Which is best for low light? Sony NEX 5N or NEX 7 and A77?

63

Comparison

(Click the ISO 12,800 sample image above to enlarge to 1:1). Original samples from http://www.imaging-resource.com

Will Sony’s decision to put 24MP in the NEX 7 and A77 backfire? It certainly seems so.

The mid-range 16MP NEX 5N has by far the better image quality at high ISOs despite being half the price. It’s the Fuji F31fd of DSLRs!

As we have already seen the NEX 5N trounces competition like the more expensive Olympus EP-3, and its Panasonic rival the GF3. It’s also cleaner than the GH2 and G3 and all of Canon’s APS-C DSLRs like the 60D and 7D.

Now for the first time thanks to Imaging Resource we have test samples of the same scene from the NEX5N versus all past cameras, like the 5DMkII and it is interesting to compare. Unfortunately their website does not have a comparison engine like DPReview do, so I have done my own in Photoshop.

Image quality at ISO 12,800 goes like this:

  1. Sony NEX 5N
  2. Canon 5D Mark II
  3. Canon 60D / 7D / 600D
  4. Sony A77 / NEX 7
  5. Panasonic GH2.

One thing to keep in mind is that these are stills and show the quality of the sensor and image processor but that might not 100% translate into video mode. Certainly with the GH2 that is the case. The GH2 has the best resolved and least compressed video of all the cameras listed above. The A77 test video samples so far have all been done by stills orientated sites and they only give a very rough outline of what video looks like, but certainly none of the new Sony cameras match the GH2’s clean video image in terms of moire, resolution, compression and aliasing. Although it will be interesting to see if the NEX 5N’s incredible low light performance translates well from stills to video shot at ISO 6400 and above – something the GH2 isn’t capable of.

In JPEG mode the NEX 5N is slightly cleaner than the 5D Mark II at ISO 12,800. Even though the 5D Mark II has a resolution advantage of about 5MP and a full frame sensor in actual fact the difference in detail and noise is very small. This is a remarkable performance from a $599 APS-C camera versus the unrivalled full frame monster from Canon, albeit a very old one in great need of an update.

The A77 and NEX 7 sensor is however rather disappointing. One saving grace is that the video mode might be better than the NEX5N but it certainly doesn’t look that way from the footage I’ve seen so far.

I think (and I hope) what Sony’s strategy is here is that they believe through market research that semi-pros put the emphasis not just on high ISO performance but a broad range of features, including handling, EVF and resolution up to ISO 1600 where the A77 certainly packs in the details relative to the 16MP and 18MP competition. The mid-range consumer that the NEX 5N is aimed at meanwhile is probably more excited by clean high ISO performance in the reviews they see on the web and like to pixel peep and benchmark the camera against others they can afford in the similar price bracket. Sony also knew that their closest rivals to the NEX system from Olympus and Panasonic were using much older 12MP sensors so there wasn’t any pressure to bump up megapixels to compete, more a desire to see those systems demolished in the high ISO stakes because that is not really the strong point of Micro Four Thirds (G3 and GH2 aside).

Do keep in mind that below ISO 1600 the cameras here are all much closer for stills and especially in video mode. In fact the GH2 has the best image in video mode even at ISO 1600 with Vitaliy’s hack giving noise a very nice fine film like grain.

Meanwhile here are some more 1:1 crops from the ISO 12,800 JPEG samples, with noise reduction set to low. They are quite revealing of the NEX 5N’s prowess in resolving the same amount of detail as the 5D Mark II yet with less noise, even though it is a 16MP APS-C camera versus 21MP full frame. Even if video quality turns out to be a moire ridden disappointment I am certainly going to pick one up for stills.

