A65 vs A77

[agorabasta]

I wonder why you seem so certain, that the RAW output is absolutely independent from the firmware. Wasn't the A700 example enough that there is more to it than just the sensor?

The a700 V1 to V4 improvement at moderate ISO800 was no more than 0.5 stop. Now I see a77 ISO400 about as bad as ISO1250 from a700. That's a difference to cover by intentional firmware crippling of early samples rather than by gradual improvement with further updates.

@agorabasta
What you call the JPEG-Engine is actually the camera built in RAW converter.

There are two logically separate steps in conversion.
The first one is converting the raw data into the basic linear RGB data; at that stage the conversion has to do the demosaicing, initial post-demosaic filtering and sharpening to compensate for the AA-filter softening, and a curve linearisation.
The second step is actually building the output format image from the normalised data obtained from the first step.

Exactly that second step should be called the 'JPEG engine' in case the cam produces some JPEG output. And you always may separate the 'JPEG engine' inside a raw converter software.

[Neonsquare]

The a700 V1 to V4 improvement at moderate ISO800 was no more than 0.5 stop. Now I see a77 ISO400 about as bad as ISO1250 from a700. That's a difference to cover by intentional firmware crippling of early samples rather than by gradual improvement with further updates.

I don't know if it was 0.5 stops - I remember quite bad first results while later V4 RAW output is still quite good - particularily when fed to a good RAW converter. The RAW converters alone make such big differences, that you can get anything from really bad to stellar. Therefore I will not start to judge a prerelease cam from random JPEGs or RAWs "developed" from mediocre converters. To me the thing is just open: perhaps the A77 delivers bad results or good ones or just "ok" ones. I can wait to judge it by the final thing.

Exactly that second step should be called the 'JPEG engine' in case the cam produces some JPEG output. And you always may separate the 'JPEG engine' inside a raw converter software.

The only thing that makes sense to get called a JPEG engine is the part that actually compresses RGB data following several steps like the discrete cosine transformation which is characteristic for the JPEG compression algorithm. A "bad JPEG engine" could be one that produces more artifacts when doing this lossy compression but this has nothing to do with pixel peeping noise performance in RAW files.

[agorabasta]

Well, actually, there's much more to the good JPEG conversion than a simple DCT over RGB data. There's also the prerequisite gamma correction/compression, then there's a lot of small pre-emphasis and filtering adjustments aimed at improving the subjective image quality as it is perceived after the decompression for viewing. If you compare the JPEGs out of a500 against those from the Nex5N, you may very actually see the huge difference in subjective JPEG image quality despite very close raw data quality from those two cams.

But I sure agree that 'JPEG engine' term is rather poor parlance. Still, people use it; after all, not all photogs are engineers/scientists.

[agorabasta]

I don't know if it was 0.5 stops - I remember quite bad first results while later V4 RAW output is still quite good - particularily when fed to a good RAW converter. The RAW converters alone make such big differences, that you can get anything from really bad to stellar.

OK, so here's a comparison of an early a700 V1 image crop to an a77 sample. Both developed with the Sony's own IDC4.0 with NR off and with all the other parameters as close as possible to produce the same output levels range. Both then cropped to same portion of image data, then upsized to 1040px width.

Resolution is very different, for sure, but that's partly from focus diff. But there's a difference of at least 0.7EV in the noise levels; so even the V1 a700 was much better with noise than the early a77.

Such a bad noise performance is either intro'ed into pre-release a77 on purpose, or it's simply an abysmally bad sensor that Sony somehow managed to create after all their successes at 14-16Mp.

P.S. Both raw samples lifted from IR.