1. Reminder: Please user our affiliate links to get to your favorite stores for holiday shopping!

We are already at the "optimal" 4/3 sensor?

Discussion in 'Open Discussion' started by hmpws, Jun 19, 2010.

  1. hmpws

    hmpws Mu-43 Regular

    177
    Apr 24, 2010
    Auckland, New Zealand
    I have just read up on digital sensor and pixel size (a bit of "leisure" reading for me, these things are certainly not new):
    Clarkvision: Does Pixel Size Matter

    It was not specifically written for 4/3 sensors, but I found it quite well written in explaining the physics behind digital sensors. It also cleared some of the myths I fell into.

    I have seen posts of people hoping for a leap in IQ with the next gen of 4/3 sensor, but as technology is at the moment, I think 4/3 sensors are at the "optimal" point already.

    Some points being:
    • Signal-to-noise ratio is directly linked to fill area for photon collection on the sensor.
    • Dynamic range is related to area and the capacity of a potential well. It is a ratio of the signal divided by the noise (mainly readout noise).
    • The pixel size at around (my guess at ~5 microns for 12mp 4/3 sensor) is at the optimum, taking noise, dynamic range and resolution into account.

    The area that can be improved on (without breaking the law of physics) from my understanding are: quantum efficiency (converting photons to electrons), fill area % and readout noise (possibly larger electron "bin").

    I don't know how close to the phyiscs limit the current 4/3 sensors are though. Either all this, or bring out some super fast lens.. but I need to read up on lens design before I understand why they can't get a f/0.5 20mm pancake out.

    In the end, is 4/3 good enough? I certainly wish it performs better in low light (say a dimly lit restaurant). Anyway, still love using my GF-1. Just throwing this out there (and maybe get a PEN with VF2)!

    P.S. I found this right after reading the above:
    http://www.wrotniak.net/photo/43/sensor-size.html
    As I understand, all things being equal, 4/3 sensors are pretty similar to APS-C sensors.
     
    • Like Like x 1
  2. Amin Sabet

    Amin Sabet Administrator

    Apr 10, 2009
    Boston, MA (USA)
    I've been reading threads for years about how one sensor or the other was as about as developed as it could be based on the "laws of physics". For example, when my Canon 5D was about a year old (2006), there were several very bright and accomplished engineers and physicists on DPReview who argued that only minimal improvements in noise were possible (citing the very same article and others). Then, when the 1D3 came along, the same fellows came out and said that read noise had been significantly reduced. Meanwhile, other new technologies have come along, such as Sony's back-illuminated sensor design.

    I'll believe that we have reached the limits of what is possible when we get to the point where a few years go by and nothing gets better. I'm pretty sure we're not there yet with MFT sensors, because some APS-C sensors outperform current MFT sensors to a greater extent than can be explained on the basis of sensor size differences. As an aside, I think the Wrotniak article minimizes the difference between APS-C and MFT sensor size by overemphasizing the vertical dimension on the sensor.
     
  3. adam

    adam Mu-43 Regular

    68
    Apr 21, 2010
    Maryland
    There's also the question of software advances. As the computing power of cameras increase, the image quality is going to get better, even if the sensor doesn't change.

    Also, the author talks about the "laws of physics" as if our understanding of the world is written in stone and can't change as we discover new things. That sorta thing sets off my BS detector...
     
  4. Mosca

    Mosca Mu-43 Regular

    103
    May 27, 2010
    "That thing will never get off the ground."

    Q: "What would happen to a man in an automobile going a mile a minute?"
    A: "That man would die."

    I remember when I was a senior in high school, oh so long ago. I was being recruited for colleges. It was 1971, and I was visiting Carnegie-Mellon (I was pretty smart back then, I've grown much stupider over the years). I remember talking to the computer science guys, and they were all excited about something they were working on. They took me into a room, and there was a 12" green screen, and a mainframe the size of a refrigerator; probably a PDP-11. And there on the screen was a little car drawn in outline, and the guy was saying, "See, when I do this, it rotates this way, and then I can rotate it that way, or zoom in...." It was an early CAD program they'd developed. And I thought to myself, "Why would I want to go here and do this? Everything's already done." (Obviously, I'd already started getting stupid.)

    I think that the sensors are optimized for the current state of knowledge, technology and economics. At the present, I believe the bang-to-buck ratio is enormously advantageous. Better ISO performance or more pixels are available, but then you couldn't sell the camera, or buy it depending on who's talking. It's not that it's optimized, it's that it's optimized for the market.

    What we can't anticipate are future needs, wants, and desires, or what future developments will happen that will create those needs, wants and desires; if future quantum research into light bending creates the possibility of inexpensive 3D photography, then we most certainly will have 3D cameras. Because that's what we do.

