Is DSLR scanning "better" than flatbed scanning?

[John51]

It should be easy enough to make one that does 6x6 in 4 shots.

A square neg carrier, when slid up against the top left corner of the light box would have the lower right portion of the neg in position to be scanned. Sliding the carrier to the top right puts lower left part of the neg in position.

Doing 6 shots would involve having the top and bottom of the light box having a notch or similar in the middle for the carrier to index to. Then it would be top left, top middle, top right etc.

 

[BMbikerider]

Just curious if a high quality enlarging lens could be used with a DSLR, assuming you could somehow fit an enlarging lens onto a DSLR?

Most certainly! I used a Rodagon 80/F4 APO on a set of Pentax bellows with a modified slide adapter (my work) and a Nikon to Pentax screw adapter, removed the optical lens needed to correct the film/subject focussing distance and used it in conjunction with a D200. The scans were good.

This was a long while ago after I sold my 1st Nikon Scanner, but it was a time consuming set up so bought a newer Nikon Scanner (Coolscan V) which I still have and this is far better and has of course the dust and scratch removing ability.

 

[JaniceNixon]

Hey there, I know this is an old thread, but I just wanted to chime in about DSLR scanning vs flatbed scanning. I've found that DSLR scanning can produce better results, especially when you use high-quality lenses and adapters. I've used a similar setup with a Rodagon lens and bellows on my DSLR, and the scans were definitely better than what I could get with my flatbed scanner. Of course, it can be a bit time-consuming to set up, so it's not always the most practical option. If you're looking for a faster and more automated solution, you could try using {redacted link} to extract information from images. It's a great tool for digitizing old photos and documents, and it can even automatically remove dust and scratches.

 

[JerseyDoug]

My best scans of 35mm b&w negatives were done with a digital camera, the second best with a Plustek 8100 dedicated 35mm scanner, and both were better than I could do with an Epson V700. I judge the scanning results by how closely the grain of the scanned image resembles that of a good darkroom print from the same negative.

The best solution I've found for scanning 35mm negatives with a digital camera is the Leitz BEOON copy stand. It is very compact and very rigid. It fits in a shoebox and sets up in seconds, with no assembling or adjusting needed (if you can dedicate a digital camera to its use). It works with either an M-mount camera and lens or an LTM mount camera and lens. Unfortunately the BEOON hasn't been made for many years and they are not cheap.

 

[beemermark]

Compare to your V550 of 4800dpi it would yield 30MP image from a 35mm film. More on 120 film. If you use a camera with more than 30MP you can have higher resolution for 35mm film but it's hard to beat the scanner with 120 film in term of resolution. Using a camera, setting it up takes time and it can be difficult to hold the film flat and you would also need a very good lens for this.

So does a 35mm negative have more than 4800 dpi/30mp of information? My gut feeling is no. I no longer have a darkroom but a 11x14 print from a 35mm negative (using an excellent lens and tripod) was about the maximum print size if viewed up close. I've been using my Epson 700 scanner for a decade or more and have some very excellent prints from it, Since my prints hand on my walls you usually can't get closer than 3 feet, so 16x20 prints looks fantastic. The human eye can only resolve so much detail.

 

[fiddle]

I dont believe the Epson's resolve anything over about 2600. from my reading, so that 30mp scan might be a bit of a stretch.

 

[Adrian Bacon]

I dont believe the Epson's resolve anything over about 2600. from my reading, so that 30mp scan might be a bit of a stretch.

For full color that's true, but if you're scanning black and white, and you're using Vuescan and you scan with just the blue channel, you can eek out closer to 3200 dpi using the higher resolution lens and careful focusing via the height of the film holder, but that's about the upper limit. In practice, it's more like 2800-3000 dpi. Even still, in terms of pictorial information 2600 dpi is only about 50 line pairs per mm, how many of us are actually putting that much information onto the film in terms of pictorial information. I'd wager that most of use aren't anywhere near that if shooting hand held and older lenses. We might be able to get close to that if we use a tripod, but in my shop, my standard baseline scan is 3200 dpi and in all honesty, the vast majority of the frames I see either have motion blur, or aren't sharply focused, so 3200 dpi is nice to have a reasonable rendering of the film grain, but scanning with more resolution isn't going to do anything but render more grain.

