Epson 4990: multisample in single pass?

[RobertoR]

I tried multisample with my 4990 and AI STUDIO IT8. It works very well, but scanning 4x5" is very time consuming. (4x)

Is not possible implement something like single-pass with multisample?

Regards and thanks for your attention,
Roberto

[RAG]

Hello Roberto,

I like the idea, but I think LaserSoft would have to make the scanner for this feature, because I believe it would require a hardware modification. If the multisampling were only done in software from one optical sample there would have to be an interpolation, don't you agree?

[RobertoR]

Hi RAG,

i was thinking something like multiple reading (speed of course is 1/4) in one scan only, not interpolated. In this situation the total scanning time should be lesser than actual "multiscan".

It may be hardware dependent, but my hope is that 4990 could do that !!!

Regards,
Roberto

[RAG]

Roberto,

I'm not sure I understand how multiple software reads from the same optically captured data would achieve the same thing as multiple optical scans. Please explain your idea.

Thanks!

[RobertoR]

It may be a stupid idea: read an array of lines, don't advance with step control, read for other n-times and then proceed ....

It may be also that my english is so poor that i can explain myself well.

Roberto

[RAG]

No,

It is not your English or ability to express your idea at all. I am just trying to imagine how the process would work, but it is not clicking for me. I think of multi-sampling as being similar to doing bracketed shots with a camera and don't see how I would get different results without any of the variables changing.

[sbosman]

VEry simple, actually. Not changing the variables is exactly the idea behind multi-sampling. The noise is random, so it changes all the times, meaning that two otherwizes identical reading would be different only by the noise, making it easier to identify what is noise and what is not noise.

[RAG]

sbosman,

Forgive my ignorance on this one. Is the implication that noise is introduced when the software receives information from the hardware? I thought it was introduced when the hardware obtains the information.

[degrub]

multisampling over time addresses the issue of random electronic noise in the sensor electronics and the fact that the ccd is not cooled to very low temperatures. If you look at CCD arrays used for astrophotography, the best use peltier cooling to reduce the electronic noise levels. This is important because the final images in astrophotography are an addition of many many images and any noise would show up as extra stars ! For us mere mortals, taking multiple samples of the same image and averaging the results cancels out the electronic noise and as a result increases the dynamic range by about 1/2 bit per 2 to the N samples. So 16x (2 to the 4th)multisampling can give you up to 2 bits more dynamic range of the capture - assuming the noise is purely random.

hope this helps.

regards,

[RAG]

For clarification, the multi-sampling you are talking about is (as I've been alluding to) done by having the hardware look at the image multiple times, correct?

[degrub]

doesn't have to. Software can drive it. But the scanner head cannot move !