A77/5N high ISO comparison

If you're looking for pretty pictures, you'd best go elsewhere...all I have is a picture of my messy desk.

I've been playing with both of these cameras for a few days and have really enjoyed what both can bring. I also have the LA-EA2 adapter for the 5N so can put the Alpha lenses on with full PD AF (along with the 1/3 stop reduction in light).

I decided to test first hand the noise comparison of the king of high ISO (5N) and the noisemaker (A77) with exactly the same lens(16-50 2.8) and exactly the same exposure (1/30 f2.8 ISO3200). They both had the translucent mirror in front, so the amount of light hitting the sensor should have been the same.

I metered with the 5N to get that exposure and it was about 1/3 stop darker than the A77 at the same exposure. This is consistent with what has been seen at all the sites...the 5N's ISOs aren't as sensitive as other cameras and will require more light for the same brightness of the image. This gives the 5N an advantage in comparisons using metered tests as has been shown here and on Imaging Resource. This is why it seems like the high ISO king, but in reality is in line with all the other newer sensors.

I normalized the brightness in raw ACR processing to bring them in alignment. I turned all noise reduction and sharpening off and processed neutrally and exactly the same. I resized the A77 to match the 5N's size for a fair comparison. Neither looks pretty, but 100% crops of ISO3200 in very low light seldom do.

Generally, the noise is virtually identical, with ever so slightly more color noise with the A77...within sample variation I'd think. There is slightly more detail shown on the A77 and more aliasing on the 5N. But none of these differences are significant in the slightest at this ISO.

All of this supports DXO's numbers. Basically, higher megapixels gives ABSOLUTELY NO penalty noise-wise when viewed at the same image size and in fact offers a slight resolution increase even at high ISOs.

It is disappointing that the review sites emphasize the "bad" high ISO performance of this sensor, since it only loses the 1/3 stop of the mirror on the A77. DPR did show the comparison with the A55 and showed it equaled it, but in the end concluded "Very noisy raw files at high ISO settings" which to me is a ridiculous statement!

I do agree that the JPG engine could use some configure-ability. But in using it more, I feel they've taken Olympus' approach of optimizing it for viewing at normal magnifications/prints (reduce very high frequency noise/details, sharpen/emphasize medium-high frequency details) and I find it difficult to duplicate in Photoshop and tend to like it better than what I can produce from RAW unless I want to view/crop at 100% which I normally don't.

A77 on the left, 5N on the right.

E x c e l l e n t ! Comparison.

Thanks so much for posting this. Read it a couple of times and also on other posters threads. The actual viewing size and pixelpeeping are two entirely different things.

Regards Perry

RicksAstro wrote:

If you're looking for pretty pictures, you'd best go elsewhere...all I have is a picture of my messy desk.

I've been playing with both of these cameras for a few days and have really enjoyed what both can bring. I also have the LA-EA2 adapter for the 5N so can put the Alpha lenses on with full PD AF (along with the 1/3 stop reduction in light).

I decided to test first hand the noise comparison of the king of high ISO (5N) and the noisemaker (A77) with exactly the same lens(16-50 2.8) and exactly the same exposure (1/30 f2.8 ISO3200). They both had the translucent mirror in front, so the amount of light hitting the sensor should have been the same.

I metered with the 5N to get that exposure and it was about 1/3 stop darker than the A77 at the same exposure. This is consistent with what has been seen at all the sites...the 5N's ISOs aren't as sensitive as other cameras and will require more light for the same brightness of the image. This gives the 5N an advantage in comparisons using metered tests as has been shown here and on Imaging Resource. This is why it seems like the high ISO king, but in reality is in line with all the other newer sensors.

I normalized the brightness in raw ACR processing to bring them in alignment. I turned all noise reduction and sharpening off and processed neutrally and exactly the same. I resized the A77 to match the 5N's size for a fair comparison. Neither looks pretty, but 100% crops of ISO3200 in very low light seldom do.

Generally, the noise is virtually identical, with ever so slightly more color noise with the A77...within sample variation I'd think. There is slightly more detail shown on the A77 and more aliasing on the 5N. But none of these differences are significant in the slightest at this ISO.

All of this supports DXO's numbers. Basically, higher megapixels gives ABSOLUTELY NO penalty noise-wise when viewed at the same image size and in fact offers a slight resolution increase even at high ISOs.

It is disappointing that the review sites emphasize the "bad" high ISO performance of this sensor, since it only loses the 1/3 stop of the mirror on the A77. DPR did show the comparison with the A55 and showed it equaled it, but in the end concluded "Very noisy raw files at high ISO settings" which to me is a ridiculous statement!

I do agree that the JPG engine could use some configure-ability. But in using it more, I feel they've taken Olympus' approach of optimizing it for viewing at normal magnifications/prints (reduce very high frequency noise/details, sharpen/emphasize medium-high frequency details) and I find it difficult to duplicate in Photoshop and tend to like it better than what I can produce from RAW unless I want to view/crop at 100% which I normally don't.