 

[Helge]

For full color that's true, but if you're scanning black and white, and you're using Vuescan and you scan with just the blue channel, you can eek out closer to 3200 dpi using the higher resolution lens and careful focusing via the height of the film holder, but that's about the upper limit. In practice, it's more like 2800-3000 dpi. Even still, in terms of pictorial information 2600 dpi is only about 50 line pairs per mm, how many of us are actually putting that much information onto the film in terms of pictorial information. I'd wager that most of use aren't anywhere near that if shooting hand held and older lenses. We might be able to get close to that if we use a tripod, but in my shop, my standard baseline scan is 3200 dpi and in all honesty, the vast majority of the frames I see either have motion blur, or aren't sharply focused, so 3200 dpi is nice to have a reasonable rendering of the film grain, but scanning with more resolution isn't going to do anything but render more grain.

It should be pointed out that grain size is not the limit of resolution for film. In much the same way tape hiss from magnetic oxide and domains isn’t the upper frequency limit of sound on tape.

You can decimate any medium down to general mediocrity with that kind of reductive thinking. Sometimes we do use great lenses and do use a tripod or high shutter speeds in good light.
And with a favorite shot, we might want to print big or simply pixel peep to our hearts content.
For those occasions high resolution scanning is it.

 

[Adrian Bacon]

It should be pointed out that grain size is not the limit of resolution for film. In much the same way tape hiss from magnetic oxide and domains isn’t the upper frequency limit of sound on tape.

You can decimate any medium down to general mediocrity with that kind of reductive thinking. Sometimes we do use great lenses and do use a tripod og high shutter speeds in good light.
And with a favorite shot, we might want to print big or simply pixel peep to our hearts content.
For those occasions high resolution scanning is it.

I don’t disagree at all, and in this forum, more resolution is totally appropriate, however, I also rarely service people in this forum on a regular basis. My average client is a college age female who rarely knows the difference between a jpeg and a tiff, and grew up taking pictures on an iPhone, and thinks 12 MP is a lot of resolution. A significant number of them don’t even have a computer, they do it all on an iPhone or some sort of tablet device, and don’t want large files.

 

[Alan Edward Klein]

For full color that's true, but if you're scanning black and white, and you're using Vuescan and you scan with just the blue channel, you can eek out closer to 3200 dpi using the higher resolution lens and careful focusing via the height of the film holder, but that's about the upper limit. In practice, it's more like 2800-3000 dpi. Even still, in terms of pictorial information 2600 dpi is only about 50 line pairs per mm, how many of us are actually putting that much information onto the film in terms of pictorial information. I'd wager that most of use aren't anywhere near that if shooting hand held and older lenses. We might be able to get close to that if we use a tripod, but in my shop, my standard baseline scan is 3200 dpi and in all honesty, the vast majority of the frames I see either have motion blur, or aren't sharply focused, so 3200 dpi is nice to have a reasonable rendering of the film grain, but scanning with more resolution isn't going to do anything but render more grain.

I scan with my EPson V850 and V600 at 2400, 16 bit grayscale. But I notice the Epsonscan software has the ability to block each of the color channels. Would I be better off just scanning on the blue channel when scanning BW film rather than using the 16 bit grayscale setting?

 

[Adrian Bacon]

I scan with my EPson V850 and V600 at 2400, 16 bit grayscale. But I notice the Epsonscan software has the ability to block each of the color channels. Would I be better off just scanning on the blue channel when scanning BW film rather than using the 16 bit grayscale setting?

If you’re happy with the output, there’s no need to change anything. I’m not familiar with Epson scan so I don’t know how it does it.

that being said, yes, the blue only channel is higher resolution than the other channels.

 

[Steven Lee]

Even still, in terms of pictorial information 2600 dpi is only about 50 line pairs per mm, how many of us are actually putting that much information onto the film in terms of pictorial information. I'd wager that most of use aren't anywhere near that if shooting hand held and older lenses.

You are not wrong, but in practice the grain from ISO 400 films looks pretty weird if scanned at anything less than 3000dpi, and that weirdness is evident even in typical screen/web use cases. Film has texture. The texture contributes to the image character. This is completely separate from resolution. Maybe I'm not being coherent here, but IMO the grain is the primary beneficiary of scanning resolution, not the image detail.

 

[Adrian Bacon]

You are not wrong, but in practice the grain from ISO 400 films looks pretty weird if scanned at anything less than 3000dpi, and that weirdness is evident even in typical screen/web use cases. Film has texture. The texture contributes to the image character. This is completely separate from resolution. Maybe I'm not being coherent here, but IMO the grain is the primary beneficiary of scanning resolution, not the image detail.

