ISO 6400 Comparison: Sony NEX-C3, NEX-5, A77, Panasonic G3, GH2

There has been a lot of discussion around the web about "sensor performance" based on in-camera JPEG crop comparisons published by EOSHD and Steve Huff, and I thought it would be a good time to try and clarify the situation by presenting some crops from processed RAW files.

The RAW files used in this comparison are all available for download from www.focus-numerique.com and processed using Raw Therapee with subsequent removal of chroma noise only (leaving behind all luminance noise) using Nik Dfine. I resized all images to the same 5040px diagonal dimension using Photoshop Bicubic sharpener, so any differences in magnification/framing are attributable to technical differences at the time of capture by Focus Numerique.

Although I do not have samples from the NEX-5N, that camera has the same sensor as the NEX-C3, and RAW performance ought to be identical. Same for the NEX-3 and NEX-5, which share the same sensor and should perform identically. The NEX-7 uses the same 24MP sensor as the A77; however, the NEX-7 should outperform the A77 in this test since the latter camera loses some light as a result of the SLT technology. I'm presenting only the center crops so as to mitigate issues related to lens performance. All at ISO 6400:



I re-did the comparison using Lightroom 3.5 RC for the RAW conversions, and the results are shown below. Lightroom doesn't support the A77 yet, so I left that out and included the Olympus E-P3 in the comparison. Again, all are center crops from ISO 6400 conversions, chroma NR only, no luminance NR applied:


Conclusions:

1. The NEX-C3 (and NEX-5N) sensor is only a little bit better at high ISO than the NEX-5 (and NEX-3). This was already demonstrated by DxOmark, and I think the results here corroborate their findings.
2. The 24MP A77 sensor does quite well when the files are compared at equal output size to that of other cameras (The NEX-7 should do better than the A77).
3. The latest Panasonic sensors (G3 and GH2) don't quite keep up with the Sony sensors but are still reasonably good at ISO 6400.
4. The E-P3 sensor lags behind the others in terms of high ISO performance (but it is still more than good enough for my purposes!)

Addendum: Some have suggested that the NEX-5N and NEX-C3 have different sensors, and one member at DPReview (KM Legacy) called my attention to the folllowing statement from the Imaging Resource: "Although the resolution is very close to that of the recent NEX-C3, the [NEX-5N] sensor is actually newly developed".

To test for myself, I took ISO 6400 RAW files for both cameras from the Imaging Resource, opened them with zero NR (chroma or luminance) in Raw Therapee, and slightly adjusted levels in Photoshop to get them to a similar apparent brightness since Raw Therapee did not match the levels. Here are the 100% crops, which to my eye demonstrate identical sensor performance (minimal differences due to imprecision of testing methods) between these two cameras:




There are lots of claims from manufacturers about what is a new sensor and what isn't. The proof is in the result, and I think we'll soon find that DxOmark gets identical results (superimposable curves) for these two cameras.


Addendum #2: I further compared the C3 and 5N sensors and got a slightly different result: https://www.talkemount.com/f2/sony-nex-5n-nex-c3-same-sensor-different-218/#post903

 

I added the Lightroom conversions to the first post, and I think the NEX-5 looks even better there than it did in the Raw Therapee conversions.

 

Nice to see you here. Thanks for registering to comment! The labeling on my comparison montages is correct. The ones labeled "NEX-5" are the "old NEX-5", and the one further down labeled "5N" is the new NEX-5N.

Those are in-camera JPEGs. I don't think you can say anything about sensor pattern noise after the image engine has intervened. There are RAW samples available and "honest" RAW converters based on dcraw to do the conversions. Using the in-camera JPEGs to judge sensor performance is like looking at a shadow of a person to see what they look like!

 

Hi Andrew, try the latest betas of Raw Photo Processor:
http://www.raw-photo-processor.com/RPP/RPP64_1454Beta.zip
http://www.raw-photo-processor.com/RPP/RPPUni_1454Beta.zip

 

I found some really minor differences between C3 and 5N sensors using RPP and the Imaging Resource files: Comparing the NEX-C3 and NEX-5N files...: Sony NEX Talk Forum: Digital Photography Review

For all practical purposes, I think they are the same and agree with you that good JPEG processing makes a world of difference.

If you try the 5D II files against 5N files from RAW in RPP, you'll agree that the 35mm format sensor still holds a significant advantage. I processed the ISO 12,800 RAW files from Imaging Resource for NEX-5N, Canon 5D II, and Nikon D3s and then downsized the NEX and Canon files to match the D3s dimensions. Here are the representative 100% crops (Nikon on left, Sony in middle, Canon on right):