Has anyone tried building a digital RAW camera around Normal 8mm / D-mount lenses?

[jzagaja]

Hi everyone,

I’m wondering if anyone here has tried to build a small digital RAW camera specifically around Normal 8mm / Regular 8mm cine lenses, especially D-mount zooms like the SOM Berthiot Pan-Cinor 8–40mm.

The idea is not to make another action camera or phone rig, but something closer to a tiny indie digital 8mm camera:

• small sensor matched to the Normal 8mm image circle
• D-mount flange distance
• RAW10 / RAW12 capture
• global shutter if possible
• simple HDMI preview
• recording to SSD
• film-like color pipeline: linear matrix, log curve, gentle highlight roll-off
• maybe later: piezo micro-shift or a controlled “organic” sensor/optical movement to reduce the fixed digital feel

I have been looking at sensors/modules such as Sony IMX900 and AR0234. The IMX900 seems more realistic as a ready Raspberry Pi 5 camera kit, but the AR0234 is very attractive because it is cheap, global shutter, 1920×1200, RAW10, and available in color and monochrome versions. The problem is the usual one: MIPI carrier board, Raspberry Pi compatibility, V4L2/libcamera driver, power rails, initialization registers, etc.

The goal would be around 1080p or 2K rather than fake “4K”, because these old Normal 8mm lenses have their own character and don’t need a huge modern sensor. I’m more interested in a compact RAW camera with a real vintage 8mm optical look than in maximum resolution.

Has anyone here tried something similar?

I would be especially interested in:

• D-mount digital camera experiments
• Normal 8mm lens coverage on modern sensors
• Raspberry Pi / Jetson / RK3588 MIPI RAW capture
• AR0234 or IMX900 RAW workflows
• custom small camera bodies
• monochrome sensor + RGB filter experiments
• piezo sensor shift / film gate weave simulation

I’m currently trying to understand whether it makes more sense to use a ready IMX900 Raspberry Pi kit, or to go deeper with a cheaper AR0234 OEM module and somehow solve the carrier board / driver problem.

Any experience, links, warnings, or crazy ideas welcome.

[eatstoomuchjam]

I think the closest thing to what you're looking for is the cinepi project.  That or there was at least some project recently where somebody removed the glass from a smartphone and juryrigged a lens mount to it.  That'd be another place to start.

 

[Clark Nikolai]

1 hour ago, jzagaja said:

Hi everyone,

I’m wondering if anyone here has tried to build a small digital RAW camera specifically around Normal 8mm / Regular 8mm cine lenses, especially D-mount zooms like the SOM Berthiot Pan-Cinor 8–40mm.

.....

Any experience, links, warnings, or crazy ideas welcome.

I'm currently building a variation on the Faux Bolex project. https://github.com/lafauxbolex

It's meant to recreate a Digital Bolex so it uses a machine vision camera with the same sensor (Basler Aviator avA2300-25gc Gig-E) which is 2K and has a Super16 sensor size but because I want a smaller camera, I'm using a different industrial camera whose sensor is a tiny bit more than 8mm. The Basler Ace acA1300-30gc. The DNG files it makes are 1280x960 pixels. So between 720p and 1080. I'm keeping the 4:3 aspect ratio. I'm thinking of something that you hold like a pistol as most 8mmm cameras were designed like.

The FB project uses a regular size Raspberry Pi but  to be smaller I'm using a Raspberry Pi Compute Module coupled with a Waveshare Nano Baseboard B for the IO. 

If you did something similar the machine vision camera likely would have a C-mount. You could easily put in a C to D adaptor and use D-mount lenses. https://fotodioxpro.com/products/d-c-p?_pos=1&_fid=d8b00a026&_ss=c

This project saves to CinemaDNG files. A single raw file for each frame.  There isn't audio implemented (yet). You can use a separate audio recorder if you want sound then sync it up later.

