Auto-ISO May Affect Buffer Size

Discussion in 'Nikon' started by shuncheung, Mar 7, 2016.

  1. ShunCheung

    ShunCheung Administrator

    In the last couple of days, I noticed that buffer size on my D7200 has dropped to 6 frames, down from the usual 15 frames for 14-bit compressed RAW. (Here I am referring to the r, remaining frame count that appears inside the viewfinder, when you half press the shutter release button.) I thought somehow my D7200 has become a D7100. :) I fired a few test shots and sure enough, the buffer filled up very quickly.
    I played around with the settings a bit. It turns out that the culprit is that I had auto-ISO on. Even so, on the D7200, as long as the maximum for auto ISO is 10000 or lower, the r frame count (buffer) is still 15 frames. However, if the maximum auto ISO is 12800 or higher, r drops to 6 frames.
    I then checked the D750. As long as your maximum auto ISO stays within the rated range, the buffer is not affected. When the maximum is set to the Hi range, the buffer drops. The problem with the D7200 is that it is rated to ISO 25600, but the auto-ISO buffer drops once it reaches ISO 12800.
     
  2. Is that slowdown still true if High ISO Noise Reduction is turned off?
     
  3. ShunCheung

    ShunCheung Administrator

    High ISO noise reduction is off.
    I went back and checked. The real problem is the high ISO setting. Even without auto-ISO, when the fixed ISO is set to 12800 or higher, the buffer size is reduced to 6 frames. With auto-ISO, the camera needs to assume the worse-case situation that ISO will be 12800 or higher.
     
  4. Even the "old" D700 shows that behavior - with High ISO NR off, the remaining buffer reduces from 16 to 13 once H.03 is selected. It happens earlier if High ISO NR is on - at the Normal setting, it happens above ISO1600.
     
  5. Is it just a matter of the number in the display or is the buffer size really affected?
     
  6. ShunCheung

    ShunCheung Administrator

    The number in the viewfinder display reflects how many frames the camera can hold before it slows down to the rate the
    camera can write onto the memory card.

    On a D500 with a fast XQD card, it can write 10 fps onto the card so that buffer size doesn't matter. On the D7200, I need
    to cap the auto ISO to within 10000 to avoid this buffer reduction. Going down from 15 to 6 is quite drastic, especially
    when I am actually using ISO 200 or 400.
     
  7. Oh - it hadn't occurred to me that it was the high-ISO NR doing it. I'd spotted this kind of thing in the past, but assumed it was just because noisy images compress worse and the buffer size claim was being adjusted based on the estimated image size. (I always shoot lossless compressed.) The NR makes much more sense. I don't usually much care because I'm usually shooting raw, but I have JPEG recorded to the second card for convenience, so I guess it would still be having an effect.
     
  8. That sounds like a "bug" to me.
     
  9. Perhaps the camera performs some kind of hot pixel suppression (or another type of noise reduction) at very high ISO settings, where some kind of reference image is stored in the buffer along with the main image, and then the processing of the noise in the main image (based on the reference image) starts after the shot is stored in the buffer but before the final file is written on the card. It could be also some kind of pattern noise that is in the darkest shadows and is removed.
     
  10. I know in the D7100 and the D3200 manuals it's made clear that some noise reduction is applied to high ISO even when high ISO noise reduction is turned off.
     
  11. A quick and dirty test with the D3200 comes out this way. Exposure was manual, with varying aperture and shutter speed of 1/160 or so, AF on back button so no shutter delay, shutter mode multiple.
    With Noise Reduction turned on: at low manual ISO, buffer size = 10; at high manual ISO buffer size = 5; At auto ISO with manual setting guaranteeing that ISO will remain low, buffer size = 10. High ISO noise reduction halves the buffer size.
    With NR turned off, buffer size remains about the same, at 11, at all settings including high manual ISO, with almost no change apparent at auto ISO. When I shot at auto with the ISO automatically jumping to 6400, the buffer size reduced from 11 to 10, but this may be a normal variation. It would appear that the noise reduction automatically applied is not slowing down the buffer by much if at all, and that Auto ISO mode by itself is not affecting the buffer.
     
  12. I just tried a similar test with a D7100, and it was pretty much the same, until I went to one of the "HI" ISO settings. At ISO 6400, there was no change in the buffer (a measly 7 at fast repeat, 12 bit Raw), and no apparent change when switched to Auto ISO when the auto result was at a normal speed, but at HI1, it dropped to about 4, even at manual ISO, with high ISO NR off. This definitely differs from the D3200. Whatever processing this system is doing for the highest ISO is eating up the buffer regardless of user input.
     
