film -> video device

[Fizick]

FredThompson,
If you understand my doc, please give me some correction of it. It was hardly for me write some therms in English.

Final comment: I am not Fred :)

[FredThompson]

Ah, I think we have a language challenge.

My original thought for the filter chain was to have the script keep some frames and discard some frames, repeating to match the pattern of the film capture. The first idea was to use a camcorder and slow film movement, pausing for each frame of film. I still think this is the best option but don't know exactly how to do it. Servo motor would be necessary.

[Fizick]

You may try various options.
I also firstly try to add electrical switch (gercon) to projector for per-frame capture.
But my plugin is not for such capture.

[dokworm]

Using an optosensor on the driveshaft of the shutter is just so easy and is only $3 worth of parts, I can post a little schematic and how to here.

It gets around complex software - you are only capturing useful frames, no need for complex algorithms then, most shutters are 3 blade, so capturing 3 frames for each frame of film is easy, or you could just capture as many frames as your webcam is capable of in realtime - just slow the motor down if you want more frames.

The beauty of a switch to tell the PC when *not* to capture is that it negates the need for stepper motors, or complex software, you only capture frames while the film is stationary in the gate.

You could also cover one of the shutter holes with a Neutral Density filter material, and do a HDR capture (i.e. effectively bracket the exposure) and average the results to capture more of the dynamic range of film than you could normally.

[FredThompson]

Yes, please throw together a schematic and how-to.

I don't understand the comment about dynamic range and ND filters. ND lowers the amount of light without altering the color, right? How would that improve dynamic range?

[trevlac]

Yes, please throw together a schematic and how-to.

I don't understand the comment about dynamic range and ND filters. ND lowers the amount of light without altering the color, right? How would that improve dynamic range?


There are more steps in the range from black to white on film.
There are 256 in 8bit video.

He suggests 2 captures grabbing the low and then high range. (I think)
Over expose the full range to get more details in the low.
ND filters would then under expose and grab more of the high.

I'm not sure if you would get better results with this, by using a simple averge especially if you just convert to 8bit anyway.

[dokworm]

You get massively better results as you can capture the full range and then compress back down to 8bit. So rather than either missing out on the shadow detail (crushing the blacks) or losing detail in the whites (blowing out the whites) you capture the detail at both ends, then with some relatively simple processing, you can keep both in the one 8bit image.
Or better still, use openEXR and keep it a HDR image.
There are a lot of photography sites dedicated to HDR imaging if you want to do further reading.
http://www.cybergrain.com/tech/hdr/
http://www.hdrsoft.com/resources/dri.html


I'll put together a simple opto circuit for you guys then as well.

[trevlac]

@dokworm

I said ... "I'm not sure ..."

The reason I'm not sure is that there are at least 3 things involved ...

1) The contrast of the scene: In this case the projection/reflection of Fred's device
2) The bit depth of the sampling CCDs (and the processing to put it into 8bit)
3) The limits of an 8bit (and eventually 4:2:0 DCT compression) target gamut

If 1 is low, 2 is good, and 3 is low ... I think "massively better" will be hard to achieve ... but I still don't know and I'd be happy to see how it works out for Fred. :)

-----------------------
On a similar issue ... oversampling the image resolution ...

I have an HDV camera and I do see some improvements in the resolution even when processed down to DV size on a STD tv. So, oversampling the analog is generally a good thing.

[dokworm]

It is more about remapping the film gamut back down to 8 bit in a sensible way, take a look at the openEXR home page to get a look at what I mean, basically, unless there is a very small range on the film, it will need remapping otherwise one end or the other (or both) will lose a lot of detail.
It doesn't really matter about 3, even at 4:2:0 compressed the idea is to put the detail into that scene, rather than losing it *before* you get to the compression phase.

A great example is here
http://www.cybergrain.com/tech/hdr/images/tone_1.jpg

The image is blown out at the stained glass windoes and the top round window.
By doing a multiple pass exposure and remapping, you can get the entire tonal range to be remapped to something that can be displayed in the 8 bit space like this
http://www.cybergrain.com/tech/hdr/images/tone_2.jpg

The problem with film getting onto DVD is the wide latitude that film has, it is very hard with a standard CCD to capture the full range in one pass, you tend to get effects like in the first image, with some details either blow out or two dark. By doing two passes (at least) one to capture the dark details, and the other to capture the light details, you can merge the two, and end up getting both represented.

I think you would agree that you would be better off starting with the second image and then dumping it down to 8 bit and then 4:2:0 etc. than starting with the first image.

From my experience it would be hard to capture the full range of film well without doing something like this.

[rfmmars]

It is more about remapping the film gamut back down to 8 bit in a sensible way, take a look at the openEXR home page to get a look at what I mean, basically, unless there is a very small range on the film, it will need remapping otherwise one end or the other (or both) will lose a lot of detail.
It doesn't really matter about 3, even at 4:2:0 compressed the idea is to put the detail into that scene, rather than losing it *before* you get to the compression phase.

A great example is here
http://www.cybergrain.com/tech/hdr/images/tone_1.jpg

The image is blown out at the stained glass windoes and the top round window.
By doing a multiple pass exposure and remapping, you can get the entire tonal range to be remapped to something that can be displayed in the 8 bit space like this
http://www.cybergrain.com/tech/hdr/images/tone_2.jpg

The problem with film getting onto DVD is the wide latitude that film has, it is very hard with a standard CCD to capture the full range in one pass, you tend to get effects like in the first image, with some details either blow out or two dark. By doing two passes (at least) one to capture the dark details, and the other to capture the light details, you can merge the two, and end up getting both represented.

I think you would agree that you would be better off starting with the second image and then dumping it down to 8 bit and then 4:2:0 etc. than starting with the first image.

From my experience it would be hard to capture the full range of film well without doing something like this.


It should be made clear at this point that the dual exposure process is only valid for use with a work printer, not a movie projector which will never run the speed twice unless it is a crytal controlled drive.