Canon 5D Mark II

5D Mark II

Sony A77 (like NEX 7)

Sony A77

Sony NEX 5N

Sony NEX5N

Panasonic GH2

Panasonic GH2

Canon 60D

Canon 60D

Disclaimer

Remember EOSHD’s Fallacy of Pixel Peeping… It only gives you a scientific benchmark not an artistic one to base your purchasing decision on. If possible before committing try the camera out yourself. Watch the video mode projected or displayed on a high quality television to determine real-world image quality. See your photos framed and hung on a wall, and handle the camera yourself to realise its potential. Never take pixel peeping and reviews as the last word. What is right for one person may not be right for you!

About Author

British filmmaker and editor of EOSHD, Andrew works in Berlin on his own self funded filmmaking and video projects.

63 Comments

  1. I found myself so excited about the A77, but now I’m strongly considering the VG20 (same sensor as the 5N), mainly because I’m a part time wedding videographer and love the form factor. I could always purchase a NEX-5N as a B or C cam.

    I absolutely love the video sharpness of my wife’s GH2 (still unhacked), so I’m hoping that the VG20 can at least come close. If so, I’m sold. I guess I’ll just have to patiently wait for some worthwhile sample footage. I find it utterly stupid that Sony releases promotional videos (“Northern Territory”) at 640×360 resolution.

  2. It should be remembered that the A77 is an SLT camera and, as such, will lose between a third and a half a stop of light. The NEX7 should perform better than the A77 on low light in the way that the A580 betters A55 using the same sensor. It’s therefore too early to produce conclusions on the NEX7 low light performance – a half a stop (or the ISO equivalent) makes a difference when it comes to pixel peaking.

  3. 5DMkII wins hands down based on your samples with more details. The NEX 5N is cleaner but less detail and smaller. If you downsize the 5DMkII to NEX 5N or upsize the latter to match the former the 5DMkII looks even better.

  4. I wish you would not state the A77 and NEX-7 together as most people do. Nobody seems to admit the fact that because a portion of the light entering the A77 is directed to the AF sensor, there may be a sensitivity decrease. I think it is shown here when comparing the NEX-5n to the A77. Despite the fact that nobody will admit that the mirrorless design may be superior in noise to the more expensive models, it is just common sense that the extra gain that is required to make up for the light loss from the semi-transparent reflection will increase noise over the NEX-7 design. We will have to wait and see, but I am pretty sure the NEX-7 will perform a better noise test than the A77. Though the two cameras may have the same sensor, they have different amounts of light reaching the sensor, they should not be grouped together as it is very misleading to readers.

  5. You’re incorrect. Same sensor and same processor = same ISO 12,800 performance. The SLT design effects aperture and exposure not noise or detail, and the effect is very small indeed.

  6. You are the one that is wrong – the SLT takes light away, and the sensor gain is turned up on all the SLT cameras to account for the light loss. That means when an A55 is shooting at (say) 1,600, the actual sensitivity of the sensor is more like that of 2,400 ISO on an A580. The extra sensor gain amplifies all the noise (and noise floor) and hence the A580 has about half a stop better DR and noise performance than the A55 using the identical sensor.

    If you doubt this, look at the DXOmark comparison figures for the two cameras. The difference is simply the SLR mirror taking away between a third and half a stop of light. It is blindingly obvious from the line displacements between two cameras with the same sensor – the A55 is worse as the gain is increased to compensate for the loss of light.

    http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/%28appareil1%29/675|0/%28appareil2%29/685|0/%28onglet%29/0/%28brand%29/Sony/%28brand2%29/Sony

  7. I have seen images from imaging resource at ISO1600. Compared between A77, Nikon D7000 and Canon 7D. Noise was very close to D7000 and better than 7D. Amount of detail was better with A77.
    Comparing at ISO12800 for me is irrelevant. I never go above ISO1600.

  8. Here is the 1080P version of the NEX-VG20 promo.

    http://www.youtube.com/watch?v=OX26IWubKbU

    Looking good :-)

    Many folks, including myself, will pay extra for all the convenience features of a video camera: external audio with manual levels, zebras, ability to use larger batteries, unlimited recording time, ergonomics, etc.

    Still, very excited to see that there is a worthy covert shooting option in the NEX-5N. Could be killer, especially with a good pancake lens!

