Epson 3200 Report and Question

[john_hennessy1]

First, I measured the resolution of my Epson 3200 scanner using a

glass usaf target and can report its resolution is about as

advertised. Vertical line pairs are easily distinguished up to 32.0

lp/mm or more, but unfortunately horizontal line are blurred past 28.5

lp/mm. To measure ppi I used the ruler in photoshop which only goes as

fine as .001" but the degree of line pair distinction is subjective

anyway. On that basis, the vertical pairs are about 4000 ppi and

horizontal about 2800 ppi. It makes sense to me that the stepping

motor is the weak link making the horizontal lines blur together sooner.

 

Then I scanned a 4x5 Stouffer 31 step wedge with VueScan, SilverFast

SE, and the Epson software and with various profiles. My theory is

that the best way to measure color neutrality is to note the

difference between the highest and lowest of the RGB info pallette

values in all visible density patches and the software that produced

the smallest differences is the best. Don�t know if that is good

theory or not. The worst was the Epson software with whatever profile

it uses followed by SilverFast SE and whatever its default profile is.

The best was VueScan before I profiled it and next best was VueScan

after I profiled with its profiler. Go figure! The most consistent but

with differences a little high was VueScan with a profile made with

Wiziwig software. The absolute best was my Microtek ScanMaker 5 (an

antique in the electronic world) with its software and a profile made

with its profiler. The high differences were about 10 to 15 256ths.

That of course is readily visible. The Microtek had no differences

above 3 and looks dead neutral. Also the default profile Epson and

Silverfast software produced differences in opposite directions from

light to dark patches. E.g., way red on one end and way cyan on the

other. Untested unfortunately is the full version of SilverFast.

 

I also have two questions for VueScan users: 1) where�s the setting

for scaling? I.e., to scan a 4 x 5 image at 100%. It is doubling the

size and halving the resolution. 2) There is no histogram with sliders

to set new high and/or low values. If its there somewhere I can�t find

it and if not I don�t how else to do that.

[leonard_evens]

Thanks for the interesting results.

 

I have one question and two possible answers.

 

Question.

 

What do you mean by checking the ppi? If you scan at 3200 ppi, then that is what you are going to get. You can verify that pretty closely by looking at the size of the scan in pixels and dividing by the dimensions of the frame. If you mean you used the Photoshop ruler and counted line pairs you can distinguish, that is another matter, but I don't see how you could have come up with 4000 ppi and 2800 ppi. Also, that is not consistent with your lp/mm figures. In this context, you should note that Photoshop can be set to use pixels as the units of measurement in the ruler. That is one way you can check resolution. (I have a 2450 where I've done similar calculations.) 3200 ppi is about 126 pixels per mm, so 1 pixel is about 1/126 mm and 20 pixels is about 20/126 mm. Count how many line pairs you can distinguish in 20 pixels, using the Photoshop ruler; say you get 5 . 5/(20/126) = 31.5 lp/mm. (That is only an illustration, not an actual measurement.)

 

Answers?

 

1) I don't know what you mean about scaling in Vuescan and percentages. Under the Device tab, choose advanced and then set the scan resolution to 3200, which is the maximum optical resolution of the scanner. You can also scan at 1600 or lower if you choose, but many of us always scan at the maximal possible optical resolution and rescale if necessary in a photoeditor. If you halve the scanning resolution, you don't double the size of the resulting image in pixels. You halve each dimension, of course. So I think you are using the terminology differently from the way I understand it.

 

2) Look at the tabs. There are histograms both for the preview and for the scan. Under the Color tab, use the white point and black point sliders to adjust how much is clipped at either end. Each time you do that, the histogram will be redrawn showing the effect of your changes.

