How much DPI do I really need from a 35mm scanner?

[darin_cozine]

I'm shooting mainly ASA 100 color and BW films. I also have a 8MP

digital that makes very nice 2400x3200 images. I would like to be able

to match that with my film cameras which I still use occasionaly. So

that would mean roughly 2500 dpi. Doesnt seem like alot to me. There

are dedicated film scanners that can go up to 8000 dpi. But I've heard

that they never deliver the resolution they are advertised at.

so..

How much DPI do I really need from a 35mm scanner?

[pjmeade]

Hello, my rather ordinary Primefilm PF3650 will give 16-bit jpg files 4968x3240 for a 1.9 MB image when set at 3600 dpi, or a 16-bit tiff, 4968x3240 92 MB. To be honest, there is no substantial difference between a jpg and a tiff off the scanner, other than 45-fold file size.

 

Both files are equally useable as an archive. If I make changes to the jpgs I always "save as" so there is no loss involved in manipulations.

 

Hope this helps. P

[alan_chan4]

To match 35mm film to 2400x3200 will be roughly 2400dpi. You can pull more details from 35mm film with consumer scanners, but if 2400dpi is all you wanted, just buy the Minolta Scan Dual IV.

[kirby_muxloe]

And in my opinion you get more than 2400dpi from film using SD IV

[ian_scholey]

The dpi you need depends on what you want to do with the image. If you intend to print the image you will need 300 dpi per printed inch.

 

Ian

 

http://profiles.colourperfect.co.uk

[mendel_leisk]

How much scanning dpi? Depends. On coarseness of grain, lens, nature of the shots, degree of enlargement you're aiming for, your storage capacities (hard drives, optical media), and last but not least, your fanaticism.

 

Mainstream scanners in the last few years have been around 2800, 3200, 4000, 4800 and 5400 dpi (or ppi, for the purists). The step from 4000 to 5400, for example, considering the Nikon V/5000 and the Minolta 5400, is a big step for file size (100megs to 200megs, roughly), but debatable in terms of of improved resolution of detail. With a very fine grained film, and a sharp, steady lens, there is some reason to go to 5400. There is *fine* detail that goes from vague to distinct, *sometimes*.

 

But it comes with a file size price. At the other end, there's really not *that* much difference, say between 2800 and 4000. Again, fanaticism come into play. When you pixel peep, you can see the difference, but...

 

Peter, regarding:

 

"Both files are equally useable as an archive. If I make changes to the jpgs I always "save as" so there is no loss involved in manipulations"

 

That initial jpeg is *already* compromised in quality. I agree, you really have to look close to see the differences, but they're there. Especially if your files are going from 92meg (tiff) to 1.9 (jpeg), there's no free lunch in that much reduction.

 

I'd say leave your master file as 16 bit rgb tiff, but do any and all downstream files as jpeg, they're all reproducable, if you record your workflow. The best lossless compression to date looks to be lossless jpeg2000, With grainy scans, it can reduce file size to maybe 2/3 of original tiff. If it's a smoother digital file, I think the compression is better, but you likely have the RAW with those, so jpeg2000 is not needed. Too bad there isn't a "raw" format for conversion of tiffs. Something where the 3 color channels are somehow merged, similar to dslr raw formats.

 

Here's one place you can get a jpeg2000 photoshop plug-in:

 

http://www.fnordware.com/j2k/

[robert lee]

You'll be able to extract 10MP to 16MP of useful information out of fine grained (Reala, Provia, etc.) 35mm film, depending on your subject matter.

 

The optics of consumer level flatbeds aren't good enough for this. However, a 4000dpi dedicated film scanner like the Nikon V or Nikon 5000 will dig most of the spatial resolution out of a frame. If you have an especially dense exposure, shot on Velvia for example, then a professional drumscan will extract more detail out of the shadows.

[godfrey]

In general, I find that scanned film images require 40-50% more ppi to image as crisply as

digital capture images. This is due to the inevitable aliasing and losses in the scanning

process, plus the influence of film grain and other emulsion attributes.

