Scan quality from Epsons?

[Kirk Gittings]

My initial tests between the Epson 4990 Pro and the Epson V750 M Pro show a small but clear improvement in sharpness and shadow detail in the V750. Though I doubt that these small images here will really show much, the difference is clear to me looking at the originals. This test is consistent with soon to be published tests by a friend of mine in a photo magazine.

All scans were done using Silverfast AI 6 with no adjustments, dry and repeated many times to check for consistency. The 750 scans were run with an "out of the box" setup with no attempt to improve sharpness by adjusting film holder height whereas the 4990 which I have owned for a long time was used "tuned" for optimum sharpness. All scans were done at 2400 DPI (which is generally considered the real resolution for the 4990).

Surprisingly the V750 is slightly slower with a scan speed of 4.5 minutes for 180MB file vs. 4 minutes for the 4990.

The real question is whether the difference is enough to matter in a print. I can't tell you that yet nor can I say what is the real resolution of the V750 (my friend says it is a real 3200). My thinking is that every little bit of resolution helps allowing me to use less sharpening. I have long agreed with the 4x enlargement limit (barely acceptable) for sub $1000.00 flatbed scanners. This scanner will not break that barrier, but it does appear to be an improvement.

The images are from left to right: Full Image, 4990, V750. The crops are at 100%.

[sanking]

All scans were done using Silverfast AI 6 with no adjustments, dry and repeated many times to check for consistency. The 750 scans were run with an "out of the box" setup with no attempt to improve sharpness by adjusting film holder height whereas the 4990 which I have owned for a long time was used "tuned" for optimum sharpness. All scans were done at 2400 DPI (which is generally considered the real resolution for the 4990).

I have an Epson 4870, not the 4990, but most tests suggest there is very little difference in actual resolution in DPI so my experince may be relevant to this discussion. Scanning at 2400 DPI gives actual resolution of about 1800 DPI, or about 75% of stated. Scanning at 4600 DPI gives actual resolution of around 2400 DPI, or about 52% of stated. In other words, there is definitely some improvement in sharpness to be gained, IMO, by scanning at the highest possible DPI setting. With large sheet film the gain may not be worth the trouble, especially considering amount of time it takes to scan at maximum DPI, but in the case of roll film it makes sense to get all of the sharpness you can from the scanner.

Sandy King

[Jack Flesher]

Here are some crops from a recent lens 8x10 test image (I was shooting with my Cooke pointed into the sun which was behind the cloud behind the tower). The film is 8x10 E100G scanned this at 1600 DPI using my above-described method. The first is the full-frame image for reference, the second is an actual pixel crop from the raw scan, however I did apply light USM during the scan.

If you use 300PPI and view the crop at "print size" you will be looking at approximately a 3" x 3.3" section from what would be a 40x50" print for the full 8x10 scan (or a 20x25 print if this were a section of 4x5 tranny scanned at 1600 DPI) -- and of course it looks better when processed in PS for final print output :

[Jack Flesher]

PS: The Silverfast Ai software gives you three dropdowns when scanning: "Reflective", "Transparency" and "Tranparency (full area)". The latter is what I meant to refer to in my above posts -- again, sorry for the confusion.

[Remigius]

Jack,

Could you elaborate on the anti-glare glass? What kind of glass is this? Where can it be purchased? Thanks for all the information. I also own a 4990, I have scanned 4x5, 645 & 35mm with it so far. I have always used the film holders, but not yet the film area guide. I scan 4x5 at 2400dpi and smaller formats at 4800. My experience is that the film grain is already visible at these resolutions and it is imho the quality limit rather than the scan resolution.

[Jack Flesher]

Jack,

Could you elaborate on the anti-glare glass? What kind of glass is this? Where can it be purchased? Thanks for all the information. I also own a 4990, I have scanned 4x5, 645 & 35mm with it so far. I have always used the film holders, but not yet the film area guide. I scan 4x5 at 2400dpi and smaller formats at 4800. My experience is that the film grain is already visible at these resolutions and it is imho the quality limit rather than the scan resolution.

I agree totally on your 2400 DPI resolution limit with grain and do scan my 4x5's at that resolution so I can print up to 32x40 at 300 PPI without interpolation. But there is just simply no need to go that high with my 8x10's and I do find 1600 to be a tad sharper than 2400.

The glass is simple clear (not green), anti-reflective coated one-side framing glass purchased from a local framing shop. My shop cut it out of a scap for me for a few dollars -- works great.

[neil poulsen]

To summarize, in Ted's test the V750 did a slightly better job on small negatives, yet a slightly poorer job with large negatives than the 4990. My test was only on a 4x5 tranny, but yielded the same result: a slightly softer scan from the V750.

How is this possible? It must have something to do with the thicker base mylar for 4x5, the difference in mounting 4x5 versus mounting a smaller format like medium format on the V750 or the 4990, or perhaps because resolution on 4x5 lenses isn't as great as resolution on a Hasselblad or some other medium format camera.

I ask, because logically every 4x5 negative contains within it a smaller format negative.

[Terence Patrick]

Kirk and Jack, thanks for posting those samples!

Jack, in the 100% crop sample, the iron lamp and some of the detailing on the white wall look like they've gone through Photoshop's fresco filter. Is that a normal artifact?

[Struan Gray]

One thing to consider when buying the Epsons, or any other consumer desktop scanner, is the tonal range of the originals you want to scan. Cheapie scanners have poor signal to noise performance compared to drums and pro-level desktop machines. This is usually mentioned in terms of shadow noise, but it also rears its ugly head if you take photos with an intrinsically limited tonal range, especially if you use negative film. Retaining the delicacy of tones in a high key portrait of a porcelein-skinned Celt or a foggy dawn down at the docks is, in my experience, much harder than with conventional printing.

An image like this portrait of my son, with a wide range of tones and colours, and a fair bit of detail, is relatively simple to scan and prepare for a good-looking print. One like this post-sunset landscape, with large areas of smoothly varying, restricted tonality is much much harder. With the second type, you may find that it is impossible to make a print that satisfies you at any size: the tonal reproduction trumps any sharpness considerations.

[David Luttmann]

Kirk and Jack, thanks for posting those samples!

Jack, in the 100% crop sample, the iron lamp and some of the detailing on the white wall look like they've gone through Photoshop's fresco filter. Is that a normal artifact?

I noticed this too, Jack. This looks like the painterly noise reduction routines used in the Kodak SLR camera combined with its lack of aa filter resulting in moire. The only other thing it looks like is artifacting from upsampling in software like Genuine Fractals.

Is this common in your scans?