There are some camera tests on Youtube.

https://www.youtube.com/@LaFauxBolexDigitalCinemaCamera

Having said that I haven't got it up and running yet. I don't have a background in programming so many things are new to me and I have other projects on the go and other cameras I can shoot with so this isn't urgent and I've set it aside for now. It involves both Python and C++. I don't have the experience to debug things properly. (I might just hire somebody to do that.) Being a passion project by an individual who is doing this on the side from working (in the film industry) and raising a family, he doesn't have time for much support. (Last I heard he had his prototype working and is out making a movie with it.)

Check it out, but it is very DIY. This might not be the right project for you and one of the other Raspberry Pi camera projects like CinePi might be easier to set up and get running. The FB connects with Ethernet. Some of the dedicated cameras for the Pi connect differently and are easier to get up and running. The Raspberry Pi Global Shutter camera has a global shutter and connects to the CSI connector on the Pi. https://www.raspberrypi.com/products/raspberry-pi-global-shutter-camera/

So, warnings:

A critical finicky thing is the lens to sensor distance. One micrometre can make a big difference in your focus. I recommend not messing with that. Get a camera that already has a C-mount and add a D-mount adapter. Don't try to create your own lens mount on to a bare sensor. It can be done but you'll spend weeks trying to get it right and it won't be fun.

The other thing would be the sensor's physical size. It should be the size of the old 8mm film gate (which is 4.5mm x 3.3mm, 5.58mm diagonally) or a bit larger.

Post on here how you're going with this. It'll be interesting to many people.

[jzagaja]

Thanks Clark, this is extremely useful.

La Faux Bolex is exactly the kind of reference I was hoping to find. I had not looked seriously enough at the machine-vision / GigE route before. The idea of using an existing industrial camera with a proper C-mount is very appealing, especially if it avoids the nightmare of designing the sensor mount, power rails and MIPI/libcamera side from scratch.

Your warning about lens-to-sensor distance is probably the most important part. I was tempted by the “bare phone sensor” experiment because damaged phones are now very cheap here in Poland — for example a Motorola Edge 50 Pro with a broken corner can be found for around 100$ — and the phone already gives you screen, battery, storage and potentially RAW video recording. But I agree that mechanically it could become a nightmare: flange distance, dust, IR filter, sensor cover glass, alignment and destroying the camera module.

So maybe the phone route is only worth trying as a cheap destructive proof-of-concept to see whether the Pan-Cinor / Normal 8mm look is interesting enough digitally, not as the final camera architecture.

The C-mount machine-vision route sounds more “correct” for a real build:

industrial camera with known sensor position

factory C-mount alignment

DNG / raw frame recording

Raspberry Pi / Compute Module control

then adapt D-mount lenses mechanically

My only concern is the D-mount to C-mount adapter. Since D-mount has a shorter flange distance than C-mount, the adapter needs to recess the D-mount lens enough to reach infinity. Have you tested whether the Fotodiox D-to-C adapter actually gives infinity focus with D-mount lenses, or is it more of a macro-only solution?

For my project the ideal sensor size would be close to the Normal 8mm gate, around 4.5 × 3.3 mm or slightly larger. That is why I was looking at sensors like AR0234 or IMX900. But your point makes sense: a slightly larger industrial C-mount camera that already works may be much more practical than a theoretically perfect bare MIPI module with no adapter board and no driver.

The pistol-grip form factor is also very close to what I have in mind — more like a digital Regular 8mm camera than a normal modern video camera.

Thanks again. I’ll study La Faux Bolex more carefully.

[jzagaja]

In the longer term, I am also watching small LOFIC sensors, because for this kind of digital Normal 8mm camera the real limitation may be highlight latitude rather than resolution. A compact 1/2.8" or 1/3" RAW sensor with better single-exposure highlight handling could be very interesting. Another related experiment would be using spectral filters, possibly with a monochrome sensor, to approximate the response of color negative film layers — so the “film look” would start partly at the capture stage, not only later as a LUT in post.