  9. But you are not showing my proof guys. That DXOMark does not show what you are claiming. On the ISO chart the cameras are pegged. In my view there is no way Sony would boost the ISO to compensate for the SLT design. Yes you lose a tiny little bit of light, but the sensor and image processor specification is IDENTICAL to the NEX 7 and calibrated at the same ISO levels. DXOMark shows that.

    Again show me proof and I’ll stand corrected, no problem.

    Cheers!

  10. A little suspect. Now the much loved GH2 has suddenly stopped out performing the other cameras?

    What I’m seeing is a noticeably different sight line NEX 5N, and focal isues on every other camera.

    I’m in the market to buy my first DSLR for video, purchasing in mid-Sept. so this is increasingly frustrated from a pre-buy standpoint. I know EOSHD has chosen this new camera as their next buy, but now it seems like this blog is advocating for it.

    The other curious thing is, I’m seeing a LOT of degraded image, moiring, stepping, ghosting, to an unacceptable degree in the video coming out of these new Sony’s.

    Reviews are ignoring this entirely.

    I’m not seeing clean video as I have with the GH2. What’s going on here? I’ll assume for now, the Emperor has clothes, and it’s just me. For now.

  11. Hi! Just wondering if you can switch the 5N between PAL and NTSC mode?

    I noticed the PAL version only shoots at 25p and 50i/p while NTSC shoots at 24p and 60i/p.

    Anyway.
    Wonderful review. Thanks. Had my eye on the 5N for quite some time now.

  12. We don’t really know exactly what video is like because it hasn’t been in a filmmaker’s hands yet. I’d wait until the camera is actually out before making a purchasing decision. Don’t pre-order it.

    Wait until my review in mid September.

  13. Thanks. Will certainly do.

    I guess the GH2 stills never really won comparison tests anyway, and it was just video right?

  14. You are incorrect. When A55 was released Sony themselves stated that sensor gain is boosted to compensate for lost light. Thus, (slightly) more noise on A55. About 1/2 stop. You can check that yourself by looking at F stop and shutter speed of A77 and compare that with other cameras. They should be the same, but the sensor gain is boosted.

  15. That’s too bad. I heard you can switch the region on the NEX-C3. Although I haven’t confirmed it personally myself.

  16. I would almost agree with Simco, to me it seems the 5D is better, BUT, i think that’s due to it having what seems like bugger -all NR allowing more detail. The 5N seems quite crushed in details from NR, in saying that, if NR was off it seems there wouldnt really be anything in it!
    Very impressed how such a little camera performs on par with the mighty 5D.

    Now all it needs is a shitload of lenses to compliment.

  17. It is a very well fact ever since A55 came out. Check any camera sites and information regarding the issue is there. It’s been mentioned over and over. It’s not something that is hard to understand either. For its AF module to work, it needs to grab that light that could’ve gone to the sensor. (BTW, that 1/3 of light loss is told by Sony themselves.)

    ISO boost algorithm has been confirmed by many photography sites and also by users. If it’s not boosting ISO, how do you think it compensates the loss of light?

    Also, ISO boost has been done to death by all the other companies. I find it surprising that you said “In my view there is no way Sony would boost the ISO to compensate for the SLT design. Didn’t you know It’s a common practice among sensor makers, not just Sony? Look at Canon 450D, thanks to its low light gathering ability of each individual pixels, it loses as much as 1 full stop of light when combined with large aperture lenses (I think it was 50mm f/1.2.) How does it compensate such loss? ISO boost.
    Even more interesting is that it also affect quality and size of Bokeh. LL wrote a nice article regarding this matter. Go check it out.

    So, when people first heard A55’s use of such mirror system, it got people worries. 30% of light loss is not huge but it’s not tiny either. However, thanks to superb sensor, the impact was minimal. However again, it seems like it’s become a barrier with this high density sensor.

