The sRaw Myth


Too bad that sRaw has the word “raw” in it ;~). That seems to be causing all the problems with people understanding what it is. 

With the D810 shipping today and far too many people on the Internet claiming that they are happy with the camera because it has sRaw capabilities, it’s time to once again remind people that you can name a “dog” a “sWolf” but it still won’t be a wolf.

For those not wanting to read far, here’s the bottom line: Nikon’s sRaw files (D4s and D810) are basically equivalent to something like 11-bit JPEGs. To be clear:

  • You get fewer bits of data
  • You get lossy compression of data
  • You get color data that is compromised, particularly in the shadows
  • The data is YCbCr (a color model), not raw data or even RGB

Yes, you can still apply “conversions” on the data, but you’re doing it on data that has already had white balance and linearization curves applied to it. Thus, your ability to correct data is somewhat compromised. In particular, there are already rounding errors in the data, so you can compound those errors during some conversions. This is particularly true in the shadows, where you’re working with a more limited number of bits.

But the problems don’t end there. Basically you saved about as much storage space as you would if you had just shot with 12-bit Compressed NEF as your setting, but you ended up with one-quarter the pixels. You also get a real tangible problem in the camera itself: the amount of time that the EXPEED4 needs to create the sRaw file means that the buffer gets sucked up. Instead of 58 12-bit Compressed NEF raw files that are 13.3MB in size, you’ll get only 18 sRAW files that are 13MB in size. 

So what are the compromises with 12-bit Compressed NEF I would recommend you use instead to save file space? Basically highlight manipulation: gross changes in highlight data can produce potentially visible effects, though in the history of Compressed NEF, the number of times anyone, including myself, has been able to show that compromising an image is countable on one hand. On the other hand, you have four times the data to work with, and it’s deeper in bit value and not compromised in color or linearity ;~). As I think I pointed out somewhere recently, if you really don’t need all those pixels, just adjust your ACR conversion to be 9mp instead of 36mp. You’ll be better off in almost every way.

Thus, I’m going to call BS on sRAW: it really doesn’t provide us with anything useful, and potentially causes other issues you won’t like. It’s a marketing check box feature to get parity with a Canon feature, not a particularly useful one. What would have been more useful? Try any or all of these:

  • 11-bit JPEGs ;~)
  • The ability to choose the actual JPEG image embedded in the NEF file (e.g. JPEG Small Fine), and then the ability to easily extract it
  • More crop options
  • 10-bit TIFF (any size) with compression

When will the Japanese find and talk to real, working photographers that have a good sense of the technical and can give them good advice on features? Let’s start with this idea: in each country, invite all NPS members to a full day meeting with a team of engineers (and translator ;~), and have someone who knows how to put together an agenda for that sort of thing get it organized and to the point. Morning: information from Japan engineering to the photographers, including more disclosure of how things really work. Afternoon: information from the photographers to the engineers, including sessions to both suggest and prioritize needed features. I’ll bet that sRaw, as defined by Nikon, wouldn’t make the cut.  

© Thom Hogan 2015 / All Rights Reserved  @bythom #bythom