Sony BJP overview

[David Kilpatrick]

You may see that dPreview has the D3X review up, and they mark their photo of the D3X sensor specially as 'Nikon designed'. However, if you go to the A900 review, grab the photo of the sensor (it's a bit small) then do the same with the Nikon review, and size them up to match, and overlay them in Photoshop, it is totally clear that the sensors are one and the same item - identical imaging area, identical infrastructure to the sensor, identical pinout arrangement. Whatever Nikon do to get 14-bit (and oversampling seems to be the answer to that question) they are using the standard connections and functions of the sensor.

I have suggested that Nikon simply get the BEST sensors - something common in IC fab where every item like this is subjected to a standard bench test. Failed sensors are rejected (something Canon got wrong with the 5D MkII if reports are reliable - rogue sensors have been found by too many new owners), those within a certain range of performance go to the A900, and the pick of the crop is assigned to Nikon - at a price. Everything else Nikon do (under-rating sensivity and applying low gain curves to an overexposed image, clipping black levels) fits with that scenario.

Of course I would be happier if all the sensors were the same, or Nikon's was genuinely 'different' not 'better grade' as that would give more hope of an improvement to the A900 via firmware.

David

[bfitzgerald]

Just as a casual observation, Nikon images do indeed look different on a tonal level, least to me.
I think I prefer the more subtle sony take in general (less obviously digital looking??). For jpegs, the biggest problem I can see sony wise, is not attacking the chroma noise enough..and doing too much to the luminance. Hence the general preference for other makers in camera processing (Canikon etc)
So is DK saying, Nikon are pretty much near overexposing the image, to cut down on noise? And that sony just tend to be over cautious, hence more noise issues.

[David Kilpatrick]

So is DK saying, Nikon are pretty much near overexposing the image, to cut down on noise? And that sony just tend to be over cautious, hence more noise issues.

That is not the whole story, because Nikon's range goes from 50 to 6400 and Sony's also only goes to 6400. Sony is applying a gain curve (like a one-stop exposure boost, compared to Nikon) to the raw files - the evidence is easy to find, to match the output curve in raw conversion Nikon needs a more of a curve at the later stage. But where Nikon has very low noise (better sensor) and can provide really clean digital data for later manipulation (boosting digitally), Sony is starting off with more inherent noise, and the curve applied to relatively underexposed image then boosts this.

It's like each one did the reverse of what you would expect. Nikon started with a quality of sensor which could have been rated at 200 to 12,800 ISO (200-3200 base plus two stop digital boost) if they had gone down Sony's route and amplified the signal early in the processing chain, inside the camera. They would then have had a camera with twice the sensitivity of the Sony, and about the same noise. Instead, they chose to clip the lower values and actually drop the exposure to 100-1600 base, and even sacrifice some highlight headroom for a 50 setting. The effect is not all that easily seen at low ISOs but at 3200 to 6400 it looks almost like a two stop benefit over Sony's approach.

I think that Sony could, even at this stage, change the firmware to make the A900 a 50-3200 ISO camera and remove the powerful gain/curve from the in-camera processing. But then the A900 would not be the same as its published specification, so all the manuals and stuff would be wrong. I think they made a mistake by working the sensor too hard and processing the image too much before the raw file gets written.

I'm also sure they keep learning from the way Nikon do things and eventually they will catch on and catch up.

David

[David Kilpatrick]

There have been a few comments on dPreview about negative aspects of the article. I would point out that my 'Lens Bait' section is not intended to diss Minolta G etc - it comments that professionals I have spoken to, and shown my system to, ALWAYS only mention the Zeiss glass. Never once has anyone (except former Minolta users) said 'Hey, you can get great Sony G glass' - it is always Zeiss they know about, particularly if they are Canon or Nikon users. Some have even mentioned using the Contax system in the (film era) past and say they are considering the Alpha 900 because of the 'same' glass (of course, it is not!). In fact this section makes the point that even the non-G and non-Zeiss line up contains some fantastic lenses but no-one ever gets excited about them.

