Friday 7 December 2012

Insane D800



The D800 broke many barriers when it was released - at the time of release/review DxO Mark awarded it the highest score in it's history (indepth analysis here); Mike Johnston at TOP made the following glowing statement following his own personal review:
For 25 years I've deliberately made it a point to experience a wide variety of cameras of most types: I've tested dozens of cameras, owned dozens of cameras, and seen and handled literally hundreds more.
[..]
Cutting to the chase: all things considered, the Nikon D800E and Nikkor AF-S 35mm ƒ/1.4G lens is the best overall photographic device of any description I have used thus far in my life.
High praise indeed although it hasn't been all positive news.


Nikon had enough concern regarding the high MP count to produce a technical guide to manage novice expectations. But more painful for Nikon and the user base was the left focusing issue widely reported on such sites as Thom Hogan, Ming Thein and Falk Lumo. An issue that still hasn't been officially acknowledged. Whilst it is now believed that the calibration issue (requiring a service centre side adjustment) has been fixed at source many forums still contain "is it safe to buy a D800" queries. Thom Hogan's D800 review was updated to suggset that D800 US serial later than 3055xxx should be free from the problem although the 'serial number safety mark' does not translate for other markets since they have different base offsert numbers: Europe for looks like it uses a 600xxxx (mine is a 609xxxx) whilst Asia has 800xxxx serial number base.

But how does the D800 compare against my old D300?


Physically compared to it's high-end compact siblings (D700/D300) we can see that the D800 is a little smaller and lighter.




The D800 is surprisingly light compared against the D700 and on par with the D300 - however, the shortened width of the grip is initially very noticeable: this has a noted change on the top LCD panel which is less wider but is taller.




Comparing ISO performance is difficult. At the advanced enthusiast level (ignoring the D3s/D4) the shootout should be D800 vs D700 but due to the size differences many people are downscaling to 12MP for fairer side by side comparisons: the logic being that the 36MP files will have more noise due to the higher resolution so 100% crops could lead to misleading conclusions.


To my eye, the D800 ISO6400 images look a little better than my D300 ISO1600 images: so the D800 would give D300 users over 2 stops of comparable noise advantage which is significant. To put this into context, consider this 100% crop from a D300 ISO3200 image (no NR applied)


and a 100% crop from a D800 ISO3200 image (no NR applied)


But let's ignore noise at high ISO for a moment - the other main factor that that makes the D800 impressive is the ability to hold detail in the ISO3200 range. The following are 100% screengrabs of the area under the focus point used.


ISO3200, 100% crop no NR

ISO3200, 100% crop default CNX2 NR

Coming from a DX/D300, the 51 autofocus points of the D800 are much more central and don't drop into the vertical thirds as per the D300.


Metering with non CPU/AI/AI-S lenses is still confined (as with D300/D700) to the central focus point.

The internal (reflective) spot metering does appear to behave a little differently: skin tones would be correct on my D300 when metered 2/3rd over but to achieve the same skin tones on the D800 I need to meter 1stop over. Following the readings of an external incident meter (Minolta IV/F) yeilds the same on both camera's leading me to conclude the previous.

Obviously with the hike in megapixels there's a direct extended requirement for additional backend storage. My 8GB CF cards would happily report capacity for ~300 14bit compressed NEFS on my D300 whilst the same cards report capacity for 100 NEFs on the D800. What's less apparent is the write/read speeds of your cards: CF cards are typically faster than their SDHC counterparts (think the equivialent SanDisk Extreme CF vs SDHC cards) due to the CF controller interface built into the card, I found my SanDisk Extreme III 8GB CF card (rated at 23MB/s) to be painfully slow reviewing images following even small bursts - 4 images over 2 seconds for instance. As an interim step, I picked up a SanDisk Extreme 45MB/s 32GB SDHC card (rated at 30MB/s) to be a noticable improvement and will go whole hog for the 32GB CF when prices eventually drop.

Rob Galbraith's CF/SDHC card-speed db provides a great starting point for evaluating card speeds.


A few final setup tips (see the user manual) based on the A1.00 B1.01 (L1.006) firmware:
  • viewfinder displays the exposure indicators reversed
  • f12: Reverse Indicators Previously, the default indicators had left being over and right being under exposure - this menu option will fix this and force exposure indicator to display like every other previous Nikon body.

  • Zoom playback to focus point
  • f3: Multi Selector ->Playback Mode set to Zoom On A very simple and quick way to validate focus is through when the center multiselector is pressed - the image will zoom to the image segment centered on the active/selected focus point used at image capture. This functionality is for AF lenses - the center of the image is zoomed for MF lenses.

  • Skip playback images via command dial
  • f9: Customize command dials -> Menus and playback set to On One useful function for playback (aside from the zoom to focus point noted above) is that the main (rear) command dial can be used to skip to other files in normal/non zoom mode - this functionality was only available on the D300 when the image was zoomed.

