6 stops of under exposure?!?!?! I've got to see this. Please provide a sample. I'll try do this with the 5DII & 5DIII. What's the best way to do this by the way? With LR4 and the new ACR I do not see a way to do a linear push on exposure. "Exposure" now only (mainly) affects the midtones. How would I do a straight linear push across the entire histogram? I have never done this before but my expectation would be that the results would be UGLY.Originally Posted by Daniel Browning
Thanks for your impressions, Daniel.
I was very seriously considering switching before the 5DIII was announced- for you the most glaring weakness of the 5DII was dr, for me it was af. But they have addressed the af issue to my satisfaction in the 5DIII.
So of course I wonder about the D800 autofocus. It is obviously better than the 5DII, but how does it compare to the 5DIII? (Just wondering aloud... I know you can't make a direct comparison)
Sure thing! I'm getting my wisdom teeth out this morning though, so it will be a while.
RawTherapee is the best program for it, I think. I think the "neutral" profile would work out of the box.
On Canon bodies it certainly will. I'll compare the same exposure when ISO is set to both 100 and 6400.
Ouch! Sorry to hear that :-(Sure thing! I'm getting my wisdom teeth out this morning though, so it will be a while.
Cool, I'll check it out
Way harsh!
Looking forward to it
Was checking the test results on DxO. Am wondering if I am missing something somewhere along the line. The dynamic range reported exceeds 14 stops. Was the number of bits in the analogue to digital converter increased from 14 for this body? If it wasn't I am very puzzled on how the test results can come back with a dynamic range greater than 14 bit data can possibly support. Looking through the information they provide they are not working with processed raw files that have been de-mosaiced when they are examining the data from the sensor.
Rereading Daniel's post at the top of this thread one other item caught my eye. Lightroom does not support 5D3 raw files as of yet. The link near the start of the post is supposed to be a 5D3 raw file that was opened in LR4. How can that be possible?
I don't mean to be critical of Daniel's thought process by any means, he clearly put a LOT of time and consideration into his decision. It is just a couple of little things that I see which I hope someone else might be able to explain for me.
Last edited by jrw; 04-12-2012 at 12:43 AM.
Because the measurement is scaled for spatial frequency (in the case of DxO, it's scaled from 36 MP to 8 MP, IIRC).
At very high levels of detail (i.e. 100% crop, or a 6-foot-wide print), the noise power is high, and so the dynamic range is less than 14 stops. But if you only look at low spatial frequencies (i.e. downsampled to 8 MP, or an 8x10 print), the noise power is lower, and so the dynamic range is higher. That is one reason why it's necessary to have an intermediate format (such as 16-bit tiff) that has higher bit depth than the original (e.g. 14-bit raw). Without it, operations such as downsampling would introduce quantization error such as posterization.
That said, it's still possible (but not desirable) to have more dynamic range than the bit depth. All it means is that the shadows will have quantization error, which manifests as artifacts such as posterization and noise. It would be better to consider those parts excluded from the dynamic range, though and in fact remove them through dithering. Fortunately, very few cameras have ever had that "problem" -- most have enough noise that they don't ever have to worry about having too much dynamic range for their bit depth.
Thanks. I've got an itchy keyboard finger. Ten of them, actually. I'm a pretty quick draw.
Same as usual - just convert the raw file to DNG with the new (beta) DNG converter.
That's too bad. I probably need some of that.
One further question with regards to banding noise. While working with Photron SA5 high speed cameras at work I do notice that I can cause banding when I bitshift the image data to compensate for a lack of sufficient light at extremely fast shutter speeds (1/75000 s and up). Does anyone know if Canon RAW files are oriented to the most or the least significant bits when the 14 bit data is put into 16 bit file format? Logical follow up question is whether the RAW converters differentiate between LSB and MSB orientation?
Since the TIFF format supports both, the decision is left up to the individual raw converter. I would assume that most raw conveters just use whatever the default setting is for their tiff library, which in turn probably just uses the default for the host CPU (LSB).
I'm sure. RAW converters have a lot of work to go through to support a new camera (building color matrix conversion, etc.), so I'd guess they wouldn't mind the effort required to support both types of endianness.
Posting Permissions
Reply With Quote