  18. There has been an update from firmware v0.58 to 1.00 for the Alpha 77. The difference is immense in many cases. Pictures presented here are from v 0.58. Would be very interested to see the results from firmware version 1.00.
    So if possible: please.

  19. Sorry, but “serious” reports or comparisons like this one, have to make their data and camera settings public for judging the conclusions of the authors. Otherwise, this kind of comparison is symply worthless.
    Could you please publish the camera settings, that you have used for this pictures? Are they OOC? etc.

  20. Of course they are pegged on the ISO chart. That’s because the gain on the A55 sensor has been increased to compensate for the light loss. It’s precisely that which causes the relative deterioration on the DR & Noise figures which trail the A580 by about half a stop. There is no such thing as native ISO on sensors – there are just a set of sensor parameters such as the sensor gain and the calibration used in the firmware (and passed to RAW converters) to approximate to ISO levels on film. In fact virtually all camera manufacturers cheat slightly on that – hence measured ISO levels are often a little lower than the camera ones – it makes the results look better if a camera set to 1,600 was actually about 1,200. (That’s what the DXOmark ISO graphs show). The DXOmark sensitivity figures are shown with the SLT mirror in place. It is therefore mathematically

    I’ll quote from an Imaging Resources review of another SLT camera – the A33

    “An increase in sensor gain to compensate for the loss may explain why we see slightly higher noise levels from the A33 versus the A560, which share the same sensor.”

    http://www.imaging-resource.com/PRODS/AA33/AA33A.HTM

    It helps to know something about sensor technology, the physics and the mathematics of this.

  21. So you agree with me that at ISO 12800 the 5DMkII has better IQ than NEX 5N? Or do you think higher in camera noise reduction tantamount to better IQ? I can easily make the 5DMkII image look like the NEX 5N but not the other way. In another word I can take away what is there in the Canon image to match the Sony but I cant give back what is missing in the Sony to make it look like the 5DMkII.
    We are not talking about prices here!

  22. Just to add to the evidence, there is one guy who was brave enough to remove the SLT mirror from his A55 and do some comparison tests :-

    http://thesybersite.com/sony/a55/index.htm

    Note that when he photographed without the SLR mirror he had to increase the shutter speed to keep the exposure the same. For example, in the first street scene he shot the SLT mirror removed at 1/4000s and the same shot with the SLT mirror in place at 1/3200s. In other words, with the SLT removed the camera was behaving as if it was operating at a higher ISO level.

    Indeed, he goes on to say that all his tests where consistent with the SLT mirror being responsible for about half a stop of light loss and the sensor gain is turned up to compensate.

  23. So will the half stop gain in low light performance on the NEX 7 be enough to bring the A77 sample above in line with the NEX 5N? If so then we have a story. I just don’t think that is the case here.

  24. What set of DXOMark numbers? DR, tonal range and resolution of course it has the chance to out perform the NEX 5N and it should do because it is positioned higher in the range, but on ISO? No chance.

    Basic physics of light capture, smaller photosites capture less and can hold less charge.

  25. I’d say it was very close. 5D has marginally more detail at 1:1 because it is a 21MP camera not a 16MP one so you are viewing a magnified image at 1:1 relative to the NEX 5N. Size them both equally to 16MP and the NEX 5N looks cleaner with a very similar level of detail, but less noise.

  26. Probably not but the case is this:

    Your saying “there is no way Sony would boost the ISO to compensate for the SLT design.”
    “Same sensor and same processor = same ISO 12,800 performance. The SLT design effects aperture and exposure not noise or detail, and the effect is very small indeed” is wrong.

  27. If you donwsize the 5DMkII image to 16mp it will become even more sharper and more solid than it already is at 1:1. On the other hand if you upsize the NEX 5N to 21mp its falls apart even more. Even if you add noise reduction to the 5DMkII to match the NEX 5N you still have more detail than the Sony whatever res you set it to so the 5DMkII has the best IQ in those test. From your sample the NEX 5N is much closer to the 60D if the latter has similar NR applied and downsize to 16mp.