    For the current discussion, as an example of how sensor design might be limited by physics yet still be able to progress, I think it might be more informative to think of the system as a whole, and the sensor as only a part of that system, and then look at how the entire system might be able to take advantage of information that the sensor might not directly acquire, but that might be gleaned indirectly from information the sensor does acquire; like being able to learn about the universe from the background radiation. There is nothing that isn't information. There is no such thing as noise. Noise is a relative term, it is only noise compared to what you are looking for. Analyze the quality and shape of the noise, and turn it into known values.

    I got a little off track, but I did it on purpose, to illustrate that how one approaches an analysis influences the conclusion. Everything is done being developed, and then something completely out of left field changes the rules of the game, and development starts anew.
     
  5. photoSmart42

    photoSmart42 Mu-43 Top Veteran

    628
    Feb 12, 2010
    San Diego, CA
    I've read that article as well, and IMO it assumes stagnant sensor technology and lack of engineering creativity by sensor engineers.

    There's quite a bit of literature on the web that talks about CCD technology being quite advanced because most of the work has been done in that arena for decades, but the CMOS technology still being immature. That means there's still a fair amount of improvement left to accomplish with CMOS technology in terms of improvements in SNR. Additionally, there are lots of processes and techniques engineers are devising in how sensors are being built - advances in filters, processors, microlenses, etc., etc.

    Going beyond existing technologies, there's work being done to introduce completely new types of sensor materials which might end up becoming disruptive technologies and change the entire game.

    I wouldn't worry too much about sensor technology becoming stagnant and reaching some theoretical limit just yet. We still have a way to go. Panasonic is in full control over their future in this case because they own the factory that makes their sensors, and by all appearances they still have a lot left up their sleeves. I suspect in a few years we'll be blown away by what they can do with a 4/3-sized sensor in a GH3.
     
  6. bilzmale

    bilzmale Mu-43 All-Pro

    Sensor technology has changed dramatically in the 20 years or less of digital imaging. I'm a physicist by training but an optimist by nature. Progress is inevitable or we may as well all pack up and go home. (Yes I've had more than one red tonight!!).
     
  7. OzRay

    OzRay Mu-43 Legend

    Jan 29, 2010
    South Gippsland, Australia
    Ray, not Oz
    If anyone says that something is impossible, sooner or later someone else will prove them wrong. I remember in the early 80s or so, suggesting to an electronics engineer that one day in the not too distant future, we'll be playing music on small memory chips, removing the need for all the complex mechanisms needed for tapes and record players. He said that would be impossible, because the memory capacity for just one song would be so great that we could never achieve such memory chips. :biggrin:

    Cheers

    Ray
     
  8. Brian S

    Brian S Mu-43 Top Veteran

    714
    Apr 11, 2009
    We had Bubble-Memory in the early 80s. And Amdahl's law. What was the guy thinking!

    My boss told me "What are you going to do when your chosen Profession makes your favorite hobby obsolete?" He was referring to digital cameras making my classic Film cameras obsolete. That was over 20 years ago.
     
  9. JoepLX3

    JoepLX3 Mu-43 Top Veteran

    509
    Jun 13, 2010
    Japan
    Also High ISO performance of the expensive non-full-frame Nikon/Canon camera's is much better than Panasonic/Olympus m43. Despite they "all" use Sony sensors (and pixel density is actually not far apart at all, or m43 even lower)...

    It all is about money to made the technology challenges happen (just like soccer besides the World Cup)
     
  10. squeegee

    squeegee Mu-43 Veteran

    403
    Jan 26, 2010
    There's always advancements in technology. I will let the next 20 years show you what can be miniaturised into a 18x13.5mm square. :biggrin:

    Even right now, things like foven have shown what some different engineering designs can bring. In a way foven is exploiting the depth of a sensor and not just the surface area.

    I have also read that nikon uses sony sensors but it is my understanding that canon makes all their own dslr / large cmos sensors. CANON

    All things being equal, we should expect that aps-c sensors should always have a 1/3rd advantage over the 4/3rds sized sensors. Realistically the technology is not too far behind each other from the different companies. I think sony/canon/panasonic all have back-lit-sensors now. So, what ever the 4/3rds camp does to make the "little" sensor better, the larger sensor people can do as well and always have a 1/3rd size advantage - or panasonic could always end up being the supplier to nikon for aps-c size sensors too using the same technology...