 

The other factor is how big a print you are considering for comparison. Digital capture

images are clean enough that they image well right up to the point where pixelation

becomes apparent. Film images tend to degrade faster as the size is increased.

 

A full-frame 35mm negative scanned at 2820ppi with the (now ancient) Minolta Scan Dual

II produces an ~10 Mpixel image that prints about as well as a full-frame digital capture

with a 6Mpixel SLR to the A3 size. An 8Mpixel camera (or 10Mpixel) camera prints up to

A3 Super (13x19) full bleed better than that, and rivals some of my scanned medium

format work up to similar size. Medium format scans at 2500ppi and greater have more

overhead, though, and can make even larger prints ... with native Mpixels after scanning in

the 20-30 Mpixel range.

 

For 8x10 to 10x15 sized prints, a 6-8 Mpixel camera provides everything you need to

compete favorably with 35mm at 2820ppi. The more Mpixels your camera has, however,

the more leeway you have for adjustment and cropping, and detailing improves nicely too

(particularly when considering wide-angle work).

 

Godfrey

[raphael_bustin2]

There's no real direct correlation. I use a Nikon Coolscan LS-8000, now nearly five years old. Its resolution is rated at 4000 dpi, and this model has been objectively measured at around 2900 dpi on USAF test targets. (That's actually pretty "honest" in the overall scheme of things, believe it or not -- and some of the worst spec inflation is found on some very high-end machines.)

 

A 4000 dpi film scan of a 35 mm frame gives about 20 million pixels. You can then put that up in a window in Photoshop, and compare it against a digicam capture, of say, a Canon 20D. For a fair comparison, you'll need to either upsample the Canon capture, or downsample the film scan, or some compromise, so that the image sizes match.

 

When you do this you discover that images from a good digicam or DSLR are "worth" a lot more than those from film scan, in terms of the detail they capture per megapixel.

 

What's the magic ratio? I can't say for sure, and I'm not sure anybody can. I will say that, since owning a Canon 10D, I'm not shooting or scanning too much 35mm film any more. If we say that the 20D "matches" 35mm film scanned at 4000 dpi, the ratio is 20:8 in favor of the DSLR.

 

Speaking from experience and having looked at many, many high resolution film scans from many, many scanners, I'm not convince that there's much detail beyond 4000 - 5000 dpi on most film captures. Let's just say, everything has to be perfect: fine-grain film, a contrasty, stationary, subject, an excellent taking lens, well focused, on a tripod, and with mirror lockup and a fast shutter speed.

 

If you've done all that, you may just see the improvement between a good 4000 dpi and a good 5000 dpi scan.

[alejandrokeller]

Hello Raphael, nice to see you arround here.

[tedharris]

Another point, you really need to run some tests on the scanner(s) under consideration.

Resolution specifications are sometimes far more marketing hype than reality and the same

holds true for Dmax. This is somewhat less true for higher end film scanners than for

consumer flat bed scanners but it is a factor to consider.

[eric_brody]

While I no longer use or scan 35mm film, my venerable Nikon Coolscan IV, otherwise known as the LS-40, served me well. I still have it stored in its box. It officially scans at 2900dpi. I used the Nikon software, and Photoshop 7 at the time. I have a number of Provia 100 images on my wall at 11x14 that seem to me to be pretty sharp and I'm not a big fan of Photoshop unsharp masking. I use it very sparingly. The newer Nikons go to an official 400dpi. I neither know how to nor do I wish to measure the resolution of a scanner, I'm interested in the resultant image on the paper.

 

To try to answer your question, I suspect it depends on how big you wish to print and how you will display the prints. A used LS-40 or a new Coolscan V or 5000 would doubtless do well for you if you scan nothing larger than 35mm. For larger film, it starts to get expensive. The Nikon 9000 sells for $1,800 at B&H and scans film up to 6x9. There are less expensive options but all are flatbeds. Reading the forums, there's lots of debate about whether these scanners are really "good enough," most feel they're less than optimal for 35mm film. I have friends who produce stunning work from the Epson 4990 from 4x5 negatives printed on a 2400 in black and white.