[kelly_flanigan1]

John; these are the same numbers I got with my Epson 2450 photo; when scanning a reference negative ; about 1 year ago. I posted the results last year; and again on a LF thread a few days ago. These results are with the stock plain jane Epson software.<BR><BR><a href="http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=0051W6">"My Epson 2450 photo scanner was used to scan a reference negative; of an old lens test that resolves 80 lines/mm . When the negative is scanned with a Epson 2450 scanner; the resolution is 31 lines/mm along the direction of travel; and 28 lines/mm 90 degrees from direction of travel. A perfect non obtainable scanner at 2400 pixels/inch should approach 2400/(25.4 * 2)= 47 lines/mm . My epson scanner's resolution is 66% ; and 60 % of the theoretical number.<BR><BR>

 

The same reference negative scanned with my Canon FS4000US film scanner reads 63 lines/mm. A perfect 4000 pixel/inch scanner should approach 4000/(25.4*2)= about 79 lines/mm. My Canon FS4000US film scanner resolution is 80% of the theoretical number.<BR><BR>

 

The Reference negative is of a high contrast USAF test target; and is Panatomic-X film; and is about 30 years old.<BR><BR>

 

Typical 4x5" negatives are not of ultra contrast perfect test targets; but of real world grey scale images. <BR><BR>

 

In many clients 4x5 color tranies; the trany is the limit; and not the flatbed scanner. For a good sharp 4x5; the Epson 2450 flatbed is clearly the limit; and a drum scan will yield a much sharper image. For mild enlargements; the print fom the scan is not much better than the plain jane flatbed @2400 ppi. I have nver had a problem with the Dmax issue; but I do use prescan curves alot to get a better scan.. The 2400 ppi 4x5 color scans I do are about 290 Megs in RGB color; if this file size scares you off; then a 4000 ppi scan of a 4x5 should be a real problem.<BR><BR>

 

-- Kelly Flanigan , May 01, 2003; 02:20 A.M. Eastern

 

--------------------------------------------------------------------------------</a>

[kelly_flanigan1]

John; it is real interesting that the numbers are the same; with the Epson 3200 nor really testing better than my Epson 2450 photo unit. Maybe I have one that focuses real well; I have not had any focus problems EVER. I read posts about all these focus problems and wonder "what the heck" ??? ; are we talking about the same scanner???

[leonard_evens]

Kelly,

 

There could be several factors explaining why his results don't seem any better than yours.

 

First, different people may differ about when they consider lines resolved. I consider them resolved if I can see them. He says they should be "easily distinguishable", which may be a stricter requirement.

 

Second, there may be sample differences from scanner to scanner of the same model.

 

For example, with my Epson 2450, I've never obtained 30 lp/mm. I've chosen a low ball estimate of 20-25 lp/mm, but it is probably slightly over 25.

 

From all I've read, the 3200 is has slightly higher resolution than the 2450 but not by a huge amount. It is apparently quite a bit faster though.

 

Until some of us who have been using the 2450 get 3200s and compare directly, I think it is still an open question as to how much improvement there is.

[john_hennessy1]

Leonard,

 

I was afraid there might a response from a math professor! You said �if you scan at 3200 ppi, then that is what you are going to get� You�re right of course; I don�t know what I was thinking or how I arrived at the ppi numbers I gave. Luckily for me I did not have to do math to arrive at the lp/mm numbers; I just matched the group and element to the table for the target. So they are right.

 

As far scaling the image in VueScan: I missed �Image Size = Scan Size� and Magnification = 100%. Under the Files tab. Image size was set 4x6 inches which I did not see. I�d rather the white and black point adjustment be in the form of something like Levels in PS with sliders and so I can simultaneously see the image. It would be good if the histograms had density levels marked so the w&b points don�t have to guessed at so much. Hopefully, I�ll get the hang of it. That is usually the only adjustment I make at scan time.

[leonard_evens]

John,

 

Thanks for the clarification. Of course, the numbers on the bars might not reflect the resolution in the negative unless the usaf target was photographed at the proper distance, but I presume you got that right. Also, the resolution of the scanner is probably slightly higher. What you see is the net result of all components of the system, including camera lens and film. The net resolution is always less than that of any component but if other components have high resolution they won't reduce the net resolution by much.

