Micro 4/3 Has m43 finally lived up to the promise?

I'd love to know the signal processing algorithm used for CMOS sensors, how much is hardware- how much is firmware/software.

The noise at base ISO is driven by "dark current", the value that a pixel gives without any light hitting it. Saturation count is the maximum value that a pixel can deliver, ie more light hitting it makes no difference. The M9 sensor is somewhere around "2" for dark current and "50,000" for saturated count. The data sheet is not available for the KAF-18500, and I am estimating that based on other datasheets published with devices in the same 6.8um detector family and comparing performance with the M8. Kodak published the KAF-10500 long sheet. So it's easy to determine the dynamic range, the ratio of dark-current to saturation. The M8 had twice the dark-current of the M9, and performance of the M9 at 2500 is almost as good as the M8 at ISO 640. Other factors such as "Quantum Efficiency" also get improved with sensor design.

I'd love to read the data sheets on the newer CMOS sensors, but they are often not published.

Now I have to admit I didn't understand one word of this...so Brian, I'll depend on you going forward to explain what this means in language used by us mere mortals... :)
 
Now I have to admit I didn't understand one word of this...so Brian, I'll depend on you going forward to explain what this means in language used by us mere mortals... :)

The difference between you and I, Chuck, is that you seem to WANT to understand. Me, I'll look at the output from a camera, work with it a little bit, and decide if I like it. I have YET to see noise at base ISO that I could really even see under normal conditions and I've never seen anything that would even remotely bother me. High ISO is a different matter but we're getting to the point where almost all new cameras with m43 sensors or larger pass my test for high ISO adequacy and the smaller sensor cams are getting real real close too. So the sensor is getting to be less and less an issue for me anymore and other features, behaviors, etc, of the camera come to the fore.

I'm glad that there are people like Brian and others who think deeply about and understand this stuff because I'm absolutely certain we wouldn't have made such great strides without them. But I'm equally glad that I'm not personally burdened by it because photography seems so much more interesting to me than sensor science. And I appreciate that they free me up to just pursue that part of it.

-Ray
 
I'd love to know the signal processing algorithm used for CMOS sensors, how much is hardware- how much is firmware/software.

It's sometimes difficult to know when there is on-chip NR/smoothing, but DxOmark seems to do a good job testing sensors, and there is a site - Sensorgen.info - that takes the DxOmark data and uses it to calculate QE, min read noise, and max saturation capacity for a variety of cameras. For the most part, I've found that the DxOmark data matches my "real world" impressions of sensor performance, and I've no reason to doubt that Sensorgen.info is doing anything wrong with the DxOmark data.
 
I think M 4/3 is a great system, I only wish that the original Panasonic sensor had seen incremental upgrades with each new M 4/3 body generation. I think the M 4/3 sensor is big enough in theory, but only recently has the actual sensors caught up to other platforms.

The Sony NEX-5N sensor is really great and something I have taken for granted- not just high ISO ability but the ability to 'torture' an image in post with exposure adjustments, fill light, etc. OTOH the NEX lens lineup is not nearly as good as M 4/3. But each platform seems to be improving in its deficiencies...
 
For me M4/3 probably has lived up to its promise, especially with the GX1 and OMD, it seems that both finally have a much improved sensor over the older Pens which is where it was lagging behind in particular the NEX.

For me M4/3 is the only Compact System which has anywhere near a wide enough lens selection to be considered as a full on replacement for a DSLR kit (save for those of you with oodles of nice legacy glass!)

IQ wise I'd say the X-Pro 1 that I've been trying for a couple of weeks is probably ahead but for me the lens lineup just isn't there (hence my decision to sell it on). I'm contemplating just going back to a DSLR again but if I don't it will be back to M4/3 as its the only setup with a really nice lens lineup for what I like to use (12, 20 or 25 & 45) which are small and relatively cheap compared to both Fuji and Sony's best lenses.

My EP2 and EP1 before that with the pancake lenses were ever present in my bag each day, easily able to be carried without any thought of them taking up space or being a hassle, thats another factor for my decision to sell on the XP1 as its just not as portable as the Pens, although that probably like the NEX has much to do with the size of the lenses.

