISO to 100 and meant only for landscape/studio work.
    But sure, where to get lens to it what that sensor would not unresolve badly?
    Having a 70-80MB files to process would be a nightmare and who would need anything like that?

    That's why we have medium format back and pros are switching to it if they need shallower depth of field, 16-bit tonal range and resolution what small format can't deliver.

    Isn't it just funny, DSLR is having hard times because m4/3 for size, weight benefit with a IQ in same class and delivering similar AF capabilities for sport shooting. And then in medium format coming back to eat small format business in studios and fine art. If needing to carry a heavy and large small format camera set-up like D800e or A7r with external flashes etc, it is no matter to take medium format.

    It is like small format could lose the purpose to exist, as the reason it won medium format among professionals and consumers was size, weight, versatility. It comes compromise without benefits from both.

    16Mpix is more than enough. But there are those who need more and ~80Mpix in medium format is for them.

    Now what we need is someone to make a 4x5 CMOS sensor with 450Mpix what can take 2-3 shots a minute (of course offer fast shutter speeds) and we can laugh how film is finally dead and get back finding the older field cameras on shelfs

    Sometimes it just feels good to be now a m4/3 owner and swapping D800/5DIII for the m4/3 isn't bad thing at all when seeing large prints you can do with so versatile system.
     
  13. Fri13

    Fri13 Mu-43 Veteran

    353
    Jan 30, 2014
    Maybe one day we can swap sensor itself in m4/3 cameras. We already have a IBIS what moves the sensor, keeps it in place. What we only need is 6th axis compensation (Z axis) and we can even have a tilt-shift via sensor . Think about angeling shot for building and press of a button levels sensor to perspective. Or pressing D-buttons and sensor shifts to that direction.

    The 100% floating sensor would allow to replace sensor module as the requirement to lock it exactly in few microns accuracy would be gone.

    So you want to go shoot stars or landscape at night? Take a 5Mpix sensor module.
    You want to shoot portraits in studio? Get the 24Mpix sensor module.
    Want to brag around? Take the 50Mpix sensor module!

    If the module swapping could be as easy as attaching a battery grip, it would change market again.
     
  14. Cruzan80

    Cruzan80 Mu-43 All-Pro

    Aug 23, 2012
    Denver, Co
    Sean Rastsmith
    Thanks for all the information, everyone. What I am seeing is that the main reason less MP is better is a faster buffer for things like sports, where internal camera speed matters, and less in the way of each individual pixel, meaning that it is able to capture more detail at a lower ISO. Is that why we have seen the creep upwards in base ISO, or is it because everyone is in a race to give the highest usable ISO, and the low end is abandoned?

    I am really trying to understand all of this, and figured that this forum, with its technical people and lack of flamers would be a better resource than say DPR. Not trying to insinuate anything about one sensor size vs another, or one generation vs another.
     
  15. agentlossing

    agentlossing Mu-43 Hall of Famer

    Jun 26, 2013
    Andrew Lossing
    Sensors are being optimized for higher ISO performance these days, and I really do think that can be at the risk of losing noise-free images at base ISO. I'm not sure MP count is at fault there, though.
     
  16. randyspan

    randyspan Mu-43 Veteran

    312
    May 24, 2014
    New Hampshire, USA
    Randy Spann
    One thing MP count consideration, solely, leaves out of the necessary discussion is the in camera processing engine. Take for example Nikon's DX line over the years D90 up to D7100: the technology in the in camera processing engine has improved much beyond what the MP count would imply. I believe my Nikon V1 will out perform the D90 with less MP!
    In the end one must take the IQ, high ISO, etc, etc of a specific camera/lens system and decide: Is this the camera for me? Maybe the best news is there is still room for improvement. Our future Mf4/3 generations will probably perform better than what we're shooting now.
     
  17. Cruzan80

    Cruzan80 Mu-43 All-Pro

    Aug 23, 2012
    Denver, Co
    Sean Rastsmith
    I completely agree that it is one piece in the puzzle. What prompted me to start this thread is people suggesting a lower MP camera in order to solve problems, and I wasn't sure what things would be solved by that.
     
    • Like Like x 1
  18. bassman

    bassman Mu-43 Top Veteran

    680
    Apr 22, 2013
    New Jersey
    Scott
    A few things which I believe to be true, all for given generation of sensor technology:

    - noise performance is dependent on the pixel size, not the number of pixels on the sensor. Larger pixels on a smaller smaller sensor will exhibit less noise. This is because a larger pixel collects more photons, therefore the constant amount of noise introduced by the circuitry is a smaller portion of the resulting signal.

    - more pixels deliver more information and are always "better" for this reason. They don't make a poor lens perform any worse, although you may not get the improved resolution from the sensor if the lens can't deliver. They may make it possible for you to see the limits of the lens if you pixel peep. It will make no difference for a normal print viewed at a normal distance.

    - more pixels take more processing power in the camera, which the camera makers usually provide appropriately. However, if you want extreme frame rates, then you ought to select one of the cameras designed for that - which will have relatively fewer pixels, like the D4.

    - more pixels take more processing power and storage on your PC. I personally don't feel this is a big deal, but then I've never used a camera with more than 16MP so have no personal experience.