  13. ShunCheung

    ShunCheung Administrator

    I also checked the D7100, which has rated ISO to 6400. If you go into the Hi range, the buffer will drop from 7, 8 frames to 3, 4 frames (depending on RAW compression and 12 vs. 14 bits). I am sure this is by design. It looks like Nikon has to do extra processing beyond a certain ISO. When it is in the extended Hi range, I think it is understandable.
    The real problem is that the D7200 has an inflated ISO range. IMO its top ISO 25600 is exaggerated and its buffer starts to shrink from ISO 12800 and up. Typically you can still get decent results at the highest rated ISO, not so on the D7200, as it is rather poor at 25600. (The D7200 has no Hi range any more.) I wonder whether the D500 likewise has an exaggerated range or not.
     
  14. Is this buffer size reduction shown in any of the manuals? If not it should be as a warning to those who use the Auto ISO feature.
     
  15. In the D7100 manual, it's on page 348, in a footnote to the buffer size entry. It mentions Hi ISO, noise reduction, and distortion control as limiting factors.
    IN the D3200 manual, it's on page 188, listing noise reduction and distortion control, but (as is my experiment shows to be the case) no reduction for HI ISO.
     
  16. Page 380-381 in the D7200 PDF manual. References above are also to PDF manuals. The printed manual for the D3200 is incomplete and Nikon should be ashamed of being so cheesy. The D7100 manual is complete and pagination is the same for both printed and PDF.
     
  17. It must be just the newer D7K cameras as with my D7000 I can shoot 15 raw frames before slowing, at 6 fps using my normal 12 bit lossy compression at H2.0 (25600 ISO). The display always shows r11 which is a little odd but the book says 15. High ISO NR is set to normal.
     
  18. Sorry if my question is off-topic : I wonder if anybody noticed how (high iso-or any iso settings) sharpness is affected by active d'lightning...If not, try to take some pictures with and without AD and compare them...
    ...and a"m sure the buffer is also reduced by high iso NR activation !?
     
  19. I rarely use ADL, because I don't care for the effect it has on shadows, and it seems easier to open up shadows in post when it's off than it is to undo the effect when it's on. But I've never noticed any effect on sharpness or resolution. In some circumstances it can underexpose slightly, increasing highlights at the expense of shadows, which it can then bring up at the expense of more noise. But I can't see any likelihood that this would change sharpness unless you use Auto ISO and the exposure change happens to be at the tipping point where Auto ISO takes effect. I would not expect much difference there either because Auto ISO makes smaller changes than manual ISO permits, so differences would also be pretty small.
    My camera offers only "auto" ADL, but taking pictures with and without it has not shown any sharpness issues.
    I'm not quite sure about the last sentence. The buffer is definitely reduced by high ISO NR activation on the cameras I've tested. On the D3200 it's halved, as you would expect since it doubles the number of shutter actuations.
     
  20. "On the D3200 it's halved, as you would expect since it doubles the number of shutter actuations."​
    I'm not familiar with the D3200 Matthew, but how does that work? As I understand it, both long-exposure and high-ISO noise reduction use a "black" image to subtract the baseline noise from the picture. Opening the shutter should definitely not be necessary to get a blank NR baseline image. In fact it would very much defeat the object. Or am I misunderstanding what you're saying?
    In fact I'm not too sure why the blank exposure appears to be tacked onto the image file and use up the buffer. Are high ISO file sizes any bigger than normal?
    Ah, right. I just realised that the buffer acts as a temporary store for the blank frame before it's processed and subtracted.
     
  21. Rodeo Joe, I misspoke when I said "shutter actuations" instead of exposures, as clearly you're right, and the shutter stays closed. Said badly.
     
  22. Can someone confirm that the maximum displayed buffer depth (rXX) for 14 bit lossless compression can only be r12 ?
    On some reviews, it is written that the D7200 have a 18 frames buffer depth for 14 bit lossless, but I haven"t been able to achieve that (at least for the displayed buffer depth).
    Thanks
     
  23. ShunCheung

    ShunCheung Administrator

    it is written that the D7200 have a 18 frames buffer depth for 14 bit lossless​
    That depends on how that is defined. If you use a 95MB/sec memory card, the D7200 can write approximately 3 frames of RAW file onto the card per second. If you shoot at 6 fps, in 2 seconds, you should capture 12 frames, but the D7200 has also written 6 frames onto the card such that only 6 frames remains occupying the buffer. Therefore, you very much could have captured 18 frames before the camera fills the buffer and starts slowing down to 3 fps, i.e. the rate it can write onto the card.
    I believe the up-coming D500 actually has not-too-deep a buffer and highly depends on the high transfer speed to a fast UHS-II SD card or XDQ card.
     

Share This Page