I'm making my own film scanner and will attempt to sell it

[gswdh]

Being able to scan up to large format would always make it more desirable!

Not if it means it'll cost 10 times more!

 

[brbo]

I have a scanner that scans a roll at 4.000dpi in 10 minutes. So, it's dog slow compared to the 2min requirement of this time and age where people shoot and scan billions of rolls per year and speed is of utmost importance...

Still, I feel that it only wastes less than a minute of my time*. This is the time that it needs to pull in the entire film in order to make previews of all the frames. After that it immediately starts ejecting the film while scanning at full resolution. I rarely finish my adjustments to every frame before the scanner is over with making full res raw scans. The final step exports final images with my corrections to desired location in desired format (tif, jpg). This step is basically limited by HDD/SDD speed. In my case it takes a couple of seconds.

* Actually, I don't consider the initial time needed to make previews a lost time. Previewing entire roll first can be quite beneficial. In case of negatives the software can make better initial inversions if it can analyse all frames on the roll and can adjust sensor exposure time for under/overexposed frames when doing full res scans.

 

[gswdh]

I have a scanner that scans a roll at 4.000dpi in 10 minutes. So, it's dog slow compared to the 2min requirement of this time and age where people shoot and scan billions of rolls per year and speed is of utmost importance...

Still, I feel that it only wastes less than a minute of my time*. This is the time that it needs to pull in the entire film in order to make previews of all the frames. After that it immediately starts ejecting the film while scanning at full resolution. I rarely finish my adjustments to every frame before the scanner is over with making full res raw scans. The final step exports final images with my corrections to desired location in desired format (tif, jpg). This step is basically limited by HDD/SDD speed. In my case it takes a couple of seconds.

* Actually, I don't consider the initial time needed to make previews a lost time. Previewing entire roll first can be quite beneficial. In case of negatives the software can make better initial inversions if it can analyse all frames on the roll and can adjust sensor exposure time for under/overexposed frames when doing full res scans.

That sounds interesting, would you mind sharing your setup with us?

 

[brbo]

My scanner is a Noritsu LS-600 with EZ Controller software. All Noritsu scanners work like that.

 

[gswdh]

My scanner is a Noritsu LS-600 with EZ Controller software. All Noritsu scanners work like that.

It does seem quite expensive and I can only find second hand equipment.

 

[brbo]

Yes, even used ones are expensive now and no new Noritsu scanners are produced anymore. There was a time when such used scanners sold for cheap. I got it for 600 EUR or something like that.

I need to replace the rollers on mine since those are now leaving marks on film. That alone will set me back some 300 EUR if I get them from China. OEM replacement parts probably cost at least three times more money, so you can guess that keeping those scanners running is not cheap and some vital parts are not available anymore (like CCDs, LEDs...) at all.

 

[mgb74]

Thanks for your help. I just ordered the card. Now I'm going to try and find a simple computer that doesn't have ghosts of Bing and HP and so on.......

Maybe I'm missing something, but why not a refurb Dell desktop. Erase the drive and install your own OS. But you make have to go to a workstation model to have room for full height cards.

 

[mshchem]

Maybe I'm missing something, but why not a refurb Dell desktop. Erase the drive and install your own OS. But you make have to go to a workstation model to have room for full height cards.

I bought a new Dell workstation tower. Cheap, nothing but Windows Pro 11. Has room for 3 full size cards. I needed a new machine. I swear if it comes pre-loaded with a bunch of crap I'm going to send it back.

 

[mgb74]

I bought a new Dell workstation tower. Cheap, nothing but Windows Pro 11. Has room for 3 full size cards. I needed a new machine. I swear if it comes pre-loaded with a bunch of crap I'm going to send it back.

If sold as a workstation, it MAY not be preloaded with crap. Every consumer PC and laptop I'm aware of has come preloaded with crap. That's why you have to erase what's there and do a fresh install.

 

[koraks]

Just remove the crap. It's not rocket science.

Anyway, could this be discussed in a different thread? Feel free to open something in the Lounge or something.

 

[mshchem]

Just remove the crap. It's not rocket science.

Anyway, could this be discussed in a different thread? Feel free to open something in the Lounge or something.

I'm done

 

[Archiloque]

As a firm believer in the power of dedicated scanners, I'll chime in and send my santa claus wishlist.

