Achievable resolution with Epson V700/V750/V800/V850

[koraks]

In Photoshop I immediately convert to 16 bit and then de-saturate to grey

OK, so you only do B&W?
Why do you compress everything into 8 bit initially (scan to jpeg) and then go back to 16 bit instead of scanning in 16 bit natively? The latter really *is* better if more dramatic contrast adjustments (possibly local) are done at a later stage as you'll avoid posterization. Notorious are posterization artifacts in bright skies when working with negative scans.

Now I am not claiming that the resolution of the images for the 4990 is 3200 dpi ( 62 lp/mm)

Yeah; agreed, it doesn't come close. In fact, I've never found any significant benefit in scanning at more than 2400dpi on the 4990 (I've had one since it just came out, so approx. 20 years). I've done the res-tests scanning at various resolutions and then downsampling, critically assessing the resulting images to see if any additional real detail is captured above 2400dpi, which I've never found to be the case (nor have other people who have systematically tested these scanners). More pixels does not result in more resolving power. It does significantly affect productivity and storage requirements.

I'm with you on the issue of disabling most enhancements etc. in Epson scan. However, for optimal color fidelity and the most flexibility in B&W edits, 16 resp. 48 bit scanning is absolutely essential; JPG doesn't cut it. JPG/8 bit is fine for B&W work where not much adjustment is foreseen and for color slides under the same conditions.

For amusements' sake, here's 3 scans of the same negative made with 3 scanners, all downsampled to the 3200dpi native resolution of the lowest common denominator (so not the 4990):

No sharpening was applied to any of these scans. They were acquired, 2 out of 3 were downsampled to 3200dpi and pasted here.
Image courtesy of @Don_ih

 

[brbo]

For my 4990 I found there was no point using 6400 dpi, but when I got my v850, I saw a detectable improvement, and 6400 dpi was worth using.

That's because max. mechanical resolution of 4990 is 4800dpi and 6400dpi for V850 (higher resolutions in those scanners use interpolation). Lenses in those scanners are nowhere near that resolution, so you get a slight* bump in (real) resolution when scanning at highest native resolution thanks to oversampling.

* it's slight, but noticeable and measurable; you do need to have film at optimal height and a suitable film (a good resolution target) to notice improvement in real resolution when scanning at higher settings; I'm pretty sure that folks that say they get the same (real) resolution from V850 at 2400 or 6400dpi are not doing that.

 

[koraks]

One of the samples of each set was scanned at 4800dpi, the other at 2400dpi (they are not necessarily presented in the same order - on purpose). Top row is both samples scaled to the same 4800dpi (so one scaled, the other left alone). Second row is the top row scaled back to 3200dpi (which is to a slight disadvantage of the 2400dpi capture, but I doubt it's relevant in practice), with some sharpening added. Bottom row is the 4800dpi scan downsampled to 2400dpi and the 2400dpi capture left alone, both with the same modest amount of sharpening applied.

There's a difference, for sure. Is it worth the extra scan time and storing up to 4x more data? That's the question I'd ask myself before pulling that slider all the way to the right.

"But your film isn't in the right place for proper focus" - this is with the stock Epson holder shimmed at optimal height for this scanner and this negative, taking into account film bulge. My scanner doesn't get any better than this.
"But your negative sucks, you need a proper test chart" - See reference Scan Dual IV 3200 dpi scan below (without sharpening, and with the same sharpening as applied to the '3200dpi' flatbed scan). Besides, what would be the advantage if the benefits only materialize with a resolution target and not with real-world images?

Not all pixels are created equal.
No innocent pixels were harmed in this exercise.

 

[brbo]

Thanks for samples @koraks. It's refreshing to actually see a demonstration with images instead of words only.

Did you also find out that there is a small bump in real resolution if you use the highest native resolution when the negative has enough detail and is scanned in optimal position?

 

[koraks]

Did you also find out that there is a small bump in real resolution if you use the highest native resolution when the negative has enough detail and is scanned in optimal position?

I hope that the examples illustrate this aptly - the bump, which I do recognize in this n=1 test, is so small that I'm not sure whether people will consistently recognize it. I doubt I would, but I had the advantage of knowing which is which. I can imagine that there are negative structures that are more conducive to recognizing it. For those who need to eek out the maximum this scanner can yield, it may make some sense to scan those negatives (and IMO those only) at the highest resolution, and then sample back down to something more sensible (e.g. 2400dpi). However, if that sort of resolution gain matters, I'd recommend looking into different ways of digitizing those negatives, allowing for cleaner, crisper results 'out of the box'.

 

[brbo]

One would hope people would see the difference if they paid attention (it's 2nd, 1st and 2nd that is sharper).

Does it really matter? No.

But, if a flatbed in this class would be my only scanner, I would scan at highest resolution (I never scan all the frames on my rolls though, I'm happy if there is one worth scanning/printing). You can set Vuescan to scan at high resolution and immediately scale down by desired factor, that way you are not wasting any disk space. And with the stupid warmup prior to every scan on 4990, the time penalty is much less than a factor of 2 if you scan at 4800 vs. 2400dpi.

 

[warden]

OK, so you only do B&W?
Why do you compress everything into 8 bit initially (scan to jpeg) and then go back to 16 bit instead of scanning in 16 bit natively? The latter really *is* better if more dramatic contrast adjustments (possibly local) are done at a later stage as you'll avoid posterization. Notorious are posterization artifacts in bright skies when working with negative scans.


Yeah; agreed, it doesn't come close. In fact, I've never found any significant benefit in scanning at more than 2400dpi on the 4990 (I've had one since it just came out, so approx. 20 years). I've done the res-tests scanning at various resolutions and then downsampling, critically assessing the resulting images to see if any additional real detail is captured above 2400dpi, which I've never found to be the case (nor have other people who have systematically tested these scanners). More pixels does not result in more resolving power. It does significantly affect productivity and storage requirements.

I'm with you on the issue of disabling most enhancements etc. in Epson scan. However, for optimal color fidelity and the most flexibility in B&W edits, 16 resp. 48 bit scanning is absolutely essential; JPG doesn't cut it. JPG/8 bit is fine for B&W work where not much adjustment is foreseen and for color slides under the same conditions.

For amusements' sake, here's 3 scans of the same negative made with 3 scanners, all downsampled to the 3200dpi native resolution of the lowest common denominator (so not the 4990):
View attachment 407721
View attachment 407722
View attachment 407723

No sharpening was applied to any of these scans. They were acquired, 2 out of 3 were downsampled to 3200dpi and pasted here.
Image courtesy of @Don_ih

I think both of the top two are usable, with a preference for the middle image. The third (bottom) is not usable.

 

[brbo]

The third (bottom) is not usable.

How so?

It's 2400dpi, which is still more than real resolution of Epson 4990.