The new Sony EX-Eiew PAL block cameras come very close to the film's dynamic range. I use the "SuperHad" Block series cameras which do very well.

Very little low light detail is lost.

richard
photorecall.net

[dokworm]

No it doesn't have to be a workprinter per se, just any projector slowed down to say less than 6fps (depending on your capture device speed of course), and fitted with a trigger mechanism (i.e. A projector that is modified to work on the same principal as the workprinter, without the expense).
Then you can capture two or three images (at least) in a single 'pass'.

Although the new CCDs come close to the film range, getting it to work in practice isn't so easy.

Regardless of the capture device, your illumination source is often the problem, and it is difficult to get the dark areas lit enough, without flaring and other problems happening in the lighter areas (especially in scenes with pure white as well as a lot of shadow detail). The bright areas can cause a washout over the entire frame, dropping the ANSI CR of the shot considerably, a multipass solution can avoid that. (good masking and light control on any surfaces is also a must to stop light reflection washing out the scene.

It isn't of course necessary to get a 'good' transfer, but if you want an excellent transfer, it can make a big difference. I'm looking at it from a professional's perspective, rather than a system for transferring 'home movies' so my design constraints would be different to others.

Depending on your sources and aims, it may not be worth your while to develop a tool any better than the current workprinter/sniper solutions out there, any system with the ccd looking directly at the film frame yeilds a better result than the other consumer level deviced that just project the image onto a 'screen' and have the camera shoot that.

But, if you want to have something that gets close to a Rank, then a multipass process (perhaps incorporating a wet system, or IR pass) will be what you want.

[rfmmars]

No it doesn't have to be a workprinter per se, just any projector slowed down to say less than 6fps (depending on your capture device speed of course), and fitted with a trigger mechanism (i.e. A projector that is modified to work on the same principal as the workprinter, without the expense).
Then you can capture two or three images (at least) in a single 'pass'.

Although the new CCDs come close to the film range, getting it to work in practice isn't so easy.

Regardless of the capture device, your illumination source is often the problem, and it is difficult to get the dark areas lit enough, without flaring and other problems happening in the lighter areas (especially in scenes with pure white as well as a lot of shadow detail). The bright areas can cause a washout over the entire frame, dropping the ANSI CR of the shot considerably, a multipass solution can avoid that. (good masking and light control on any surfaces is also a must to stop light reflection washing out the scene.

It isn't of course necessary to get a 'good' transfer, but if you want an excellent transfer, it can make a big difference. I'm looking at it from a professional's perspective, rather than a system for transferring 'home movies' so my design constraints would be different to others.

Depending on your sources and aims, it may not be worth your while to develop a tool any better than the current workprinter/sniper solutions out there, any system with the ccd looking directly at the film frame yeilds a better result than the other consumer level deviced that just project the image onto a 'screen' and have the camera shoot that.

But, if you want to have something that gets close to a Rank, then a multipass process (perhaps incorporating a wet system, or IR pass) will be what you want.

Well yes a triigger device would pertty much make it a work printer. I modify all my projectors with new 150 watt halagen bulbs soft foucs to lessen hot spot, connected to a separate 21.5 volt transformer. A wall dimmer is connected to the transformer primary so the lens stays set at it's best sharpness. I control the exposeure with the dimmer. The projector has a tone generator that puts a burst on the audio channel so if I have to re-expose a section of film, I have the tone code to show where to delete those spots.

The 21.5v transformer for the bulb, lets me unload the motor so the speed is not effected by change of the resistance of the bulb as it warms up.

After software dymanic and color correction, the end result is a 2-5:1 enhancement over projected film with just one pass in the projector.

This seems to be an easier way to go about it.

Look into the specs. of the Sony EVI 371 PAL or later cameras. You can control the camera and setup with software via RS 232 port.

EDIT: You can get OLD Revere 718 8mm and Revere AZS 830 Super 8 projectors on E-Bay for nothing. I just got the Super 8mm today for $0.99

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&ssPageName=ADME:B:EOIBSAA:US:11&Item=7563732127

The big deal with these is that the entire transport,shutter,and gate come out as one assembly letting you put the light source where the lens once was, and the capture device were the lamp was. it's backward but makes super easy access. You all so get a speed contol pot and a super motor, plus one the best lens out there if you want to go the other way.

richard

[hepi]

Hi there Happy New Year
I am following this with interest Thanks Fred for starting it.
Would any of this Cameras be able for this, some have build in Infrared LEDs Wonder how that would work for telecine. http://www.protectiondepot.com/bulletcameras_menu.asp
Herb

[morsa]

two stupid suggestions from an stupid guy who works with stupid film stuff every day.

Light source, just use a calibrated fluorescent tube (bulb) with a flicker free power source.(flicker free in fact means 100HZ and above).
Phillips and Osram manufacture them.Just look for their codes.I guess they cost around 4 dollars here.
Small variations in color balance from the light source can be easily compensated with the (in)famous "White Balance"

Camera.Get any USB or Firewire one with the resolution you want.Then you can make as many exposures as you want from the same frame.
For blacks =l onger exposure time.For highlights=shorter exposure time.
Then use masktools or simillar to merge both exposures and compress the dynamic range to fit it in 8 bit.

Frame steadyness.Use a register pin.Film perforations have two uses.
1) Film transport
2)Frame registering.

hope this helps

[videoFred]

Oh my God!!

I missed this very interesting thread!:scared:
But here I am :D

Working on this for two years, now.
Some things to consider:

1) capturing 1:1 (frame accurate, one filmframe=one AVI frame) or not?
I work 1:1, but then we need frame rate conversion afterwards.
Fizicks MVFlowfps does miracles, here.

2) the camera: I started with the Philips Toucam Pro: a real CCD webcam with a resolution of 640 x 480. The Philips is USB1.. comes with fine software. This is a very cheap step-in solution: 50 Euro's for the cam. It MUST be the ToUcam Pro, all the others are CMOS!.