A77 on the left, 5N on the right.

Thanks, I have been waiting for someone to put this to the test with the new LA-EA2 adapter and identical exposures.

Confirms what many suspected/noticed before in the case of the RAW files.

But I also agree that there is room for atleast more adjustability in the case of the jpeg engine.

It's sad that somehow the higher resolution cameras still get (unevenly) penalized by Dpreview (see also D3X vs D3 for example) in their ratings for "more noise at high ISO", when in reality for a given shutterspeed and F stop combination , the main bonus is more resolution in some cases, not more noise at a given output size in any case.

Thank you,
This is wat I want and have, A77 normal apc camera with higher 1600-3200 iso.
And a full frame use at lower 50-800 iso.
Cropping 24 m is wonderfull, and stays nice.
And making a raw better in Lightroom, dan jpg, is not easy.

Cheers Willem.

The only remaining potential problem is the constant reports of the EVF being relatively dim in bright light conditions. However, I used the EVF on my KM A2 in the AZ sun plenty, and don't recall problems in that regard, and this EVF must be a lot better.

I've had that problem with EVF's in the past too though and bought a cheap but large generic (stretchable rubber so it could be pulled over the existing one) eyecup which helped a lot , even when wearing glasses.

That said, a friend of mine (also "active" here) is receiving the A77 probably this week and still owns the A2. I will ask him to make a direct comparison in this regard between the 2 cameras in all kinds of lighting conditions.

The 5N without the adapter is clearly better than the A77 according to you.
Thanks.

Also, it looks like if there were a 5N sensor version of the A77 for a lot less, it would be a better option because the IQ would be equal unless you absolutely needed 24 MP for something...even then according to you at higher ISOs the resolution is the same.

RicksAstro wrote:

It is disappointing that the review sites emphasize the "bad" high ISO performance of this sensor, since it only loses the 1/3 stop of the mirror on the A77. DPR did show the comparison with the A55 and showed it equaled it, but in the end concluded "Very noisy raw files at high ISO settings" which to me is a ridiculous statement!

I think what's happened here(with DPR) is that they based their opinion on manufacturer specifications. ie. If the camera was designed to claim ISO3200 then it will be tested accordingly. And in this way, we are looking at design flaws rather than that of the testers.

Having said that, I've always found noise to be rather pervasive when it came to evaluations. And the main reason behind this is where we often find noise patterns to be subjective in terms of the type of noise patterns were dealing with and the effects those have on an image.

ie. uniform chroma noise can is most often addressed with little impact on IQ. Whereas low frequency noise is most difficult to address with heavy impact on IQ.

Likewise... when we consider the exponential nature of signal noise, establishing baselines can often lead to misleading results. And one example of this is where we might consider where the A77 and 5N noise thresholds come together on the sensitivity scale. ie. Based on noise patterns, it would appear as though both units exhibit similar noise patterns at or around ISO2000. However we a divergences both prior too and following this threshold(ISO2000). Which will results in less noise and more noise for the 5n in contrast to the A77.

Anyways.... all this to say that though I think the A77 can in fact prove to be in par with the 5n at some point along the curve, I also think the A77 ultimately exhibits more noise as a whole toe to toe. Which doesn't mean the A77 is inferior, though I do think that these characteristics will ultimately be more visible if/when we examined both units throughout the entire sensitivity scale. And one way that I think might help establish whether or not this is the case, would be to produce two sets of files ranging from the lowest to the highest sensitivities(ISO100 - ISO12800) and pixel peeping both sets at similar resolutions. Though I would lean toward upresing the 5n over that of downsizing the A77 given that downsampling can affect noise patterns also(ie. indirect NR).

Hope this helps.

There's no reason to believe the 16mp sensor is any cheaper to produce than the 24mp one.

and start giving normalized crops equal time with 100 percent ones.

In the early days of DP Review, when cameras typically had 6 MP or less sensors, Phil Askey's 100 percent displays had greater significance.

Now, when production cameras have as much as 24 MP (and, probably soon, 36) and as few as 10, DP Review's continuing emphasis on 100 percent views disadvantages higher pixel count sensors because so many pixel peepers (and trolls) take the 100 percent views at face value.

Without also including more normalized crops and data, and reminders of what is representative of the "real world", many readers are increasingly veering away from reality and a level playing field.

Based upon the resulting brouhaha on this and other forums, Andy Westlake's response that DP Review prefers to assume readers are knowledgeable enough to keep 100 percent views in perspective simply isn't working often enough. It also fails to address an obvious problem ostensibly because it's "complex" and they're "still discussing it".

Honestly, I always thought expertise is most useful as a basis for solutions, not amplification of the problem. Could pixel peeping have other analogs here?

