Epson 4870 & NGC 5139

Telescope: Pentax 105SDHF (D=105mm, f=700mm)
Tracking: Takahashi NJP mount, 6cm/F12 guidescope, manual guiding
Camera: Pentax 67
Film: Konica Centuria 400
Exposure Time: 55 min

Resolution

This is the very central 1.86cm x 1.43 cm of the 67 frame. Image is scanned by Epson 4870 Photo scanner at 4800 dpi and 16 bit per channel, with Digital ICE off. The above image is downsized to 1600 dpi and processed in PhotoShop CS with Curves and Levels only. A 2400 dpi version can be found here. The stars have FWHM of ~15 um and are oversampled at 4800 dpi (~5 um per pixel). In this case, 2400 dpi is pretty enough. On the other hand, I can imaging some pictures taken with fine-grain films and shorter focal lengths may require 3200 dpi. Nevertheless, both 2400 dpi and 3200 dpi seem to be piece of cake for this scanner.

Noise

To characterize the scanner noise, I made three scans of an identical region. Any common features in these three scans must be something real on the film (real celestial objects or film noise = grains). Signal that changes in these three scans should be considered as noise coming from the scanner. Such noise signal should cancel out after the three scans are averaged. More precisely, the amplitude of the scanner noise should decrease by factor of 1.7 after the three scans are averaged. If the single scan looks identical to the averaged one, we can say that there is no scanner noise. If the single scan and averaged one look very different, we conclude that the scanner noise is very strong. Below shows the single scan (one of the three) and the averaged images.

Shadow	Highlight
mouse on: averaged image; mouse off: single scan

The scanner noise appears to be stronger in the highlight area. This is not surprising at all because Centuria 400 is a negative film. From the scanner's point of view, there is less photon coming from the highligh on negatives. If we are scanning transparencies, we will see the opposite.

The scanner noise of Epson 4870 at 4800 dpi seems negligible for most applications. However, I will still scan the film forat least two times and average them to get the best results. The above noise test images were scanned with 4800 dpi. I did similar tests at 2400 dpi. There is a barely noticeable decrease in noise, comparing to 4800 dpi. At 2400 dpi, more than one scans are still needed to achieve the best results.

Color Accuracy

I don't care.

Other Impressions

At >1200 dpi, Digital ICS doesn't work on negatives and reflection documents. Slides are not affected by this. Although I mostly use slides, I am not too happy with the fact that Digital ICS sometimes doesn't work.

Scanning speed is OK, although I always hope it faster. But, sometimes it takes one or two minutes to warm up before it scans. This is the part I hate.

The film holders are nicely built.

The Epson scan software is nice. It provides most of the controls I want. There is a function called "densitometer" but it is certainly not a densitometer. It shows pixel values, but not diffuse densities. From a photographer's point of view, it would be nice to have a real densitometer. This function should not be difficult to built. Anyway, the "densitometer" of the Epson software is still useful for pre-scan adjustment.

I haven't tried SilverFest yet, and probably I won't.