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

[mshchem]

Another film transport system.

 

[Steven Lee]

A scanner shouldn't need software. "Scanning software" is a fossil from the 90s. I too have been thinking about building my own, and in my design a film scanner implements USB mass storage protocol class: press a button and you get a folder with DNGs for each scanned frame. Want live preview? USB supports webcam protocol. Want ICE on/off? Press a hardware button. No need for PC software.

 

[mshchem]

A scanner shouldn't need software. "Scanning software" is a fossil from the 90s. I too have been thinking about building my own, and in my design a film scanner implements USB mass storage protocol class: press a button and you get a folder with DNGs for each scanned frame. Want live preview? USB supports webcam protocol. Want ICE on/off? Press a hardware button. No need for PC software.

Sounds nice

 

[koraks]

But how can one ASSUME that the spectrum of light from any given LED source is truly as continuous as electronic flash or incandescent source?

It doesn't have to be continuous at all.
And this is where manufacturer specs come in. Reputable brands publish this data in their datasheets. Even many Chinese resellers have simple spec sheets with this sort of data for their products.

 

[Richard Man]

A scanner shouldn't need software. "Scanning software" is a fossil from the 90s. I too have been thinking about building my own, and in my design a film scanner implements USB mass storage protocol class: press a button and you get a folder with DNGs for each scanned frame. Want live preview? USB supports webcam protocol. Want ICE on/off? Press a hardware button. No need for PC software.

What makes those frames appear as files?! ;-)

Oh yes, by magic.... programming code, otherwise, known as software

 

[koraks]

What makes those frames appear as files?! ;-)

Oh yes, by magic.... programming code, otherwise, known as software

I think you misunderstood what @Steven Lee suggested. He proposes to do that part in embedded software in the scanner itself. He specifically mentions "no PC software".
There's some merit to his idea in terms of long-term usability of a device, but I doubt it balances out well with the initial investment in making the embedded software. It's generally easier and quicker to engineer the same solution on a PC/desktop platform than in an embedded context due to the limitations of the latter.

 

[Chan Tran]

I would use a camera for the digitizing because current software package can still open files from very old cameras. So future image editing software would be able to open the files from camera I use today. Most cameras are capable of 14 bit depth. With a typical 24MP camera would be more than 4000dpi for a 35mm frame.

 

[brbo]

It's generally easier and quicker to engineer the same solution on a PC/desktop platform than in an embedded context due to the limitations of the latter.

Yes, and even if scanner basically had some sort of a "PC on a stick" inside and you could run basically any higher abstraction level code on it, in these days you will still need to provide at least Windows and macOS software interfacing the scanner*. At that point you are basically doing software for three platforms without any significant savings in time/effort and just additional costs in hardware.

* Unless you really want to go with the idea of just dumping raw files onto poor user and having him figure what to do next. You must know that with such workflow you are only targeting people that already enjoy the horrible workflow of camera scanning so your 1.5k "scanner" is now competing with endless stream of camera scanning holders.

 

[Steven Lee]

There's some merit to his idea in terms of long-term usability of a device, but I doubt it balances out well with the initial investment in making the embedded software. It's generally easier and quicker to engineer the same solution on a PC/desktop platform than in an embedded context due to the limitations of the latter.

There are no limitations. Modern ARM-based Linux boards are no different from a PC from just a few years ago. One of my past jobs was building machine vision algorithms running on "the edge", i.e. right inside a camera. It was no different that building for a Linux box, and that was 12 years ago! These days you can package a lot of power into a small form factor for cheap.

Besides, I do not believe a scanner (or scanning software) should be doing any kind of complicated image processing either. Uninverted linear DNGs is what the output should be.

A more advanced version of the same idea runs a webserver and offers a web-based UI.

 

[gswdh]

There are no limitations. Modern ARM-based Linux boards are no different from a PC from just a few years ago. One of my past jobs was building machine vision algorithms running on "the edge", i.e. right inside a camera. It was no different that building for a Linux box, and that was 12 years ago! These days you can package a lot of power into a small form factor for cheap.