In this aspect of the article, as with the bits about the flash shoe being unpopular, I am reporting general working pro opinion - though I add my own comments (some more based on typical dPreview enthusiast opinion, like the lament for the lost macro lighting - few pros care a toss about macro or macro ring flash unless they are forensic, scientific, medical and biological IMBI members etc).

Ref the 16-35mm - I've studied enough results being shown from final production versions to know that the pre-production model I tried was not substandard. It has a compound distortion pattern at 16mm, while the Nikon 14-24mm has almost pure barrel distortion at 14mm. In this respect, it is a design similar to the Canon 16-35mm MkII which I have used a lot. It's the Nikon which is the exception, the CZ performs much as you would expect, and I am sure that when it was designed the Nikon lens had not been seen or studied. The Nikon lens is so good it surprised everyone. Even the new 18mm Zeiss has compound (wave-form) distortion.

David

[braeside]

David, any comment on the DPR review regarding speculation about Sony using a stronger AA filter than Nikon? How does this square with the apparent better resolution results with the Sony?

[David Kilpatrick]

Both makers claim that a new type of AA filter is used. Sony states that the filter is further away from the sensor surface than normal, resulting in dust having less visible effect. Technically, this requires a weaker AA filter (it's easy to see why, get an anti-Newton slide mount, place it on a page and read type, then lift it a couple mm - further away it gets, the more blur - extremely crude compared to an AA filter which uses a diffraction-based thin layer coated pattern, but demonstrates the effect).

Nikon has not bothered to build in an anti-dust system on the D3X. Why? Simple enough, the sensor is identical, and that includes the AA filter. They look identical in-camera and of course both companies claim a unique AA filter advance. But the AA filter is actually part of the assembly and it is clear that the sensors are physically identical (no matter what Nikon enthusiasts may say about it being entirely different). In my studio tests with a normal 3D subject using angled planes, where focus variations can be judged even if the exact point of focus is slightly different, I can detect no material difference between the Sony and the Nikon in terms of detail rendering or the AA filter effect.

The 3600 versus 3700 resolution finding is insignificant, dPreview's test shots show visible variation in actual focus point between the cameras. My own tests showed, with two cameras both offering micro AF adjustment and two similar Sigma lenses, I was unable by any adjustment to get consistently identical (to within 1mm or whatever) focus on a close target. There was more difference between a few successive AF refocus locks on the same spot on one camera, than between two cameras. The focus confirmation systems showed that a fairly wide range of movement on the focus ring was all 'within focus'. If one camera will focus slightly differently on a few shots in a row, on a tripod, what chance is there of exactly matching focus ever being achieved with a range of cameras at different times using different lenses?

The differences in extinction resolution are actually more likely to be down to differences between the lenses used and the true focus point, than the AA filter.

Nikon results may look sharper in the field because the black is slightly more intense (especially in sRGB modes) and the midtone slope of the contrast curve is steeper (this also applies to Canon). Sony has opted to show more shadow detail and also more highlight detail, simultaneously - this means the output curve is a bit closer to linear. This gives an impression of flatter tone and colour and less visual acuity. It is easily corrected by using the 'Strong Contrast' curve in Adobe Camera Raw instead of the default 'Medium Contrast' (I'm considering making a new even stronger custom curve after using the 5D MkII). Raw conversions from the A900 can look muddy, and are not helped by the very warm colour balance from ACR/LR.

David

[braeside]

Thanks David.
One more question:
Does the live view contrast detect in the D3X give more accurate and repeatable focussing than the normal phase detect AF (for static subjects)?

[David Kilpatrick]

Thanks David.
One more question:
Does the live view contrast detect in the D3X give more accurate and repeatable focussing than the normal phase detect AF (for static subjects)?

Yes, it's dead accurate. All Live View focusing with magnification tends to be more accurate than all phase detect AF, depending on the user.