I hear what you're saying, however, there's a lot of variables that can affect the appearance of grain on the scanning side, and how much you're enlarging it on the output side will affect to along with the processing chain between the two, so it's not just a simple "you need a minimum of x DPI or it looks weird".

Maybe with your particular setup, you find that ~3000 dpi is where things start to look OK to your eyes, which is totally valid. I've done scans of 35mm BW at my standard 3200 dpi with my setup and have gotten comments about how much more detail and sharper my scans are compared to the same frame scanned by another system at supposedly a higher resolution, and the only thing I can say is none of my scanning setup applies any sharpening, and how you handle the raw samples after they are read off the image sensor matters a lot with how it's going to look in the end.

 

[L Gebhardt]

I've read all these mp numbers in this thread. Are you all referring to 8-bit jpeg files?

When I scan a 35mm color negative or slide with a Plustek 7600i scanner at 3200 dpi, the 48 bit (16 bit per channel) TIFF file is about 105 mp. Don't you want 16 bit files to save the maximum color information? (I recall a long thread debating this topic 1 or 2 years ago.)

when talking about mp (megapixels) it’s just the two dimensions of the image multiplied together. For example a 6000x4000 pixel image is 24mp, or 24,000,000 pixels. Ideally each of to pixels would contain 16 bits of color information per channel as you mention. Are you perhaps confusing MP with MB (megabytes)?

 

[Helge]

This is twelve years old, but it hasn’t aged a day because scanning mostly hasn’t and because of fundamentals.

https://www.onlandscape.co.uk/wp-content/uploads/2011/12/microscope.jpg

It is interesting for many reasons.

- It shows us that drum scanning is not the be all, end all perfect scanning we’ve been told it is. But still better than most types of dedicated film scanner in use today.
- It shows the difference between diffuse light and condensed light, and why condensed/collimated light is probably not ideal for scanning.
- And, it also shows that film has more resolution than even the best traditional scanners.
And the artifacting we can expect from scanners when scanned at even 8000 dpi: Optical beating, interference, streaks etc.
Even something like Portra 400 has very impressive resolving power with the right kind of scanning (a digicam on a microscope in this case).

 

[GLS]

There's quite a lot of variance in the quality of those microscope captures, which isn't doing the comparison any favours (the Delta 100 8x10 drum scan looks much better than the microscope image, for example).

 

[Helge]

There's quite a lot of variance in the quality of those microscope captures, which isn't doing the comparison any favours (the Delta 100 8x10 drum scan looks much better than the microscope image, for example).

That only works to the advantage of the overall point AFAICS.
Even a non-optimal microscope grab is better than the drum scan.
The 8x10 is choppier simply because it’s segmented. But the resolution is better on the microscope, no discussion needed.
And important to point out: This is not about format. This is about those particular emulsions at grain level. We can only infer scale.

 

[Adrian Bacon]

This is twelve years old, but it hasn’t aged a day because scanning mostly hasn’t and because of fundamentals.
View attachment 332262

https://www.onlandscape.co.uk/wp-content/uploads/2011/12/microscope.jpg

It is interesting for many reasons.

- It shows us that drum scanning is not the be all, end all perfect scanning we’ve been told it is. But still better than most types of dedicated film scanner in use today.
- It shows the difference between diffuse light and condensed light, and why condensed/collimated light is probably not ideal for scanning.
- And, it also shows that film has more resolution than even the best traditional scanners.
And the artifacting we can expect from scanners when scanned at even 8000 dpi: Optical beating, interference, streaks etc.
Even something like Portra 400 has very impressive resolving power with the right kind of scanning (a digicam on a microscope in this case).

The IQ180 and Nikon D3X also illustrate that what you do with the samples afterwards matters a lot. There is no color information there, so quite a lot more resolution can be unlocked by white balancing the raw samples, then just treating them directly as a monochrome image. This will immediately eliminate the color artifacts that are caused by the interpolation when demosaicing the CFA, and because you’re now not doing any interpolation of the samples, a whole pile of fine detail is not being erased by that process, and the resolving power and contrast response of the sensor comes through. Digital sensors tend to have a 100% contrast response all the way up to the nyquist of the sensor, so this results in a massive boost of perceived baseline sharpness.

