m43 sensor evolution

Discussion in 'Open Discussion' started by With_Eyes_Unclouded, Oct 17, 2013.

  1. With_Eyes_Unclouded

    With_Eyes_Unclouded Mu-43 All-Pro

    Apr 17, 2012
    Vassilios
    I was checking the DxO sensor comparison tool the other day and an interesting pattern emerged:

    1. There seems to be noo new major development in :43: sensor performance since the E-M5 was introduced. The GX7 seems to be roughly equivalent to the E-M5 and latest Pens (in fact, slightly inferior, but no big deal) and we have no reason to believe the E-M1 shall improve things very much. Improvements have been marginal; interestingly, the E-PM2 seems to have the best ISO performance among all :43: cameras, but indeed such differences are nothing to talk about. Also, any real world improvement seems to come from software upgrades in the image processor.
    2. There is a certain pattern between different formats. :43: seems to be about between 1/3 to less than 1/2 of a stop behind the best APS-C sensors in ISO performance (fairly trivial in the real world). APS-C is also better by about one bit in color depth (best sensors compared). The same pattern exists in comparing APS-C with FF. The best APS-C sensors trail the best FF by about 1 1/2 to 1 2/3 stop in ISO and 1,5 bit in color depth.
    3. Where things differentiate is in maximum dynamic range. :43: is 1 1/2 stop behind the best APS-C where the latter is not only comparable with FF but in many cases much better. Furhtermore, the best 1" sensor seems to match the best :43: in color depth and dynamic range, only being inferior in ISO performance (by about one stop).

    Cameras used for this comparison were the Sony RX100ii, Olympus E-P5, Pentax K-5iis and Nikon D800E. I don't know if this link would work for you:

    http://www.dxomark.com/Cameras/Comp...d2)/Olympus/(appareil3)/830|0/(brand3)/Pentax

    I understand this is not the whole story about final image quality: apart from the obvious lens choice issue, there are also image engine matters, etc. I'm just trying to understand how sensor size relates to certain sensor performance parameters, more specifically (max) color depth and dynamic range.
     
  2. emptysensor

    emptysensor Mu-43 Veteran

    435
    Dec 8, 2011
    Virginia
    Joe
    I think the differences you note will always be there as dictated by the physics of the sensor area available. As new sensors come on line there will be instances where some members of the group advance or retreat, but in general the members within the three groups will always be clumped together, with all three groups advancing slowly with technology. The interesting thing is what do these advances mean in practice? I looked at some shots I took with my old film camera and they were pretty terrible, much worse than even a current point and shoot. So at what point does the rearward most group (mu43) reach the point that it is more than enough for anyone, and at what point does the front group (FF) become totally needless overkill? To get back to your original comments, I'm a little concerned that the GX7 only matches the EM5. You would think that Panny could have overtaken the EM5, but I guess not. Only matching the current best with a new product doesn't seem like a great direction to go.
     
  3. dougjgreen

    dougjgreen Mu-43 All-Pro

    Jun 5, 2013
    San Diego
    Doug Green
    The simple fact is, sensor technology is a step function driven by semiconductor process technology.

    The first 2 years of Micro 4/3 didn't see any significant upgrades in sensor technology either. Then Panasonic made an incremental improvement in their first 16MP cameras, followed soon after by Olympus making a BIG step forward by moving to the Sony 16MP sensor. Panasonic has now made it's second small step, essentially matching the move Olympus made earlier when they moved to the Sony 16 MP sensor.

    But we won't see another step forward until Sony moves to it's next process technology - those moves typically happen every 2.5 to 3 years. So that's roughly a year away - and chances are that Sony's own new Full Frame and APS-C sensor lines will get those new designs onto the new process first, followed by the M4/3, perhaps trailing those by 6 months. Panasonic seems to be about 15 months behind Sony in terms of the maturity of their sensor technology, but they may catch up as Sony necessarily focuses on their own Full Frame and later APS-C products needs first.

    In the meantime, the existing process technology adds features, such as the PDAF Pixels. The next moves for Panasonic is to drive their new sensor into the entirety of their product line. The next moves for Olympus will necessarily be incremental ones as they wait for the next sensor step function.

    It is what it is.
     
  4. With_Eyes_Unclouded

    With_Eyes_Unclouded Mu-43 All-Pro

    Apr 17, 2012
    Vassilios
    I completely understand this; my question was more about the technical issues about DR (and, secondary to this, color depth) and the disparities observed between :43: and APS-C. Differences in ISO are rather scalar to the sensor size. OTOH, DR and CD are not; indeed several smaller sensors are about the same as the average of :43: sensors and sometimes comparable to the best.

