Optimal Scanning Resolution

[widevista820]

Simple and to the point. I am having my 8”x20” negatives drum scanned to fine tune for Platinum Palladium printing. The files are tweaked in Photoshop then printed on Ultra Premium OHP Transparency Film. The drum scanner operator wanted to know what resolution to scan on. After some research and a few emails I determined that the 780 dpi setting on the drum scanner was ideal. He had initially suggested 1440 yet I think that is overkill. The negatives I am printing on the Epson are 8”x20”. So I would gather that 780 would be good as I am not enlarging. I did output one 8”x20” negative from the first scan and the sharpness and detail look fine. I will not really know until my Amergraph frame arrives in a week. Any one else out there drum scanning 8”x20” negatives?

-Bruce

[Bruce Watson]

You aren't making any enlargement, yes? So the print is the same size as the film. In that case, 780 spi should be fine. You could probably easily get away with something in the range of 300-360 spi, or 450-540 spi if you want to over sample a bit. You'd have to try them out and see what the final prints look like to know for sure.

As a drum scanner owner/operator I have to ask: Why do you want drum scanning for this application? I would have thought this a perfect application for a flatbed scan. Stitched if you need it. An Epson 7xx scanner could do this in two 8x10 pieces (three if you want a fair amount of overlap for stitching) and you could stitch them back together in Photoshop or some other application. I'm asking because it's expensive and I don't imagine you'll see much if any quality improvement since we're talking about a 1x enlargement here.

[Ben Syverson]

Fascinated by this analog -> digital -> analog "round trip." Will you post some pictures of the finished product when it's done?

[Don Hutton]

You aren't making any enlargement, yes? So the print is the same size as the film. In that case, 780 spi should be fine. You could probably easily get away with something in the range of 300-360 spi, or 450-540 spi if you want to over sample a bit. You'd have to try them out and see what the final prints look like to know for sure.

As a drum scanner owner/operator I have to ask: Why do you want drum scanning for this application? I would have thought this a perfect application for a flatbed scan. Stitched if you need it. An Epson 7xx scanner could do this in two 8x10 pieces (three if you want a fair amount of overlap for stitching) and you could stitch them back together in Photoshop or some other application. I'm asking because it's expensive and I don't imagine you'll see much if any quality improvement since we're talking about a 1x enlargement here.

Microcontrast of a drum scan will reveal all the nuances contained in the original - a consumer flatbed will lose some... It's not all just about resolution.

[PenGun]

Microcontrast of a drum scan will reveal all the nuances contained in the original - a consumer flatbed will lose some... It's not all just about resolution.

PMTs are nice but "all"?

[IanMazursky]

I always tell people to scan a little bit more then the current application requires.
Ive ran into this a number of times this month with a few clients. One wants an LVT 8x10 from a 100mb scan we did a while back.
I did tell them to scan for the future! You never know when you may have to repurpose it for a larger print or for archival purposes.

I would tell the operator to scan the film at the optical resolution (in relation to the aperture) to the spi you want. Depending on how sharp you want the grain, the aperture can be adjusted accordingly.
Im not sure which scanner they are using, maybe a hell 3000 series or a howtek 7500 (my vote and my favorite scanner next to the aztek premier).
The best results are usually obtained when youre not interpolating the scan by picking an arbitrary spi. Especially on smaller scan sizes when youre not near the far end of the scanners capabilities.

[Bruce Watson]

Microcontrast of a drum scan will reveal all the nuances contained in the original - a consumer flatbed will lose some... It's not all just about resolution.

For 780 spi you're talking an aperture in the range of, say, 32 microns. Such a scan is not likely going to recover all the information contained on the negative. In particular, it's going to leave some of the fine detail out of the scan file. This includes some of the microcontrast information. That's just the laws of physics.

Such a scan file will likely make a print that's nearly indistinguishable from a contact print at normal viewing distances. But you'll get very similar results from a flatbed scan -- nearly indistinguishable from a contact print at normal viewing distances, and nearly indistinguishable from a print from a drum scan, at normal viewing distances. Much depends on the operator, techniques used, etc.

[Lenny Eiger]

I always tell people to scan a little bit more then the current application requires.
I would tell the operator to scan the film at the optical resolution (in relation to the aperture) to the spi you want. Depending on how sharp you want the grain, the aperture can be adjusted accordingly.
Im not sure which scanner they are using, maybe a hell 3000 series or a howtek 7500 (my vote and my favorite scanner next to the aztek premier).
The best results are usually obtained when youre not interpolating the scan by picking an arbitrary spi.

Ian, I agree with you - it's always better to scan for not only printing but archiving as well.