 

[Helge]

The IQ180 and Nikon D3X also illustrate that what you do with the samples afterwards matters a lot. There is no color information there, so quite a lot more resolution can be unlocked by white balancing the raw samples, then just treating them directly as a monochrome image. This will immediately eliminate the color artifacts that are caused by the interpolation when demosaicing the CFA, and because you’re now not doing any interpolation of the samples, a whole pile of fine detail is not being erased by that process, and the resolving power and contrast response of the sensor comes through. Digital sensors tend to have a 100% contrast response all the way up to the nyquist of the sensor, so this results in a massive boost of perceived baseline sharpness.

That’s because you can pull up microcontrast almost as much as you want in demosaicing, because there is no physical substrate structure to deal with, only regular samples.

You’ll need at least three samples in a real world situation to place a point, line or edge accurately. Preferably quite a bit more, because of Bayer filtering and a possible AA filter (physical or not).

When someone (have they already?) figures out a convolution/transformer etc. network that can “see through” grain the way we humans kind of can, then you can pull the contrast of negative film with loads of sub 20% MTF detail up all you want.
That will need a high resolution scan though. The more of the grain detail the AI can see the better.

With slide film the problem with grain and colour masks lowering low contrast detail is less, for various reasons.
Probably part of the reason it was so popular in print. First with raster patterns and later with scanning.

 

[wiltw]

This is twelve years old, but it hasn’t aged a day because scanning mostly hasn’t and because of fundamentals.
It is interesting for many reasons.

- It shows us that drum scanning is not the be all, end all perfect scanning we’ve been told it is. But still better than most types of dedicated film scanner in use today.
- It shows the difference between diffuse light and condensed light, and why condensed/collimated light is probably not ideal for scanning.
- And, it also shows that film has more resolution than even the best traditional scanners.
And the artifacting we can expect from scanners when scanned at even 8000 dpi: Optical beating, interference, streaks etc.
Even something like Portra 400 has very impressive resolving power with the right kind of scanning (a digicam on a microscope in this case).

It was indeed interesting to see things only from a detail content consideration,with resolution targets reproduced.
.
Yet, for many, we are not simply replicating B&W images, we are copying color prints and color transparencies and color negs...and in the case of color neg duplication, the quality of the color positive result can make the use of [camera + lens] a far interior result, because of color tint and loss of contrast. We need to not only get right the mechanics (film flatness, lens choice, eveness and color balance of the illumination), but the negative-to-positive conversion software is very important.

Years ago I took a Kodacolor negative shot 40 years ago, and scanned it with a Canon photo/film scanner and also shot the negative with a Canon camera and lens using full spectrum Daylight balance lighting, to compare the positive image which resulted...
First the scanner with the scanner software's built-in neg-to-positivie conversion

Now the camera-shot image sequence...with postprocessing software's color inversion function, then attempts to improve contrast, etc.

There certainly are some available software conversions that do a better job of the neg-to-positive conversion than the result of the postprocessing software that I used. The above sequence illustrates the importance of obtaining good neg-to-positive conversion software...do not assume that your postprocessing sofrware will do a good job simply because you see it can Invert image!

 

[Helge]

It was indeed interesting to see things only from a detail content consideration,with resolution targets reproduced.
.
Yet, for many, we are not simply replicating B&W images, we are copying color prints and color transparencies and color negs...and in the case of color neg duplication, the quality of the color positive result can make the use of [camera + lens] a far interior result, because of color tint and loss of contrast. We need to not only get right the mechanics (film flatness, lens choice, eveness and color balance of the illumination), but the negative-to-positive conversion software is very important.

Years ago I took a Kodacolor negative shot 40 years ago, and scanned it with a Canon photo/film scanner and also shot the negative with a Canon camera and lens using full spectrum Daylight balance lighting, to compare the positive image which resulted...
First the scanner with the scanner software's built-in neg-to-positivie conversion

Now the camera-shot image sequence...with postprocessing software's color inversion function, then attempts to improve contrast, etc.

There certainly are some available software conversions that do a better job of the neg-to-positive conversion than the result of the postprocessing software that I used. The above sequence illustrates the importance of obtaining good neg-to-positive conversion software...do not assume that your postprocessing sofrware will do a good job simply because you see it can Invert image!

Yeah, that is a big problem not faced with B&W and slide.

