Wolverine-Hawkeye Telecine

[Bruce Davis]

Hi Stan thanks for your ongoing work, now I am confused, in VideoFred script there is an option to change the final output frame size.
For example:- W=1920 H=1080
This does actually change the output frame size, the script comes with the default set to 720x576, but this can be changed.
Internally I do not know whether the script works on the original input frame size or whether it scales to 720x576 first before any processing.
One would assume that this function allows for up scaling and down scaling.
Have included an example of output frame sizes below (1256x924 was the original input size).


Regards - Bruce

[Stan Jelavic]

OK let me try to explain Bruce. I thought that you were starting with the default values of for H and W and then resizing after somewhere in the script but I did not see it.
So, what you do is change the default values. The script then resizes the original video to whatever values you put in and then applies all filtering on this new size.
I did not like that because I was concerned that resize would introduce artifacts. So I removed the resize and try to process the original size but then I got these nasty artifacts.
If I use the resize function then everything is happy. But now the question is how much distortion is introduced by resize. I will create a script that runs the resize function only and then put the output
side by side with the original. If this works then we are good to go. I will still have a version of the script that uses my color shift for blue and red. This time it will not have that grainy look because I will not be skipping any of the steps.

OK, got the resize comparison. This is 1280x1024 original vs original cropped by 64 and then resized back to 1280x1024. I did not see any noticeable degradation.
Also ran the comparison vs the original and it looks quite good. Thanks Bruce for your help. I think this is pretty good the way it is.
Here is the resize compare:
And here is option4 vs the original:



BTW - I set the gamma to 1.0 in the script since the camera already does the gamma. With default set to 1.3 the image is a bit low on contrast in my opinion.

[Stan Jelavic]

V8 speed control works. Hopefully, will do the frame sync and takeup tomorrow.
https://photos.app.goo.gl/obKZk2QpFumA5gG96

[Bruce Davis]

Stan, the resize comparison, as you stated, the frame image quality does look the same.
Original vs Option4, there seem to be sharpening artifacts (if that is the correct term) introduced into the Option4 frame.
I agree that Gamma 1.3 does reduce contrast but I only use it because it looks better on my LCD TV, I could adjust the TV but then would have to adjust back for other content.
Thank you Stan things are beginning to look very promising.

[Kamel Ikhlef]

Ah!!! I was looking forward to the v8 tests ......To continue to my front control panel.
Thanks for sharing, Stan. If you need 3D parts, tell me if I can help (new capstan motor support, pulleys, lens holder,...)

BTW: Please,would it be possible to know the resistors values provided in the v8 schematic ?.
Thanks in advance. I am very impatient to test the new speeds and the new camera synchro (I think, a new MSP code ?).

[Stan Jelavic]

Bruce, option4 output is definitely sharper than the original. Sharpening takes place in a few places along the processing pipeline.
MVflowfps() adds additional frames (without changing the frame rate) to make the motion more smooth. It does that by estimating the pixel travel between two or more adjacent frames
and estimating where they should be for the injected frame. You can see the added frames if you step through the output video.
Kamel, I will provide the resistor values from the assembly steps. I.e. power components image. Speed control image etc.
Here is an example:
Speed control:

[Kamel Ikhlef]

Sorry Stan, I'm always too impatient! ... And, as I like to understand all, I ask myself a lot of questions!
there I am trying to understand the new speed connector. I made a diagram but I don't know how to connect pin 4 (RW takeup).
I would like to put a separate button. Thank you for all the time you take to respond to us.

[Stan Jelavic]

Kamel,
pin4 grounded - takeup off
pin4 open - takeup on


Here is the new sketch that will help you. You need a 3 position switch for speed.

[Stan Jelavic]

The frame sync new cct works OK based on a quick test. I left the transistors dangling like that so that it is easy to attach a scope probe. Needed an extra cap to prevent the circuit from oscillation. Without it it
oscillates and generates high voltage due to resonance. Took me a while to figure out what was going on.