Besides, I do not believe a scanner (or scanning software) should be doing any kind of complicated image processing either. Uninverted linear DNGs is what the output should be.

A more advanced version of the same idea runs a webserver and offers a web-based UI.

It's a nice idea but not the path I will take. I think you can achieve a far better user interface on the screen and the advantages of moving the control software to the scanner doesn't gain that much in the fight against obsolescence - USB still needs supporting etc. The scanner will be "dumb" just outputting line data in a 16 bit format. I don't see the software maintenance being a big deal either.

 

[Richard Man]

My apology to Steven Lee for misunderstood his proposal ;-)

Sure, in that case, write the firmware as a web server, and write a webapp that talks to it to control the film format (e.g. xpan vs. normal), file qualities (JPG or DNG), and transfer to the client PC. The embedded code grabs the frames and does minimal processing.

But that's not what the OP wants, still could be interesting idea though.

 

[albada]

Hand-offs

Based on talks with mechanical engineers involved with scanning film, and my own experience, I know that hand-offs can be a problem with this type of design. A hand-off occurs when the film enters or leaves a roller.
Assume that you have a design like the one I posted a picture of above: A large rubber-lined roller with three small rollers on top. Let's call the small rollers L (left), M (middle), and R (right). Film moves left to right, so it will be fed using roller L, and scanned using roller M. There are two hand-off problems with this design:

1. Part way through the frame, the film will touch roller R. This disturbance can create an anomaly (artifact) in the scan.
2. Further in the frame, the film will come out from roller L, a second hand-off, creating a second anomaly.

I envision two possible solutions to these anomalies caused by hand-offs between rollers:

1. Use only one roller, eliminating hand-offs. This solution works well for an entire roll, but might make scanning the edges of a strip difficult. Strips are often cut poorly, cutting a bit into the image area, making the user want to scan to the edge.
2. Put a rubber belt around all three rollers, similar to the tread-belt on a tractor. This solution also eliminates hand-offs, but the mechanics are a little more complex.

Mark

 

[brbo]

These days you can package a lot of power into a small form factor for cheap.

But... why?

Why would you want to put a lot of power into the scanner when you don't want to do any data processing on board? That is just another unnecessary complication, cost and point of failure.

 

[Hasslebad]

I applaud the OP’s project. I also agree that with the market size for film scanning, the design should be supportable going forward. Maintainable hardware and open source software give all of us who love photography a chance of continuing to use these tools.

Steven Lee’s point, I think, is to make it easy for the user. He is proposing to partition the system such that you get DNGs out (without an app on your PC). It simplifies what a user would need to have in the future to use the scanner. Instead of having the correct app to run the scanner, the scanner appears like a disk full of DNGs.

Putting on the builder’s hat, there are already a lot of challenges and in my experience “keeping it simple” is the best way to proceed. It is easier to write the client on the computer. But I definitely understand Steven Lee’s point.

 

[Alan Edward Klein]

Where are you getting the capital to finance development and marketing? How much do you need? How do you intend to market?

 

[mshchem]

I can use the firewire scanner. If I can get your friend scanner cheap it would be great.

Chan Tran,
Other than a 15 year old Mac, what's a (ideally PC) computer that I can install a Firewire port? Can I just call up Dell or a local store and say make me one?
My computer/Firewire guru is a brilliant fellow, who literally, this morning flew out to his Arizona winter place. Iowa City is supposed to have snow tomorrow

 

[Richard Man]

Chan Tran,
Other than a 15 year old Mac, what's a (ideally PC) computer that I can install a Firewire port? Can I just call up Dell or a local store and say make me one?
My computer/Firewire guru is a brilliant fellow, who literally, this morning flew out to his Arizona winter place. Iowa City is supposed to have snow tomorrow

You just need a PC with a PCI-e slot, which is "all of them". Make sure you get ones with real TI 1394a compatible chip. Other chipset might not work as well. I have one to use with my Flextight and it works great. Worth that $50

 

[mshchem]

You just need a PC with a PCI-e slot, which is "all of them". Make sure you get ones with real TI 1394a compatible chip. Other chipset might not work as well. I have one to use with my Flextight and it works great. Worth that $50

Any source you recommend? B&H probably has something. Last couple computers I purchased are laptops, I want a tower like the good old days.