  28. Does anyone Know if the A77 will autofocus in manual video mode? I read a comment in another forum that it will only autofocus when in P mode!
    thanks

  29. I’m getting conflicting info on this PAL/NTSC switching thing. I’m pretty sure the TV output is switchable, but in regards of the regarding format, some sites claims it’s not and some claims that it is, like from the imaging-resource for example. http://www.imaging-resource.com/PRODS/NEX5N/NEX5NA.HTM says you could switch the video format from PAL to NTSC:

    QUOTE: “The NEX-5N’s Full HD video is recorded using AVCHD Version 2.0 compression, with Dolby Digital (AC-3) audio, and a wide range of progressive-scan and interlaced frame rates are available. When set to NTSC mode, the available progressive-scan rate are 60 fps (28Mbps) or 24 fps (24 Mbps or 17 Mbps), and you can also opt for an interlaced 60 fps (24 Mbps or 17 Mbps). If you switch to PAL mode, the options are the same, except that the 60 fps rates are replaced by 50 fps equivalents, and the 24 fps rates by 25 fps ones.”

    Can I ask for a favor for you to confirm this with sony or something? Would be greatly appreciated. :3

    Thanks in advance. And sorry to keep bugging you about this. I live in a PAL region and is wondering if sony is going to screw us with this 24p thing again.

  30. Higher NR? More like less noise to begin with.

    Yes the above samples are JPEGs but NR on all cameras is set to the lowest possible setting. Can’t open the NEX 5N RAW stills on my Mac unfortunately, new kind of format not supported by Camera Raw.

    Of course 5D has edge on resolution it has a 21MP sensor versus 16mp on the NEX 5N. More important issue here is does one need 21MP? I certainly don’t.

  31. Hmm IR are no video experts I’d take that with a pinch of salt. Who knows? Camera not out yet, wait for my review.

  32. Actual ISO 12,800 will look much worse in low light, with poor illumination and these test samples are shot under bright studio conditions, a continual folly of many reviews today.

    Noise reduction is turned to ‘LOW’ on both cameras in the samples above, so the 5D is performing noise reduction and smearing detail as well yet the noise level is higher.

    The difference in noise between the 5N and 5D will be larger in dim conditions. CMOS sensors behave totally differently with the lights on.

    Yes sure it is also a case of 21MP versus 16 on the 5N so the 5D does resolve very slightly more detail but there is no getting away from the fact that you A) don’t need it and B) the 5N is less noisy to start with even before moderate in-camera NR is applied.

    In fact other than the full frame sensor for fast wides the 5D only has one other advantage over the 5N and that is if you are printing a billboard size graphic with lots of fine detail that will be viewed on a wall within 30cm of the viewer. An unlikely scenario.

    Is 16MP not enough for you? It is for me. In every practical reality known to photographers the resolution from both cameras will look the same and certainly the NEX will look cleaner because it has less noise to begin with before applying NR.

  33. Only in P mode since AF needs aperture control apparently – in other words AF doesn’t function very well when the lens is stopped down so in P mode the aperture will likely be wide open. Not great.

  34. Well I can answer as a physics graduate. Firstly the DXOmark ISO graphs are simply how close the camera’s settings map to measured ISO. One way that all camera manufacturers cheat a bit is to overstate the actual sensitivity of their camera ISO settings – quite simply, if you set your ISO to 1,600 and the actual sensitivity is 1,200 the camera will gain an advantage. All camera manufacturers cheat a bit on this measure.

    DXOmark does not include ISO accuracy as part of it’s rating, but it is taken into account when producing the overall rating and the individual measures, such as dynamic range, noise and so on. That’s because they plot results against measured ISO, not what the camera nominally says.

    A further very important thing to note is that the overall DXOmark number, which attempts to combine all the technical measures into a single rating, is normalised to a standardised output. There is a mathematical formula which essentially calculates an image quality (a combination of DR, Noise etc.) which would be seen if the image was output to a 8 x 12 size at 300 DPI (a bit u8nder 9MPix). Note that’s not a real output print, but a mathematical projection.