All still frames on my website are taken with the Philips.
This was about two years ago...
In the mean time, I learned a lot...
Now I can take much better frames with this Philips.
Time for an upgrade of my site.

I had to re-align the CCD of the Philips, and made a new box for it.
Everything is on my website. I'm working on a new Regular-8 transfer device, and I'm going to use the Philips for it.

Then I bought an Imaging source machine vision camera with a resolution of 1024 x 768 pixels. This is a fire wire camera and it gives great results. The clips on my site are taken with this cam. I made these clips before I discovered MVFlowfps(). So quality of the clips can be discussed..movement is not 100%. I have better ones, now. I can use some tips for optimal WMV compression, too.

3) the lens: I look straight at the film frame.. I use a 25mm lens for the Philips (1/4" ccd diagonal), a 35mm lens for the machine cam (1/3" ccd diagonal). Both lenses with extention tubes, of cource. Otherwise it is impossible to focus on this distance and this small (6x4mm) filmframe.

You find an original, not manipulated 1024 x 768 bitmap frame here:
taken with the machine camera and the 35mm lens with extention tubes.
This is the real sharpness of film, it can be artificial sharpened with the sharpness setting of the camera, and with limitedsharpen() afterwards.

http://users.telenet.be/ho-slotcars/testmap/sailing.bmp

In the beginning, I tought there was something wrong with my system, but I'm very shure by now: this is how film looks. Of cource it depends on the used filmcamera: The 1970-1980 models of Canon had very good optics. But this specific 'sailing' 1976 frame was taken with a Eumig Vienette camera.


4) the backlight: I have the best results with a standard 12V 35W halogene spot, at full brightness, powered by an old computer power supply. Because the machine camera runs at 15fps, I need this stabilised power supply for the spot. Otherwise I get 50-100hz flickering. For diffusing I use opal (white) glass. The best! Heat is no problem with the 35W spot.

5) dynamic range: film has more dynamic range then digital can capture. But in most cases, with good exposed film, it is possible to capture most of it in one pass. A histogram while capturing is very useful. The software of the machine camera has a histogram option. Hovever, for difficult cases, I have developed a double capture system. Someone wrote a small application for me to use with the machine cam. Every time the projector gives a trigger, the cam takes two bitmaps of the same film frame, with different electronic shutter speeds. As you all know, importing the stills and overlay them is very well possible with Avisynth.

For normal capturing I use Cinecap, this program creates an AVI file, and fills the file with frames. The trigger is a switch on the projector, connected with a modified mouse. Capture speed is slow: 4fps.

Richard goes another way, but he has lots of expierence, too, and he is always very helpful and ready to share everything with others. I agree his system works faster.

So, Fred, if you want to discuss anything, here I am.;)
I still have to learn so many things myself.
Mostly about color space, resizing etc...

The script I use for converting my (huge) Huffyuv 1024 x 768 AVI files:

-resizing - MVDenoise - MVFlowFps - Depan - LimitedSharpenFaster.

All in one script, it runs at 3fps. I save the 720 x 576 result with Huffyuv again, then I do some color correction with Virtual Dub and save this result with the Panasonic DV codec. These are my masters for my NLE program.

But, again, I have lots of questions myself to find the best possible workflow and also to (maybe) automise levels, white balance etc...

Here are some frames, taken from the Mpeg2 file, created from the DV masters with Magix video de Luxe.

1972, kodak film:
http://users.telenet.be/ho-slotcars/Frames/1972_kodak_003.jpg

http://users.telenet.be/ho-slotcars/Frames/1976_Kodak_002.jpg

I still have a long way to go.. I need help with upsizing, downsizing, colorspace conversion, overlay etc...

Working with real film is FUN! It's like a time machine...
Quality of more recent (1985) and even today's film (yes, some guys are still filming the old way) is incredible, and even beats digital! (OK, it's softer.. but the colors and the latitude!)

More WMV clips here: (temporal)
http://users.telenet.be/ho-slotcars/WMV/

The problem with WMV is: I must tweak contrast and sharpness before converting to WMV, otherwise the result looks very flat. But I have tweaked sharpness a little to much on these clips. Any help is very welcome.

One more last thing to consider:
Do we want to make film look like digital?
I don't think so...


Fred.

[videoFred]

Now, about the double capture principle:

dark:
http://users.telenet.be/ho-slotcars/Double_Cap/WIM_DARK.jpg

bright:
http://users.telenet.be/ho-slotcars/Double_Cap/WIM_BRIGHT.jpg

used mask:
http://users.telenet.be/ho-slotcars/Double_Cap/WIM_MASK.jpg

overlay:
http://users.telenet.be/ho-slotcars/Double_Cap/WIM_OVERLAY.jpg

I used the normal overlay() Avisynth command.
To create the mask, i used invert() and greyscale()

Something like this:
mask = greyscale(invert(clip1)).tweak(bright=-90,cont=1.2).blur(1.0).blur(1.0)

I bet with masktools we can do a much better job!
(but I do not know how to use it :confused: )

Fred.

[rfmmars]

Excellent post Fred. How mush time would you say it takes to capture a 3" reel and process if it is good quality film stock?

I and Fred are super users of Magix's "VideoDeluxe NLE" for our film work. When ever I have to back up the projector for re-exposure, my modified projector puts a tone code on one of the audio channels for quick post editing. Fred did you know you can multi frameserve from Virtualdub(xxx) to Magix?

Richard
photorecall.net

[videoFred]

Hi Richard,

Well, the projector runs at 4fps, the script at 3fps. It takes houres...
If I would have to live from it, I'd go for a setup like yours.
To transfer films for home use, there is no need for 'super' resolution either.
400 - 500 TV lines looks great on a stand alone TV-DV player.

Or I (my banker:scared: ) would buy a Flashscan8...(25000 Euros)
This new developed machine has everything we can think of.
It works with continue run, and flashes the frames with a Led matrix.
It can run at 25fps 1:1 or slower with automatic pulldown patterns.

But nothing beats the pleasure of making something of your own, with your own hands!