 

[Steven Lee]

But... why?

Why would you want to put a lot of power into the scanner when you don't want to do any data processing on board? That is just another unnecessary complication, cost and point of failure.

It is a huge simplification, because you don't have to build Linux, MacOS and Windows versions. It is also cost efficient, because the cost of supporting downloadable desktop software over time is enormous (speaking from experience).

 

[Richard Man]

Any source you recommend? B&H probably has something. Last couple computers I purchased are laptops, I want a tower like the good old days.

I either built my computers or purchased Dell for my daughter. It really depends on what else you use the computer for, and what SW you want to run with. Pretty much all PC will have at least one PCI slot. Simplest to do is go to Dell and find a tower that suits your budget and performance requirement, then double check that they have at least one slot

Specifically, this is the TI based card that I have:

 

[bfilm]

Any source you recommend? B&H probably has something. Last couple computers I purchased are laptops, I want a tower like the good old days.

FireWire 400 1394a

SIIG DP 2-Port FireWire PCIe

SIIG DP 2-Port FireWire PCIe for purchase at CDW

 

[Adrian Bacon]

Simply put, if you make a scanner today that only uses the dedicated USB3 contacts at the back of the plug, it won't talk to any computer with a USB2.0 interface. The computer will have to be upgraded with a USB3 interface card, which in the case of laptops may prove impossible or not feasible. On the other hand, if the scanner has a USB2.0 interface, it's usable on pretty much any computer in use today on this planet and new computers for the years to come.

I realize that I'm very late to the party. This is not true. A USB3 device works just fine with a USB2 host. You simply use a USB cable that has a USB 2 Type A connection on one end that plugs into the computer that has USB 2 and the other end of the cable has whatever USB 3 connection it has (USB 3 Type B, USB-C, etc). They will negotiate the appropriate speed. I know this because I actually have several computers that are USB2 only and regularly plug in USB3 devices to them with the appropriate cable. It works just fine, albeit at USB 2 speeds.

 

[mshchem]

I either built my computers or purchased Dell for my daughter. It really depends on what else you use the computer for, and what SW you want to run with. Pretty much all PC will have at least one PCI slot. Simplest to do is go to Dell and find a tower that suits your budget and performance requirement, then double check that they have at least one slot

Specifically, this is the TI based card that I have:

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.......

 

[koraks]

@Adrian Bacon it works in your case because your usb3 devices are engineered to be made backwards compatible with older usb2 hosts. This backwards compatibility is however not a given in building a new device; it's compatibility that needs to be added specifically along with native usb3 support. Please have a look at the differences in hardware connectivity and protocols. Usb3 adds complexity for the engineer of in this case the scanner while not constituting any added value in return for it.

Usb2 support is relatively easy; use any of the e.g. stm32 platforms with USB support and you're pretty much set. Usb3, not so much; count on having to add peripheral hardware for your uC of choice, be restricted to the few uC's that have native usb3 support (I guess those exist ...?) or be forced to not get away with a uC and use the sledgehammer approach of integrating an actual computer into the device as Steven Lee suggested. Which is massive overkill for again very little benefit (at least in terms of the USB connectivity).

Don't be fooled by your user-perspective experience which is NOT representative of the engineer's perspective!

 

[prado333]

It depends doesn't it? If it only takes a few minutes to scan the whole roll, why not? The export time is trivial for modern computers and well written software. I know because my prototype already does so. Scanning the whole roll in one go is obviously much more efficient than feeding in or mounting strips of film.

I really don't think the USB version will make any difference, it would probably add another €50 to the cost of the scanner for literally no purpose.

I think only labs wants to scan a whole roll, people usually scan frame by frame looking for maximum quality all least is my case. What i search in a scanner really exists , for example the now difunt hasselblad scanners or the heidelberg topaz drum scanners.