    The important thing about this normalised output is that down-sizing will increase the dynamic range and reduce noise. This is basically because noise is a random statistical value and the very process of downsizing reduces the contribution of noise as a proportion of the output. As it is essentially the “floor noise” of the sensor which dictates the ultimate dynamic range, the downsizing also increases DR. The upshot of this is that a sensor with more, but slightly noisier pixels can outperform a sensor with fewer, but slightly less noisy pixels when viewed at any given output size. That’s because the higher density sensor has a higher spatial frequency.

    If you look at the DXOmark measures, you’ll see there are (misleadingly named) “print” and “screen” measures. The former is the DR, NR etc. measure for the normalised output. The second is the “per pixel” measure.

    This brings us to the “myth of the bigger” pixel issue. It has entered legend that sensors with larger pixels have better DR and noise characteristics. In fact for “equal” levels of sensor technology, this is not the case. What is true is that the “per pixel” DR and noise will be worse for the smaller pixels. However, when transformed to any given output size, the extra spatial frequency of the smaller pixels will cancel this out whilst giving (ultimately) higher resolving power (and some technical improvements on colour rendition on bayer-pattern sensors with their 2×2 mosaics).

    At their heart, photo-sensor sites are simply buckets for collecting electrons excited by the absorption of a photon. The key metrics here are the quantum efficiency (QE), that is the proportion of photons which generate an electron, and the number of electrons the photosite can absorb before overflowing (colloquially called he “well depth”). The very best sensors are achieving a little over 50% QE. Now the theoretical dynamic range that a single photosite can reach is defined by the ratio between the maximum number of photons that it can accept and the “noise floor”. The noise floor is where it becomes impossible to distinguish different levels of grey as the randomness becomes dominant. This is an inescapable statistical effect of the random arrival of rate of photons from dark resources. This is called “photon shot noise”.

    Now the reason smaller photosites exhibit worse DR and noise is not that the wells are smaller in proportion to their size. To a first approximation you can say a photosite of half the area will accept half the number of electrons before saturating. However, the important point is at the “dark end”. As the number of photons is now halved for any given light level, the point at which randomness (the noise floor) is reached comes earlier. Hence smaller photosensor sites have worse DR and noise. However, and it’s an important one, if the outputs of the “fine” and “coarse” sensors are mathematically combined to a normalised output size, the DR and noise characteristics become equivalent again.

    Note that there are other sources of noise – like “read noise” when the electron charge is read off the photosite in the form of a voltage, amplified via a gain control and then digitised. However, great strides have been made in cleaning this up on the latest sensor designs, of which that in the Sony A580, Nikon D7000 and Sony NEX-5n is the best available.

    That brings us to ISO sensitivity. In essence, this is (mostly) done by amplifying the voltage in each photosite before going through the A/D conversion. What this means is that a “half full” well can gain one stop by doubling the gain. That sounds wonderful – however, it also amplifies the “noise floor” too. The result is that for each doubling of sensitivity there is a loss of one stop of dynamic range. Note that for older sensors, other noise sources (like read noise) meant that this relationship rather broke down. However, look at the A580 & D7000 DXOmark values for DR and you’ll find that it holds true for most of the ISO range offered. It also means that ISO will become less important – you can “under expose” and “push” in post-processing and the results will be almost identical. Indeed there’s an article in Amateur Photographer in April called “the death of ISO” where they demonstrated exactly that over a 6 stop range on a Nikon D7000.

    So what does this all mean? Firstly, the most important factor is not the photosite size, but the total area of the sensor. Bigger sensors (with appropriately scaled lenses) basically collect more photons. It’s ultimately the number of photons collected and the spatial frequency which dictates IQ. Secondly, ISO is an illusion on digital sensors. It’s simply a gain control one the sensor to amplify the charge. With an idealised sensor (and some are close) you could heavily underexpose and “push” the result several stops to get pretty well the same result. If you want the best possible DR then you should run at “base” ISO and expose to just avoid clipping. That way DR will be maximised.