Times have changed. DXO Mark and many other review sites have normalized data and output in order to make their data and analysis more truly useful.

Even Ken Rockwell has derided excessive emphasis on pixel peeping. I would hate to think DP Review is increasingly falling behind him in this regard.

Rick, thanks very very much for your work.

Regards,

Tony

Anthony Cheh wrote:

and start giving normalized crops equal time with 100 percent ones.

In the early days of DP Review, when cameras typically had 6 MP or less sensors, Phil Askey's 100 percent displays had greater significance.

Now, when production cameras have as much as 24 MP (and, probably soon, 36) and as few as 10, DP Review's continuing emphasis on 100 percent views disadvantages higher pixel count sensors because so many pixel peepers (and trolls) take the 100 percent views at face value.

Without also including more normalized crops and data, and reminders of what is representative of the "real world", many readers are increasingly veering away from reality and a level playing field.

To be fair, they did that on page 19 (IIRC) of the A77 review (altough probably still not at identical exposures). But it would be very welcome to see the comparison tool updated, so people have a choice to compare different camera ISO's for a given amount of noise at pixel level or compare different cameras for noise at a given out put size. And again at equal exposures.

Andy did highlight a problem with equal output size, namely the question which size that should be. IMHO that should be the smallest resolution of the four (or less) cameras as chosen by the user. Based on EXIF info (resolution), the tool should resize on the fly. Which will no doubt have quite a few practical problems (computing/load being one). And in the meantime for its current state, it should atleast be strongly advised to users to compare at equal output size themselves to get an idea how they really compare in this regard, using the originals. And the difference in sensititivies should be noted in or along side of the tool. Now we can't even see the shutterspeeds directly.

That would be considerably more useful than the current review format.

Although, many readers will be disappointed to see how similar files from most DSLRs look under real world conditions nowadays. Less real fuel for brand and camera model tribalism. That's another significant change from the early days of digital photography and DP Review.

Regards,

Tony

The older 16MP R&D and other costs are paid off now.

It is also in at least 5 cameras (including Nikon) so the large numbers make it cheaper to produce also.

Good stuff. I really fear photography going down the line of audiophiles where a group of "subjectivists" have gone down the line of claiming they can hear differences which they cannot reproduce under controlled conditions. These are the people that believe in all sorts of voodoo nonsense like direction speaker cables, audible difference in SATA disk cables, silver plated mains-power leads, speaker cable "burn-in" and the like.

There is a role for subjective evaluation in photography of course, but it's about aesthetics; framing, colour, timing, subject selection, perspective and much else. The aesthetic trade-offs to be made in tuning JPEG production to give the most pleasing output is also a subjective issues. However, sensor technical measurements are not subjective. The parameters that describe the technical capability of a sensor to capture an image accurately are generally fairly well defined.

For those that care about these issues, it's amazing just how well, after things have settled down, the real-world performance of a sensor correlates with the technical measurements such as those from DXOmark.

nb. not all technical measurements are equal - DPReview's DR methodology being something rather ill-defined in the area of noise floors. There's something in there tests stating that DR is essentially the same for different ISOs.

Just Having Fun wrote:

The older 16MP R&D and other costs are paid off now.

It is also in at least 5 cameras (including Nikon) so the large numbers make it cheaper to produce also.

Not really, according to Sony the 5N is a different design than found in previous cameras. (power consumption being one, possibly offset microlenses to improve corner results in the NEX bodies a second).

Just Having Fun wrote:

The 5N without the adapter is clearly better than the A77 according to you.
Thanks.

Also, it looks like if there were a 5N sensor version of the A77 for a lot less, it would be a better option because the IQ would be equal unless you absolutely needed 24 MP for something...even then according to you at higher ISOs the resolution is the same.

Sure, without the mirror the 5N and presumably the 7 will perform about 1/3 stop better if you had the same spec lens. In this case, the lens is 2/3 stop brighter at the wide end and 2 stops brighter at the long end than the NEX kit, and performs similarly wide open, so you'll still get better performance, albeit with a lens/adapter which dwarfs the camera and doesn't have IS with the 5N. This lens combo would be better than the kit for moving kids indoors where the extra stops and faster AF would help you. If I were going on a hike, I'd certainly just take the kit or 18-200 which both offer excellent IQ.

It's awesome Sony has given us the options. The 5N is a wonder of a camera with IQ as good as anything out there. The NEX7 will as well and will give the user a choice to trade off resolution for noise. The 5N doesn't offer that choice to the same degree.

I guess I don't consider it an advantage to have a camera that limits my choices. But if the 5N's resolution is good enough for you in all foreseeable cases, then I'm happy you have found a camera that meets your specific needs so well.

If you check the NEX C3 sensor is better at high ISO.
I think they all use the same sensor, but have different processing in camera.

They may have tweaked it and maybe not, but it looks like you can get different results depending on the camera (see Nikon and Pentax too).