However, my results don't agree with yours on matching the ppi to the scan aperture. I don't want to hijack this read... but it is generally regarded that a scan do do better than Photoshop in moving things up to size. Scanning a med format neg, for example, at 13 microns, would convert to a ppi of only 2000. I get exceptional quality scanning that neg at 8000 ppi. There is a slight advantage to doing it at 2000, but I think it is only a percent or two. Looking at the grain interpolating in PhotoShop will degrade one's image by 30 to 40% in comparison. (These are very subjective numbers, meant to be a comparison of my impressions only) It is always worth it, in my opinion, to set the aperture to match the grain effect you want and generate the amount of pixels you want.

Lenny

[widevista820]

Thanks to all who replied. I am indeed taking an 8"x20" negative and coming back out of the Epson with another 8"x20" negative. The reason I am scanning is to gain the ability to increase contrast for producing a Platinum Palladium print. The first negative thus far has produced a negative on the same quality as the original negative. Not real sure if the resolution is doubled if the results could be seen on an 8"x20 contact print.

-Bruce

[Ed Richards]

> Microcontrast of a drum scan will reveal all the nuances contained in the original - a consumer flatbed will lose some... It's not all just about resolution.

I thought the dynamic range of PT printing was about 1:64 tops, and more like 1:32, plus he is upping the contrast. I do not think micro contrast matters for this application. If money does not matter, might as well get the drum scan. But I am with Lenny on this - might as well also get a higher rez scan and get the operator to give you a downsampled scan as well if you cannot handle the full scan on your computer. You never know when you might decide you want to do a really big negative and a huge print. If money does matter, go with Bruce.

[SCHWARZZEIT]

How does the micron value of a drum scanner's aperture actually translate to the physical aperture of the scanner? The resolution aperture is obviously a circular hole. Is the micron value the radius or the diameter of that hole, or is it just a way of expressing the smallest resolvable spot size of the corresponding aperture?

-Dominique

[IanMazursky]

Ian, I agree with you - it's always better to scan for not only printing but archiving as well.

However, my results don't agree with yours on matching the ppi to the scan aperture. I don't want to hijack this read... but it is generally regarded that a scan do do better than Photoshop in moving things up to size. Scanning a med format neg, for example, at 13 microns, would convert to a ppi of only 2000. I get exceptional quality scanning that neg at 8000 ppi. There is a slight advantage to doing it at 2000, but I think it is only a percent or two. Looking at the grain interpolating in PhotoShop will degrade one's image by 30 to 40% in comparison. (These are very subjective numbers, meant to be a comparison of my impressions only) It is always worth it, in my opinion, to set the aperture to match the grain effect you want and generate the amount of pixels you want.

Lenny

Hi Lenny,
For medium format i definitely agree with you. I do the same here but the large and ULF negatives that i can do on my Howtek 7500 (16x20" max) don't need 5000ppi at 5uM.
Although i can do it, not that i would want to spot that file :D Also, stopping the aperture down will only exaggerate the grain in this case.
Usually i find that for those kind of negative, matching the ppi to the aperture works extremely well. I have seen no real advantage to make the aperture smaller.
I probably did not explain my point in my first post very well. My mother always said, don't post hungry ;) or something along those lines.

[IanMazursky]

How does the micron value of a drum scanner's aperture actually translate to the physical aperture of the scanner? The resolution aperture is obviously a circular hole. Is the micron value the radius or the diameter of that hole, or is it just a way of expressing the smallest resolvable spot size of the corresponding aperture?

-Dominique

Hi Dominique,

The stated aperture value is in Microns. It directly relates to the laser drilled hole in the aperture wheel.
Usually they have 13, 17 or more apertures but most of them are at the larger/less useful end (lower ppi ie. 100-600ppi).
The 7500 (13 steps), has steps on the small end from 20.32µm (1250 ppi) to 15.24µm (1666.67 ppi) to 10.16µm (2500 ppi) to 5.08µm (5000 ppi) .
The premier (17 steps) has steps on the small end from 12.7µm (2000 ppi) , 9.5µm (2667 ppi) , 7.9µm (3200 ppi) , 6.4µm (4000 ppi) and 3.2µm (8000 ppi).

I am not 100% sure but i think its the diameter of the hole. I have a few extra aperture wheels from my howtek 7500 (spare parts) but i have no way to measure something that small...yet!
I find it amazing that someone actually figured this out!