    Having said that, I don't mean to belittle the 4/3rds sensor. There will come a point in time when the quality is so high that being 1/3rd better makes no difference to the end user (professionals always excluded in conversations). Kind of like the computer GHZ wars... you can currently buy 32 processor 4 ghz computers but really... cutting the time to open your web browser from 1/4th of a second to 1/200th of a second... *shrugs*
     
  11. JoepLX3

    JoepLX3 Mu-43 Top Veteran

    509
    Jun 13, 2010
    Japan
    Canon EOS 7D - CMOS- 18.0MP - 3.32 cm2 - 5.4 MP/cm2
    Panansonic DMC-GF1 - CMOS - 12.1 MP - 2.43 cm2 - 5.0 MP/cm2

    Which camera has best High ISO performance (despite resolution)? I thought 7D...
    - Only pixel density doesn't seems to be the performance limiting factor

    And low light performance also dependents on the speed of lens (F-number)
    - So the High-ISO rat-race doesn't make all sense to my, but although build-in EVF is more important (compactness&price compared to GF1 with EVF) as well as faster 50 mm prime lens, still I hope the GF2 will be improved on this High ISO performance aspect, that is much more important to me than 18.0 MP over 12.1 MP.
     
  12. Djarum

    Djarum Super Moderator

    Dec 15, 2009
    Huntsville, AL, USA
    Jason
    Wasn't it Bill Gates who said we'd never need more than 640k of RAM! LOL.

    I remember when I started in the current business I'm in now. I remember when production companies only needed video routers to pass 1.5gpbs(720p, 1080i). 4 years later, we were designing routers to pass 3gpbs(1080p).
     
  13. squeegee

    squeegee Mu-43 Veteran

    403
    Jan 26, 2010
    hahaha well don't get me wrong, I'm not saying cameras don't need to get better or "more". What I'm saying is... as an example

    A 300mp camera v.s. a 200mp camera... how important will the MP difference be at that point in time to the average consumer or will they just buy the cheaper one?

    Or what if the dynamic range captured 3 billion different shades v.s. 2 billion different shades? will people care anymore?

    If you want to compare it to the computer/cpu mhz wars. Just look at the last two years when no one has heard much of a peep about mhz/ghz anymore and all the craze is about the slowest machines around - netbooks and tablets... most are running at only roughly 1ghz. (Lets not get into the details that mhz doesn't actually reflect speed...) Most machines can now take up to 128gb of ram but most people are buying / using less than 4gb still.

    Anyways, back to the original point of the 4/3rds hitting it's limit, I think the back lit sensors will be quite a leap in quality over the existing sensors so I don't think we'll have to wait long to show that what's existing can't be dramatically improved with existing technology let alone future technology.
     
  14. OzRay

    OzRay Mu-43 Legend

    Jan 29, 2010
    South Gippsland, Australia
    Ray, not Oz
    The thing is 12MP or thereabouts is already giving you quality that few will ever exploit to the fullest. It wasn't all that long ago (E1) when just about everyone said that 4/3s sensors would never achieve decent image quality over about ISO400, if not less. There are imaging technologies and general improvements being developed all the time that simply overcome problems of yesterday's technology.

    Yes, anything that improves the quality of a small sensor can be equally applied to to a large one (notwithstanding IP issues), but at some point in time, the only difference will likely be MP and probably even that won't be a big deal. That's when the days of the massive DSLRs will be numbered and that day isn't all that far away, and in some respects much of it is already here.

    I can conceivably see a day when even 4/3s will be considered too big and the only limiting factor will be the practicality of ergonomics and, to some extent, lens design. Just look at mobile phones, we can make one now to fit into something the size of an average wrist watch or smaller, but it's the practicality of operating it that ultimately determines the best form factor. The same will apply to future cameras.

    Cheers

    Ray
     
  15. JoepLX3

    JoepLX3 Mu-43 Top Veteran

    509
    Jun 13, 2010
    Japan
    Dear Ray,
    Some phone / camera's are already too small to hold practically, but to my opinion over the last 50 years relatively not much optical improvements to create real shallow depth of focus control with a smaller lens. A smaller sensor actually also makes this more difficult. But on the other side "faster" sensors will in the future allow even "slower" lenses...
    In other words, in Memory / Processors business smaller transistors are economically interesting, but for Image Sensor in a camera addressing photography applications including shallow depth of focus control it might at a certain point in time not be smart to move to even smaller sensors (given it will result in more complicated optics to maintain some optical performance). I am not sure the size of m43 is close to that balance point and probably that point will continue to slight to smaller combined sensor & lenses over time as well.

    The systems used to make these Memory, Processors as well as Image Sensors are typically getting bigger and bigger, especially the optical system which determines the critical dimensions/design. So in principle smaller sensors require bigger/better optics, but as far as I can see the mirrorless design of m43 gave Panasonic and Olympus a kick-start towards mainliy smaller body & lens mounting.

