Want to Build a Scanner but where to Start

Good Day, I want to digitise about 20 reels of 8mm/S8mm in the shortest time, covid is the cause. I have read through lots on this forum and am as confused as I had started. Let me put it as I understand it in component level.

1: Projector - that take 8mm and S8mm film, that is to be reengineered.
2: Stepper motor - to take the place of the projectors motor, normally 17 nema stepper motor.
3: Controller - Arduino uno3 or Raspberry ? (which system is entry level)
4: Camera - a digital camera or Raspberry Pi ( which camera and which Raspberry is entry level)
5: Lens - which is the entry level that will work with the raspberry Pi ? camera
6: Software is either Python for Raspberry or Arduino for the digital camera to do the capturing.
7: Software to do all the processing ( not sure what to use)
8: Good fast PC to build the digital images back to a movie.

There are the 8 steps I would like to follow and have your input to make the right choice of equipment. I have been advised on the raspberry Pi HQ camera but it is just out of my budget for now.
What I have bought so far is a Eumig projector and a Minolta maxxum lens AF50/2.8 macro, not sure if it will be good or not. Now for the rest of the components, I do have the PC .
Maybe we can start a new category called - "
Starting Point".

There exist dual format projectors with a swapable film gate you can take out. This helps in enlarging the film gate in case you want to scan the area usually not seen in a normal projection. This is however a challenging operation - you want to be careful not to introduce any scratches in your film gate while enlarging it. You might stick with the projection window as it is, much easier.

There are lots of examples here in the forum about projector-based Super-8 scanners. Also, youtube is certainly your friend. Try “telecine DIY” for example. And yes, a lot of people do use NEMA-17 motors. Be aware that this is only the mounting specs - it has nothing to do with the power a motor can deliver. Be sure to source a sufficiently strong motor. Also, do not drive this motor with a stepper driver connected to 12 V. Use at least 24 V.

Be sure to throughly clean and re-lubricate at the appropriate place the projector you obtain. These things usually have been sitting tens of years unused in some storage. This eases the amount of work your stepper has to perform.

Again, browse the forum. Both machines have been used for driving the stepper motor and triggering a frame capture. Normally, the Arduino or a similar microprocessor is better suited for time-critical task (like driving a stepper), as the Raspberry Pi’s Linux-based operating system is no real-time operating system. However, there is software available on the internets for both machines. From a system design point of view it might be a good approach to drive the projector/stepper via an Arduino and to take the scans with a Raspberry Pi 4 and a HQ camera connected to that Raspberry Pi. Again, search the forum - I know of at least three people (@dgalland, @PM490 and @Manuel_Angel) who published their approach to a film scanner in detail here on the forum and on github.

If you want to go high res, choose a digital camera. But be aware the storage requirements and processing time increase exponentially with the pixel count. Note that even the best Super-8 footage I have seen never exceeds HD resolution. Because you might want to opt for doing overscan (include part of the sprocket hole for example for frame stabilization) and you might want to have some head room for post processing, you should opt at a scan resolution for Super-8 of at least 2k.

Note that picture taking with a normal digital camera is a quite slow process.

Which brings the Raspberry Pi 4 equipped with a HQ camera into play. A lot of people here in the forum are using that combination, as it is able to capture 2k-like scans with quite some speed. Again, there are ready-made software solutions available on the internet (see links above).

depends on your overall system design. If you do not want to spend a sizeable amount of time for your film scanner, simply use existing software. That will most probably be written in C++ for the Arduino and in Python for the Raspberry Pi.

A lot of people obtain quite good results with the Schneider Componon-S 50 mm. It is a lens actually calculated for the 35mm format. So if you are using that lens with Super-8 film stock and a Raspberry Pi HQ camera, you will be using only the sharpest part in the center of the lens. Best performance is achieved with f-stops between 4 and 5.6.

There is no question that the best software for post processing is daVinci Resolve. There’s a free edition which is very powerful; daVinci Resolve is especially reowned for excelent color grading - which you will have to do a lot of when scanning Super-8 material. Only if you already familiar with another video editor stick with what you already have mastered - daVinci Resolve comes with a steep learning curve. But again, the internet and youtube are your friends (lots of tutorials there).

