Scanners - current status

[Worker 11811]

The tipping point in the Law of Diminishing Returns comes where the resolution we ask the scanner to produce goes below the resolving power of its lens. (Or optical system.) It would be better to scan at or below that tipping point then use other software methods to scale the image to the size we need instead of "overscanning" and scaling down.

Am I finally reading correctly?

I have an idea to, hopefully, find out what that is. I have test film for cinema projectors which was printed on a pin-registered step printer using a glass negative. I'll bring a piece of that film home and scan it and see what comes out.

At worst, it will be an academic exercise but maybe I'll find some useful information. I'll let you know.

 

[Rudeofus]

The optimal scanning procedure was discussed here a few threads ago: scan at the max resolution your scanner gives you, then take the huge file and downscale it to match the effective resolution of your scan.

What the effective resolution of your scan is, depends on many things: sharpness of the image capturing process (lens quality, motion blur and camera shake, focussing accuracy), the film you use (high ISO film is coarser grained than low ISO film, B/W is finer than color film, ...) and the scanner.

The numbers I've posted before are the limit for most flat bed scanners. Film itself is usually not the limitation unless you use professional drum scanners.

 

[Loris Medici]

For sake of clarity, is that "...scan at the actual resolution limit of your scanner..."? Like, for instance, scanning at a resolution not higher than 2400dpi with V700/750? That's what I would do exactly. I would never use higher resolutions with a V700/750 simply because that would be simply loosing time (scan time and labor to rez-down) and storage. I don't bother scanning @ 2400dpi with my good-old Epson 2450 for instance, I prefer to use it @ 1600dpi which is the max. it will give at its best.

Regards,
Loris.

The optimal scanning procedure was discussed here a few threads ago: scan at the max resolution your scanner gives you, then take the huge file and downscale it to match the effective resolution of your scan...

 

[cupcake_ham]

For sake of clarity, is that "...scan at the actual resolution limit of your scanner..."? Like, for instance, scanning at a resolution not higher than 2400dpi with V700/750? That's what I would do exactly. I would never use higher resolutions with a V700/750 simply because that would be simply loosing time (scan time and labor to rez-down) and storage. I don't bother scanning @ 2400dpi with my good-old Epson 2450 for instance, I prefer to use it @ 1600dpi which is the max. it will give at its best.

Regards,
Loris.

With the V700 and 750, the 6400spi setting engages an optic with a wider aperture and higher rez. You can pull about 2600ppi or slightly more from it. Thus, it's better to scan at 6400spi, and then downsample to 3200ppi to make sure all the rez is captured. Scanning at 2400 is not quite enough to pull everything from this scanner.

 

[Loris Medici]

Thanks David, I wasn't aware of that peculiarity. What is the limit with the smaller aperture?

 

[sanking]

With the V700 and 750, the 6400spi setting engages an optic with a wider aperture and higher rez. You can pull about 2600ppi or slightly more from it. Thus, it's better to scan at 6400spi, and then downsample to 3200ppi to make sure all the rez is captured. Scanning at 2400 is not quite enough to pull everything from this scanner.

I have tested two V700s and one V750 with high definition targets and the most I could get out of any of them was 2300 ppi, even when adjusting the holder for optimum plane of focus and scanning at 6400 ppi.

There is some improvement in resolution depending on whether you scan at 2400 ppi, 3200 ppi, 4800 ppi or 6400 ppi but the improvement is relatively small beyond 3200 ppi. I don't have my test data on hand but as I recall scanning at 3200 ppi gave effective resolution of about 2225 ppi, while going all the way to 6400 ppi only increased resolution to 2300 ppi.

For the record, the V700 and V750 have a two lens system. One lens is called a "high resolution" lens, the other a "super high resolution" lens. The HR lens is engaged whenever one chooses film area guide in the scanner software, the SHR lens is engaged when you choose film holder in the software. Optimizing for plane of best focus is important with both lenses, but more important for the SRH lens because it has a wider aperture and consequently has less depth of field than the HR lens.

Sandy King

 

[cupcake_ham]

The lowr figure is entirely possible Sandy. Sometimes my guesstimates are a bit optimistic. That said, I did notice enough of a difference when scanning at 6400 with the higher rez optic that I made it habit of always using that....except for 4x5 at 16bit.

 

[Loris Medici]

Thanks for the quantification Sandy.

In light of this information, I personally won't bother scanning at 6400spi resolution with a V700/750, if I ever obtain one. (I was in the market for one as I was planning to buy a Bessa III and was in need of a better-than-2450 MF film scanner. I just dropped the plan of buying the camera in favour of a professional UV lightsource with vacuum frame...) Because you gain only ((2300 / 2225) - 1) * 100) = 3.4% more resolution in expense of doubled file sizes and longer (I presume?) scan times. OTOH, David says it makes enough difference to choose that path. Well, can't comment before I try it myself. Thanks much for the insights!

Regards,
Loris.

 

[Worker 11811]

Okay, I did a little experimentation.