    I guess my question is, why doesn't :43: perform closer to APS-C in these departments? Is it simply because of current :43: sensor design (in other words, a desing decision) or a technical limit?
     
  5. alex66

    alex66 Mu-43 Top Veteran

    715
    Jul 23, 2010
    Well I would say that 35mm sized tech reached the point of overkill with the 36mp Nikon D800 and in many ways ยต43 is at the good enough and has been since the first gen depending on what you shoot, certainly my G3 does all and more that I require from a camera.
     
  6. Repp

    Repp Mu-43 Top Veteran

    506
    Jan 27, 2011
    Oak Harbor, WA
    I'm more concerned with base iso performance than higher. For me the GF1 produced cleaner images at base iso than my GX1 or my friend's EM5 does. Now I'm actually considering downgrading back to it instead of upgrading to the GX7 or EM1, or maybe switching systems. For what I enjoy shooting, I'd much rather have a clean, grit free iso 100 or better yet 50 than a pretty usable 6400.
     
  7. With_Eyes_Unclouded

    With_Eyes_Unclouded Mu-43 All-Pro

    Apr 17, 2012
    Vassilios
    I am on the same boat actually. I would very much like to have native ISO 80 or 50 output, with extended DR and color reproduction, esp. when shooting with flash. But also because I live in one of the sunniest countries in the World; indeed there are numerous occassions I can't shoot with an aperture larger than f/2.8 with the E-M5, 1/8000th shutter would rectify this, but still no substitute for "real" low ISO.
     
  8. hookgrip

    hookgrip Mu-43 Regular

    150
    May 21, 2013
    A 2-3 stop ND filter over the lens would solve your problems with big apertures in daylight. Not the most ideal solution but it's the only thing you can do on a camera without ISO 50 or 1/8000 shutter, like the EM5.
     
  9. dhazeghi

    dhazeghi Mu-43 Hall of Famer

    Aug 6, 2010
    San Jose, CA
    Dara
    You could just as easily ask why it is that Canon's APS-C doesn't measure closer to Sony's APS-C. Bearing in mind that DXO is at best a loose approximation for real-world performance, I would argue that for the same generation and pixel pitch, performance is actually remarkably close - i.e. the DR and SNR of the 24MP APS-C D7100 vs. the 16MP E-M5 are quite comparable if you compare the DXO numbers in 'screen' mode. In 'print' mode they're lower of course, but that's because the APS-C file has more pixels. 50% more which equates to roughly 1/2-2/3 stop...

    The main thing that's notable to me is that per-pixel sensor performance for any camera, as measure by DXO, really hasn't seen any meaningful increases in the past 18 months. Admittedly they leave out the Fujis, so it's possible there's something going on there, but I think the reality is that we have already hit the point of diminishing returns, both for m4/3, APS-C and full-frame.
     
    • Like Like x 1
  10. krugorg

    krugorg Mu-43 Hall of Famer

    Jul 18, 2011
    Minnesota USA
    The only comment I would make is that I am not sure m4/3 sensor evolution has been slow compared to APS-C, for example. How many years was the 12 mpx Sony sensor used in APS-C and then how many years for the new 16 mpx sensor? Even though camera makes have claimed that sensors are new (okay, and maybe there were minor changes such as micro lenses, PDAF sites, etc.), they really have been using the same basic sensor for a few years.
     
  11. mattia

    mattia Mu-43 Hall of Famer

    May 3, 2012
    The Netherlands
    Well, up to a point. I would take a 24 megapixel 43 sensor with the same per pixel performance of the current generation as a pretty significant improvement. More megapixels with worse per pixel performance not so much, but I like having a lot of resolution. DxO is keeping the X-trans stuff at arms length (no support in their software either), so it may simply be that they don't feel they can compare the output to Bayer array sensors in a meaningful numeric way.
     
  12. With_Eyes_Unclouded

    With_Eyes_Unclouded Mu-43 All-Pro

    Apr 17, 2012
    Vassilios
    Sure, the 16Mp APS-C sensor used in the D7000, K-5, Sony NEX, etc is more than 3 years old. The :43: one used in the OM-D and GH3 is roughly 2 years old. I would expect, by now, :43: sensors to be somewhat closer to APS-C for DR and color output; I'd say closer to 13.5 stops DR and some more color depth would be more logical, for the same ISO performance.