I am not sure a 2000dpi scanner - or even 3000 dpi - would be an instant seller for me. I get the appeal for small labs necessitating a fast turnaround and fair quality. But for enthousiasts like us, ready to spend 1000€+ on a scanner, wanting something on par with 20+ years old legacy scanners doesn't seem too exagerated.

And yes : I can see a pretty significant difference between my 4000 dpi coolscan and my 2000 dpi Pakon. As fast as my pakon is, I seldom use it because I'd rather have the best resolution available to me when archiving my negatives. Will I really need the resolution : probably not. Do I want it anyway ? Of course I do !

I may never print large, but I'd like to be able to do it if necessary. So, in my case, I'm fine letting my Coolscan 5000 scanning my roll of film for 45 minutes instead of waiting 5 minutes with my Pakon. For users like me who shoots a few rolls a month, there is absolutely no difference whatsoever. I am not in front of my computer anyway.

So, if it's not too late, I'd be in your shoes, I'd try to equal or surpass the old 4000dpi reference. But I understand it's total wishful thinking, I just had so say it.

Oh and ICE is mandatory, I've ready you're already considering it but I'll insist. Let's be real, ICE is the last fallback trench of the scanner world, without ICE, i'd have switched to digital camera scanning a while back.

Everyting else has been said already.

 

[srb383]

It's always preferable to use at least a 16-bit sensor, especially for black and white film which can reach a D-max of around 3.0.

If we assume a 14-bit sensor, there's very little code value resolution for highlights.

CV 16383 - 100% transmittance - 0.000 density
CV 8192 - 50% transmittance - 0.301 density
CV 4096 - 25% transmittance - 0.602 density
CV 2048 - 12.5% transmittance - 0.903 density
CV 1024 - 6.25% transmittance - 1.204 density
CV 512 - 3.125% transmittance - 1.505 density
CV 256 - 1.562% transmittance - 1.806 density
CV 128 - 0.781% transmittance - 2.107 density
CV 64 - 0.390% transmittance - 2.408 density
CV 32 - 0.195% transmittance - 2.709 density
CV 16 - 0.097% transmittance - 3.010 density

16-bit scanners can keep the highest densities above 128. So there should be a noticeable difference in noise.

 

[Adrian Bacon]

It's always preferable to use at least a 16-bit sensor, especially for black and white film which can reach a D-max of around 3.0.

If we assume a 14-bit sensor, there's very little code value resolution for highlights.

CV 16383 - 100% transmittance - 0.000 density
CV 8192 - 50% transmittance - 0.301 density
CV 4096 - 25% transmittance - 0.602 density
CV 2048 - 12.5% transmittance - 0.903 density
CV 1024 - 6.25% transmittance - 1.204 density
CV 512 - 3.125% transmittance - 1.505 density
CV 256 - 1.562% transmittance - 1.806 density
CV 128 - 0.781% transmittance - 2.107 density
CV 64 - 0.390% transmittance - 2.408 density
CV 32 - 0.195% transmittance - 2.709 density
CV 16 - 0.097% transmittance - 3.010 density

16-bit scanners can keep the highest densities above 128. So there should be a noticeable difference in noise.

Yes, 3.0 can be reached by BW films, but in my experience, that's pretty aggressive exposure and/or processing. Fogged film base can reach that, but if exposed and processed reasonably, most pictorial information tops out at ~2.4. I use a 14 bit sensor, and put the film base plus fog at ~8000-10000 for the green channel, and rarely see code values lower than 32 (the rough equivalent of 2.4 density if FB+F was at CV 16383), and that's for extreme specular highlights. Regular highlights are in the CV 128-64 range.

That being said, is 16 bit better? You betcha. It gives you way more discrete tone values to work with, and if you keep the FB+F as close to ADC saturation as possible, then you have enough SN ratio in the denser parts (and discrete tone values) that noise just isn't much of an issue.

 

[koraks]

That being said, is 16 bit better?

If done well.

It's easy to request 16 bit resolution instead of 14 bit.
It's not so easy to make a signal path that actually yields the additional data and not just noise when going from 14 bit to 16 bit.

People forget too easily that an A-D system like a scanner really starts in the A world...that's analog.