I got out my SMPTE test film for the movie projector and scanned it. I also found some film put out by Lucasfilm for the THX Theater Alignment Program.

JPEG scan of SMPTE film........ JPEG scan of THX film:
]

On both test films I was able to get approx 40 lines per mm of resolution off the film:

SMPTE.............................. THX:


If I was using this information to judge the quality of a movie projector, I'd be disappointed. The minimum resolution you should expect to produce is 48 lines per mm. On average, a projector should be able to produce 56 lines per mm and, if you are willing to invest in good lenses, you can expect to see even more.

If these were projected images in my theater I'd be telling my boss that I've got work to do!

 

[Rudeofus]

40 lpmm is not so bad! Unless you use a sturdy tripod, mirror lockup, careful manual focussing and a damn sharp lens, the scanner won't really be your limit! The nice thing about film is that you can always move to a larger format if you really need more detail.

 

[pellicle]

Sandy

The HR lens is engaged whenever one chooses film area guide in the scanner software, the SHR lens is engaged when you choose film holder in the software

now this one is worth noting! Thanks for that!

 

[pellicle]

Hi

40 lpmm is not so bad! Unless you use a sturdy tripod, mirror lockup, careful manual focussing and a damn sharp lens, the scanner won't really be your limit! The nice thing about film is that you can always move to a larger format if you really need more detail.

Oh I agree completely and this of course brings up something that has many confused. I have a few cameras here and of course have done lens testing and film testing. There is lens resolution and film resolution. People often get these two confused.

However in a nutshell my point is that with larger film area you can have less of both and get sharper images.

Less film resolution is assisted with larger formats for the obvious reason of increased capture area and therefore decreased magnification on a given print size.

Less lens resolution is not as obvious, but comes down to how lens testing works. Targets are photographed at a function of lens focal length. So a 100mm lens will be photographed at a greater distance than a 35mm lens. Now as it happens on 6x9 film a 100mm lens is about the same angle of view as a 35mm lens is, but the working distance of reglar photography means that you will never stand as close to your subject to get the same view as you would locate your camera for lens testing.

When you place both cameras at the same distance (6x9 and 35mm) you'll end up with more lens line pairs per mm on the film or sensor.

24mm lens on a 10D camera:

which will give an effective view of a 100mm lens on a 6x9 camera.

but with both cameras at the same working distance that was required for testing the 100mm lens ...

we get a usable sort of image for this sort of photography... the smaller format ended up only showing about 1/2 the detail captured by the 6x9 camera:

so there are other advantages in larger formats, you need longer lenses for the same angle of view.

This is extracted from a blog post of mine here.

 

[Loris Medici]

I'm sorry but 40 lines per mm (l/mm) is nothing near of something you would expect from a good (not even high quality) 35mm system (even in a casual shooting context). Anything giving resolutions below 77 l/mm (lines, not line pairs!) in reasonable conditions, I'd call a toy camera... (This is the starting point, I'd normally only pay for something significantly better...) Do the math, and you'll agree with me...

40 lpmm is not so bad! Unless you use a sturdy tripod, mirror lockup, careful manual focussing and a damn sharp lens, the scanner won't really be your limit! The nice thing about film is that you can always move to a larger format if you really need more detail.

 

[Rudeofus]

I'm sorry but 40 lines per mm (l/mm) is nothing near of something you would expect from a good (not even high quality) 35mm system (even in a casual shooting context). Anything giving resolutions below 77 l/mm (lines, not line pairs!) in reasonable conditions, I'd call a toy camera... (This is the starting point, I'd normally only pay for something significantly better...) Do the math, and you'll agree with me...

You mistake MTF number from controlled lens tests with real world hand held shooting. If I chase my daughter around with my camera, hand held, lens wide open, no flash, auto focus, well, you get the idea. No chance I'll get real 40 lpmm, no matter what camera and lens you'd give me. Needless to say that my EOS 3 + 85L can give me higher resolution if I used that camera for carefully set up landscape work, but not for the style of shooting I do with a 35mm format SLR.

 

[Loris Medici]

I use RF's (w/ fast quality lenses) and don't chase children. That's quite not the usual / mundane hand-held shooting context anyway, it's a special case... OTOH, AF sensors have usually a sensing capacity of at least 50 l/mm even in case of entry level cameras, and wide open usually means better resolution in case of high-end lenses.

40 l/mm means you're limited with 4x enlargement, I mean before prints (roughly 4x6" in this case) start to get blurry to the naked eye. 35mm film/format isn't that much bad...

Regards,
Loris.

 

[Rudeofus]

I use RF's (w/ fast quality lenses) and don't chase children. That's quite not the usual / mundane hand-held shooting context anyway, it's a special case... OTOH, AF sensors have usually a sensing capacity of at least 50 l/mm even in case of entry level cameras, and wide open usually means better resolution in case of high-end lenses.

Despite the ravings of many 85L owners, according to MTF measurements the 85L is not particularly sharp wide open, and most fast lenses are rather soft unless stopped down. Add to that the fact that the AF takes no advantage of apertures larger than F/2.8, and you get quite inaccurate focussing at F/1.2, too.