[widevista820]

So I am wondering where the point of diminishing returns is here? I ran a drum scanner many years back. A DS Screen. Very large drum. At the time all the scans were for halftone film output so calculating optimum resolution was not a huge issue. In this instance I have no idea where the point of diminishing returns comes into play. The difference between a 780ppi and say 1560ppi is a huge hit in the wallet. The charge for 780 res is $150. Double the resolution and we are into $450 per scan. This is on a Howtech scanner. From what I can tell the 780 res coming out of the Epson at 16 bit with 2880 res looks good. I will know here within a week. I will also run a test this week to see if doubling the res on the scanner will make any difference. Updates to follow.

[SCHWARZZEIT]

Hi Ian,

thanks for the explanation.

The values you calculated are the exact step or sample widths for the equivalent resolutions. If the micron values were the diameter of the sampling spot then the scanner could optically resolve a much higher ppi number because the sample spots can overlap and still resolve detail. I imagine that it works similar to diffraction limited optics where according to the Rayleigh Criterion the smallest resolvable detail is one half of the diffraction spot diameter.
Overlapping sampling spots reduce contrast in these high spatial frequencies but it also reduces aliasing. Thus I think it's not that important to exactly match the spot size and scanning resolution but there are less optimum combination, e.g. the spot size is smaller than the sample width causing aliasing or the spot size is too large reducing sharpness and optical resolution.

-Dominique

[sanking]

So I am wondering where the point of diminishing returns is here?


I have scanned many ULF negagtives of 11X14, 7X17 and 12X20 in size and printed from them with digital negatives in various processes, including carbon transfer, vandyke, kallitype and pt/pd. For making same size digital negatives scanning at 720 spi will give you way more than enough information and scanning at higher resolution would be overkill. What some of these folks who are advocating scanning at higher resolution don't seem to grasp is that your negative on OHP will limit resolution to about 360 spi, no matter how much resolution you have in the file. It is just like when you are printing on paper in that the printer limits the resolution to 360 spi, or 480 spi at the highest. And if the negative did not limit resolution the texture of your final support with pt/pd will also limit resolution to no more than 480 spi.

Sandy King

[widevista820]

Thanks Sandy! That is what I needed to hear. The explanation of printing on the OHP is a bit confusing yet I had I had calculated that a higher resolution would not make a bit of difference. The initial scans looked a bit soft in Photoshop. The main reason is that there was no sharpening on the drum scanner. The scanner operator recommended sharpening afterwards in Photoshop as to allow for more flexibility. Now I need to get deeper into the sharpening techniques in Photoshop. On a side note I will also get back on the carbon discussions. I did make a go of some early Carbon work and had varying levels of success. I still have the carbon tissue from B&S. Soon as I get my Amergraph ULF28 I can try some carbon work. Thanks again Sandy!

-Bruce

[Lenny Eiger]

The difference between a 780ppi and say 1560ppi is a huge hit in the wallet. The charge for 780 res is $150. Double the resolution and we are into $450 per scan. This is on a Howtech scanner. From what I can tell the 780 res coming out of the Epson at 16 bit with 2880 res looks good. I will know here within a week. I will also run a test this week to see if doubling the res on the scanner will make any difference. Updates to follow.

This is precisely why I charge a set fee based on the original so that people can get a high-rez scan for a reasonable amount. The difference between a 2000ppi scan and a 4000ppi is a few more minutes on the scanner - its doing the work.

There are reasonable limits to how much ppi one needs, of course. I think going past 4000 ppi with a 4x5 is rarely necessary, for example. Going to 8,000 yields a 6 Gig file. Lately, I have been running all my 8x10's at 2666. It gives me 3 Gigs of color, or a Gig of b&w. I run all med and small format film at 8000.

Contact printing is very efficient. One needs very little in comparison to enlarging.... to much greater sizes...

Lenny

[Vertex Ninja]

... What some of these folks who are advocating scanning at higher resolution don't seem to grasp is that your negative on OHP will limit resolution to about 360 spi, no matter how much resolution you have in the file. It is just like when you are printing on paper in that the printer limits the resolution to 360 spi, or 480 spi at the highest. And if the negative did not limit resolution the texture of your final support with pt/pd will also limit resolution to no more than 480 spi.

Sandy King

Everyone here has more experience than me... everyone. However, I just wanted to throw my $.02 in. It is my understanding that on Epson printers, the native resolution is 360PPI. Anything bigger than that gets downsampled by the printer and anything smaller gets interpolated up. There is a check box for "finest detail", If this is checked then the printer accepts 720PPI and again anything smaller or larger gets interpolated. HP and Canon supposedly use 300/600. There is probably near 0 visual difference, but it might be better to send the printer what it wants rather than allowing it to decide for you. I have no idea what 3rd party rips do. I can only say, that on my 3800, I've noticed slightly better results prepping the file myself vs letting the printer handle it.