  • Reduce size of RAW/avoid embedded jpegs
  • 14 Bit lossless compressed NEFs are approx ~50MB and full sized 36MP jpg ~18MB. This means that post editting in CaptureNX2 could result in ~70MB NEFs. To avoid huge embedded JPGs, recent CNX2 versions (v2.3.4) have the new option to embed only 320x213pixel thumbnails:

    I'm doing this for all D800 NEFs to essentially strip out the embedded jpeg and bringing down my 'out of camera' 14bit lossless compressed NEFs to the ~37MB region.

  • Assign non-CPU
  • The D800, as with the higher end bodies, allow the use of the older non-CPU AI/AI-S lenses. The camera needs to know the focal length and max aperture value for each non-CPU lens that can be entered into the body's internal non-CPU bank - doing so, allows the EXIF to include the focal length and aperture along with other features documented in the user guides. The camera body needs to be told which non-CPU lens you are attaching (it has no way to tell) so the user has to manually select the lens from the pre-entered bank. To make this more convenient, the selection can be tied to a custom setting + command dial. See f4 Assign Fn button or one of the others available (AE-L/AF-L and DOF preview) in the "Assign non-CPU lens number selection to camera control"

Unfortunately, the current D800 A1.0 B1.01 firmware does not allow users to reallocate the ISO button unlike the D4 which would allow for single hand manual adjustment of ISO and the ability to trap focus has been dropped.

Furthermore, the initial D4/D800 lockup issues don't appear to have been completely fixed: I personally have experienced the lockup issue with the A1.00 B1.01 firmware where when I turned off the camera the top LCD panel was still active. I was able to turn the camera off by rotating the on button past "on" to "illumination" and then back to "off" and this appeared to clear down the issue.


Update May 2014 - A1.10/B1.10 firmware was released that included fixes for the lockup but more importantly trap focus functionality was restored as per previous high end models along with the new ability to assign ISO setting to the movie record button on the right hand side and thus allowing single handed manual ISO adjustment.

Dead Pixel

My D800 body came out of the box with an apparent 'dead pixel' - it appeared at the exact same pixel position and showed up as a white/green burned pixel: this was visible only at 100% zoom on both the RAW (in CNX2) and the jpeg preview on the back of the camera and was on all images, regardless of lenses/ISO/shutter speeds. Interestingly though, images shot at 1/250th with a lens cap on (black image) did NOT have the said troublesome pixel.

So one problem pixel only noticeable at 100% was slightly annoying but not enough for me to exchange for what would have been most likely a refurbished camera.

However, I came across this DPreview thread which claimed such 'dead pixels' could be remapped by forcing 2x sensor cleans in quick succession. Whilst this sounded like an implausible fix, I attempted this nothing to lose.
Sensor clean from the menu, Switch off and back on and then force another sensor clean from the menu
Incredible, the subsequent images do not have the problematic pixel!

Amazing. I don't know how/why this would fix this - especially since I have sensor clean enabled on startup/shutdown - but it worked.

AF Fine Tune

It has been noted on the review sites that the larger MP count uncovers problems with user technique and also focus. With this in mind, it is useful to perform basic checks that all your lenses are reasonably within tollerance and in sync with the D800.



Nasim Mansurouv has an artcile regarding AF evaluation but the technique is simple:
  • find object for focus test
  • I have a WhiBal that just happens to have a Seimens star focus object printed on it so I use that. Others suggestions range from the full ISO 12233 focus chart to some simple focus charts and another here or finally the (apparently defunct) focus test chart

  • align camera on steady tripod
  • The focus point in the viewfinder is not necessarily a true representation of the EXACT location of the focus sensor. Therefore it is suggested that the test/focus object is parallel to the image sensor to ensure that you don't overlook back/front focusing issues.

  • take manually focused reference frame
  • Put the D800 on LiveView, put the lens and body on manual focus setting the lens wide open. Use centre (most accurate) focus point, zoom in 100% and manaully focus the lens til subject is in focus. Remote release for acquiring the RAW image - delay/self timer can also be used but beware of slight movements to the camera rig.

  • take AF frames and evaluate
  • Put the D800 back on normal mode and rack the focus to inifinity. Initiate autofocus and acquire the RAW image. On a computer, evaluate at 100% the reference and AF frames.

  • adjust focus with AF Fine tune
  • Dial in AF fune tune for your lens and retry the AF frame to verify improvement. Repeat til satisfied.
My 50mm and 85mm always seemed reasonably well calibrated for my D300 but the D800 did show there were some minor focus issues that were easily resolved with minor +12 AF (50mm 1.8G) and +1 AF (85mm 1.4D) fine tune parameters and +4 for the 24-70mm. Whilst lenses and camera bodies are manufactured to certain tollerances, you may find that one combination of tollerances (my D300 + 50mm) work better than another (D800 with same 50mm) so it's advisible to take make these checks to remove such barriers that would otherwise hamper the optimal output.

No comments:

Post a Comment