 

[srb383]

Yes, 3.0 can be reached by BW films, but in my experience, that's pretty aggressive exposure and/or processing. Fogged film base can reach that, but if exposed and processed reasonably, most pictorial information tops out at ~2.4. I use a 14 bit sensor, and put the film base plus fog at ~8000-10000 for the green channel, and rarely see code values lower than 32 (the rough equivalent of 2.4 density if FB+F was at CV 16383), and that's for extreme specular highlights. Regular highlights are in the CV 128-64 range.

That being said, is 16 bit better? You betcha. It gives you way more discrete tone values to work with, and if you keep the FB+F as close to ADC saturation as possible, then you have enough SN ratio in the denser parts (and discrete tone values) that noise just isn't much of an issue.

Yeah base fog typically sits around a transmittance of 0.5, so that's appropriate. 14-bit would be fine for colour negative especially. It must also be noted that base density is an average, there's a good stop in transmittance above and below contributing to granularity for D-min.

 

[srb383]

If done well.

It's easy to request 16 bit resolution instead of 14 bit.
It's not so easy to make a signal path that actually yields the additional data and not just noise when going from 14 bit to 16 bit.

People forget too easily that an A-D system like a scanner really starts in the A world...that's analog.

Exactly. For color negative in particular, most people use high CRI white illumination and neutralise the base via white balance, but that's effectively analog gain, polishing a turd signal. If instead the orange mask were neutralised with light (cooler spectra), it would allow for a more consistent S/N ratio between channels.

 

[Romanko]

For color negative in particular, most people use high CRI white illumination and neutralise the base via white balance, but that's effectively analog gain

Does anyone know how white balance is actually implemented in digital cameras? Adjusting the gain of individual channels before digitization seems more logical than applying the corrections in post-processing.

 

[koraks]

Adjusting the gain of individual channels before digitization

That's what digital cameras do NOT do. White balance can be adjusted after the shot had been made on the raw (digital) data. That's one reason why people shoot raw.

 

[Romanko]

That's what digital cameras do NOT do. White balance can be adjusted after the shot had been made on the raw (digital) data. That's one reason why people shoot raw.

If I take two shots of the same grey card under the same conditions using different white balance settings of the camera will I get two raw files with identical pixel values but different white point levels?

 

[srb383]

Depending on the scanner. I had a monochromatic scanner in mind which can adjust the analog gain of each channel exposure made via custom filtration. But in both scenarios, amplifying the sensor voltage or manipulating the raw input code values can only work with the electrons captured. Think about what it means when you expose your sensor to orange light from the masking, more electrons will fill the well of the red and green filtered photosites than the blue filtered photosites. So before even moving to the digital domain, you will have less head room for the red and green channels than you will for the blue channel above D-min. So post white balancing is polishing a turd in essence.

If the backlight spectral power distribution multiplied by the masking spectral transmittance and sensor spectral sensitivity per channel achieved a trichromatic response which neutralises the orange mask (produces the same code values per channel) without the need for any post white balancing, and then the overall exposure were adjusted such that D-min transmittance sits at a code value of (2^bit_depth-1) / 2 per channel. That would allow for the cleanest capturing from D-min to D-max.

 

[koraks]

If I take two shots of the same grey card under the same conditions using different white balance settings of the camera will I get two raw files with identical pixel values but different white point levels?

Yes, indeed. The only difference is that each raw file will have a different color temperature tag embedded in it. This is why the images may look different once you load them in your raw processor, since that will display the image using the "as shot" white balance.

 

[Romanko]

Yes, indeed. The only difference is that each raw file will have a different color temperature tag embedded in it. This is why the images may look different once you load them in your raw processor, since that will display the image using the "as shot" white balance.

Thank you for clarifying this. I wonder if all digital cameras use this approach. I wish the manufacturers tell us more about what our cameras are actually doing with our images.

 

[gswdh]

Thing is there aren't any image sensors (apart from huge sensors designed for space or special industrial applications costing hundreds of thousands) capable of anything anywhere near 16 bits or even 14 bits. Sure, you can digitise at any resolution you like but you'll end up just digitising the noise.

The datasheet for the TCD1209DG, for example, says the dynamic range is ~2000 typically which is around 66dBV signal to noise. This can be captured in around 11bits, 12 just in case. This is a sensor with very large 14um pixels with a minimal signal path.

 

[koraks]

I wonder if all digital cameras use this approach

To the best of my knowledge, yes. I don't believe there's anyone out there setting gain per color channel depending on separately collected white balance data, although technically I imagine this could be possible.