No, I did not know one could frameserve to Magix.
This would save several compression steps!

But then going forwards in Magix is very slow, I guess.

But I must go, now.
I'l be back next monday.

Have a nice weekend!
Fred.

[rfmmars]

Hi Richard,

Well, the projector runs at 4fps, the script at 3fps. It takes houres...
If I would have to live from it, I'd go for a setup like yours.
To transfer films for home use, there is no need for 'super' resolution either.
400 - 500 TV lines looks great on a stand alone TV-DV player.

Or I (my banker:scared: ) would buy a Flashscan8...(25000 Euros)

No, I did not know one could frameserve to Magix.
This would save several compression steps!

But then going forwards in Magix is very slow, I guess.

But I must go, now.
I'l be back next monday.

Have a nice weekend!
Fred.

Not slow at all, infact loading is very fast. I have used 12 Virtualmpeg2 frameservers at one time feeding one Magix NLE. On our forum, the record is 80 plus at one time. It would be great if we could frameserve out of VideoDeluxe. You can "Smartrender" export mpeg2 at 15,000 bps lossless just like "Direct Stream Copy" in VD(xxx).

See this long post.

http://support.magix.net/boards/magix/index.php?showtopic=12122

I am working a show this weekend for my telecine work, so I will crash Monday.

Best to you,

Richard
photorecall.net

[burnselk]

Perhaps the use of an old projector, then all the mechanicals of moving the film will be taken care of.

Of course, the motor would have to be replaced with a servo/stepper and the bulb would not need to be as bright.

But I get what you mean - feed the image direct into the camcorder, not via a screen, and control the frame rate of the film from the PC.

Nice idea - I'm currently getting about 3000ft of 16mm converted, and it is costing me about a grand. A "home remedy" would have been an attractive option, even at half the price.

MM.
Mustardman.....how did the dvd's turn out? Was the quality any better than the film projected on a screen?

I'm thinking about having my regular 8 mm film put on dvd's. Do you have any suggestions as to where I should go to get that done?

I like Fred's idea. But, I bet the cost is going to be pretty high....guess we'll have to wait and see. I'm certainly interested....I must have 2,000 ft. of regular 8 I want to put on dvd.

I'd like to hear about your experience with the company that put your film on dvd....would you share that with me?

[videoFred]

I like Fred's idea. But, I bet the cost is going to be pretty high....guess we'll have to wait and see.

You do not have to wait...
My transfer unit is running very fine.

About the cost (if you make it yourself):

Old Eumig: 25 Euro
Philips Webcam: 50 Euro
Old 25mm lens: free
Slow drive engine: free (old Telefunken casette player motor)
All together with some additional other parts: about 100 Euro.

If you want a better system: the cost of the machine cam and lens was about 800 Euro..

If you want a company to do the transfer for you:
Try Richard ;)

Fred.

[johnmeyer]

I own an 8mm/Super8 Workprinter. Great device, but slow. I now have to do 16mm conversions, so I bought my own projector. With very little work, I now have a device that produces frame-by-frame film to video transfer (like Workprinter) but at full 24 fps, and with no need for the cam/switch/mouse approach that uses Stop Motion animation capture. Benefits include the ability to simultaneously capture fully synced sound, much faster throughput (capture is real-time), and the ability to capture to HDV (which cannot be created with stop-motion).

The projector modifications only require removing the projector shutter, and, of course, getting a diffuse light source (if you are going to shoot back into the projector using an aerial lens or similar arrangement). I just replaced the MR16 ELC projector bulb in my Eiki SSL-O with a 30W 24V MR16 (similar color value, costs $2.50), and stuck some frosted glass on the end.

The secret to the whole thing lies in the knowledge accumulated in the AVISynth section of this forum, and in a few critical insights I gained while trying to understand why real-time transfers using a 5-bladed shutter (something I didn't want to do because of the lousy quality) were better than with a 3-bladed shutter. After careful study, it became clear that most of the reasons usually offered were incorrect and misleading. I started to sketch out on a timeline what happens at each moment in time when a film is projected (on-off-on-off-on-off-on-off film moves etc.), and when an unsynchronized CCD video camera photographs that mess, and does so with various video camera "shutter speeds."

I have to go to a meeting, so I'll leave you hanging for a moment, but if you're interested, I'll give you the rest of the story. There has to be a downside to what I'm doing, but so far I haven't found it. Some of the post processing ideas for getting closer to film gamma are still going to be extremely useful, although I can do some of this in my FX1 when I capture.

[rfmmars]

I own an 8mm/Super8 Workprinter. Great device, but slow. I now have to do 16mm conversions, so I bought my own projector. With very little work, I now have a device that produces frame-by-frame film to video transfer (like Workprinter) but at full 24 fps, and with no need for the cam/switch/mouse approach that uses Stop Motion animation capture. Benefits include the ability to simultaneously capture fully synced sound, much faster throughput (capture is real-time), and the ability to capture to HDV (which cannot be created with stop-motion).

The projector modifications only require removing the projector shutter, and, of course, getting a diffuse light source (if you are going to shoot back into the projector using an aerial lens or similar arrangement). I just replaced the MR16 ELC projector bulb in my Eiki SSL-O with a 30W 24V MR16 (similar color value, costs $2.50), and stuck some frosted glass on the end.

The secret to the whole thing lies in the knowledge accumulated in the AVISynth section of this forum, and in a few critical insights I gained while trying to understand why real-time transfers using a 5-bladed shutter (something I didn't want to do because of the lousy quality) were better than with a 3-bladed shutter. After careful study, it became clear that most of the reasons usually offered were incorrect and misleading. I started to sketch out on a timeline what happens at each moment in time when a film is projected (on-off-on-off-on-off-on-off film moves etc.), and when an unsynchronized CCD video camera photographs that mess, and does so with various video camera "shutter speeds."

I have to go to a meeting, so I'll leave you hanging for a moment, but if you're interested, I'll give you the rest of the story. There has to be a downside to what I'm doing, but so far I haven't found it. Some of the post processing ideas for getting closer to film gamma are still going to be extremely useful, although I can do some of this in my FX1 when I capture.