 

As to Vuescan, you are right that it would be more convenient to have a histogram interface like that in Photoshop, but you get used to using the white point and black point sliders and flipping between the image and the histogram. In practice, it doesn't take very long, particularly if you use the values shown in the panel below the image.

[leonardo 144]

John, I think your resolution numbers are correct. I get approx. the

same numbers. Disappointing when compered with what I get with my old-fashioned analog enlarger.

This evening I scanned an old APX 25 test negative with my Epson 3200

and after that I went into my B&W darkroom and made an enlargement of the same test negative.This enlargement I scanned with the Epson.

The numbers represent the double resolving power, so for the actual resolving power (in line-pairs per mm) one has to DIVIDE THE NUMBERS

BY TWO!

The first image is the analog enlargement(50mm Rodagon),the second the scan from the Epson at 3200 dpi<div></div>

[leonardo 144]

The Epson scan at 3200 dpi<div></div>

[kelly_flanigan1]

My numbers quotes way above are in Line (pairs) /mm. In optical engineering; we just call it "lines/mm". With the USAF charts; I have been reading this data for about 3.5 decades. Since we are always reading the same darn pair of lines; the "pairs " lingo is understood by workers in the art. NOT until the internet came out; did I see ANYONE try to frig numbers by a factor of 2; by twisting the formal definition of the 50+ year old chart. To make a lens appear twice as good; is marketing BS; and not real world Engineering. <BR><BR>Shoes are similar; when a lady goes buts shoes; she buys them in pairs; most all people have two feet. <BR><BR>The absolute maximum resolution of a film scanner is (pixels per inch) divided by the 25.4 mm/inch; then divided by 2. The "2" makes it for line/pairs per mm; the lens testing standard for over 1/2 a century. Dropping the "2" is BS; and has no place in actual optical engineering; but usefull for Marketing; Quackery; and Hucksters. The a perfect 3200 ppi unit would approach a 3200/(25.4*2) = 63 lines (pairs) /mm.<BR><BR>The

[kelly_flanigan1]

My "factor of two" rant is NOT directed to anyone above. I got a really weird email; mentioning my data was off be a factor of two; and my numbers should be divided in two. <BR><BR>My lens test negative was shot 30 years ago; and has been used for alot of lens; enlarging lenses; and now scanner tests. I never dreamed that the internet would cause confusion in a 50+ year old standard test; in which the pairs of lines are always counted. <BR><BR>The "pair police" must hang out at mails; and make sure people are actual buying shoes in pairs for the left and right foot; instead of single shoes; for one foot..........; or making sure that bras have two bumps; or books are printed on both sides of the page. The absurd abberation to make lenses appear twice as good is very recent; and causes confusion.

[leonardo 144]

Whow!What a reaction.

Kelly, I did not intend to correct or insult you, in fact I agree to what you and others wrote.

My only intention was to show the different results I got between the Epson and an analog enlargement.

I think the confusion is about the "divide by 2" and "line-pairs/mm."

The only reason I said to divide the numbers (as shown on my (Dutch) test card) by 2, is because I photographed it at half the intended distance.I did not intent to make a connection to the difference between line pairs/mm and lines/mm or in any way criticise you.It's just my writing habit.If that was confusing from my posting, I'm sorry.

I also do not imply to say that the Epson is worthless because it has less resolution than my analog (B&W) enlarger, it's has numrerous

other advantages and resolution is reaonable for 6x7.My disappointment lays in the fact that the advertised 3200 dpi number

(as you and others also have shown) absolutely does not deliver.

 

Cheers,

[leonard_evens]

Lee,

 

From what you posted, I think we can conclude that your scanner is resolving more than 30 lp/mm and less than 40 lp/mm.

 

As to the necessity of dividing by two in what you exhibited, I think that is pretty clear. Otherwise, your analog results are showing a phenomenal 140 lp/mm. 70 I believe. 140 I would be skeptical about.

[kelly_flanigan1]

Leen; sorry for the rant!<BR><BR>Leonard ; John and the group; I posted the scans of the my negative I shot; on the other Epson 3200 thread.