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

[Hassasin]

I don't think so. At this point, opting for 3.0 only connectivity would incur a bigger penalty in terms of sales hurdles than opting for 2.0 will do for the foreseeable future. This is due to the hardware compatibility of both versions which will guarantee compatibility with 2.0 devices in the years to come, whereas the installed base of 2.0 machines won't be able to interface with 3.0-exclusive devices.
Furthermore, the 'regardless of throughput requirements' statement ignores the reality of the timing of the analog capture and A/D conversion and the resulting data streams, which is likely to be limited to well withing the capabilities of USB2.0.

Simply put, there's no compelling reason now to limit the design work to USB3.0. It seems @gswdh realizes this perfectly well and he's right in doing so.

I'm not an IT guru, how does backward compatibility stack up to what you say? I use 3.0 devices on older 2.0 machines without any issues.

And I only said what I said, because for general consumer, 2.0 in 2024 is like Ford T displayed at Tesla dealer.

 

[gswdh]

Looks like at least the B&W array has some sensitivity around 900nm and probably the red array does so, too. This means you could at least theoretically sample in the near-IR band and then use that data for dust & scratch removal. How feasible it is to write (or borrow) the algorithms for that recombination of data, I don't know, and I imagine that's where the real challenge is.

Cool project overall and yes, it seems feasible. I can't judge how the quality would stack up against what's available on the second hand market especially, but as you pointed out, that's a finite and limited supply. I think there are plenty of small labs interested in purchasing a new scanner to replace aging and perhaps temperamental Pakons, scavenged Frontiers etc. And the amateur market will also be present.

I think your main challenge really will be to stay ahead of Chinese competition. As soon as they figure out what you're doing, they're going to do their own version and likely undercut your price point significantly. So from a business standpoint, I'd certainly consider the possibilities of doing a 'hit & run', i.e. make the product profitable right off the bat with the first production run, so that anything that comes after it is a bonus. The main concern here would be the after sales support, since it might represent a commitment that trails long after the profitable part of the product life cycle.

Thank you for your feedback and well spotted on the spectral response there. I understand the premise of ICE, it would really require some experimentation to reverse engineer an algorithm and test it. I think it would be possible, my expertise is actually in signal processing.

 

[koraks]

how does backward compatibility stack up to what you say?

The USB3 interface is designed to remain backward compatible with USB2 devices. It's highly likely that any interface that follows after USB3.0 will retain the same policy so that even USB1.1 devices will remain usable.

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.

my expertise is actually in signal processing.

Ah, that helps. Besides, I can imagine open source libraries are now becoming available that offer functionality that could be used to implement ICE.

It's a very interesting and nice project; good luck with it! I'm looking forward to seeing how you'll manage the optical part; I'm not too enthusiastic about the idea of using a dSLR lens for this task since those lenses are not optimized for these working conditions (extreme close focus). I'd be tempted to either try and find a catalog part somewhere that fits the bill, or to team up with someone well-versed in optics design who can design a custom lens from stock elements for this purpose. This would of course have influence on production batch sizes, up-front investment etc. and as such it will place more emphasis on what I said about the possibility of competition. The Chinese market is both an opportunity (both in terms of supply and demand/sales), but it's also a major threat to your venture.

 

[Hassasin]

The USB3 interface is designed to remain backward compatible with USB2 devices. It's highly likely that any interface that follows after USB3.0 will retain the same policy so that even USB1.1 devices will remain usable.

So in other words, if I plug a 3.0 device into a 2.0 port it will NOT work? Meaning when I plug a 3.0 device into a 2.0 and it DOES work, designers got it all wrong?

Correct mine above as I am perplexed at what you are saying. From my reading, a 3.0 device will simply go slower on a 2.0, not that it will NOT work at all.

All I said was 3.0 is simply a "modern" interface found on pretty much every new device, hence anything less is NOT a good selling point, irrespective of data flow needs.

