Is softness at higher DPI normal?

[magicianhisoka]

Heya all, just picked up a canon 9000F with the intention to cut my film bill. Unfortunately, my first 24 hours has been pretty much marred by disappointment, until i accidentally set my DPI setting at 600 instead of 1200. Strangely enough, my photos finally turned out to be acceptable in terms of detail and sharpness. However, it seems rather silly for me to be buying a scanner to use it at the lowest setting. A buddy of mine shared that he has no problems getting my results on 1800 dpi using a flatbed too and said I should get mine checked out.

Anyone else suffers from this problem? Will post up some photos of what I mean when I can. Should be soon once I convert them.

P.S Currently I'm encoding them as .tif files using Vuescan.

 

[efix]

In a review of my Plustek 7400 scanner, I read that using 7200 dpi only brings a slight advantage in resolution over the 3600 dpi setting, with an actual resolution somewhere around 3500 dpi, as opposed to realistic ~3000 dpi at the 3600 dpi setting. It is also my observation that 3600 dpi scans aren't pixel-level sharp, but downsized by 50% -- which still is ~ 10 megapixel quality -- they are very good.

THUS: I suppose scanners promise more resolution than they will finally yield. If it doesn't take too long, I'd scan at the highest setting and then downsize the result.

 

[Brian Legge]

I'd expect somewhere around 1600 to be as the best resolution with the scanner. Definitely higher than 600.

Any chance you have some automatic settings enabled? Stuff like dust cleanup, etc?

 

[dct]

[...snip]
THUS: I suppose scanners promise more resolution than they will finally yield. If it doesn't take too long, I'd scan at the highest setting and then downsize the result.

Not only Canon scanners. All market higher dpi then really can be achieved. Good background and model related scanner dpi information.

 

[magicianhisoka]

Heya all, didn't expect this many replies so soon 🙂

I know they like to play up their capabilities. But mine lists itself as being 9600. It would suck for it to be 600 >_>

@efix : my results at 4800 look HORRIBLE. Just changed them to jpg. gonna upload them to flickr now.

@Brian : I have automatic settings but i maintained them for all the different DPIs. Shouldn't be the variable that's causing the problem right?

 

[magicianhisoka]

http://www.flickr.com/photos/geraldkishuachew/sets/72157628403395583/

If you guys don't mind, you'll have to look at it in the lightbox. Look at the title to tell which DPI setting was used. They all seem to look the same when looking at it in the default flickr window.

 

[sevo]

I know they like to play up their capabilities. But mine lists itself as being 9600. It would suck for it to be 600 >_>

Well, it will probably be somewhere between 1200 and 2400 like all other scanners of its class...

Scanner makers love to give the figure for stepper pitch. But that is just a figure they can increase almost arbitrarily, without improving the scanner. There are technical limits as to what resolution you can achieve in a device of the rigidity of the regular 20 pound desktop scanner, even if the makers would bother to invest into optics that have more than the $30 production value I'd estimate for the average scanner lens assembly (given that spares are below a hundred).

There are (or were - Kodak seems to have killed the branch that made them) a few flatbed scanners that reached 4000dpi in terms of actual resolution - but those are professional pre-press devices the weight and price of a small car, and even these were soft at 9600 (if they did it at all)...

 

[magicianhisoka]

I was ready to accept about 1800 given the price 🙂 I'm not looking for a dedicated film scanner in a flatbed haha. Just something I can hopefully blow up to 6R but fingers crossed on that.

 

[Brian Puccio]

Hmm, looking at your flickr post, it looks like that even when scaling to the same size (as opposed to looking at the images at 100% at various resolutions) we can defintely see what you're talking about. The only thing I can think of is that the stepper motor, when needing to move in smaller increments, is horribly inprecise, as compared to moving in larger increments (that's a lower resolution so "bigger steps").

Hopefully you bought it new and can either exchange it or send it to Canon for warranty service. I know it stinks to get something that is (for most intents and purposes) DOA, but it happens. 🙁

 

[magicianhisoka]

Where I stay, exchanges don't exist sadly. If I can, I will try and beat canon's head in with it tomorrow. Though I don't know what I can tell them to make my case.

Could you elaborate more about the stepper motor? I need to make a credible case for tomorrow.

 

[Brian Legge]

Gut response, the photos all look fairly similar to me.

Slightly backfocosed in camera perhaps? The softness would show up in larger images while the smaller scan may hide it a bit. This would be particularly true if you have automatic sharpening turned on.

There is some variation in the amount of noise/grain. Again, with a smaller image, automatic options may effectively hide this compared to the larger image. Or the scanners calibration may be off? Out of curiosity was this slightly under exposed?