Your video editing program should be able to read single frames and output these into a movie. daVinci Resolve certainly is able to do that. You might want to get the fastest PC you can affford, equipped with the best graphics card you can get (as daVinci Resolve and most other video editors nowadays use the graphics card for computing). Your other bottleneck will be the hard disks. Be sure to use SSDs if possible, large ones. Things can speed up if you have the input frames on one fast disk and output the movie onto another one. Also, put the chache of the video editor on yet another drive.

well, the Rasperry Pi-combo is about the cheapest option you can use for digitizing Super-8 material. I think the Raspberry Pi 4/HQ camera option should set you back at most 150€, all in all. A used projector (if it is in good shape, and for the sake of your movies, you probably want it to be in good shape) will run between 50€ to 400€ on ebay. A new PC with which it is fun to postprocess/edit videos will take you back at least 1000€. But daVinci Resolve is free :wink:

2 Likes

Thank you for the time you have put into answering my question path, I hope it can also put other newbies on the right path of susses . I feel more at ease with my selection of equipment and now need to order in and do maintenance on the projector. I am really grateful for the feed back, Rolf. I must also acknowledge Manuel_Angle for his input as well.

Hi Lee, welcome. I would like to share a few tips, and perhaps an additional perspective.
In broad strokes, the machines divide into two main categories: continuous-motion (real time), and stop-motion. Based on your level of expertise, pick one… if you don’t know which, stop-motion (slower than real time) is a place to start.
Some common building blocks:

  • Light source
  • Film transport
  • Film gate
  • Optics
  • Sensor (camera)
  • Storage/processing

Another great friend: Youtube. There are a lot of great videos showing machines both build from scratch or Frankenstein-ing a projector.

8mm is hard because of the sizes are less tolerant. Perfection is enemy of the good. Get something working, and then improve one block at a time as you go.
Improve/iterate with these blocks as necessary. For example, I am working on my fourth light source, the first one was good, just making it better.

Where do you start? Where you feel more comfortable is a good point, and small steps/successes will get you going. Getting your optics and camera configured is a good place to start.

Good luck and glad to share any information I have.

1 Like

Thanks for the support and still in the research phase, but time is not on my side.

Hi,
I have built this T-Scann 8 - project page and im just getting it to work nicely. I think it has all of what you ask. This guy, Torulf, that designed it is really helpful, the only thing i miss is a forum like this to connect with other builders.
It is a really fun project with affordable parts. I had no knowledge about 3d printing and electronics and very litte knowledge in programming beforehand, and now it is standing here. I am very happy with it…

3 Likes

You are asking the right questions so here are my ideas on the decision tree of successive choices.

1- Projector
Modified projector or independent mounting like Kinograph ?
The modified projector allows to get to the result faster.
A Kinograph type mounting is more professional and has some advantages but requires a lot of 3D printing

2- Stepper Motor
Some people try to keep the projector motor if it can run slowly enough. Very bad idea in my opinion

3- Controller 4- Camera 6- Software
The three choices are linked with two completely different solutions

Solution A : a PC with a Camera or a DSLR USB Software for example Digicamcontrol and to control the motor an Arduino see the post of @PM490
Solution B: a Raspberry PI with the HQ camera and a client-server software with a PC see my project or the one from @Manuel_Angel

The choice is not easy
Solution B, the one I implemented, guarantees to get a satisfactory result with a reasonable cost.
Solution A allows in theory to use DSLR or high end camera. Faster ? Better quality ? That’s up for discussion

There are endless discussions about the right resolution to scan 8mm or Super8
The Raspberry HQ camera can scan in 4056x3040 2028x1520 or 1920x1800
But the maximum resolution is impracticable with a much too slow capture, here the PI hardware shows its limits. The other resolutions give a very good result but my tests show that it could be slightly better in full-res
For the control of the HQ camera the best solution is to use Python.
There are some evolutions in progress for the control of the cameras on the PI but they are still a bit experimental

5- Lens
CCTV type lenses for the HQ camera are not excellent with blur in the angles.
I think it is the same for the 16mm proposed with the HQ
The Schneider Componon is a very good lens but your Minolta should also do the trick.
Attention also to the assembly of all this, well stable and adjustment in X-Y-Z

7-8 Post processing
Da Vinci Resolve in free version has already very good possibilities
I’m a bit traditional and a supporter of Avisynth, difficult to master but very good results

2 Likes

Thanks for the detail feedback, My first pass the camera will have to be Raspberry V2. Once I have the system up and running, then might get the HQ camera.

No, especially not V2!
The V2 camera with a non-stock lens has a strong vignetting which makes it completely unsuitable for this project.
On the other hand, the V1 can be used and gives a result almost as good as the HQ (see various posts in my project group).You can find a V1 with a C/CS mount

I nearly made a huge mistake, thanks. I will defiantly go for the Raspberry Pi v1 camera with the C/CS mount. I had a look at the Raspberry Pi on eBay and Banggood, there are quite a few so now I should research which one will be suited for the v1 camera, would the ram play a big part seen that it has to take lots of pictures.?