 

[gswdh]

USB 3 will just enumerate as a USB 2 HS device in a USB 2 port so it will be limited to USB HS speeds. If you want to use the USB 3 functionality somehow (higher throughput), this would then make the product incompatible with USB 2 even if data can be transferred. As @koraks says, the sensor readout and the digitisation is the limiting factor here, not the USB throughput so USB 3 is useless.

 

[koraks]

So in other words, if I plug a 3.0 device into a 2.0 port it will NOT work?

It depends on how it's implemented. If the USB3.0 device is limited to USB3.0 protocols only, it will NOT work on a USB2.0 host. USB3.0 uses an additional array of pins not present on a USB2.0 plug, and the transfer protocol is also different. I don't know on how many of the OSI layers the differences are, but since there's e.g. also (in addition to the hardware difference) a full-duplex provision on 3.0 that's missing on 2.0, the differences are likely to be quite fundamental.

A backward-compatible USB3.0 device will basically act as a USB2.0 device on a USB2.0 host; it will revert to using the USB2.0 pins and protocol. This means that a backward compatible USB3.0 device in reality is two more or less separate USB devices rolled into one. This additional complexity of course has implications for a development effort and from the perspective of someone like @gswdh, I'd imagine they'd want to focus on getting the essentials right (i.e. signal processing) and not add unnecessary complexity (and cost) by implementing an interface that yields no added value.

All I said was 3.0 is simply a "modern" interface found on pretty much every new device, hence anything less is NOT a good selling point, irrespective of data flow needs.

I know what you're saying and I think it's not accurate. There may be people in the future who balk at the lack of USB3 interfacing (although I doubt they'll walk away from the purchase in the face of a lack of alternatives!), until they learn that such an interface wouldn't have added any advantage, the device will still work fine with USB3.0-equipped computers etc. So it's a matter of adding a single line to the sales sheet: "Connectivity: USB2.0 High-speed (the scanner is fully compatible with and will also work at its full speed on computers with USB3.0)". This should settle the minds of all except for the one or two who are irrationally stubborn, and frankly, there's no way to win them over so I wouldn't ever focus my engineering efforts on them.

 

[Richard Man]

Great idea! I have a Pakon for fast scans and a Flextight for high quality scans but a good alternative is a great idea!

The Python code seems to be very basic. How do you detect the frame edges, or do you plan to post process a whole strip? How do you accommodate exposure variations? It would also be nice to be able to read the DX code if available.

Color is of course a huge problem. Besides color correction, make sure it works well with something like Negative Lab Pro is important. Also supporting some kind of dust removal would be nice.

Anyway, it would be trivial to convert the existing Python code to C++ code, but I suspect it's only about 10% of what is needed at best, not even counting the UI/UX part. If the HW works well though, it would be a wonderful platform.

 

[gswdh]

Great idea! I have a Pakon for fast scans and a Flextight for high quality scans but a good alternative is a great idea!

The Python code seems to be very basic. How do you detect the frame edges, or do you plan to post process a whole strip? How do you accommodate exposure variations? It would also be nice to be able to read the DX code if available.

Color is of course a huge problem. Besides color correction, make sure it works well with something like Negative Lab Pro is important. Also supporting some kind of dust removal would be nice.

Anyway, it would be trivial to convert the existing Python code to C++ code, but I suspect it's only about 10% of what is needed at best, not even counting the UI/UX part. If the HW works well though, it would be a wonderful platform.

I hope this would perhaps be able to replace both assuming one is only using 135 film.

It is basic at the moment, it's something I wrote in an evening basically. For my prototype I think I will write a port processing script to extract the individual frames but that would be included in the SW for production, of course. I think it can be done mostly with a high pass filter on the averaged line values combined with a fitting algorithm, it could even be done in the FPGA on the scanner.

I don't think the colour is a difficult problem to solve. The capture of the data without loss is the most important part for the scanner itself. This is done by setting the appropriate gain and offset values for each channel. Then, some LUTs can be used to convert this data based on the film type entered.

The DX code reading could be very interesting and maybe not too hard to do... thank you for that suggestion!