David

[braeside]

If Sony can improve the Intelligent Preview mode in the A900 so that the user can zoom into the preview for focus checking, that would be really handy in MF tripod Macro work etc.
I wonder if they have sufficient hardware within the A900 to even offer real off-sensor liveview as an upgrade?

[David Kilpatrick]

If they improved intelligent preview to that level, it would involve storing at the least a complete JPEG in buffer - or a complete raw. Since the camera is so fast to permit reviewing of a normally shot and stored result, why not just make a test shot for focus? IP allows adjustment of colour, contrast, DRO, and exposure but it could not really adjust focus (think about it!) and you would still have to take additional IP shots.

David

[David Kilpatrick]

Here's an amusing dPreview thread rant from someone called Gary who hasn't used live view in the studio - at least, not with a properly HDMI feed equipped camera like the D3X (or the A900 if they had enabled it):

"He doesn't know what he is talking about. He should explain how live view would help precise studio work. I could use laugh. If you are doing precise studio work you would have the camera tethered to a computer so you could shoot and check the shot at 100% on a real screen, not some live view on a 3" screen on the back of the camera. LOL!"

Well, here is one for Gary:



When connected to an HDTV using HDMI in the studio, the D3X switches from its rear LCD to the connected TV. The resolution is excellent, and Live View including 100% focusing is transferred to the large screen. Unlike tethered shooting and testing on a PC, no software at all is needed and you can connect to any HDMI equipped TV, projector or monitor (or to a DVI interface with an adaptor). The image is very clear and stable, not like VGA video feed and the only accessory needed is a regular HDMI cable - the D3X, unlike some other DSLRs, uses a full size HDMI connector not a mini one. The camera still must be controlled from camera position, so it's ideal for working with a suitable monitor (a 22 inch like this one is about the optimum) close at hand.

This photo is from my report on the D3X for Master Photo>Digital, where the use of an attached monitor (a rotating one is really needed for portrait shooting!) has applications in portraiture with live view.

And, as it happens, it's easier to check focus at full magnification on the D3X rear screen than it is on many computer screens - certainly a great deal faster. And, as with the HDTV screen rather than the shoot and check process, you adjust the focus finely at maximum magnification. It is a live process. The HDTV screen allows a better method for checking depth of field.

David

[braeside]

That's neat with the HDMI and the Nikon.

My thinking re having "zoom in" with te IP of the A900 was that it would save storing and deleting multiple MF focus test shots, certainly you still have to take the 'shots' with IP and wear out the shutter.

[David Kilpatrick]

And, for what it's worth, this is one of the problems I have with dPreview - they run a series of standard tests which cover all the normal functions of each camera. But they omit a whole load of vital tests, like hooking a camera up to HDTV and showing the result. They do not test SSS properly, they have never analysed DRO+ and the various levels of DRO in comparison to Nikon D-Lighting or Canon's highlight control and tone compensation. I have not actually published anything myself comparing these, but I have used all of them, and it's a close run between the DRO+ and Active D-Lighting with Sony/Apical winning in most situations. Canon's variant is not clever by comparison, and they make little of it - no wonder, as it on a camera like the 50D (which I was using last month) it demolishes their noise advantage.

dPreview does not even list flash modes. I bet that hardly anyone is even aware that no Canon EOS camera offers second flash sync with any studio flash strobe or non-Canon-dedicated flash. Even with film cameras, all the way back, their system never permitted rear curtain/second curtain sync unless you are using battery TTL/e-TTL flash. Both Nikon and Sony offer assignable second curtain sync with studio/mains strobes and manual handhelds like the old Metz 60CT-4 which I have.