Waiting for your next post with details of your system.

Richard
photorecall.net

[johnmeyer]

I'm a little shy about posting all the details until I run a lot of film to verify that all is going well. I will share, however, a few of the insights that led to my "invention" (I am doing nothing that hasn't been done before, but no one -- I don't think -- has put it all together this way).

1. Why is a 5-bladed projector shutter better for doing full-speed film to video transfers?

I had no desire to go this route because of the fuzziness that comes from frame blending, but I was fascinated by all the discussion about needing a 5-bladed shutter if you wanted better results. Most discussions referenced the flicker you perceive when watching the actual film projected onto a screen with a 5-bladed shutter. This clearly isn't something that matters to the video camera.

I also noted that all the "cookbooks" on capturing video directly from a projected image all made it clear that you had to use a shutter speed of 1/60 of a second on your video camera.

That got me to thinking, and before I could formulate a complete answer to my first question, I ended up asking a second question:

2. Why do you need to use a camera shutter speed of 1/60 of a second?

This one I figured out pretty quickly. If you use a really fast camera shutter speed, then you might capture a moment when the projector shutter is completely closed. That means that field would be completely black.

3. But wait, what does it even mean to have a shutter speed on a video camera? Isn't this thing scanning?

I was hung up on this for a long time. Obviously a traditional TV set plays the video one scan line at a time for each field, and then does the same for the other field, 1/60 of a second later. Each scan line occurs at a slightly later instant in time than the line above. Older TV cameras used to do the same thing when capturing. For them, the idea of a "shutter speed" was meaningless. It took me awhile to have that big "duh" moment and realize that the CCD in modern video cameras is no different from the CCD in still cameras (which is why video cameras usually have some sort of still photo capture capability). Thus, each field is captured in the same manner as an image is captured on traditional film, with the entire image area EXPOSED AT THE SAME MOMENT IN TIME. The light actually "imaged" onto the sensor is the average of the light that falls on that sensor during the time each pixel in the CCD array is enabled.

This then, at last, gave me the answer to the first question about why a 5-bladed shutter is better.

It also lead me to the discovery that is at the heart of the invention.

First, when taking a picture of a series of still images that are flashed on the screen for brief periods, the actual image that gets transferred to the sensor is the AVERAGE of the intensity for the time the sensor is enabled. If, during the capture, the shutter closes for awhile, and if the sensor is enabled for the entire period of one field of video (1/60 of a second for NTSC), then rather than getting a black image on some of the scan lines (like an old camera would have recorded -- this is what hung me up for awhile), instead, the black pixels get averaged with whatever light appears during the time the projector shutter is open. The longer the duty cycle (i.e., the larger the percentage of time the shutter is open), the brighter the picture. Since the video camera and the projector cannot be synchronized, there is no way to know at what point in the projection of each frame the video camera will actually make its 1/60 second snapshot. Depending on where it starts, and on how many times the projector shutter opens and closes for each frame of video, the CCD may see a very large percentage of all black (shutter closed) or a very small percentage.

Here's the thing: the more blades, the less the possible variation in that duty cycle, and therefore the less the flicker.

Great, so where does this all lead? I don't want any flicker, and I want to capture individual frames, just like my 8mm/Super8 Workprinter. I have this new (to me) Eiki 16mm projector, why don't I just slow it way down and proceed to pretty much mimic what was done on the Workprinter?

But I have cans and cans of 16mm film and this project is going to take forever if I do it that way. What's more, I find out that this projector uses a synchronous AC motor, and there is no way I can electrically slow it down like was done with the DC motor in the 8mm projector. I can try to use belts and pulleys, but that is going to be hard because the gear ratio results in pulley sizes that won't fit on the machine. I can use another motor (I briefly ran the projector using my electric drill), but this is getting ugly fast.

There has to be a better way.

Well, if I remove the shutter, I can get rid of all that time when the frame is black, and I won't have to slow things down that much. I start sketching things out on a timeline, and indeed, I can run a lot faster. However, I keep running into timing issues, because with a 1/60 of a second camera shutter and 24 fps coming out of the projector, depending on how long it takes for the film to advance, I have to always have enough time left to snap two successive video fields. Without going into all the math, I conclude that I can run the projector no faster than 12 fps. When I look at other factors, about how the stop motion capture software works, I realize why the Workprinter is only rated at about 6 fps.

And then it dawns on me.

Why the heck do I need to use a 1/60 shutter? The only reason for that shutter speed was to keep the shutter open for the entire time it takes to capture one field of video in order to provide the maximum averaging for the projector shutter opening and closing during the capture (this is why I gave you that long explanation above).

But I've removed the projector shutter!

Why, instead, not use a camera 1/1000 shutter. If I do that, then I can get two samples only 1/60 of a second apart. Think about it, the camera captures one field right now: blink 1/1000 of a second and its got the picture. From that moment, exactly 1/60 of a second later: blink 1/1000 of a second, another frame is captured. Even if the projector is running at full 24 fps, no matter when the video camera snaps a frame relative to the projector I can be absolutely positively guaranteed to get two fields (i.e., one complete frame) of video of a motionless frame of film. After those two fields, depending on timing, I will either get another good field of the same film frame, or I might get a snap of the film as it is moving to the next frame (which will now be a clear snap rather than a blur, although it will be out of registration). What I will get is 3 good fields followed by 2 suspect fields followed by ...

Can you hear it coming?

Yup. It's 3:2 pulldown. What I've captured is ALMOST the same as what 24 fps video looks like when pulldown has been applied so it can be shown at 29.97 interlaced.

And what do we do, class, when we want to recover the 24 fps video from 29.97 interlaced video?

IVTC.

So, folks, that is my invention. It is so simple that I guess maybe it shouldn't be called an invention. Others have implemented various parts of it before, but none of them -- I think -- were familiar with IVTC software to have the solution seem so obvious.

