I’m starting a somewhat large process of digitizing old analog photos. Someone has offered me a free scanner dating from 2008: the Epson Perfection 4490. 4800 dpi optical resolution.
I’m wondering how a scanner of that vintage (assuming everything is working) compares to a new scanner; e.g. a Perfection V39, also 4800 optical.
I don’t care at all about associated software stuff, just the raw scanning fidelity, speed, contrast, color accuracy, etc.
Have scanners changed all that much? I can get a used V39 on ebay for like $80, so if it’s going to be significantly better than the old scanner I will just do that.
That old scanner was like 800 bucks and cites dynamic range as part of the specs… I think it will be much better than the 80 dollar one… I don’t think much effort has been put into flatbed scanners since 2008 so I bet it’s pretty solid…
Hello, I can’t comment on the last part, as I still use my trusty Epson Perfection 3200 Photo – the predecessor of your 4490 – bought somewhere in 2004 or so.
First off, the optical scan resolution you mentioned is only interesting when scanning negatives. They’re smaller than photos thus need more enlargement.
You’re not going to scan a 4x6 inch (10x15cm) color photograph at 4200ppi. Because if you do that, you’ll end up with a (8-bit, jpg) file size of about 1 or 2 GB!
Rule of thumb is: scan such a photo at 300ppi if you you want to print it at the same size, use 600ppi if you want a 8x10 inch print. For use on screen / social media, 100ppi will do.
Back to my prehistoric scanner. I used it last year to scan lots of photos, color and black&white, from the 60s and 70s, for an exposition about the history of a local pottery/ceramic plant. Many were printed (by myself) at 8x10 inch, some on A1 (23x33in, 60x80cm) and even a panorama of 24x78in/60x200cm. All that with a very good result! I made all the scans as 16-bit tiffs btw.
I really can’t imagine that a new scanner would make a big difference, so I would say, stay with your “vintage” 4490!
Probably not that useful for you as you already have a scanner but I found using this film copier from Nikon (see link below), attached to a macro lens on my Panasonic G9 was quick and effective. First was scanned on a flatbed back in 2000s, second with the G9. Though, to be fair, I did some correction in post on the new version and I know slightly more what I’m doing.
Thanks everyone – got the (old) scanner today – I’ll see how I get along.
I’m using linux so drivers aren’t too much a concern. And yeah I’ll be scanning at 600 or 1200 max.
I guess speed might be an issue – the newer scanners are faster for the same quality, right? Anyone have an idea of how much faster? When I look up the V39 it sounds like it’s actually slower per scan line than this old 4490 (specs are given differently though, so hard to tell) but I guess the 4490 was a more upmarket scanner so that’s maybe not such a surprise.
My last encounter with an Epson scanner (a V19) revealed that it seemed to “miss” scan lines every now and then, at all resolutions. If you put graph paper on the bed, turned it at 45, and did a scan, you’d see these regular skips in the lines where the whole vertical column of pixels was apparently gone:
Heck you might even be surprised how good a recent smart phone is… google has a photo scanning mode / app that corrects for uneven lighting during the process of capturing the photo…it does a pretty good job. But I think that old scanner is pretty solid and most of them seem to be USB2 so no speed up there either…
Thanks – I do use my phone for basic document scanning. I’m getting this going for photo archiving.
Re: jerks in the mechanism – yeah but the missing pixels were on the other axis, i.e. the sensor axis not the transport axis (I explained it poorly in my previous post.) I wondered if maybe a pixel goes dead, the firmware detects it, and rather than tell anybody about it just silently skips it thereafter.
I got the 4490 going in linux – wasn’t easy but it’s working. Unfortunately the linux drivers don’t support 16 bit color (or 4800dpi scanning, AFAICT). Going to try in windows and see how I get on. Apparently there is still a driver for this in Win 10.
Thanks! Looks like an impressive piece of work. For now it seems the windows version is working fine (they apparently have a Win 11 driver too), so I’ll use that unless I decide I need to ante up for VueScan. VueScan does support this scanner, though, which is good to know.
Hello, Xsane does 16-bit scanning. Don’t use the Xsane plugin for Gimp, because scans are always reduced to 8-bit files, even if you tell Xsane to scan in 16-bit. Just use Xsane stand-alone and choose 16-bit.
Just know they are never really even close to the optical resolution .
At least for film scanning , most flatbeds top out at like 1600 to 180p foi max, whatever it says on the spec sheet.
A dpi of 600 and more seems more than reasonable for me to get good scans for printed photos . So if the mechanism is still going , and software is working good (can be an issue , but vuescan helps a lot ) i wouldn’t see much harm in there.
Had this once with a film scanner . Might be that the dpi selected is not evenly divisible with the native dpi. And it then achieves the target dpi by skipping.
See i as the same as line scanning in a sensor when recording video.
Think about it like this : resizing down is resampling , which basically means ‘get all samples and get a representation of combining those samples into one’.
But if you don’t want to ‘get all’ - you want faster speeds - what is a scanner to do if you enter an uneven count ? It can only come up with patterns like ‘skip 1 skip 1 skip 1 skip 2 skip 1 skip 1 skip 1 skip 2’ or something to get your target dpi.
If scanners could choose their skip distance freely , it might fix things , but often they are limited to their original max dpi resolution , and operate in lines from there.
So, even though it might not apply to every scanner out there , i always make sure to use an even divisor of the original native dpi . Divide by 2, 3 or 4… But not 2.5 or 3.2.
Or , as i do with my film scanner, just scan at max but immediately size down to a target size/dpi and archive that.
But that’s mostly because my film scanner isn’t any faster if picking less dpi than native .
@paulmatthijsse – thanks for that tip. In this case it’s apparently the epson drivers for the scanner that don’t support 16bit (the GIMP plugin doesn’t even work for this scanner, unfortunately, but I can run iscan directly). Using xsane directly would still use the epson drivers in the background, right?
@jorismak – yeah I tried 9600 just for fun in Windows (supposedly the scanner has a overscan feature that can generate 9600) and it looked identical, if not slightly worse, than 4800 in terms of detail. I haven’t tested the 2400 vs 4800 or 1200 vs 2400, etc.
I like the aliasing theory in terms of missing pixels; but it would mystify my why they would choose a number of pixels in the CCD that was close to but not quite the scanning resolution… it has been a few years since I was poking around at that, but my memory is that there was only like one or two such lines per 8.5" width. At any rate, I still need to check this scanner to make sure it’s not happening with it as well.
Missing scan lines can be due to the calibration zone being dirty. A scanner may have a small clear strip (or window) behind the scanning area that it scans to calibrate white. If there is dust, dirt, Windex, etc in that area, the calibration will be off and you’ll ger a dark or light line in the direction the platen moves.