The streaking on the left side of the image is there in all scans. I assume thats on the negative?

I'd suggest disabling automatic sharpening, noise cleanup, etc and scanning at 600 and 2400. After scanning, resize the 2400 image to 600. I'm guessing they'd appear more similar.

 

[magicianhisoka]

Turned off a few of the auto functions and doing a rescan now. Hopefully something changes. Thanks for the help so far Brian 🙂

If the camera were backfocused, wouldn't it show up in all the wide open shots? The cow toy was shot wide open and the smallest file still seems fine.

I think it was slightly underexposed. The white cow may have fooled the meter. Think the streaking was a misalignment. Using auto crop too.

 

[magicianhisoka]

Posted a new photo, still gives the same results despite disabling features, scanning at 2400 dpi and resizing. If you move back and forth between the photos, you'll get what I mean. It is very obvious in lightbox mode.

http://www.flickr.com/photos/geraldkishuachew/6506087275/in/photostream/

This is so very troublesome >_>

 

[charjohncarter]

For years I used a digital SLR mounted with a Pentax bellows and slide copier. These images were satisfactory for me in 35mm but I had trouble with 120 as I had to make a special holder there always seemed to be light piping. I solved this but the process was a hassle even though the actual getting the files was very quick. So I decided to get a scanner. It works fine, but I think that the 35mm negatives are not as sharp as my DSLR setup, but the 120 is similar in quality. I definitely have to sharpen every image with the scanned images in PSE. Neither seem to me to be as good as a wet enlargement. Maybe, I'm crazy as I don't understand scanners very well.

I did get the "betterscan" holder for 120 which did make some difference even though it was minor (every bit helps)

 

[Brian Puccio]

Where I stay, exchanges don't exist sadly. If I can, I will try and beat canon's head in with it tomorrow. Though I don't know what I can tell them to make my case.

Could you elaborate more about the stepper motor? I need to make a credible case for tomorrow.

Wikipedia is a pretty good starting off point. My knowledge of stepper motors doesn't go much beyond that, I messed around with them several semesters ago in one of my classes at college.

http://en.wikipedia.org/wiki/Stepper_motor

What I think could be happening here is that when the scanner needs to move over to the next "step", if it's a large step (low res) then it moves where it should. It seems that the 9000F has a 12 line CCD sensor, so it must position the sensor dozens and dozens of times to get 600 pixels per inch. But these steps are large. If you tell it to scan at a high res, it must position the sensors hundreds and hundreds of times per inch, being that much more precise.

If it cannot do it precisely, the sensor images a "slightly off" part of the negative/print.

To completely eliminate software issues, I'd also try some third party software. You can download the Vuescan demo and give that a shot. It would also be helpful on the phone to tell them "Not only did I use your software, but third-party software as well, and they both exhibit the same problem, which leads me to believe it's your hardware ."

 

[Pikapig]

i think he is scanning using vuescan for all..

I was thinking the same as brian...could it be the lens...where we downsize the pic at 600dpi..we cant tell the flaws....unsharp images at the smaller dpi could seems to be more sharp than viewing it at a 1600dpi..

hmmm..best bet is to try scanning another satisfactory picture..where the lab has done it for u..and compare the 2 results...

alternative...it does no harm to visit canon and ask them wat is happening..

 

[cabbiinc]

If you're using Vuescan try tuning off just about everything you can. You may even reset the program to Default Options (File> Default Options).
On the Input tab.
Scan at full optical resolution. This is pretty easy to do with Vuescan, it's the largest number available for your scanner in Vuescan. 48 bit per pixel (64 bit RGBI if you want IR dust removal, but keep in mind the IR channel is negated by traditional black and white film by the silver content of the film itself).
On the output tab set it to Tiff, 48 bit, compression of 2. This will halve the number of pixels after the scan, but the compression will help with the sharpening.

You may also try placing the negs straight on the glass. More chances for newton rings but the focus of the scanner is closer to the glass surface. After all it's a flatbed designed to scan printed material directly on the glass, why film holders are made so thick is beyond me.

http://www.stockholmviews.com/canon_8800f_review/8800fpage1.html This review of the 8800f has some images from the 9950f as well. I've also seen other reviews of the 9000f that use the compression method with good results. I believe they all used 4800 dpi and a compression factor of 2.
Dan

PS, this is an example of the stuff I can get with an HP Scanjet G4050 using this method. http://www.flickr.com/photos/cabbiinc/sets/72157627903149543/ All but a few of these are mounted slides. The Canon 9000f is regarded as a much better scanner and the HP scanner isn't even considered in the same class.