[David Brown]

I think my V7 board moves slower than the V8 at .5fps. Didn't time it, just counted 1001..1002... That only matters if the file save takes longer than 2 seconds per frame. If not, maybe it's time for a new computer!

My Hawkeye sensor is on backorder at OEMcameras and I will not see it before mid February at the earliest.

I found a webcam board with the exact same sensor, and can take snaps with AMCAP. I also bought 2 16mm lenses. One is made for 1/2.5" and the other for 1/2". The point was to compare the projected image that would be cast on the sensor. The 1/2" is clearly larger.

I placed the lenses together and used a flashlight to compare the circle. Clearly larger, didn't measure. The 1/2.5" lens does not fill the frame with the 8mm film's width at a reasonable distance from the gate. The 1/2" does, and takes a better image too.

So I'm ready to order and only need to choose between Scorpion (no cut filter) and Alibaba. About the same price even with a $10 filter from Amazon. Just hope it's the right spectrum and size!

The picture below is not an exemplary image on fim. It's a piece of scrap I use to test with, but I focused on the film's grain. It can be seen to soften in the periphery. The lens used is not a macro and is F2.0.

Even more, I don't think the sensor and the film are parallel. The M12 mount on the webcam has a different bolt spacing so I could not use Kamel's longer mounts. I used duct tape to mount the board and to join another M12 mount to the end, giving more spacing between the sensor and the lens.

Picture looks OK considering all these flaws!

[Stan Jelavic]

V7 board does run slower. I had an issue with the MSP430 firmware. If I set it to 2 seconds per frame it would run out of steam because I left the debug mode on. When I subsequently went to the release version, I could run it at 2 seconds per frame no problem. One FPS also runs but has issues if capstan code is in. Still looking at this. For 2 FPS I run the FW in 1 FPS mode and increase the motor speed by changing micro-stepping.
For the lens selection it is a hit and miss to a certain extent. Scorpion vision has a good chance. I will chat with them when chat becomes alive and see what they say.

David, you say that the 1/2.5" did not work for you. Is the circle too small? That actually is a very good test. We are going closer to the film than what the lens spec is for minimum distance. That could make the circle smaller. So it does make sense to go with a 1/2" or larger if we want the whole sensor to be covered.

[Kamel Ikhlef]

David, if you need specific 3D parts, I can design them. I will be happy if I can help. Stan, thanks for your work on the V8.

it's very motivating and inspiring !

[Stan Jelavic]

Thank you Kamel. You are always there. Really a ppreciate your help.
I just finished the frame sync tests and added the capstan code. One issue. The capstan code is too long for fast mode. So I decided to enable capstan in slow mode only.
But then we have to think how to wire the switch.

[Kamel Ikhlef]

Stan,
In my case, I will use most of the time the slow mode because I want to privilege the quality !...
And compress the captured pictures too much to gain speed is to return to the original Wolverine artefacts.
Not the goal at all. With all the work that this project required. We are very close to get a high quality telecine !

Long live the new v8 Hawkeye !
I will patiently wait for the end of your tests to finish my cover front panel. it's wiser ...

BTW: Your new connectors speed concept is very intelligent.
A "turbo" button is very well thought out (very clever). As it may suit everyone.

[David Brown]

I know this may be too late in the game for hardware changes. Does the alarm still sound if the capstan is activated?

Can the capstan advance the film by only one frame, or is that too much precision to ask of it? If there are multiple torn sprocket holes, multiple frames would not be captured. Am I right?

Stan: Yes the two lenses, when positioned at the same distance, project different sized circles. Interesting they are both 16mm, yet are different lengths. I'll post a picture later.

Kamel: Thank you, I don't need anything right away. The webcam won't be used after this, I just needed to get an estimate. Later I might need an extended s-mount if the two you already designed are not in range.
I don't know how badly torn your film is. I know I have some, I remember seeing projector problems years ago. It might be interesting to create a single frame advance (hand operated ) behind the gate for that problem. That might be a lot of work for a small amount of problems. It would need to be engaged and activated faster than .5 frames a sec!

Just a thought.