This technique can be used whether you project onto a screen, use a telecine box, or capture looking directly into the projector lens. It can be used with any projector, sound or silent, once the projector shutter is removed (although most IVTC software is "tuned" for 24 fps operation).

The post processing techniques and exposure issues still all need to be addressed, as discussed earlier in this thread. The key thing this approach provides is:

1. Throughput. You capture at the full speed of the projector.

2. Frame-by-frame capture. You get the same benefits of a Rank Cinetel or Workprinter capture (although the Rank has lots of features you can't get with a projector).

3. HDV capture (maybe). If you want to capture 16mm using HDV, the 5-bladed projector approach is a bad idea because the blended frames will make the MPEG encoding go nuts. I am not sure (because I haven't yet tried it) whether the spurious frame captured while the film is moving (which will eventually be discarded by the IVTC software) will screw up the HDV encoding or not. Hopefully not, because HD capture of 16mm is a holy grail.

4. Capture of synchronized sound. With the Workprinter, if you have sound film you have to capture the sound separately and then sync up in your editing software. With this approach, you are capturing the sound at the same time as the picture, so they are -- by definition -- in sync (assuming your projector is in sync). After applying IVTC, if you set the flag in the AVI file to 24 fps, everything will playback in perfect sync. You can then encode directly to DVD, setting the 24 fps flag, and you are finished.

Let me know what you think.

[videoFred]

Thank you for this very interesting explanation and for sharing it.
The big question: what do you do with the out of registration fields/frames?

Second question: if I do this with my 15fps progressive machine camera, I still get blurred frames.....

Maybe I should try even higher shutter speeds?
I can go up to 1/4000. Of cource the projector must run at 15fps also.

Fred.

[rfmmars]

Thank you for this very interesting explanation and for sharing it.
The big question: what do you do with the out of registration fields/frames?

Second question: if I do this with my 15fps progressive machine camera, I still get blurred frames.....

Maybe I should try even higher shutter speeds?
I can go up to 1/4000. Of cource the projector must run at 15fps also.

Fred.

Yes they are GREAT details and fast coming,

Haven't try this so far, my concern would be reduced light sensistivty with a fast shutter. Using a 3CCD camera, I am at 15 lux. F 1.2

Richard
photorecall.net

[FredThompson]

Boy, this thread is turning into a real dream come true :P

@VideoFred, I've been meaning to drop you a note after a buddy and I stumbled across your discussions in another forum. Glad to see you're here, too. Great name, btw. Us Freds gotta stick together.

@all recent posters, thanks. Great ideas. I'm still enamored with the idea of multiple captures of each frame but realtime audio capture is also important.

[johnmeyer]

Actually, the shutter speed really doesn't need to be all that high. I used 1/1000 just to illustrate the point. The main reason for reducing it to less than 1/60 is to reduce the total capture time so that you can always be assured of getting two good captures (one for the top field of video and one for the bottom field) for every time each frame of film is at rest in the gate. Actually, even 1/120 would probably be sufficient. Remember that the only captures of interest are those where the film ISN'T moving, so having a higher shutter speed has nothing to do with "freezing" the action. I probably mislead some by making the off-hand remark about being able to freeze the film if you happen to capture a frame while it is moving.

As for IVTC, I didn't go into much detail (that's the part I'm withholding for now until I make sure it works 100% of the time), but if you've ever used any of it, you know that it compares fields, and keeps those that have very little difference.

[videoFred]

As for IVTC, I didn't go into much detail (that's the part I'm withholding for now until I make sure it works 100% of the time), but if you've ever used any of it, you know that it compares fields, and keeps those that have very little difference.

I begin to understand...
So IVTC removes bad fields/frames?
Those with to much difference? (moving)

EDIT: no, I do not complete understand.. What's IVTC?

Fred.

[FredThompson]

No, IVTC is related to removing the padding frames which are added to film source when it is converted from 24 fps to 29.97 for NTSC TV format. It has nothing to do with using an NTSC camera to record film real-time display. Real-time playback of the film should give proper audio but the frame rates are different so you get corupted frames in the video.

[videoFred]

No, IVTC is related to removing the padding frames which are added to film source when it is converted from 24 fps to 29.97 for NTSC TV format.

I see... But why remove them if you are making a NTSC DVD of it?

(experimenting with 4200K halogene light source, now.)
(also comparing Workprinter Sony 3CCD transfer with my system)
(hope my system is not too bad)
(also using special testfilm to test sharpness etc..)

Fred.

[FredThompson]

If you make an NTSC DVD of a film source you can encode at 24 fps and set the pulldown flag. The playback will look proper on NTSC equipment or a software player. The padding frames are inserted when a 24 fps film source is converted to an actual 29.97 fps NTSC stream so it looks like the proper playback speed. (I can't, for the life of me, comprehend why history shows don't properly pad early film so the movement isn't jerky...) PAL has a progressive mode and most of the time film is broadcast as the original frames but at 25 fps instead of the original 24 fps. Progressive images are going to be easier on the eyes and you don't lose bandwidth storing padding frames.

If you use an NTSC camera to record film at real-time speeds, you'll have video frames which don't exactly correspond to film frames. The hope is to get rid of the junk extra frames with a result of the proper framerate. This is why I am interested in slow capture or synchronized computer camera capture instead of NTSC video capture. Make sense?

[johnmeyer]

Answer to the IVTC question.

Film is 24 frames per second. Video is 60 fields per second, where two fields make up one frame. Since you only have 24 pieces of motion each second, yet you need to display 60 fields every second, you have to come up with a scheme that duplicates each of the 24 frames. Since 24 doesn't go into 60 evenly, what is done is to repeat one frame of film for three fields of video and then repeat the next frame of film for two fields of video. You then repeat this over and over with subsequent frames of film. Thus, you get two frames of film spread out over five fields of video. Since 60/24 = 5/2, this makes the film project at the proper speed. Your eye doesn't detect the fact that some frames are shown for a fraction of a second longer than others.

