Reducing A900's noise with filters

[01af]

Yesterday I tried using filters to reduce the A900's low-ISO noise in typical landscape shots. My result: Yes, filters can reduce low-ISO noise indeed. But no, it still doesn't make any sense in real life.

I compared shots taken with no filter, with a Kodak Wratten CC30M colour-compensation gel filter, and with a Minolta #85 colour-conversion glass filter, at ISO 100/21° and at ISO 200/24°.

Looking at the histograms as displayed by Rawnalyze, without any filter the green channel has the widest range of tones (i. e. it will clip first on over-exposure), the blue channel covers a medium range, and the red channel covers the narrowest range of tones. This means the red channel basically is under-exposed when the green channel is exposed to the right, which leads to some noise in the red channel even at low ISO settings.

A magenta filter blocks green, so it's possible to give more exposure which will benefit the blue and red channels without blowing the green channel. With a CC30M, the blue channel now lines up nicely with the green channel but the red channel still is low, albeit not as low as it was without filter. In the ISO 200/24° shots, noise in the blue sky is less than in the unfiltered frame. Of course, with filter the exposure is about the same as at ISO 100/21° without filter---and at ISO 100/21°, there is virtually no apparent difference in the noise levels in the blue sky. So the benefit of the magenta filter is purely academical; it's mostly pointless to use the filter for practical intents and purposes.

A warming colour-conversion filter reduces blue and green, and blue more than green. Consequently, when exposing the green channel to the right through the #85 filter, the blue channel is lower and the red channel is higher than without filter. In fact, blue and red now line up. The effect apparent in the image is similar to the magenta filter: at ISO 200/24°, the blue-sky noise is reduced, however not as much as with the magenta filter. The effect is about half-way between no filter and magenta filter. And at ISO 100/21° the effect again is virtually zero.

From looking at the histograms, the CC30M and the #85 filters combined should bring all channels in line which in theory would mean maximal dynamic range and minimal noise. But I didn't try that, and I don't plan to. The benefit is not worth the effort, and it's impractical to juggle all those filters when shooting out in the field. Gel filters are extremely delicate; they are meant for studio use, not for field use.

Stronger than the effect on low-ISO noise are the filters' effects on colour rendition. When adjusting the colour temperature and the green/magenta tint in Camera Raw's white balance so that the rendition of the green of the conifers is about equal then the blue of the sky is significantly different! Compared to the unfiltered shot, in the CC30M shot the sky appears much greener (Nikon style?), and in the #85 shot the blue of the sky is clearly purplish. I don't like either so I will keep shooting without filters.

-- Olaf

[David Kilpatrick]

Useful research Olaf. I have been wondering whether to use an 85, but also wondering whether exposure and raw conversin adjustments can achieve the same result.

Back in the days of scanning backs, each camera had a special filter to line up RGB perfectly in daylight. I still have two of these - a sort of green filter used by the Kaiser/Practika scanning back, and a pink one for Leaf. They are both rather small (52mm) for modern lenses as the standard lenses for these modules at the time were Nikon, and Nikon used 52mm as their standard thread.

I think that if anyone is going to issue custom pack RGB lineup filters for modern DSLRs, it will be Singh-Ray - snake oil merchants at the best, but they could do it.

David

[Mark K]

Thanks David.
The noise issue is really bugging me. I will take a very cautious decision for my own A900. I really wish Sony will give us another surprise by firmware upgrade like V4 for A700

[David Kilpatrick]

Try different raw converters. ACR is really pretty bad and so is IDC. Capture One Pro and Raw Developer, Raw Photo Processor, Raw Therapee may all make a big difference.

David

[artington]

David, would you mind elaborating on why you (apparently) dislike ACR? From my perspective it is free whereas Capture One Pro costs EUR 299 (and Capture One v 4 EUR 99). Now clearly you pays your money and takes your choice but as an amateur am I going to get worthwhile benefit from C1 vis a vis ACR? (on an A700).

[David Kilpatrick]

I use ACR and almost nothing else, but then, I hardly ever shoot over ISO 800. When I do, for Sony something different is needed. ACR produces lots of noise, coarse splodgy grain, low sharpness - it's very bad at high ISOs. The A700 firmware fix improved it a bit, but it equally improved the better results obtained from other converters. I use Iridient Raw Developer normally for high ISO. I have an NFR license for C1 Pro and I have attended press workshops etc, it's very impressive but I just can't use it for my workflow.

David

[bakubo]

David, would you mind elaborating on why you (apparently) dislike ACR? From my perspective it is free whereas Capture One Pro costs EUR 299 (and Capture One v 4 EUR 99). Now clearly you pays your money and takes your choice but as an amateur am I going to get worthwhile benefit from C1 vis a vis ACR? (on an A700).

You might want to consider using ACR for photos up to ISO 1600 and for those that are higher try the free Raw Therapee.