For me M4/3 and in particular the latest bodies (GX1 / OMD) are the fullest system of mirrorless camera, I'm sure Sony will improve their lens lineup as hardware wise the NEX 7 seems outstanding, particularly if you use legacy glass, and I'm also sure that Fuji will eventually release a lot more lenses that would satisfy my wants, but for now if Olympus and Panasonic can keep improving the camera hardware then I can see M4/3 going from strength to strength as the lenses are already in place.
 
Photodetectors, CCD or CMOS, generate electrons -measures as voltage- when light hits them. They generate voltage even when no light hits them- which is noise. They "saturate" after a certain level of light hits them, and cannot generate higher voltage after a certain point. That is how you blow highlights. Voltage at saturation versus voltage with no light hitting the detector is the dynamic range. Signal processing techniques are used to smooth over the noise.
 
It's sometimes difficult to know when there is on-chip NR/smoothing, but DxOmark seems to do a good job testing sensors, and there is a site - Sensorgen.info - that takes the DxOmark data and uses it to calculate QE, min read noise, and max saturation capacity for a variety of cameras. For the most part, I've found that the DxOmark data matches my "real world" impressions of sensor performance, and I've no reason to doubt that Sensorgen.info is doing anything wrong with the DxOmark data.

Quantum Efficiency is the efficiency of the sensor in converting photons to electricity. It is the measure of total number of photons coming in, across the entire spectral response of the detector. CMOS and CCD sensors used in digital cameras are sensitive to more than visible light, because the material they are made of is sensitive to Infrared. Those photons get filtered out in most cameras, except "Full-Spectrum" like my old Kodak DCS200ir. After the IR is filtered out, about 70% or so of the Photons make it to the Color Mosain Filter. About 1/2 of them get filtered out by the color filters. Of the remaining 35% of the photons- about 90% get converted.

The QE numbers of 60% are only possible for a monochrome visible camera, or for full-spectrum (No IR filter like my DCS200ir and Full-Spectrum EP2) cameras with color mosaic filters in visible. CMOS and CCD detectors based on Silicon (Just like Star Trek) have about 30% of their sensitivity in the Infrared. This is eliminated by the IR absorbing coverglass. That and the real QE of the sensor- gives about 60% QE for visible monochrome cameras. Add in the Red-Green-Blue dye for the Bayer Pattern Mosaic filter: and QE is closer to 30% for a color camera in the visible range.

I fired up the DCS200ir and stacked two IR cut filters in an attempt to do some visible-only pictures. No where near close enough, lots of IR coming through. I will try stacking four IR cut filters and retry.

Real detector QE can be found in the manufacturer's data sheets. Kodak, now Truesense, is one of the few manufacturers that publishes them. I've been collecting them for over 20 years. Kodak CCD arrays, 6.8um family- reached 85% QE for the full-spectrum monochrome ~1999.
 
The QE numbers of 60% are only possible for a monochrome visible camera, or...

If you follow the links from Sensorgen about how they calculate their figures, the reference indicates that the calculated QE is really a relative QE. Perhaps all the values there are 50% above what they ought to be, but they are billed as being internally consistent.
 
Brian, what you are saying suggests to me that the Leica 9M may be taking advantage of the physics you describe. If that is so, then the high quality of some of the 9M images I have seen is not just due to the quality of the Leica lenses. And if that is so it suggests that other manufacturers, at more reasonable prices, could do the same if they were to see a market for a mirror-less monochrome camera.
 
A few points:

[*]While base ISO noise remains an advantage for Leica M8/9 over 4/3, the latest MFT sensors (GH2, E-M5, G3, GX1) have surpassed both the M8 and M9 when it comes to high ISO noise. At least that is my impression in shooting my bother's M9 against my GH2 and E-M5, processing all RAW files in Lightroom. My brother feels the same way shooting his M9 and E-M5.

...

For some reason that I can't fully identify, the E-M5 appeals to a whole group of people that were never taken with MFTs before. I still think the differences from GH2 or G3 are pretty modest, but the reaction to the E-M5 has been remarkable.

Amin, as someone who owns the GH2 and the OM-D, do you think that the OM-D is that much of a step up? If you had the choice, would you get an OM-D instead of a GH2 now? I know that you are probably still flush with the pride of new ownership, but what are your thoughts?