Of course video is really 59.94 fields per second, so there is a slight adjustment that gets made, but this is the basic idea.

Inverse Telecine software merely removes the duplication and "recovers" the original 24 original frames. This is extremely important software to use if you want to encode a movie for DVD, but your only source for that movie is a tape you made over the air or a VHS tape you purchased. If you merely encode that, the encoder goes nuts with the staggered fields. This is especially true because every frame of video is made up of two fields. Because five fields of video are used to encode two frames of film, one of the frames of video that results from blending two of those fields together must contain video from two frames of film. When projected, this looks just fine, but the encoder sees this "blended" frame as having abnormally high motion. It takes more bits to encode. What's more, the encoder must encode 30 frames every second, instead of 24. If you only have to encode 24 objects a second, you'll have more bits available for each frame, which results in a higher quality encode.

Back to my film project. When you use a video camera to capture film in the manner I described, you end up with exactly the same 3:2 pattern. The only difference is that on some captures, depending on the timing between the camera and the film, that one blended frame (which is the one you want to throw away) may include a field that was captured while the film was advancing. All that does is make the frame even easier to detect and throw away.

[tedkunich]

@videoFred

I was looking at your site for making your own works printer and am interested in the obective lens you used - the site translation was difficult to understand. Would you know of any sources for such a lens?

Thanks


T

[videoFred]

@videoFred

I was looking at your site for making your own works printer and am interested in the obective lens you used - the site translation was difficult to understand. Would you know of any sources for such a lens?

Thanks


T

Sure, here:
http://www.1394imaging.com/en/products/optics/megapixel/

But if I find the money, I might try a macro lens, too.

Fred.

[videoFred]

@ Fred and John:

Thank you both for explaining!
Living in Pal Europe, I never have to deal with this.

(Experiment with 4200K Halogen spot:
Yes, it's better. More easy to white balance, and better colors, too.)

Fred.

[smok3]

well, tnx all for an interesting thread (iam still with 'how to use my 35 leica lens for dv cam - to decrease the dof" thought, but this thread did give me some ideas . i think...)

[hepi]

Just to keep this alive
A link I found to bad it is only for 35 mm and could not find how much this camera cost. http://kmpi.konicaminolta.us/eprise/main/kmpi/content/cam/cam_category_pages/DigitalFilmScanners
Herb

[hanfrunz]

I think there is a super8 gate for this baby (http://www.thomsongrassvalley.com/products/film/spirit_4k/):D

[videoFred]

Hi, FilmFolks ;)

Any new developments?
I changed my URL: www.super-8.be
Added some new filmframes on my site, too.
If the English intro has bad errors, please let me know.:scared:


I'm working on a new regular-8 unit:
Same capture camera, but this time I'm gonna use a Sony 45mm macro lens.
Pointed straight at the film frame.
Gives a very sharp picture, also in the corners.

The camera will be mounted on a precision machine slide, for optimal focussing.

This projector will be able to run stepless between 1 and 20 fps.
And I'm going to make removable shutter blades.
Because I still want to try to capture at high speed.
This way, I can test 1 - 3 - 5 blade shutter setup.

With a one blade shutter, there are no moving or blurred frames.
But a darker frame, now and then.
Depends on the shutter speed from the capture cam, of cource.
And the speed from the projector.

Within a few weeks, I can begin the experiments.
I do not have to worry about fields, because my cam is progressive.
I hope Avisynth can help to remove eventual duplicate/dark frames.

And if it's not good enough, I still can capture frame by frame.
This has proven to be the best system for now...
But slow...:eek:

Fred.

[johnmeyer]

New developments? Well sort of.

I have worked out all the kinks in my film to video transfer. It was a good thing I didn't rush right in and try to do the whole thing back in January, becuase some newer versions of software, and also a little reflection on my design and technique have allowed me to perfect the whole process. The results truly are to die for.

First, the newer versions of TIVTC (which includes TFM) really made a difference in the process of removing the pulldown and extra frames. That process now works very well. I also was able to solve the problem of adjacent fields being offset because of residual "squirming" of the film as it comes to rest. With other types of film to video transfer, you would not see this, but because I am "snapping" the film image using a 1/1000 shutter speed. the residual movement as the film comes to rest is not blurred out over time. I solved the problem by using motion estimation software, which almost perfectly aligns the two fields horizontally and vertically. I then vertically shift the second field to put it spacially back in the correct place (one scan line lower).

Finally, I decided to eliminate the front mirror and aerial lens that I took from my Workprinter and was using for this project. What were we all thinking? What a horrible thing this device is. It introduces all sorts of chromatic aberrations (colored fringing) as well as keystoning and other distortions. I now just point the camera at the projector, about six inches away from the lens. The picture is upside down, but that requires exactly one single line of AVISythn code to fix. Boy, the picture sure is sharp.

I also spent a lot of time learning all the controls and settings on my new Sony FX1 HDV camera, and found various settings that really make a difference for film transfer. I auto white balance on the white light itself. I use the ND filters to get into the mid range of exposure. I make sure that the auto-gain is set to 0dB, so that all exposure is made with the aperture and the ND filter. I use a 1/1000 shutter speed. Most important, I use a combination of the "spotlight" feature, coupled with the autogain reduction. With the zebra pattern turned on, I adjust the autogain reduction until the zebra pattern just disappears. This makes sure that I always get proper exposure in the highlights. I then adjust the shadows and midtones using the correction tools in my video editing program (Vegas). I set up six different levels of gain reduction in the "PPV" menus, so I can instantly dial in exactly the amount of reduction required to just make the zebras disappear.

As I said, the results are unbelievable. And, I can capture at full 24 fps, and yet get Workprinter/Cintel frame-by-frame results.

[rfmmars]

@ Fred and John:

Thank you both for explaining!
Living in Pal Europe, I never have to deal with this.

(Experiment with 4200K Halogen spot:
Yes, it's better. More easy to white balance, and better colors, too.)

Fred.