Now, from my point of view, this is a HUGE minus for Canon. Second curtain sync is almost mandatory for some FX and techniques, often used with Lumedyne, Quantum, Norman, and many type of manual/AC/location/studio flash gear. Yet Canon's entire range including professional bodies up to the 1Ds MkIII does not allow it. Minolta, KM, Sony has always done so - for years. Does this get mentioned? Never. It is out of the radar for dPreview staff, because they have (presumably) never worked professionally. I have not done so for nearly 20 years but it does not mean I have forgotten everything I used to use and need.

I see that Gary on the thread mentioned suggests I have no idea about shooting packshots. Well, my company was the first ever to do so in bulk as far as I know. In the 1970s we produced catalogues the traditional way - sheet film and dividers (you turn the back to expose on two halves of the sheet, or four quarters, with cut-down darkslides - every subject is traced to scale on the groundglass with Kodatrace overlays and shot to 'in pro' size for assembly - which we also did, by hand, on lightboxes with acetate). In 1996, we did the Paterson Photographic catalogue using the Nikon E-2 (http://www.mir.com.my/rb/photography/co ... /index.htm" onclick="window.open(this.href);return false;) proto-DSLR. Shirley shot 800 products in under one week earning us £10,000 in shoot fees, linked to a Mac Powerbook of an early generation. We also had a Leaf Lumina and ended up shooting the cover on this - actually, we shot the cover during a demo to a university photography dept. It was not fast enough for bulk packshot shooting.

On this sort of subject I am not talking out of my hat! Before we moved to Scotland in 1988, we won the first ever PIRA DTP Awards (UK) outright - mainly because we were the first in Britain to own and use an imagesetter to create page film for a commercial client magazine (Ilford PHOTO). Around 1990 (I think) Apple featured me in international publicity as one of the pioneers of this technology (Gestetner, an Apple reseller, hired me to sell systems and most of the UK national publishing groups bought consultancy before adopting their first DTP installations). It all began to fall apart for me around 1997-9 when the systems moved out of my range (£250,000 instead of £50,000) and the market shifted to major investors. We had the first digital product photography studio in Britain - one of the first in the world - in 1995.

Today, I do no commercial work. There is no point. Back then, my work supported at least 10 and maybe more adult wage earners, but I never made very much personally. Today I support me, my wife, and my son - end of story! We all do just as well as we did back then, but we have a bit of free time and no big worries about our responsibility to staff and suppliers. I have stuck with my current magazine printers for over 25 years. I value people I know more than a cheaper quote. Our work is not worth a lot to them now, but they still treat it the same way as a priority.

If someone came and asked me to do 100 or even 1000 packshots in a day, I would know how to do it - I'd probably suggest they go elsewhere...

David

[David Kilpatrick]

I see I have misunderstood Gary's post on dPreview - he was suggesting that shooting in volume (packshots) didn't amount to precise studio work. Well, it used to, as everything had to be 'in pro' scaled and exactly matching the designer's layouts, and before digital arrived the standard tool for catalogue production was a 5 x 4 with rollfilm back (or divided darkslides, which many preferred as the sheet film was thicker than rollfilm, making it easier to handle when assembling the repro montage).

Live view is essential for precise studio work if tilt-shift lenses are used, as I found myself (too late, as it happens, to produce a perfect example). The focusing screens on current DSLRs don't really allow any accurate judgment of depth of field or precise focus point, and in a 'deep' studio product shot with a tilted lens, sharpness needs to be checked at all points of the frame. Trying the judge how sharp the image is at the top and bottom of the focus screen, even on the A900 which has the best screen of any on the market, is almost impossible. You may think you have tilted the lens correctly but be way off. Live view with a movable focuspoint, and magnification, allows every part of the shot to be examined just as critically as you would use a loupe on a groundglass. It is possible to fine-tune the focus and tilt of the lens while moving the focus point round and checking all parts of the plane you want to be sharp. Doing this on the back of the camera is perfectly practical, and enables totally precise focusing.

Since Sony has no perspective control lenses, this particular use of LV is not relevant to the A900 but - as commented - if they are to compete properly they will one day need both.

David