    - maximum acceptable print size is based on both the number of pixels and the final viewing distance. Sensor size is irrelevant.
     
  19. OzRay

    OzRay Mu-43 Legend

    Jan 29, 2010
    South Gippsland, Australia
    Ray, not Oz
    I think what Sony has done with the A7 series is look at the different needs of photographers and produced three, effectively, identical camera (bar sensor) that are designed for three different purposes. Each camera is quite reasonably priced, so a professional could easily have two or three in their kit for specific tasks, at a price that is possibly well below just one single-purpose camera from the opposition. That said, without a decent selection of lenses, that cameras aren't going to get much traction.
     
  20. Fri13

    Fri13 Mu-43 Veteran

    353
    Jan 30, 2014
    I that would be true then it would not matter what architecture, design and materials would be used as noise would always be same if pixel size is same.

    Proof for it is that we have different size pixels and smaller ones can generate less noise to image. Not because pixel size, but because differences in electronics through digital camera parts responsible to generate the image.

    Example, let's take a few generations older 4/3 sensor and compare it to latest m4/3 sensor. Same size but older is 10Mpix while newer is 16Mpix. 60% more pixels and still many stops better noise handling.

    Agree in some point. Even improvement to see the worse quality can still result worse image. After all it is visual side. And difference can be huge can you see pixels or is the large pixels scaled so transition would be smooth. Closer look would reveal that image is blurry in lateral but with pixels visible it can look sharper because better contrast.

    It is like so many wants a totally noise free photo, but the noise, like film grain does add sharpness and details what otherwise would had gone or look plain.


    D4 has 16Mpix like EM-1 does (16,2 vs 15,9) and D4(s) has 11 frame per second speed, and EM-1 has 10 frame per second.
    EM-1 seems to have a 8GiB RAM for image buffer. Based just for average RAW file size and then cache size before camera stops taking new photos when using extremely slow SDcards for test purposes (like 2MiB a second writing speed). The digital camera does capture image signal uncompressed but compress it before caching it. The resource impact (processing speed, memory consumption and so on battery consumption) is usually smaller when data is compressed and manipulated as such than manipulating it as uncompressed data. Meaning it is faster and more efficient to compress signal from sensor, then start manipulating that data trough compression algorithms as RAW before caching it, where to do final compression from RAW to JPEG (or kept as is) before flushing cache to storage medium like SDcard.
    6-7GiB would be enough for most situations with EM-1 but when considering all real time filters in 120Hz speed etc, there must be more than reserved for cache. So 8GiB seems logical amount as it would have About 1GiB free for everything else processing in camera than image cache.

    The point is, there is no huge impact to collect, compress and re-compress data before flushing cache to long term storage when camera has dedicated processors designed for specific kind calculations and memory is ultra fast (like DDR2 on typical computers) so it can read, manipulate and write ~300MiB a second. And that is nothing if comparing it to speeds what DDR4 or DDR5 memory can so today's graphics cards. Speed is in gibibytes in a second.

    I would quess that the EVF/Rear display are the main resource consumers. 120Hz display (120fps) a high resolution image with real time filters and same time pushing sensor data between frames or from frames to cache instead /Dev/null.

    The battery is the problem when it is smaller and as DSLR has a OVF it gives a edge.

    The impact is pretty small for conversion if it can be done background like what Adobe products do. They sample the large image to smaller one what you see on screen. Then the edits are done to the preview version in real time to areas what you manipulate. And then at background the edits are done to actual data in slower pace. And if you need larger data what isn't yet processed, it gets priority and sampled again to correct size. Very neat tricks what gives a illusion to user edits are in real time to full image while it isn't.

    The storage part isn't huge problem unless shooting a lot like 1-2k a day. Then the amount of photos becomes the problem as who has time to review, rate and tag 1000-2000 photos a day, getting to state where 100-200 would be worth to save and 10-30 worth to edit?

    Shooting a 32GB card a week, is only a 1.7 TiB a year. That is less than $100 for HDD (+backup one). And that is the result if not removing a single file.

    Time is money and amount costs way more than the file size as searching and organizing photos has largest impact to process.
    Why I wish there would be a option to bind a tag to photos in camera when shooting so you mark photo as important. Sure we can use lock function, but not exactly easy to use or same thing. But on trips it is crucial as one can go trough photos in lock good shots at every evening/pause and then hit "Card > Remove All" and every photo what is not locked gets deleted.

    I would like to know what kind prints do people order and then at what distance they really look them?

    Even when oil paints often have larger brush used than photographs has details, the brush shape and oil texture adds details. And that's why I think it is silly to avoid noise, as noise is same thing as brush shape and the texture it leaves on oil, giving a own unique detail to areas what otherwise would be often perfect gradient between colors without small details.

    Long time ago I had to do 60x100cm prints from 5Mpix camera with a soft lens and even edited with uncalibrated display at uncalibrated room light and sent the file recompressed JPEG with worse compression than best and print what I got looks nice from 2-3 meter distance where they were hanging. Even when colors are off, it was good enough unless you knew....

    It would not helped to have even higher resolution camera as the whole processing chain was what it was done in emergency situation with third class devices.