The OM-D is appealing on a lot of levels. It has that groovy and gorgeous retro styling, whereas the Panasonic cameras look like pygmies got some DSLR's and shrunk them like heads. The build quality seems to be excellent and it is weathersealed, a major first for a m43 camera. The operation is amazingly fast and really bridges the gap with DSLR's for everything except for sports/action/tracking shots, and that capacitative touchscreen shooting means that you are much less likely to accidentally trigger the camera than the GH2's resistive screen. I think that the market has become used to the idea of higher priced m43 cameras now, where the GH2 plus 14-140 kit was very expensive at the time.
 
The operation is amazingly fast and really bridges the gap with DSLR's for everything except for sports/action/tracking shots, and that capacitative touchscreen shooting means that you are much less likely to accidentally trigger the camera than the GH2's resistive screen.

This should be the other way around. The resistive touchscreen will require more effort to operate because you need to apply pressure. The capacitative touchscreen on the Olympus only requires the same light touch as a smartphone and supports dragging guestures in some circumstances (during image review for example), although sadly it doesn't recognise pinching to zoom in and out.

So far I've managed to trigger the touchscreen when it touched against my arm, and also when I absent mindedly went to brush a piece of dust off the screen. Good news is that it can't be triggered by random, inanimate objects pressing against it.
 
Nic, thanks for that. I was wondering if the skin resistance thing might be an issue with accidental shooting. It's just that I've read people say that the GH2 is fairly easy to trigger by accident if it bumps into something like hard clothing protrusions or buttons.
 
Amin, as someone who owns the GH2 and the OM-D, do you think that the OM-D is that much of a step up? If you had the choice, would you get an OM-D instead of a GH2 now? I know that you are probably still flush with the pride of new ownership, but what are your thoughts?

It's a tough one to answer. I don't have to GH2 anymore, and I am still pretty new to the OMD. Overall, I prefer the OMD to the GH2 for a few reasons:
-I really like body image stabilization so that my primes can be stabilized
-I prefer the looks of the OMD
-I prefer the flip display to a fully articulated one
-I prefer capacitive touchscreen to resistive touchscreen


A few things I miss about the GH2:
-multi-aspect ratio sensor
-better lowlight autofocus hit rate
-larger EVF
 
Brian, what you are saying suggests to me that the Leica 9M may be taking advantage of the physics you describe. If that is so, then the high quality of some of the 9M images I have seen is not just due to the quality of the Leica lenses. And if that is so it suggests that other manufacturers, at more reasonable prices, could do the same if they were to see a market for a mirror-less monochrome camera.

The quality of the M9 Monochrome images is largely due to the absence of the Color Bayer pattern filter, and the absence of the Anti-Aliasing filter. The Color filter means that interpolation takes place to produce RGB for every pixel. Ant-Aliasing filters work by blurring the image.

My M9 usually has a 1930s Sonnar on it. Sharp enough to out-resolve the Bayer filter even at F1.5.

If you follow the links from Sensorgen about how they calculate their figures, the reference indicates that the calculated QE is really a relative QE. Perhaps all the values there are 50% above what they ought to be, but they are billed as being internally consistent.

They should not call the measurement "QE"- which is usually an entry in the manufacturers spec sheet. If the numbers given are relative efficiency, the numbers below 40% are truly awful.

The KAF-18500 used in the M9 and KAF-10500 used in the M8 are from this family of detectors:

http://www.kodak.com/ek/uploadedFiles/Quantum.pdf

It also shows that detector noise is lower when the detector is cold. Bet you always wondered why pictures taken in the cold look better than those at the beach.

I found a comprehensive data sheet for a full-frame CMOS sensor.

Image Sensors

The QE curves shown are for the monochrome version of the detector.
 
A post I stumbled across on the topic, interesting read -

LensRentals.com - The Rashomon Effect and My Small-Camera Choice

I'm glad you're reading a lot and sharing links - that's the second very good read in a half hour - so THANKS! And I think he's dead on and this explains our little skirmish early in this thread - we clearly value very different things in a camera. You put a much higher premium on image quality than I do and you scoffed at the notion of "good enough", where good enough was the only salient point I can EVER see regarding IQ. Rashomon indeed! And even though the author and I came down to the same place in terms of the OMD (he rejected the X-Pro while I also chose that), our reasoning in getting there was not at all the same.

So, as I said before, to each their own...

-Ray
 
My second Digital Camera took four flight racks and sucked the generators of a P3 Orion dry to the point that the crew could not use the Coffee Maker when we were running.

My DCS200 was tiny.
 
Back
Top