I would like to keep the scanner SW as simple as possible. Mainly to make it easier to ensure a good UX experience and to leave any superfluous bits to the likes of Photoshop etc.

 

[Paul Ozzello]

It really needs to be able to scan medium format film, and preferably all formats including 617 in one pass. And nothing lower than 4000-5000 true optical resolution. And film holders that hold the film perfectly flat, with the option of fluid mounting.

 

[koraks]

It really needs to be able to scan medium format film, and preferably all formats including 617 in one pass. And nothing lower than 4000-5000 true optical resolution. And film holders that hold the film perfectly flat, with the option of fluid mounting.

Sounds like you're looking for a fundamentally different kind of development than the one in progress here. Perhaps put out an RFQ somewhere and see if someone bites. It'll help if you could bring $250k seed capital.

 

[xkaes]

It depends doesn't it? If it only takes a few minutes to scan the whole roll, why not?

Varies from person to person. Right now, it normally only takes me a few minutes to scan the few 35mm shots on a roll that I want to scan. (After processing, I cut my film into strips to save in protective sheets & produce a contact sheet) Why spend more time on the computer deleting the vast majority of shots I don't want to scan -- and paying a lot for the privilege to do it?

 

[mgb74]

Because it is typically a waste of time, resources and post scan dumping requiring that extra time. I only see full roll capability for labs where they do it, obviously, and that is all they care about for mass scanning.

I also agree that in this day and age dumbing down connection to USB 2.0 is a bad design/business decision, regardless of throughput requirements. It simply will kill a lot of potential customers, simply because some things need to go with times. Digital has never been about nostalgia, it's all about next best, fastest etc.

I would argue that by scanning the entire roll, you’re just changing the point in the workflow where you evaluate the negatives. And it takes less time to scan an entire roll then to scan 3 (or even 2) strips of film. And you have to differentiate your product from inexpensive flat bed scanners.

 

[Andreas Thaler]

- Terrible software
- Are legacy hardware only working with legacy PCs
- Poor scan quality
- Poor user experience
- Extremely long and tedious scan times

I can't understand that with a Nikon Coolscan 5000 ED under VueScan. Stable software, manageable results, excellent quality with a quick scanning process. Other scanners may work slower, but if the situation were actually that bad, I would have heard about it.

With VueScan, which probably supports most scanners, the problem of outdated scanning software has long been solved.

 

[xkaes]

...it takes less time to scan an entire roll then to scan 3 (or even 2) strips of film.

You're obviously not using my scanners.

 

[mgb74]

Why wouldn't there be warranty or after sales support?

USB 2 is plenty for a scanner, is cheaper and is much easier to implement.

It’s one thing to design a great product, it’s another to manufacture it, and yet another to have an after sale support and repair. A startup is at a disadvantage as consumers will be uncertain about how long they’ll be around to honor a warranty. Yes, an established company can go under too, but the odds are better.

 

[mgb74]

You're obviously not using my scanners.

Obviously I’m not. But I’ve used flatbed scanners, Coolscans, and Pakons.

And it’s not just the time it takes, it’s the active time it takes.

 

[gswdh]

I can't understand that with a Nikon Coolscan 5000 ED under VueScan. Stable software, manageable results, excellent quality with a quick scanning process. Other scanners may work slower, but if the situation were actually that bad, I would have heard about it.

With VueScan, which probably supports most scanners, the problem of outdated scanning software has long been solved.

So I did say they have at least one of those issues, not necessarily all of them. I’d say your setup would fall into the legacy hardware category that can’t be bought new, or supported by a warranty or be officially supported. It’s also expensive for an old piece of equipment.

Edit: It seems that scanner is very slow, too.

 

[gswdh]

Varies from person to person. Right now, it normally only takes me a few minutes to scan the few 35mm shots on a roll that I want to scan. (After processing, I cut my film into strips to save in protective sheets & produce a contact sheet) Why spend more time on the computer deleting the vast majority of shots I don't want to scan -- and paying a lot for the privilege to do it?

Would you mind sharing your setup?