The main problem is that the peakiness of the colour couplers in RA4 paper is not replicated in the scanning process.
I don’t know how it is dealt with in “good old scanners”?
Whether it is the light source that is peaky (the proper solution) or some careful custom channel tweaks?

Notice how much more shade detail is in the digishot version though.
Even better consumer scanners, have a nasty habit of drowning out fine detail in the clear parts of the negative. And also at the other extreme, have a hard time punching through dense slide.
That is very rare with camera scanning.

 

[runswithsizzers]

I am in the process of using my Fuji XT-1 digital camera to copy about 300 Ektachrome slides. I had previously scanned this same set of slides using a Minolta film scanner. When I have finished, I will try to compare my results from the two different processes, and comment on some of the differences. Since both my camera (from about 2013) and my film scanner (from about 2000) are antiquated, my results may not be that relevant to those who can afford state-of-the-art gear.

On the other hand, I think the whole topic of scanning slides has an old-fashined, d.i.y. vibe to it - similar to restoring a classic automobile, or building a boat in your basement - so maybe my results will be of interest to someone?

 

[Despite_it_all]

I am in the process of using my Fuji XT-1 digital camera to copy about 300 Ektachrome slides. I had previously scanned this same set of slides using a Minolta film scanner. When I have finished, I will try to compare my results from the two different processes, and comment on some of the differences. Since both my camera (from about 2013) and my film scanner (from about 2000) are antiquated, my results may not be that relevant to those who can afford state-of-the-art gear.

On the other hand, I think the whole topic of scanning slides has an old-fashined, d.i.y. vibe to it - similar to restoring a classic automobile, or building a boat in your basement - so maybe my results will be of interest to someone?

I'm interested in your results, Gary.

 

[PeterTredrea]

I am in the process of using my Fuji XT-1 digital camera to copy about 300 Ektachrome slides. I had previously scanned this same set of slides using a Minolta film scanner. When I have finished, I will try to compare my results from the two different processes, and comment on some of the differences. Since both my camera (from about 2013) and my film scanner (from about 2000) are antiquated, my results may not be that relevant to those who can afford state-of-the-art gear.

On the other hand, I think the whole topic of scanning slides has an old-fashined, d.i.y. vibe to it - similar to restoring a classic automobile, or building a boat in your basement - so maybe my results will be of interest to someone?

I have been following the same path and just completed a vertical slide copying system. Have concluded the following after a few trials:

1. 24 MPx camera (Nikon D7200) will capture the grain structure of Kodachrome.
2. JPG format is not satisfactory, grain is translated into JPG artifacts Used RAW format and image was fathfully captured.
3. LED lamp (Blue + Phosphor) gave surprisingly good results. Commenced with B&W negatives where WB was not an issue.
4. White Balance based on a 35mm kodalith slide when corrected to grey gave a good represtentation when transfered to the Kodachromes and Ektachromes
5. Larger (6x6) transparencies could be scanned into 4 segments using a square fabricated guide. The stitched image was first class
6. Dust was a problem. Built a cleaning system with a small budget vacuum cleaner filled with clean toy stuffing. Blows and sucks in a small chamber. Combined with a soft brush removed most

More to learn, will report again.

 

[Cholentpot]

I have been following the same path and just completed a vertical slide copying system. Have concluded the following after a few trials:

1. 24 MPx camera (Nikon D7200) will capture the grain structure of Kodachrome.
2. JPG format is not satisfactory, grain is translated into JPG artifacts Used RAW format and image was fathfully captured.
3. LED lamp (Blue + Phosphor) gave surprisingly good results. Commenced with B&W negatives where WB was not an issue.
4. White Balance based on a 35mm kodalith slide when corrected to grey gave a good represtentation when transfered to the Kodachromes and Ektachromes
5. Larger (6x6) transparencies could be scanned into 4 segments using a square fabricated guide. The stitched image was first class
6. Dust was a problem. Built a cleaning system with a small budget vacuum cleaner filled with clean toy stuffing. Blows and sucks in a small chamber. Combined with a soft brush removed most

More to learn, will report again.

Can you post a photo of your vacuum cleaner setup?

I've also been doing 3 or 4 stitched photos of 6x6. Software does a good job stitching these days. I am using a lower MP camera than yours. A Canon 550D 18mp. It does the job as far as I can tell. I may attempt to switch to a 6D full frame but I don't think a larger sensor would be the difference maker at this point. Might be worth finding a used higher mp DSLR.