Hi @Lee,

I wanted to tell you that the V1 camera, if used with lenses other than the original, (as it seems to be your case) presents the problem known as “lens shading”. It consists of the appearance of a vignetting of different tones in the image captured by the camera.

This defect can be corrected via software, although it is always a partial solution, it is impossible to eliminate completely.

When I started in this world of Super8 digitization, I also used the V1 camera (the HQ camera did not exist), the truth with very satisfactory results. The lens shading problem was practically unnoticeable. Later I made a change in the lighting in order to reduce the exposure times and at this point the problem became very evident. All attempts to remove it were unsuccessful and the results were frustrating. For this reason I decided to switch to the HQ camera, which does not have this problem, along with a quality lens.

Regarding the Raspberry Pi, I use Raspberry Pi 3, although I have also tried Raspberry Pi 2 and in both cases the operation is correct.

To use the HQ camera, you need to reserve 256 MB of ram for the Pi GPU. 128 MB is sufficient for camera V1.

You don’t need to have the machine built to test the Rasperry Pi based server and camera. You can test everything logically except the motor and motion of the film. For this you can use the software Yart of @dgalland or my own software DSuper8.

Regards.

Hi @Manuel_Angel - thanks for taking time out for me. I started to look for the Raspberry v1 camera and wasn’t certain that I had found the right one, so I have convinced the bank manager that HQ is the way to go. Will be placing the order soon for the Raspberry HQ and the controller Raspberry Pi 3. Once the order has been placed then I sort out the stepper.

Had a look now on a site where the HQ camera is advertised, they have the Raspberry Pi 3 (A+) board is that the same as Raspberry Pi 3. These updates are a bit confusing.

No the 3A+ is not suitable, it is a reduced version of the 3B+ but without ethernet. Not very good for a client-server project !
For the project the 3B+ is quite sufficient. I don’t know the price difference with the 4B
It is not the CPU or RAM that are the blocking factors for performance but the camera and possibly the network speed. The Raspberry PI4 would be slightly better in networking.

Will have a look at both the 3B+ and the 4B and make the call. Shipping to South Africa take about 3 months, yes not a good postal service in SA.

Raspberry products are often not really more expensive at local official resellers than in chinese stores. There is no PI store in your country?

Note that the RP4 improves certain aspects of older RPs which might be of interest in a film scanning application. The RP4 has two USB-3 connectors, the RP3 does not. More important: the RP4 has a real Gbit ethernet link, the RP3 maxes out at 300Mbit/s (there is a bottleneck in older RPs including the RP3 - the network is piped internally through a USB-hub). If you are following the usual approach, using a HQ camera and streaming the images to a more powerful PC via LAN, you might encounter that actually your network performance might be the bottleneck. One drawback of the RP4 is that you definitely will need an active cooling solution for the board - it gets much hotter than the RP3 usually does. There are RP case solutions with heat sinks and one or two cooling fans - that is what you most probably want to use. The RP3 will, as @dgalland rightfully remarked, do the job. But if you can afford it, buy a RP4 (54€) instead of a RP3 (38€).

With respect to the v1/v2 cameras - they have all a lens-shading problem when the original lens is replaced with another lens. Same issue can be noticed by third-party products, which pair for example the v1-sensor chip with a C/SC-mount. The issue is caused by the mismatch of the microlens array in front of the sensor with the actual lens mounted - the microlens array works good only for the original lens (or lenses with a comparable focal length).

As soon as you swap the lens, you are going to experience two things: color shifts between the center of the image and the borders, plus desaturation (less intense colors) towards the borders.

And there is not way to counteract these issues in a way that it will not become visible sooner or later in your scans. It might not be noticeable in most of your scans, but as soon as light orange or brown tones are covering a large area of your frame, it will be noticeable to the trained eye (maybe not your viewers). I have used v1 and v2 cameras in film scanning projects and since the introduction of the HQ camera I have never looked back (v1 and v2 cameras are based on a sensor intended for mobiles, the HQ camera chip is aimed at consumer video recording).

4 Likes

Great thread and thanks to all the people who have helped answer the questions from the OP. My only addition is to check out the list of existing machines we know of. Some of them have documentation that may be helpful. You may even be able to contact one of the makers to assist you in building yours.

@Henke_Holmqvist I’ll add yours to that list. Looks really great!

1 Like

Had to attend to home issues, but back again, thanks for the guidance, more or less got the camera and controller sorted. now to look at the stepper and the stepper controller/driver, here I will use the Arduino uno3 to control.

I am going to use the raspberry HQ camera, you kind gentlemen have persuaded me in that direction. The company that is selling the camera also has the lens, here are the specs; CGL camera lens C-Si-2 with 10 megapixels res. note, the lens is just as expensive as the camera module.