The Backlight


Maybe you are right, I do not know, I am not the expert in chemical film processes and a simple increase of the amplitude could be enough.
I am also here to learn.
I just assumed from black an white material, that can shift in intensity due to oxidation of the silver.
In the colour layers there is no silver anymore, but maybe other chemical processes which affect the spectrum.
And another thing is, that yes there are layers, so a 3 dimensional structure, but in most cases the film layers seems to form a kind of irregular bayer pattern itself, for directed light.
That is what I saw mostly under the microscope.
So if the illuminated particles are mostly beside each other and not above (technically they still are), the G and B or C and Y parts will only affect the corresponding particles and not the red ones. In six passes you would be able to reconstruct the right colours, if there is no crosstalk between the R, G, B, C, M, Y bands of the illumination.
Film was ever projected with a continuous spectrum and that gave the best results for colour reproduction in projection, so it may be a also the right way for digitisation.


Seems like it is an aliasing effect of the PWM and a rolling shutter sensor.
You already hat the idea to use a RC filter. That should smooth the current through the LEDs.
Direct PWM can damage LEDs because they are intolerant to fast current changes.
It may also help to increase the PWM switching frequency.
The artefact can be a form of EMI from the PWM module to the sensor too.
On our Korn film scanner we had an EMI problem between the illumination and the magnetic audio head, resulting in a high frequency noise in the recorded audio signal.


Another approach to avoid PWM would be to use 3 I2C DACs (MCP4725). They could be multiplexed together using a TCA9548A I2C mux connected to your uC. The max current output of the DAC is 25ma, so if additional current is needed, an op amp or transistor circuit could provide the boost.


I was able to make a test system using a 3W RGB LED and a White LED, with the help of a friend. It seems to be OK. (I will post a video soon with details),

Now I need to build the lightbox. With a single RGB LED the colors are not mixed perfectly, even with a diffuser. Your ideas are welcome.


For diffused light, you might want to build a diffusion dome, like this one:


Thanks for the link, Martin.


And found this thread on the forum.


Thanks for linking that thread. I re-read the info on Frank Vine’s webpage and it was immensely helpful. Take a look and let me know what you think:


  • R, G, and B LEDs on separate constant current circuits
  • When the shutter opens on the camera, all the circuits are on, then each one shuts after a set time according to the user’s color balance settings.

Is anyone interested in building one of these and testing it?


You might consider a WS2811 controlled RGB driver like this one. WS2811 is easy to control with 1 data line and gives you individual RGB/brightness control. No need for handling PWM or rolling your own constant current drivers. Rob offers 10, 20 & 100W boards and these can be chained if you want to have more than one for mixing. These are popular for use in flood lights for x-mas light shows.


@todd1814 thanks for the suggestion. I had mentioned a WS2812 LED option further up in the thread. I looked into the datasheet and see that the integrated driver is also constant current. To me that means that the PWM used to control the color balance is just a data signal. There shouldn’t be any flicker due to the constant current circuitry. That means the Adafruit Neopixel ring could still an option. But I’m an amateur and learning as I go so if anyone sees flaws in this logic please let me know.

To the group - can someone help me understand why something like Frank Vine’s system (see link in my post from earlier today) is superior to the Adafruit setup?

Thanks to all on this thread. Really helpful.