DK! Post Some IS0 1600-6400 a900 Samples Soon!

[lock]

Thank you Sonolta, thanks Dave for the abstract.

I'm going to try this one for sure. But essentially it's about taking in more light than default, hence reducing the basic noise levels. Two stops max can be a lot but I will try and vary this and see how well this works.

[David Kilpatrick]

It can just be about making sure the camera's tendency to underexpose doesn't get in the way of good high ISO pix. Remember, the Nikon D3 and D700 both typically give one full f-stop more exposure, in a given situation, compared to the A700 or A900. I have found this when using them. The exposure is simply more generous; does this mean their ISO 12,800 is not really 12,800? No - the images are brighter (if the camera gave one stop more, and the images were the same density as the A900/A700, then Nikon's claimed ISO ratings would not be true).

So, I have done a few tests using both generous (+2 reduced by -1 in ACR at 6400) and correct (+1, in-camera JPEG and ACR at 1600) tests with a situation comparable to Don's cats - my cats on the bed, with reflected sunshine 'bounce lighting' them:



http://www.pbase.com/davidkilpatrick/image/103702323

This is the first of a sequence of five files, two of them very large full size, with captions explaining the shots. Use the link and then view the sequence using the NEXT tab.

Here is one of the samples - 6400, filesize reduced to exactly match the A700 dimensions, crop to eye only:



David

[David Kilpatrick]

Don, I already did that. My Papa Joe shots were in pretty much exactly the same light as your cats. I referred you back to them again, but you had completely screwed up my thread posting images from photokina by littering it with arguments and promo pix from Canon and Nikon, and ignored these entirely. I hate to 'do a Sonolta' and post something repeatedly but here goes, repeated, between the dashed lines:

-------------------------------------------------------------------
Here is the Nikon D3, section of a 12 megapixel image shot at f5.6 and 1/125th using the 24-70mm at 24mm. Now it's not stunningly detailed even if the noise is very even, and the noise is not zero either. Processing conditions using ACR were 0 sharpening, 50 chroma, 0 luminance. The focus point in the shot is the copper cauldron. Checking many other shots, this degree of detail softening is typical of the Nikon at 3200:



Now look at a section of one of the shots I just posted, this time processed identically to the Nikon (the full image actually uses some luminance NR and sharpening 25 because it can take it, this omits the luminance NR and sharpening):


--------------------------------------------------------------------

The light level I shot the latest cat pix in is typical of many uses for 6400, and here is a comparison with the Nikon D3:

http://upload.pbase.com/image/103706023

For reference, a small version of the pair:



The illumination levels are within 1/3rd of a stop for these two shots. I just don't get where you are coming from. What are you trying to say, or prove? Absolute light level is not as important as relative exposure level, within certain limits, for getting a decent sharp result with moderate noise. I shot this pic mainly because I realised I didn't have any keeper shots which showed our cats doing their usual all-day stint getting the bedclothes furry, and when passing the light looked good. Your cat pix had reminded how few 'good' shots we have of our cats - once grown up, they stop being chased by the camera.

Within the limits of its ISO range - since it does not go beyond 6400 - the Alpha 900 is going to prove itself in the hands of users, no matter how much you jump up and down protesting that it won't and can't. You would not be advised to get an A900 anyway, except perhaps for your shots of graffiti and similar fine art copy/trouvé work - you benefit more from crop factor format shooting because of the distances involved, and the subjects really don't need 24.5 megapixels. Nor do a lot of mine, and I've decided that the A350 with CZ 16-80mm attached will be second string for 'walkabout' situations.

But, so far - the A900 will be used for high res AND for high ISO shooting. Apart from anything else it focuses much more accurately with my faster lenses (50mm f1.4, 28mm f2, 100mm f2..

These are no doubt not the last 1600-6400 samples I will post, so I am sure further light conditions around the hand-held limit at 3200 will come up.

David

[David Kilpatrick]

OK, night time, 100W bulb lighting the entire same room (and not directly either, shade under the bulb uplighting the ceiling):



1/13th at f2 with the Sigma 24mm f1.8 lens, ISO 3200, ACR, no sharpening, no luma NR, 50 chroma NR; 2400K for the very warm tungsten light; image reduced to exactly match A700 12 megapixel height; 100% clip.



The whole shot - http://www.pbase.com/davidkilpatrick/image/103720732 to view at 24 megapixel size (it's cropped at the end for a better composition but is otherwise the full image height). This cat is unflappable, putting the Sigma right up to his face then firing the Alpha 900 no movement, no blink (twice as loud as a 700, if you have not heard one yet - extremely chunky mirror sound, like closing a car door every time you take a shot). I took some at 1/20, 1/30, even one at 1/10 and he was sharp enough for all of them but focus varied quite a lot with the Sigma, probably small movements on my part. SS worked perfectly.

David

[David Kilpatrick]

Oh, and ref the KM D5D result from Don:



This is a 6 megapixel size 100 per cent crop on the same basis - but I have applied web USM, as the really fine noise structure would take it OK:



You can tell how low and non-directional my lighting is - look at the slit pupil in Don's cat's eye, and the wide open pupil in mine (Birman cats often look wider-pupilled though and cats vary a lot in their iris/pupil reactions to light).

David

[David Kilpatrick]

Yes, I removed them. They were seriously underexposed shots in conditions where I should not have posted them, and your comments on low light exposure were correct - it is better to avoid that sort of lighting. My only concern was that some people were picking them up and referring to them as if they were trying to show how good the A900 could be, not how bad it could get. I don't really want to be feeding trolls bait for dPreview etc so I removed them (I also removed some other files, at the request of Sony, directly - I broke their embargo on using examples from the 16-35 and 70-400mm by showing them).

