sonyalpharumors

I’ve been through this – there’s some fundamental flaws in the analysis. The most important is that it ignores the point that those lost bits in the least significant bit areas are essentially noise and don’t contain image data. It’s an example of a simplistic approach. Indeed Nikon have use such RAW compression techniques for a long time.

As for the higher ISOs having fewer levels; those higher ISO levels have a higher noise floor and there is no extra image data in the bottom bits (look at the DXOmark DR graphs where DR has a negative correlation to ISO). The fact is that as ISO rises there is less image information in the RAW files.

Provided that the A/D conversion is done accurately, then there is no IQ difference between analogue gain and the same applied digitally. Indeed there is something usually referred to as ISO-less shooting. Tests have demonstrated on the cameras using the 16MP Sony Sensor, such as the K5 and D7000 that, instead of shooting at a higher ISO you can shoot at base, dial in under-exposure and compensate by “pushing” in PP and produce a virtually identical image. So instead of shooting at ISO 6400, shoot at ISO 100, dial in -6 EV of exposure and PP +6 EV into the RAWs. The results are very similar indeed. Of course all PP does is apply a mathematical multiplication.

If you do preserve extra bits it contains nothing but noise – not image data.

Simply at (say) 6400 ISO there are only about 6-7 bits of useful data in each position (maybe 128 levels) compared to 12-13 at ISO 100.

For anybody with a bit of background in maths who wants to actually understand this stuff rather than produce hysterical reactions and think they know better than engineers who design this stuff, then I’d recommend reading this stuff.

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/

The conclusions that QX drew were wrong, and he told so already on 23rd of October. I do not understand why you, Andrea, post this as a “news” item on your site.

QX posted on dpr: “I stand corrected on the issue of RAW compression. There is indeed sense in the nonlinear encoding algorithm.” (http://forums.dpreview.com/forums/read.asp?forum=1037&message=39661323)

Please make sure you read and understand the whole story behind a thread if you choose to cross-post

Admin, so if I said something about the A77 and then produced a load of crap techno-babble you would simply post it on your blog without checking the accuracy of it?

Alright, I was the OP. So I’ll summarize what I found here.

To answer the posters above, yes, there is logic in the compression of highlight range of RAW values. However,if you look at the graph, it is not the smooth parabola which is mathematically ideal, but has a few kinks in the curve. Whether it causes problems will only be seen later on.

1) If you’re shooting JPEG, choose whatever ISO you want. For High ISO try NR Weak, Clear -2,+2,-2 settings. Those worked for me at 6400 in 6MP mode. 24MP JPG high iso isn’t really… acceptable. You’ll see what I mean.

2) If you’re shooting RAW, stick to ISO 1000, and underexpose by whatever you need to get the equivalent ISO. For example, ISO 1600 means ISO 1000 and -2/3EV compensation. You can bring exposure up later .

3) If you are shooting in good light, don’t bother exposing to the right, as it seems that the RAW file stores almost the same number of levels per stop, so there is no nett gain by exposing to the right. This was my beef in claim 2 in my original post – there is no way to buy yourself any more processing headroom by ETTR (stuff like increasing clarity and contrast will increase noise, and if you tried to boost a properly exposed, low contrast subject, you will get more noise. Traditionally, if you ETTR, reduced exposure in ACR and then boosted contrast, you’d get better results).

4) Make sure you tune the focus correctly for your lenses. The A77 is really high res – it will make you pay for any small front/back focus. I had to tune my 16-50 kit to +3 to make it spot on, and it’s been behaving ever since.

QX

I still can’t believe you guys are arguing about the merits of this camera. I’ve had it now for a few weeks and I’ve already made several thousands of dollars with it, so in my mind there’s really nothing wrong at all! The only thing I can’t stand is the record button is placed as I often hit it back accident at a crucial moment when I need to take a picture and then need to wait for it to cancel the recording and get itself ready to take pictures again. So far I’ve missed a few golden opportunities to capture special moments because of this. But otherwise, everything including the raw format seems to be working just fine. Peace out.

I had a great laugh when this thread was actually posted on dpreview some time back. While I was seeing people at Dyxum post wonderful images from A77, dpreview forum members were busy shootnig sh** with their new cameras and making all these pointless graphs. “Partial knowledge is dangerous” is what applies to the gearheads on dpreview who think they know more than the engineers at Sony or anywhere else for that matter. If you are that good send in your application to Canikon or Sony and you might as well get hired to work for them, rather than being armchair engineers.
Like the first post says here, doesn’t matter if the graph is a line or loop or anything, just go out and shoot some stuff. I can see the results with my own eyes in the dpreview studio comparison and I think the A77 files look really good for a 24MP camera. I also believe in comparing jpegs because that is what the engineers thought would be the best from that sensor and it makes sense to compare end points rather than data upstream(RAW) especially when raw is cooked as well

You’re no expert, my f2.8 landscape shooting friend.

Also, if you don’t use adobe camera raw, nobodys interested.