    Concerning resolution, with 24Mp APS-C sensors available for about 3 years now, I'd also expect a 18Mp or 20Mp :43: sensor to be available by now.

    There are some third party test results published, using the DxO Analyser system which shows the X-Trans to actually perform lower than the best :43: for DR and color. So perhaps their software can't usefully measure these sensors?
     
  13. nstelemark

    nstelemark Mu-43 Hall of Famer

    May 28, 2013
    Halifax, Nova Scotia Canada
    Larry
    This is likely base process improvement. The question is when the next jump will occur.

    Per pixel performance is really the benchmark for me. If the m43 sensors are significantly behind, their added reach is negated by the better APS-C sensors.

    Put another way if you shoot the same scene with a 200 mm lens on both APS C and m43 and the cropped version of the APS C version is noticeably better than the native m43 version, then m43 is less attractive to me (all other things being equal). This assumes that you shoot a lot of telephoto of course and I do.
     
  14. With_Eyes_Unclouded

    With_Eyes_Unclouded Mu-43 All-Pro

    Apr 17, 2012
    Vassilios
  15. dougjgreen

    dougjgreen Mu-43 All-Pro

    Jun 5, 2013
    San Diego
    Doug Green
    Until the Sony and Panasonic foundries move to a subsequent generation of mixed signal silicon process, with smaller feature geometries, the sensor resolutions at any given sensor size can't be increased without decreasing the size and with it the quality per pixel. It would be unusual to see a small jump in resolution from 16 MP to 18 MP. At the next process technology jump, you'll likely most initially see a resolution jump to 24 MP (which is the same jump Sony's APS-C sensors experienced during their last jump in silicon process technology) Although it's possible that they might only jump to 20 MP while adding on-sensor features, such as PDAF pixels and more processing logic such as programmable anti-aliasing logic at the pixel level. BTW, you'll see APS-C sensors and Full Frame sensors jump in resolution either somewhat earlier or no later than at the same time - because they are driven by the same silicon process technologies, and Sony would probably target the sensors that go into their own cameras first before targeting M4/3 sensors.
     
    • Like Like x 2
  16. dhazeghi

    dhazeghi Mu-43 Hall of Famer

    Aug 6, 2010
    San Jose, CA
    Dara
    I suspect that's the explanation. That said, the real-world results from the X-A1 suggest to me that it's not so much X-Trans as Fuji's processing (and perhaps ADC) that's the cause of their apparent high ISO improvement.

    I agree that a 24MP m4/3 sensor with equal per pixel performance to the current sensors would be a significant improvement. My point is simply that we have seen no such improvement. Nor has APS-C seen that - per pixel the older 16MP Sony sensors are better than the newer 24MP ones. The last significant jump was 3 years ago or so, when they transitioned from 12 to 16MP on APS-C.
     
  17. Kiwi Paul

    Kiwi Paul Mu-43 Top Veteran

    729
    Aug 15, 2011
    Aberdeen Scotland
    Do we really want higher pixel sensors for mu43 though? The current 16Mp sensors seem fine to me in resolution terms and diffraction isn't a major issue but if they increase to 24Mp diffraction may begin to be a real issue at smaller apertures and that negates any gain in resolution by more MP's. By using the next generation sensor technologies but keeping the pixel count the same surely even greater ISO, dynamic range etc can be achieved and that would be more beneficial to us mu43 users than any resolution increase.

    Paul
     
    • Like Like x 1
  18. With_Eyes_Unclouded

    With_Eyes_Unclouded Mu-43 All-Pro

    Apr 17, 2012
    Vassilios
    Theoretically speaking (I don't have the E-M1 to evaluate these claims), Oly's latest processor is capable of diminishing the effects of diffraction, actually optimizing for the specific lens fitted.

    But, personally speaking, in some cases, I'd prefer a 12 or 14Mp sensor with ISO 50 capability and much extended dynamic range and color at base ISO. To tell you the truth, perhaps a micro four thirds FoveOn sensor might be the best deal for these purposes. Probably the system ought to expand with more specialized offerings, but there's not always enough market for them.
     
  19. nstelemark

    nstelemark Mu-43 Hall of Famer

    May 28, 2013
    Halifax, Nova Scotia Canada
    Larry
    Given the disparity of the RAW and JPEG results I would not give this too much weight.

    The JPEG results look consistent with the reviews - slightly better performance at higher ISOs, otherwise a wash.