f64

[Kirk Gittings]

In short words. The noise you see in the Epson is probably related to digital manipulation vs the most likely analog nature in the drum scanners.

What digital manipulation? You can see it in the shadows of a simple linear scan from an Epson? The noise in the deep shadows in scans from Epson scanners is well known (and other scanners at this level) and obvious to see. That is why Silverfast offers the multi-exposure sampling feature http://www.silverfast.com/highlights...mpling/en.html What you are describing-that Epsons have less noise than Howteks-is contary to my experience. I'm sorry I don't get it. Even less so comparing Epson scans to an Aztec Premier.

Also, FWIW I have a fair amount of experience scanning. I have done my own scanning on Epsons for many hundreds of my color images for magazine articles, and probably a couple hundred on Imacons of my b&w images that have appeared in museum shows additionally I have scanned and printed a show of my sister-in-laws at the Smithsonian and I teach scanning on Epsons and Imacons as part of my classes at SAIC. If nothing else I am a pretty knowledgeable and demanding consumer of scans and know when I am getting a decent scan from myself or a drum scan operator.

[onnect17]

Kirk,
Digital manipulation does not imply is done by the operator. Could be done, as in this case, by the driver or the firmware and the manufactures will not tell you about it.
When I say linear I'm not referring to a curve in the screen, I'm talking about the output signal in the CCD before quantization and AFAIK the user do not have access to it in any Epson model, not even RAW data.
Multiexposure and multisampling are two different things. Silverfast is making money selling multiples scans or exposures as multisampling. The name is clear, takes multiple continuous samples and decide using average or any other criteria which is the final value of the pixel (Nikon coolscan uses it).
Multiexposure involves another scan introducing mechanical error and affecting resolution.
Another way to improve the SNR is control the power of the light source (analog gain in Nikonscan) or extend the duration of sampling (Adaptive light in FlexColor)

Manufacturers are in the business of making money and processing data by software is a lot cheaper and flexible than by hardware and 99% of the users do not see the difference in the product. A typical example are the digital cameras. Saving in optics and a lot of post processing.

I owned and own several scanners. Clicking behind a keyboard is not enough for me. I have to open it, see what's inside, understand the working principle, analyze the output and all have pros and cons, in hardware and software.

Glad to hear you teach the subject and I'm sure your students love the class. A simple and interesting task (if you are not doing it already) could be to scan a step wedge as a positive without any correction in two scanners using different technologies and discuss the results.

[Don Hutton]

I hope I can help the conversation with my opinion.
The level of noise in the howtek is a lot higher than the Epson. Period.

If that's the case, it's due to the Howtek operator setting the aperture inappropriately for the film which is resulting in extensive grain aliasing coupled with the fact that the Epson's lack of extreme sharpness (probably the result of very economical optics combined with fairy crude stepping mechanics) causes a smoothing blur to the finest of details (i.e. anything beyond the actual optical resolution of the Epson in question). At comparable optical resolution, the Howtek with a competent operator will produce a scan with less noise than an Epson flatbed. Once the "blurring" effect is reduced to a similar level of sharpness that the RAW output of the Howtek should be capable of, the Howtek scan will have obviously less noise.

[onnect17]

Don,
I'm not familiar with the RAW output in the howtek. I'm using DPL and tiff is the only option to me.

[Don Hutton]

Don,
I'm not familiar with the RAW output in the howtek. I'm using DPL and tiff is the only option to me.

By RAW, I simply meant unsharpened, unmanipulated scan (make the scan with a film profile - and no sharpening).

[onnect17]

The film profiles (not color profiles) are curves which are downloaded into the 12 bits LUTs of the scanner and is not longer raw data.

[Don Hutton]

The film profiles (not color profiles) are curves which are downloaded into the 12 bits LUTs of the scanner and is not longer raw data.

Fine - use a film profile and no sharpening - call it whatever you like - and you'll have a better result for a given optical resolution IF you have the aperture set appropriately AND your scanner is in good mechanical and electrical condition. More than half the drum scans I have seen from amateurs and professional operators have had the aperture set incorrectly (there's a tendency to stop it right down as small as possible - 6 microns on a 4500 - because operators don't appreciate the optical differences of "oversampling" a wider area). Some testing can reveal some very interesting facts about an optimal aperture for a given emulsion and subject matter. Evidence of testing seems to be lacking in a lot of the scans I've examined.

[onnect17]

I would say mine is OK usable condition, not better than that. I use the 13 microns setting because the amount of noise at 6 is just too much, at least in the D4000.
Also when I select a number above 13 the focus feels it. All this at 4000dpi with fresh, clean and aligned lamps using manual focus.
Never thought of taking the subject in consideration.

[sanking]

Peter,

Do you have any idea why the grain become less obtrusive as you scanned at higher resolution? And can you provide some details of the scan? Were you scanning in grayscale, and was it a dry or fluid mounted scan? And did you keep the scan at the scan resolution or down size before evaluating the grain.

Sandy

Recently, I've been scanning some of my old 35mm film with my Cezanne scanner. Some if it is on Kodak HIE film, which was very grainy film. I did comparison scans at 2000, 3000, 4000, 5000, and 6000 dpi. I sincerely doubt that any of the scans left out scene detail, since the film's grain was so coarse, and I doubted that there would be any benefit with the higher res scans, but there was! As the res got higher, the grain became much less obtrusive, and the 6000 dpi scan was clearly the best.

[Peter De Smidt]

Hi Sandy,

It might be due to grain aliasing, although I really don't know. I scanned "dry" using a clam shell holder. I scanned in 16-bit RGB as a positive, and I down-res'd the larger scans using Photoshop's bicubric interpolation to match the lowest one.