David

[bfitzgerald]

Oh well, never mind, we all saw them anyway

I guess that means no more sony stuff for David...lol

Are they upset?

[David Kilpatrick]

It's a bit frustrating when several sites have plenty of 16-35mm prototype shots up on line. I think, as ever, I was revealing a slightly negative aspect of the lens (distortion type) rather than just trying to take 'wow' pix to show off. Have you seen Uschold's A900 pre-test in the BJP? He praises the 24-70mm but says it has strong barrel distortion at 24mm and needs stopping down to f4 to improve sharpness - and that it has sudden vignetting fall off in the corners at 24mm on full frame, much like the 16-80mm CZ on APS-C. I am feeling happier by the day with my 17-35mm and 28-75mm KM D lenses!

David

[Javelin]

Oh oh... Your going to be written about in 100 notebooks at the next Sony board meeting

"I am feeling happier by the day with my 17-35mm and 28-75mm KM D lenses!

David

[lock]

"I'm going to try this one for sure."
That 's what I said and I did it before sending in the body.
I tried plain iso 3200 (1), iso 6400 +1 stop (2), and iso 6400 +2 stops (3). The last two were pulled back in Sony software with 1 or 2 stops to correct the exposure.
I used the 18-70 to take the pictures as my Tamron also went back to the factory for adjustments. A-mode, F6.3 at 1/50 sec, center weight metering.
I did use a lot of light/dark objects, clothes spread across the room/chair, to get as much noise into the scene given the exposures. I think these pictures do challenge the sensor much more than many of the cat shots presented here.

First the pictures without any NR, straight from raw but cropped in the order mentioned:
1)
http://farm4.static.flickr.com/3058/291 ... 96dc_o.jpg
2)
http://farm4.static.flickr.com/3273/291 ... 89ed_o.jpg
3)
http://farm4.static.flickr.com/3014/291 ... 44cc_o.jpg

Now with a default NR mode in Neat Image, the 'blotchy noise' option in the advanced menu:
1)
http://farm4.static.flickr.com/3288/291 ... f943_o.jpg
2)
http://farm4.static.flickr.com/3023/291 ... 8e96_o.jpg
3)
http://farm4.static.flickr.com/3010/291 ... 8d74_o.jpg


I'm not going to judge the pictures in terms of retained details, as in these conditions the lens is surely a limiting factor compared to e.g. my Tamron. If I look at noise however, the iso 6400+2 stops is the best, followed by the iso 3200. After NR the order is much the same but the differences between the two do become very small.
So it seems to work, but I never get the low noise levels Sonolta is showing us with his A700.

There are two questions I would like to ask you guys. I want to know why more light at a higher sensitivity seems to be more effective than using a lower sensor sensitivity. In fact, it does not seem to work that well with only 1 stop extra, but it does with 2.
Is it because pulling back the exposure hides some noise in the darker scenes but uses the extra light to gain detail against noise ? Or is there a non linear relation between noise and iso going from 1600 to 6400 in the A700 ?

lock

[lock]

Oké I did some make some mistakes in typing it. I changed it now. I sent the camera in because of FF with any lens faster than F/4.

If I should have used iso 3200, presumably because it's better, then why isn't the iso 3200 picture better than the +2 stops iso 6400 picture ?

[lock]

Thank Sonolta. The NI noise reduction was set to a default mode to simply make things comparable. Surely it isn't max. effect.

I was posting iso6400 +2 EV is a said I did. The factt hat you never posted something like this is very well possible. Most of your fantastic iso 6400 pictures are evenly lit or the crops of evenly lit parts. The same accounts for DK's catshots. The one cat lying on the bed and the 100crop at page 4 does not contain that many conrasting dark and bright areas as mine do. So I understand that mine are worse.
But they still serve the purpose of checking what the effect of the overexposure trick may be as there is 'enough noise' in the relativwely underexposed parts.

You said with iso 6400 it would lose fine detail compared to iso 3200. That was one of the surprises actually. I've got more detail in the iso 6400 picture with +2 EV than I got in the iso 3200.

If the camera comes back I will try more of this, as my two cats are quite a challenge: tricolore and a chocolate brown one. That's going to test some noise...

[InTheSky]

I will say that is the best picture i have been able to get from ISO6400 so far ... and this is probably what is missing in the A900, High ISO capability.



Full Sized image there :
http://www.pbase.com/nadeauf/image/1042 ... iginal.jpg

For ISO3200, tonight i have try some shots , and this is my best so far, i think this can become a good 6MPix noise free to eye ... but no more .


Full size again : http://www.pbase.com/nadeauf/image/1042 ... iginal.jpg

For ISO1600, it's depend of the situation, but we can get very good result.

Regards,

Frank

[bakubo]

I've got more detail in the iso 6400 picture with +2 EV than I got in the iso 3200.

This seems a bit strange to me. Why shoot at ISO 6400 with +2 EV instead of just shooting at ISO 1600?

[lock]

Correct. But Sonolta also mentioned he overexposed up to 2 stops and I simply did the same. It did it to find out how iso 6400 + 2 ev pulled back in the convertor with two EV (to get a normal expusere again) would compare any standard iso 6400. But in stead of comparing it with the standard iso 6400 (0 EV), I compared it with a standard iso 3200 picture and an iso 6400 +1 EV. The iso 6400 +2 ev was at least equal or better than the iso 3200 picture. I found this pretty weird.

But now that you said it: if you can do iso 6400 +2 EV you effectively mimic an iso 1600 shutter speed picture give that specific aperture, but with a two stop extra sensitivity. This causes the overexposure but you can pull that back with the convertor. This is why the iso 6400+2 ev was better than the standard iso 3200 picture: it's effectively an iso 1600 picture.