 

Good luck.

[eric_brody]

Quick correction, the new Nikons go to 4000 dpi not 400.

 

Sorry for any confusion.

[mr._smith]

I've printed quite a few FP4+ negatives to 12"x16" that were scanned with a Dual Scan III. Those printed images look sharp and natural and get the best comments at art fairs.

 

I've sold that little scanner and converted to full digital and don't miss the time spent developping film. Digital doesnt get me better printed images, just a faster workflow.

[scott_eaton]

<I>When you do this you discover that images from a good digicam or DSLR are "worth" a lot more than those from film scan, in terms of the detail they capture per megapixel. </i><P>You do realize you're wasting your time trying to be logical with this lot? .<P>This is why there exists larger formats than 35mm film, and why we scan them at much lower resolutions than these consumer desktop scanners. With my 6x7 tranny work I *refused* to scan higher than 2000dpi because beyond that film start to look like absolute crap and worse than any consumer digicam. Yet, 35mm shooters here are talking about 5000 and 8000dpi. Anytime Les want's to compare his 5000dpi wonder files scanned from 35mm against my 2000dpi scanned files from 6x7 I'd be happy.<P>A sentient and logically minded person would then ask the question 'gee, if larger peices of film scanned at lower resolutions looks better, then maybe a higher resolution scanner isn't a good thing because 35mm doesn't really have all that clean information.'<P>It doesn't, which is why that 2400x3200 digital will spank the best scan from 35mm unless you the only thing you care about is micro fringe resolution (or Les is using the digital camera).

[daniel_taylor]

"Scott, I would want you to shoot the Rand McNally map with your MF film and scan it at 2000dpi for comparison. That way the "DSLR beats MF" crowd can finally see too. BTW, I believe we've put the "spanking" criteria to rest grossly infavor of lossy scanned 35mm film over all DSLRs. Hey, maybe the soon to be released 1Ds MKIII can finally match it . . . ;-)"

 

Must I remind you again...

 

Attached is your latest 5400 scan put next to a 10D image shot at the same distance from the map as a 35mm or a full frame camera would be shot. The light falling on the 10D sensor was the same light that was falling on your film, the crop covering the same physical size on sensor as on film. It's absolutely clear, even to you, which has better image structure and resolution, which crop "spanks" the other one. Which is why whenever I post it you have a temper tantrum.

 

The 5D and 1Ds mkII each have a higher pixel pitch than the 10D sensor, which means that they would, without question, out perform your "lossy scanned 35mm film", even on the sole criterea of resolution.<div></div>

[Nicholas Barry]

<p>

OK. I think I see it now... Les is either a troll or a <a href="http://en.wikipedia.org/wiki/Crank_(person)">kook</a>.

 

<p>

Les, your last response a) doesn't make sense and b) doesn't address Daniel's attached image. Further, your "Film and Digital Resolution Album" is miserable in its lack of information. As Wolfgang Pauli once said "This isn't right. This isn't even wrong!"

 

<p>

But as I recall, your tests were addressed in another thread when you first posted them. I got tired of reading misinformation so I gave up.

 

<p>

I shoot film and have an LS-4000 scanner with which I scan Provia. But the fact remains that film grain != useful image information. There was a reason why large format and medium format was used in the analog world. Scanning 35mm film at increasing resolutions will only get you more crap. Not more image information. And Daniel's post shows that non-miniscule sensor images (i.e., APS-C and larger) can and do beat scanned film. Not to mention are faster to process. Scanning sucks.

[Nicholas Barry]

<p>

<blockquote><em>That is <strong>aggravating</strong> isn't it! </em></blockquote>

 

<p>

And the word you're looking for is <em>irritating</em>.

[Nicholas Barry]

OK. I'll bite again (though I don't know why). I can see where you're getting your information. And that's fine. It would seem you're only interested in pure spatial resolution based upon R.N Clark's site and tests. At this point I'm going to have to delve deeper before I feel a full authority on this subject.

