Does the Alpha 700 really do ISO 12,800?
I HAVE set up a pBase gallery with a large number of comparison studio shots, all the way from ISO 100 to the maximum, using the A700 and Nikon D300. It has some minor discrepancies in shooting conditions which I need to explain to anyone visiting the gallery.
First of all, the lens used on the Nikon was a 24-70mm f/2.8 AF-S G ED . It has been suggested that using this lens, at f/11, while the Alpha 700 was fitted with a Konica Minolta 28-75mm f/2.8 (D) at the same aperture, amounts to favouring the Nikon. The Nikon lens sells for a few pence short of Â£1,200 in most places and the 28-75mm was acquired for Â£190 used. But Sony has no lens in any way similar to the Nikon, and this is the closest I could find.
It has also been suggested that to show resolution, the very best glass should be used. If the 24-70mm Nikkor at 50mm and f/11 and a comfortable shooting distance does not perform well, I have no idea what better glass might be suggested! It is the purpose of this lens to perform staggeringly well on digital cameras. That is why it’s huge, heavy and costs as much as the DSLR body it was mounted on.
Secondly, my framing for the Nikon is slightly closer. Both lenses were set to exactly 50mm and report exactly 50mm to the cameras in EXIF. The Nikon has a 100 per cent viewfinder. I have tested it, and it’s true. It really is accurate. The Alpha 700 has a 95 per cent view. Through the finders, everything looked identical. If I repeat the test I’ll use the review screen to adjust the image size more accurately.
Finally, the Nikon camera gave what appears to be double the exposure despite the ISO 100 and 200 flash shots showing the Nikon D300 sensitivity (at least to flash) is lower than the Alpha 700. The files when converted to the same colour space are very similar in density. It looks as if Nikon 6400 by tungsten light may be equal to Sony 3200.
Here is a comparison between the ISO 6400 NR ‘NORMAL’ frames from the Alpha 700, top, and Nikon D300, bottom. For this, the AdobeRGB JPEG has been opened with conversion directly to sRGB (Bridge/Photoshop) including black point compensation, perceptual modelling. No other adjustment of any kind has been made. Since sRGB is a smaller colour space, the conversion may compress gamut but does not artificially change the original Nikon shot:
Click on this to load my full size screen shot (sRGB JPEG) from a 24 inch iMac screen. For the sake of helping you judge, the A700 colour is dead accurate – the D300 colour is a very warm, saturated version of the real subject.
Please note that the exposure for the Alpha 700, top, was 1/60th at f/11; for the Nikon D300, 1/30th at f/11. Conditions were absolutely identical. Far from being lighter than the Alpha 700 shot, the Nikon file correctly viewed with the same colour profile is similar in density, with better colour saturation and a warmer balance (not such effective tungsten WB, both set to the manual preset). It actually has denser shadows (the ‘black eye’ of the mask) and the Alpha 700 has greater lift in the three-quarter tone area (the black frogged thread surrounding the eye holes). The Alpha 700 file has mor artefacts and more noise, but concealed in this is a good level of printable detail.
So, which camera is giving the correct exposure? Fortunately, the entire set-up is still intact (and will be for testing my next Alpha 700 body when it arrives – my own not a loan!). So I was able to use the Minolta Flashmeter IV and take a meter reading. The incident light reading is 1/60th at f/11.2, which agrees perfectly with the Alpha 700 auto aperture priority exposure.
By this standard, the Nikon D300 is exposing at 3200 ISO not 6400 – even though 6400 is set on the camera, the auto exposure system is treating things as if the sensitivity is 3200, and the resulting file certainly has a density which matches the exposure.
I am not here to find fault with other makes (I spend enough time finding out what’s right or wrong with our own chosen system) but it’s been my job for many years to spot things like this and make them known, even if I do not have an explanation for them.
What is clear is that the Alpha 700 will probably give the same actual shutter speed and aperture at the ISO 3200 setting as the Nikon D300 does at 6400, which means that the doubtful or troubled Alpha owner, wondering whether they should have waited for a D300 for the superior high ISO NR, should be comparing 3200 with 6400. In Nikon’s exposure and sensitivity terms the Alpha runs to 12,800. In Minolta Flashmeter IV terms, it’s an accurate 6400.
A possible reason why it’s happening
Setting any WB preset other than Daylight may affect the sensitivity of a DSLR, and the noise level. Using tungsten light always increases visible noise. Why? When you set Tungsten WB, the raw file remains untouched. It stays just like a daylight raw file, the channel levels are not altered. All that takes place in JPEG conversion. The blue channel is boosted relative to red (green remains about the same) to correct the colour balance. Changing the ‘levels’ on the blue channel increases blue noise in the darker tones. Alternatively, it is possible to CUT the red channel level, if the camera has a system which instructs it to give extra exposure.
To balance for Tungsten by cutting the red channel, an exposure increase of around one stop is needed. You can only do this if the camera is using a true 14-bit conversion, and the Nikon D300 is. With 8, 10 or 12 bit conversions there is not enough headroom to give a stop more exposure without risking burned out highlights. So such cameras must always boost the blue channel and keep the exposure unchanged. This results in the classic shadow noise look you often see in tungsten room lighting when an attempt is made to remove the colour cast. It’s also why such cameras only go half way to correcting the cast!
Is Nikon doing something radical to eliminate noise by increasing exposure when Tungsten balance is set, then using red channel cut instead of blue channel boost? The digital equivalent of fitting an 80B conversion filter?
Well, I’ll have to find out. If they are it is clever but deceptive, because it would make a hand-held meter reading inapplicable to the D300. The TTL metering reading clearly works.
Sony, for certain, is not altering the physical exposure when you use 6400 in tungsten light. It certainly is boosting blue to make the correction, as the noise in the dark ‘black eye’ (grey, due to chromatic noise reduction) shows. For stadium sports, concerts and similar jobs you might routinely be getting twice as fast a shutter speed with the Alpha 700 – or the ability to set half the ISO.
– David Kilpatrick