 

[gswdh]

It’s one thing to design a great product, it’s another to manufacture it, and yet another to have an after sale support and repair. A startup is at a disadvantage as consumers will be uncertain about how long they’ll be around to honor a warranty. Yes, an established company can go under too, but the odds are better.

I am very passionate about enabling people to only have to buy things once. This means the product will be fully serviceable with a proper repair manual with all electronic documentation freely available. This is all in addition to a warranty of course. I’d also make sure replacement parts are easily purchased at a reasonable price. If the company ever had to close I’d do my best to make all of the design files publicly available, too.

 

[Andreas Thaler]

So I did say they have at least one of those issues, not necessarily all of them. I’d say your setup would fall into the legacy hardware category that can’t be bought new, or supported by a warranty or be officially supported. It’s also expensive for an old piece of equipment.

It's simple: If you put something on the market that's better than what you can buy, and the price is right for your target audience, you'll be successful.

Edit: It seems that scanner is very slow, too.

That would be new to me, because the 5000 is still one of the fastest 35mm scanners.

 

[gswdh]

It's simple: If you put something on the market that's better than what you can buy, and the price is right for your target audience, you'll be successful.



That would be new to me, because the 5000 is still one of the fastest 35mm scanners.

How long does it take to scan? I read around 40 mins for a full roll which needs manually feeding unless you have the additional hardware?

 

[L Gebhardt]

USB 2 is plenty for a scanner, is cheaper and is much easier to implement.

The Canoscan FS4000 came with SCSI and USB2. The USB2 interface is much slower than the SCSI interface. I'd want USB3 and at least 4000ppi 12bit for a 35mm scanner.

 

[Andreas Thaler]

How long does it take to scan? I read around 40 mins for a full roll which needs manually feeding unless you have the additional hardware?

Here is one of the well-known longer reviews of the 5000 with data:

Nikon Super Coolscan 5000 ED Filmscanner Slide Scanner: Review, test report, experiences, tips and tricks for scanning use

www.filmscanner.info

Following this you'll get about 40 minutes if you scan at maximum resolution with ICE plus spool feed.

Now one can discuss what is long and short, it will depend on the specific requirement.

Anyone who needs all the shots of a 135 film in the highest resolution ICE filtered will think about whether it couldn't be done faster. But who should that be? Aside from index scanning for the archive, what's the point of scanning a film in its entirety?

At least I choose the images that I then want to edit in Photoshop, which takes up most of my time. The scanner can run at the same time, which doesn't bother me.

I don't understand the decisive advantage here.

All of the points you mentioned do not apply to me, not even individually.

But if my two 5000s give up, I have to look around for replacements. And if you have something that convinces me, I'm your customer. Until then, I'm very relaxed about the whole thing

 

[xkaes]

Would you mind sharing your setup?

I have different scanners attached to different computers for different formats and for different reasons. Too complicated to go into here. That's one of the reasons I said every one's situation is different. Some shutterbugs may only be interested in scanning newly developed rolls of film. I'm just as likely to be scanning film that is already processed and cut into strips -- and only looking for one particular shot on that strip. The time-consuming part is finding the image. Putting in the scanner and getting it into the computer takes only a minute.

I'm in the same boat as Andreas (above):

"I choose the images that I then want to edit in Photoshop, which takes up most of my time. The scanner can run at the same time, which doesn't bother me. I don't understand the decisive advantage here. All of the points you mentioned do not apply to me, not even individually."

 

[John Wiegerink]

I can't understand that with a Nikon Coolscan 5000 ED under VueScan. Stable software, manageable results, excellent quality with a quick scanning process. Other scanners may work slower, but if the situation were actually that bad, I would have heard about it.

With VueScan, which probably supports most scanners, the problem of outdated scanning software has long been solved.

Yes, Vuescan is the answer to many problems people are having trying to get an older scanner to work on a newer PC or Mac. I really liked my NikonScan software for running my LS8000 scanner, but there was no way I could use it on my MacBook Pro or even my newer Dell PC with Windows 11. Problem solved with Vuescan.