 

But you're completely ignoring (perhaps just for this discussion), the difference in image quality that a DSLR generates vs. film. Conclusions that R.N. Clark himself makes. If what you wish to discuss is pure film resolution, then I'm sorry I interjected (boy, am I sorry) (at least for now). The fact remains that for most applications DSLR image quality is better than film. And scanning film is a laborious, only semi-productive process. A process that has made me almost give up on film altogether (I've shot about five rolls in the past year). When I scan, I scan once, using 16x multi-sampling and Light ICE. That's a 30-45 minute per frame process. That's a pain in the arse. And that doesn't take into account the post processing that I have to do to get the colors to be neutral. And to remove the dust specs that inevitibly make their way onto the film.

 

Les, you have this notion that film is way better than digital. If you like film better, then that's you're game. But saying that it's just flat better than digital is incorrect. And by cornering the discussion to spatial resolution to back your viewpoint you're doing yourself and others here at this site a disservice.

 

And by the way, how did this thread get to this point? It started out as a discussion of how much res do you need from 35mm scanning to match an 8MP (presumably point-n-shoot) digital camera. I apologize if my posts seemed antagonistic. I guess I'm just getting tired of seeing threads that beat a dead horse. I'll have to do a better job of skipping them in the future.

[daniel_taylor]

"Oh yeah, you have the magical 10D that can grow its sensor . . . ;-)"

 

You do realize that everyone who reads the post and views the images understands what I've done. I've just compared 10D sensor level technology to 35mm film on a square mm basis rather than frame vs. frame. It's a perfectly valid answer to the question "What if the frame sizes were equal?"

 

When your only response is to try and horse laugh the comparison, all you do is prove to everyone that you don't have a valid argument, that you know the test proves what I say it proves, and that you're nothing more than a horses...fill in the blank.

 

"BTW, I do have the 1Ds MKII images and it is interesting that it doesn't outresolve even the 4,000dpi scan of 35mm Velvia, much less the 5,400dpi scan."

 

You are a liar. Let me repeat that for everyone to read and see. YOU ARE A LIAR. You have been claiming that you would "get around" to posting 35mm vs. 1Ds mkII crops for months. Every time you make this claim I've asked: where are the images? Has never happened. My guess is that if it does happen, you will have hobbled the 1Ds in some way so that film comes out ahead, the same way your 20D images were but a fraction of the quality that could be expected from a 20D, which I proved.

 

"what happened to the 3MP D30 outresolving Imacon scanned Provia 100F? Or what about that other link you posted about the 6MP 10D (your DSLR) outresolving a Provia 100F scanned by a multithousand $$$ scanner?"

 

Again: YOU ARE A LIAR. Nobody has ever made the claims you put in their mouths. This fight started because I tried calling you on that when it came to Reichmann, a man who has achieved more in photography than you could ever hope to.

 

"To clarify for you, Daniel's attachment is based on him shooting only a portion of the target and believes that directly correlates to what the resulting resolution would be if his 10D's sensor were to magically grow to the size of 35mm film."

 

Don't misrepresent it. It's not a matter of "belief". I placed my 35mm body on a tripod, framed the map in the viewfinder, put my 10D on the tripod *without moving it*, and shot.

 

It is exactly what the image would look like if the 10D sensor were full frame, and is an exact comparison of 10D vs. 35mm on a square mm vs. square mm basis.

 

"In his conclusions, he is essentially stating that the Canon D60 has the same resolution as 4000 ppi scanned Fuji Velvia 50 *at the surfaces of the respective media*. Got that? That's not overall, that's millimeter by millimeter at the film/sensor surface."

 

Which is exactly what my test shows! Why then don't you concede the point and validity of the test instead of mocking it?

 

"Don't come here projecting your presumptions because as you now carefully back peddal you have realized that I've made no such representations of broad generalization of "better image"."

 

What a hypocrit. How funny to see you lecture someone else on how to debate and behave in an Internet forum.