9.9.08

B&W Scale Tests

After a break I'm back. Wasting gigabytes and pixels in abundance. As usual i have not even left my front room to perform these experiments. No point in going 'live' until a secure working method is resolved. In this case i am seeking to increase speed of capture and get super detail images. The speed part is controlled much like a shutter except each line of the scan is expressed as Line-Time. With a stopped down lens and the infrared filter installed this often is around 1/15th to 1/50th at approx 1000+ ISO, the cumulative effect is often five plus minutes per image scan. The ISO seems high but with the scanning back you can go quite high before any serious attention to noise is required thanks to the purity of the data. The newer BetterLight backs boast a faster (natively more sensitive) sensor that can operate at higher ISO's without introducing image noise. Alas Im not buying a new back. But by removing the infrared filter Im able to up the line-time dramatically. This comes at the expense of a colour image and normal contrast. This is only an early image, but it is constructed from six scans that were taken within about thirteen minutes. Open this image in a separate tab - it's over 7mb.
Window: 1/65th : f32 : iso758

This image is very much like an Orthochromatic image, except it's made primarily from red data. This seems to effect the contrast adversely requiring serious tweaking with curves in PS to get an image that looks vaguely normal. Even with these limitations, when viewed at 100% the level visual information is awesome. This mostly thanks to the Rodenstock Sironar S 210mm lens and delicate sharpening. This is yet to be refined even more but the core contrast of the image effects the sharpening, this symbiotic relationship makes editing a slow and ram stealing process. Another minor limiting factor here is Im shooting through a double glazed window.

The initial idea was to be able to use the sensitivity to IR as an advantage, so i could shoot in dimmer condition or possibly at night at the cost of colour and outright image quality. But then i noticed the lower scan times during daylight.
This demonstrates the benefits of IR at night. An IR filtered version would be virtually black and take considerable averaging and editing to get something visible.
Leaving the camera overnight to shoot the scene 30+ times offers enough data to create a well averaged image thus having little noise. There is also the possibility of adding the images together to simulate a super long exposure. A good overview of this technique and the software involved is here. Hopefully i'll have some tests up by the end of the week.