    Best regards,

    Joep
     
  16. OzRay

    OzRay Mu-43 Legend

    Jan 29, 2010
    South Gippsland, Australia
    Ray, not Oz
    If I could find the article, I'd post the refrence, but there's already a process to manipulate depth of field from shallow to full on. This was demonstrated at a recent imaging technology seminar. In fact, effectively one shot allowed you to do what seemed impossible and things only get more magical year after year. Whatever anyone says can't be done, can be done, and it's happening faster every year.

    Cheers

    Ray
     
  17. JoepLX3

    JoepLX3 Mu-43 Top Veteran

    509
    Jun 13, 2010
    Japan
    Yeah, maybe/probably your are indeed 200% right, because I kind of overlooked that the digital correction for distortion is also a main contributor for smaller m43 lenses, it just takes me time to get used to all these innovations...
    - But most of these inventions are actually electronic / software leaps forward instead of optics, look at the 3D television, optic-wise it ain't "rocket"-science :wink:...
     
  18. Sam Roberts

    Sam Roberts Mu-43 Rookie

    11
    Apr 18, 2009
    I agree with a lot that has been said on this, how big a print can you produce with 12mp? I'm sure a decent, in focus low noise shot on my GF-1 could easily print at A4, yet the biggest I've printed so far is 7x5. Some of those are from my old 4mp HP compact, some from a D40 (6mp APS-C) and some from my GF-1.

    Most of the time my photos are displayed online where I generally have to downsize anyway.

    I think CMOS has a way to go, particularly when the new back lit sensors become the norm and make low light noise problems a thing of the past. Fufifilm made great advances with DR with their super CCD (I think that was what they were callled) sensors
    which doubled up highlight pixels with normal ones to capture more highlight range. Sigma's Foveons have much better per pixel sharpness and purer colour than bayer sensors, but currently suffer from problems at high sensitivities and need a fast camera to make a dent in the market.

    What I'm getting at is there's more than one way to skin a cat, if I had a an ideal sensor I suppose it'd be a back lit true 12mp (36mp/3) foveon, with high ISO performance and DR on par with or better than a Nikon D90, with Olympus JPEG output in a m4/3 Oly/Panasonic compact body with built in flash, EVF and stabilisation.

    Am I asking too much? ;0) In the mean time I'm loving having a camera compact enough to take with me that's a joy to use and has output as good as a DSLR in most conditions.
     
  19. Michael

    Michael Mu-43 Regular

    It is perpetually arguable which is the best sensor technology, CMOS and its variants, CCD or Faveon, but the fact is that Canon more than most like Kodak before them had surplus profit to put into R&D not just from its photography wing but from its very diverse portfolio of industries. There is also the crossover technology factor from developments within Canon. And as such it will always be, the company that has the most revenue and invests in itself will deliver the technological breakthroughs. It must also be remembered that Canon also probably benefitted greatly from its association with Kodak when it produced the EOS DCS series in the 90's

    Right from its very beginning the Faveon sensor has shown so much promise. The images from the DP-1 and 2 are excellent but the cameras are slow to function compared with even the the most modest P&S. Its my guess that Sigma does not have the resources to take it where it should probably be

    Nikon and Leica were in trouble a few years ago until they diversified and recapitalized. Then there is Fuji, (not forgetting the Xerox connection) now there is a sensor/software I would really like to see in a M4/3 camera but for some reason despite its technological innovation Fuji has preferred to spend its R&D money in other product areas. The sensor in the S2 and even some of the Fuji P&S cameras have brilliant qualities the like of which are not seen in other manufacturers products.

    Olympus has also had its problems but has managed to maintain a degree of independence from Matsushita (Panasonic) since its merger. I'm guessing a bit here, but it looks like this came about because Matsushita did not have the same market penetration in medical and scientific imaging which are far more lucreative than mere cameras! Olympus cameras/sensors have greatly benefitted from being associated with Matsushita.

    In the print / publishing industry bigger images are always in demand and thus the market for larger sensors. I have not come across an art director yet that has asked for smaller images… They want more to crop and more detail and so on. Every format and sensor size has its use and place, M4/3 offers a small eminently portable and adaptable system camera. Twelve MP on a M4/3 can give a good workable sized A4 print from a single frame and its ideal for web and screen. I bought an EP-2 for these reasons with the bonus of using legacy systems digitally.

    Cheers
    Michael
     
  20. LisaO

    LisaO Mu-43 Top Veteran

    798
    Mar 18, 2010
    New York Metro Area
    Lisa
    I don't think m 4/3 is at it's best yet. First thing that needs improvement is high ISO performance probably 16-18 clean (not noisy at high ISO) megapixels would be optimum.

    I think the perfect camera would be the size of the new Sony NEX, High ISO performance like a Nikon D3s, Shoot video like a Canon 5D MKII, have controls of a Canon I DS MK IV, have lens quality of Leica M lenses, lens selection of combined Nikon and Canon DSLR collections and for the price of an Olympus E-PL1.