8mm DIY Telecine

Results and the future

Results

There is nothing much to be said about the current results. Some short clips are shown below. Note that these are completely 'as is' - no attempt has been made to sharpen, remove embedded dirt and scratches, reduce noise etc:

These are not especially good or bad results, just typical.

Conclusions

A few points that come out of this whole process:

1. The whole project has cost around £600-700 (though could have been less without the mistakes), so it is cheaper to do this than pay for professional conversion with in excess of 100 50' reels, or equivalent
2. A good light source is key
3. The Imaging Source camera seems to have worked out well although its dynamic range is not brilliant - but then neither is that of the transparency film in the first place
4. The original lens didn't - Frank Vine's suggestion of an enlarger lens is the most cost effective
5. CMOS seems to be no disadvantage over CCD - though see point about the slow 'wave' below
6. The slow 'wave' mentioned below makes me think that a Sankyo Dualux may not be the ideal platform, but it does work
7. Cleaning and lubricating the film tends to make the transport more consistent
8. A good fast PC is required, codecs are performance hungry, and if saving uncompressed or with a lossless codec, the disk bandwidth is very high

The future

The following are just some ideas that might be explored in future:

1. Post processing for the images: noise reduction, spot removal, shake removal, contrast, brightness or colour balance changes and sharpening are all possible, maybe by using AviSynth scripts such as that developed by Freddy van de Putte. I must admit to being in two minds about this as it seems that some of the essential quality of the 8mm film might be lost.

2. Whilst the quantization does not appear to cause a visible problem at low light levels, it would be better if it could be masked in some way. One possibility is to average the errors in the gamma / colour conversion over time, ie frame to frame.

   One way to do this is to keep the error (per colour pixel) following the lookup table conversion (the lookup table would be fractional rather than integer values) then add it into the conversion on the next frame, recalculating the error. This would effectively average out the error over time. Another way is to add noise equal to 1/2 a quantization level.

3. A better light source based on RGB LEDs. This would make better use the the camera spectrum, and also - see below - opens the way for separate RGB exposures.

   The brighter the better - allows more diffusion and shorter exposures.

4. Dynamic range in the camera is paramount if one wishes to handle corrections in post-processing, rather than stopping and starting capture.

   A camera with more usable ADC depth would be good, but expensive, providing one could get at the larger depth.

   Without a new camera, there might be three possibilities: (a) capture each frame, look at the maximum level, reset the exposure to bring it to the top of the ADC range, capture again and modify the result by the change in exposure (effectively more ADC bits, might give better black level response), (b) capture each frame three times (red green and blue in monochrome) and then combine, and (c) capture each frame several times and average each pixel over time to reduce noise (oversampling).

5. The 'wave'. In some samples a very slight slow 'wave' can be seen in the moving image. It has been determined that this can only be caused by the film moving slightly in the gate when it is supposed to be still. That might not matter with a CCD global shutter camera, but it does with a rolling shutter although it is quite likely with a global shutter that it might just be seens as 'jerk' rather than a 'wave'.

   There only seem to be two options here: (a) get another projector that doesn't do this, or (b) try to use a camera with a global shutter. (This wave effect would also complete kill the RGB approach above as huge amounts of colour fringing would be seen, and with oversampling, the image would seem blurred).

   The best answer is to stop it happening in the projector in the first place, especially as it also seems to be the case that there is a slight positioning difference frame to frame anyway.

   This has now been somewhat improved by the film cleaning regime described earlier, and also by taking care to keep the film path scrupulously clean, although it has not been completely eradicated.

6. Most of the above requires much better control of film transport. It would be better to have a controllable motor as the Sankyo speed control leaves a lot to be desired.

   This would allow slower frame rates if necessary (can then do more real time processing if desired, or use higher resolution, and also would allow the mechanism to be completely stopped while the single or multiple exposures are occurring). It might also stop the 'slow wave' mentioned above.

   A DC servo or stepper motor could be used. They both have pros and cons: steppers are cheaper, but require more complex driving electronics and are very easy to stall, especially when torque requirements vary. However, precise positioning control is automatic (just step to where you want to be). Servos don't suffer from stall problems, are easy to drive, but to stop the transport in a precise place might well need an encoder as well.

So that is it at the moment.

It is an interesting project, but only in its really early stages so far. A lot of lessons (some expensive) have been learned, and no doubt there are more to come. In particular the advice from Freddy van de Putte and Frank Vine was appreciated.

If you have any comments, questions or suggestions, feel free to email.

ADDENDUM

It is now October 2019, and I have done nothing more with this. Main reason - no more films to convert! However, just in case anyone is interested I have made available the Windows application I wrote and used as part of the conversion process - a download link for it is on the links page. Anyone may use it, however, bear in mind the following:

1. It is copyright to me and whilst you may use and redistribute it freely, you (a) may not modify it, and (b) may not use it for commercial gain or as part of a commercial product.
2. It was written a long time ago, it ran under Windows XP. I see no obvious reason why it would not run under Windows 10, but I have not tried it.
3. It uses a library that same with the Imaging Source camera, so I assume it will only work with their camera.
4. E&OE!