40 l/mm means you're limited with 4x enlargement, I mean before prints (roughly 4x6" in this case) start to get blurry to the naked eye. 35mm film/format isn't that much bad...

Given that my daughter (2 years old) won't cooperate with any tripod based shooting, "mundane" hand held shooting is my only option. If I want higher resolution and have more time to get everything right, I use my RZ67.

 

[pellicle]

Hi

Add to that the fact that the AF takes no advantage of apertures larger than F/2.8, and you get quite inaccurate focussing at F/1.2, too.

a common misunderstanding. Recall that AF is completed with the lens wide open (unless you're using one of the rapid shooting AI Servo modes where the stopping down remains stopped down to speed the things up). Even still the camera does its first focus with the lens open, as it only stops down during the moment of taking.

 

[Rudeofus]

a common misunderstanding. Recall that AF is completed with the lens wide open (unless you're using one of the rapid shooting AI Servo modes where the stopping down remains stopped down to speed the things up). Even still the camera does its first focus with the lens open, as it only stops down during the moment of taking.

From what I heard the AF sensors do not take advantage of light rays from further to the outside as what would be F/2.8. So while AF takes place with the lens wide open, the extra light does not help.

 

[sanking]

I can not speak to the working habits of other photographers but for my own work in medium format the equivalent of 40 lp/mm would be totally inadequate. My equipment is Mamiya 7II and Bessa III and the type of work I do is landscape and architectural. I work with the camera on a tripod when possible, unless the film and subject conditions allows me to shoot at optimum aperture for the scene at a shutter speed of 1/125 or higher. I test my lenses with both targets and in practical picture taking situations and I am getting the equivalent of at least 60 lp/mm on the film, and in most cases a lot more than that.

BTW, when one tests a resolution target with a camera there is a formula that takes into account the actual focal length of the lens being used, and the distance of the lens from the camera. So regardless of whether one is using a 6X9 folder or a 35mm camera the focal length of the lens and the distance to the target is part of the equation to calculate real resolution in lp/mm.

Sandy King

 

[pellicle]

Hello Sandy

BTW, when one tests a resolution target with a camera there is a formula that takes into account the actual focal length of the lens being used, and the distance of the lens from the camera. So regardless of whether one is using a 6X9 folder or a 35mm camera the focal length of the lens and the distance to the target is part of the equation to calculate real resolution in lp/mm.

understood, for the charts I was photographing it was 51 x "focal length", so a 100mm lens is a 100mm lens, but a 100mm on a 6x9 covers a larger image circle than that on a 35mm camera. My expectation is that as it happens the test distance for such a lens works out to be a useful photographic coverage while to cover that same view on a smaller format camera that one would need to be using a shorter focal length lens and thus its testing resolution would not be equal to what was captured in scene for scene.

if I've made some false assumptions I would be pleased if you would point them out.

 

[sanking]

Hello Sandy

if I've made some false assumptions I would be pleased if you would point them out.

I was not suggesting that what you wrote was incorrect, just wanted to add the clarification about the relationship in resolution testing between distance and focal length for those who may have missed the point. My intention was not to question your assumptions. Clearly more real estate in terms of film size has its advantages and 40 lp/mm from a 6X9cm folder is going to give a much better print at a given size than 40 lp/mm from a 35mm negative.

Sandy

 

[Loris Medici]

In my view, there's definitely a problem in the assumption of 40 "lines per mm" (not line pairs! I wouldn't jump into the discussion that way if the figure in question was lp/mm...) is adequate and can't be surpassed in casual shooting conditions (casual = hand held / no tripod / no mirror-lock-up), if I know something about 35mm systems...

Regards,
Loris.

 

[pellicle]

Hi

I was not suggesting that what you wrote was incorrect,

understood. personally I am always reviewing and questioning what I write. Particularly with a complex issue I am never totally confident that I have not made some error in coordinating the logic of my assesment.

so I genuinely sought any clarification from somone whom I expect knows more about this than me.

 

[nsouto]

Not any longer. That scanner has been discontinued.

Don Bryant

But the 9000 hasn't and is still available.

 

[Rudeofus]

In my view, there's definitely a problem in the assumption of 40 "lines per mm" (not line pairs! I wouldn't jump into the discussion that way if the figure in question was lp/mm...) is adequate and can't be surpassed in casual shooting conditions (casual = hand held / no tripod / no mirror-lock-up), if I know something about 35mm systems...

Oops, I guess you discovered my error in thinking. Yes, I indeed and mistankenly read 40 line pairs per millimeter, where the original post actually wrote about lines per millimeter. I would still hope that the Canon scanner can do better than 40 lines per millimeter, since that would amount to no more than roughly 1000 dpi, a pretty poor value even for flat bed scanners. 40 line pairs per millimeter sounded a lot more like typical numbers from decent flat beds, I guess that's why I misread the post. My bad.