… well, Frank Vine’s color control system is quite powerful. It switches the LEDs on during a short time when the camera takes the frame. Otherwise, the LEDs are off. So it’s a pulsed system, and as Frank remarks, “not suitable for use with rolling shutter cameras.” The Raspi HQCam is a rolling shutter camera, among others.
With a help of a FET, a microprocessor can switch on-off any LED quite precise. That is what is used in Frank’s system for exposure control. He even describes the electronics used at his page.
The system monitors the temperature of the LEDs and applies a temperature correction. I have my doubts that something like this will have noticable effects in this setup, as the switching of the LEDs leads to a continous temparature change during operation. Maybe it’s operating on a time-averaged setting.
The special twist of this system is that the image from the camera is analyzed and the LEDs are driven in such a way that the maximum dynamical range of the camera is used in every color channel. In this way quantization effects and noise are reduced. Of course, the frames captured will have a weird color balance which needs to be corrected in the post.
Well, if you design a system from scratch, you have many options available. Here’s my take on a illumination system:
-
Will your film be running continously and you need to freeze the frame by flashing? If so, you will need a global shutter camera and a LED system which you can trigger fast and reliably. Simpler and cheaper is to use rolling shutter cameras and a continously operating illumination system. For this to work, you obviously need to stop the film during frame capturing.
-
Do you want to scan film material which is severly under-exposed or even off-color (maybe forgot to switch on the daylight-filter, for example). If so, you will need an adjustable light source with a sufficient power reserve. On the other hand, a lot of people have obtained resonable results with a simple white-light LED, compensating with the camera’s whitebalance and exposure settings. Of course, in doing so, you will not be able to use the full dynamic range of your camera.
-
Depending on the film material, your camera might not have the power to resolve the full dynamic range of the film. To optimize here, you have two options: one, as noted above, is to use your cameras whitebalance and exposure settings. The other is to employ an adjustable light source to maximize results, while keeping the camera settings to fixed values. The later option has the advantage that LED-light sources can be switched usually much faster than usual cameras. You can basically vary the current of the LED (not much used), or the duration the LED is switched “on” (PWM - defacto-standard nowadays, big advantage in heat dissipation, among other things). If using PWM, you need to make sure that the switching is done with a high enough frequency in order not to run into beat patterns or flicker problems.
-
Film material which is/was used for projection is tuned to the spectrum of bulbs, which is very different from the spectrum a LED light source produces. Actually, there’s even a difference spectrum-wise between a light source composed of red, green and blue LEDs and a “normal” white-light LED. In any case, there is also the spectral filter characteristic of your film, which will vary between different film stock, and the spectral sensitivity of your camera. All these spectral characteristics interact and make it kind of difficult to arrive at an optimizal light source. And an optimal LED combination for Kodachrome might not be perfect for Agfachrome or Fuji stock. It’s probably easier to tune all that in post production.
-
What you can optimize is the dynamic range your camera is operating on. Here’s my take on this: all film stock I have encountered (Super-8 home movies) have occationally some part of the frame which is so heavily overexposed that it is pure white, i.e., the pure light from the illumination source is shining through. Clearly, this gives you the upper range of intensities you not want to be clipped by your camera. Now, consumer-grade cameras (including the Raspi HQCam) are somewhat optimized for daylight operations. “Daylight” corresponds to certain white balance settings of the camera, and it is that setting one should use in manual whitebalance mode. If you do so, these clear patches in your film material probably won’t be really white (actually, you have to check this with an exposure setting low enough so that these light patches turn grey, in order to stay away from clipping effects). If you now tune your light source so that these patches appear grey again, you are going to have nearly equal amplitudes in the red, green and blue color channels of your camera. Not perfect, but close enough in my experience.
-
Film will have scratches. Some are minor and can be handled by appropriate light source design. Basically, you want to shine light through the film frame from every angle possible. There’s only one design which does that perfectly, and that is an integrating sphere. However, for this to work as intended, you need to place the film as close as possible to the port hole of the sphere. Any further distance reduces the incident angle of the illumination, reducing in turn the desired effect.
Well Matthew, I probably would not really know what to reverse-engineer…
The basic operation of the unit is quite well described at Frank’s webpage, even the circuit diagram he’s using is there. From what I understand the system is basically a hardware-based unit with dials for adjusting the red, green, blue and an additional “IR”-channel.
The temperature-compensation Frank is mentioning is quite interesting, but highly probable tied to the specific LEDs he’s using. Since he’s driving all LEDs of a channel in series, variations of the LEDs respondance to a certain current (which happens with real LEDs) would not be covered by such a temperature compensation. Also, at least in my experience, temperature drift of LEDs has barely a noticable effect in real captures (given, I have done only a few experiments about this and never looked back - maybe I should revisit that theme again…)
Well, continueing: presumably one additional microprocessor in Frank’s unit is acting as a kind of sequencer, receiving a trigger input itself and than triggering in turn the different LEDs and the camera, for example to realize separate red, green and blue captures of a given frame with a monochrome camera. He’s using PICs for that, I would certainly use some Arduino-clone nowadays.
The hardware unit features for every channel a simple constant current source, plus a simple FET-switch for the realizing the pulse timing. The LED current seems not to be adjustable - for a redesign, that would be certainly an option to consider adding. If so, you could use the unit in both ways, either in varying current mode, or in varying exposure time mode. Or even in a combination…
The software he’s describing seems to be actually split in two parts - one software is a capture software for firewire-cameras. Nice, but, in sense, firewire probably had it’s days. New interfaces (USB3, CameraLink, GigE and CSI) offer much higher transfer rates and in turn potentially higher scan resolutions.
The other piece of software he’s decribing is a application which is just a software recreation of the real dial hardware, enabling you to turn the knob on your screen without turning to the real hardware. The only thing I can imagine important here is that watching the real-time histogram (of his capture software) with selectable region-of-interest can be combined with the on-screen dials to optimize the light source for a given scan (for using the full dynamic range in all color channels, for example).