    Finally, back to the NEX-7. I would pretty well guarantee that, for and given output size, the NEX-7 will have slightly better IQ than the NEX-5n. That’s basically because I can’t imagine Sony producing a top-of-the-line camera with worst sensor performance than their earlier model. Note, not necessarily better per-pixel measures, although I expect them to be close, but better output.

  35. I tried downsizing processed A77 RAW files to match the image dimensions of processed NEX-C3 (same sensor as NEX-5N) RAW files, and the A77 results were only slightly behind. I think the NEX-7 will do quite well at high ISO and that your presentation of pre-production in-camera JPEGs is not relevant for photographers interested in shooting stills in RAW format.

    My comparison is here: http://www.talknex.com/f2/iso-6400-comparison-sony-nex-c3-nex-5-a77-panasonic-g3-gh2-205/

  36. hmmmm – does the the nex 5-n have the same limitations or is it able to autofocus across the aperture range in manual mode?
    thanks for your help

  37. Dont’ forget, the A77 has Multiframe noise reduction which works great, you get sharper pictures and with less noise.

  38. Thanks Steve. However the problem with scientific theory is that reality has a habit of disproving it.

    “I would pretty well guarantee that, for any given output size, the NEX-7 will have slightly better IQ than the NEX-5n”

    Well that’s easy to test.

    See below samples, both ISO 12,800 resized to just 800×533 in photoshop.

    http://www.eoshd.com/uploads/nex5n-800.jpg
    http://www.eoshd.com/uploads/a77-800.jpg

    The extra resolution of the A77 does nothing to help DR, colour, tonal range or noise when downscaled to the same resolution as the NEX 5N.

    At EOSHD I am interested in real world performance. NEX 5N is superior, Sony have made a mistake, and no amount of science, charts or analysis can change that fact for me.

  39. There’s a (significant) difference in sensitivity because of the translucent mirror, but other tests point out that the mirror also means a general loss of IQ of maybe 5 per cent. These are the reasons, many exptect the NEX7 to perform better than A77 even with a similar sensor. And because of this, unfortunately, your headline referring to NEX7 while based on A77 data is misleading and should be corrected.

  40. Downsizing does help. Ie, if you print a noisy image small enough, it can still look okay. It’s the reason DxOmark gives you an option to compare things in “print” mode (and defaults to that option). What 24MP gives you is higher resolution, particularly at low ISO. At high ISO, you lose out in S/N.

    Try this experiment: Download ISO 12,800 RAW files for Canon 5D II and Nikon D3s and process them in RPP without noise reduction or sharpening. Compare them at 100% pixel view, and you’ll see that the D3s files look dramatically better. Then downsize the 5D II image to match the dimensions of the D3s file, which is equivalent to comparing a same-size print from the two cameras and compare them again. This time the 5D II will look less bad by comparison (although still considerably noisier, especially in the shadows.

  41. You are right but what I mean is that downsizing doesn’t help the A77 overcome it’s overpacked sensor retaliative to the NEX 5N.

    At high ISOs the 5N also has better tonal and dynamic range.

  42. Now I see what you mean. Yes, most likely the NEX-5N will have better tonal and dynamic range at high ISO. However, the A77/NEX-7 sensor will have better resolution at low ISO and may or may not prove to have better tonal range at low ISO. It’s analogous to D3x vs D3s.

  43. It’s scientific theory that brings you these sensors and cameras.

    However, as far as the test goes, forget trying to do this with JPEGs. Quite apart from the issues of pre-production software, there is too much cooked into the processing. It has to be done from raw using properly debugged and optimised software. We’ll see what happens when there are some properly controlled tests.

    Also the SLT cameras will be worse for any given sensor. The proper comparison with an A77 would be the A55 or an NEX-7 vs an NEX-5n using raw files.

Leave A Reply