What I do is to install a 300 watt halogen 82V bulbs to all of my modified projectors with a separate mains transformer with the primary side having a wall dimmer. Yes they say you can't use a dimmer with a motor or transformer, Yah right.

This allows me to 'Light Blast" the dark scences" and also allows the projector light level to be set for the sweet point of the pickup camera, which I am using a 2/3" 3 CCD Sony DCX 755 with analog color balance and black clamp and iris control.

What I see happining here is that people are getting caught up in thinking that they are dealing with professional shot home moves, which they are not.

These old cameras had wind up spring motors or battery operated drives. Trying to worry about pulldown and other like things is a waste of time, you are dealing with non standard, varying film rates. Wait until you have to deal with "Sound Super 8 or 16mm. All those details goes out the window when you are dealing with sound film shot with weak batteries. Now your dealing with as low as 6 fps and varying across the entire clip.

Richard

photorecall.net

[dokworm]

No I think you misunderstand, we are talking about using a pulldown style software side to allow us to capture the film in at full speed, and yet end up with one video frame for each film frame (without having to resort to single frame advance and capture)

John, are you just pointing your Sony video camera at the *lens* of the projector?!?
I have been removing the lens altogether and pointing the camera directly at the *film*.
I'm suprised it works shooting through the projector lens as well, what sort of zoom level are you running on the video camera?

[dokworm]

Yeah I wish the FX1 had a true progressive mode, I hit the same problem with film movement between fields even when experimenting with an old TRV900E about a year ago when I tried coming in full speed and shutterless much in the way you describe.
I'm surprised the motion estimation works so well, can you share your setup?

[johnmeyer]

John, are you just pointing your Sony video camera at the *lens* of the projector?!?Yup. I set the camera about 6-10 inches from the lens. I focus the projector lens about midway in its travel and then zoom in with the camera. It doesn't require much zoom. Like I said in the earlier post, I have no idea what I was thinking, using that aerial lens, other than that was what Roger Evans used with the Workprinter I bought from him. I think the reason he used it is that some of his products are designed to be used without any computer involvement, and therefore flipping the image right-side-up would be a big deal.

I am surprised you can zoom in far enough to get the 16mm frame full size. I am quite certain that my camera would not be able to get something smaller than a postage stamp to fill the frame all the way.

Yeah I wish the FX1 had a true progressive modeActually, if it did, I couldn't use it. You'd have to draw out the diagram to see what the timing looks like, but if the video camera is not synced in some way to the projector (which it is not in my arrangement), then there is no way you could ever get a 24p camera to get a single frame of film, without the pulldown (if you remove the shutter) or without getting part of another frame (if you include the shutter).

I'm surprised the motion estimation works so well, can you share your setup?Well, I haven't cleaned up the code, so what follows is REALLY ugly. However, it works. This is the code for the second pass. In the first pass, where all the fun stuff happens, I actually figure out which fields I can discard. I put this information in a text file, and then massage that text file externally to make sure all the decimation decisions are correct. I then read that file back into TFM in this pass and do the actual decimation. The decimated frames get read into the motion estimation. Since the motion estimation unfortunately corrects in the vertical direction as well, I have to shift the fields (complementparity). The thing that I was stumped on, and which someone in these forums (in Germany) helped me, was to figure out how to ONLY have the second field compared with the previous field, and then have the motion estimation reset. Otherwise, you end up with motion stabilization which is not a desired result at all.

OK, here it is. Don't blame me if it doesn't make sense. When I'm finished with all this, perhaps I'll post the entire "recipe." Note that I use "autogain" to get better contrast for the TFM and motion estimation algorithms, but I actually then operate on the original, unadjusted footage. Neat trick (at least I thought so) to improve the quality of these filters.

I put this project on hold for most of last month, and I can't remember at the moment whether the normal operation of the script is to return "k" or to return "i". I was playing around with a lot of things. I think "k" is the final result. I think I was trying to fix a residual problem where the result is shifted by one frame. Not a big deal, but I wanted to fix it and then got side-tracked.

# Script to recover film frames from film projected on shutterless 16mm projector.
# Second Pass
# Copyright 2006 John H. Meyer
# Revision May 9, 2006
#-----------------------------

loadPlugin("c:\Program Files\AviSynth 2.5\plugins\TIVTC.dll")
loadPlugin("c:\Program Files\AviSynth 2.5\plugins\Depan.dll")
loadplugin("c:\Program Files\AviSynth 2.5\plugins\MultiDecimate.dll")

AVISource("D:\OPRFHS\Intersquad 1st reel)0002.avi")

FlipVertical()
converttoYUY2(interlaced=true)
AssumeBFF()
tfm(display=false,micout=2,mode=0,cthresh=45,mi=1600,\ pp=6,metric=0,field=1,micmatching=2,slow=2,\
blockx=256,blocky=256,sco=-1,\
debug=false,input="d:\tfm.txt")

a=MultiDecimate(pass=2)
b=AssumeFrameBased(a)
c=b.separatefields()
d=c.colorYUV(autogain=true)

e=interleave(blankclip(c),selectodd(c),selecteven(c),blankclip(c))
f=interleave(blankclip(d),selectodd(d),selecteven(d),blankclip(d))

mdata2 = DePanEstimate(f,range=1,dxmax=32,dymax=32,pixaspect=0.9091,info=false)
g=e.Depan(data=mdata2,matchfields=false,offset=1,info=false)
h=g.selectevery(4,2,3)
i=h.ComplementParity().weave().AssumeFrameBased.AssumeFPS(24)

j=i.ComplementParity().DoubleWeave().SelectEvery(6, 0, 0, 2, 3, 4)
k=j.AssumeFPS(29.97) # not a problem, since you have no audio
k
#i

[rfmmars]

No I think you misunderstand, we are talking about using a pulldown style software side to allow us to capture the film in at full speed, and yet end up with one video frame for each film frame (without having to resort to single frame advance and capture)



What I am saying is don't get caught up in the details when your source has a varying frame rate, because is was shot with crude equipment by today's standards.

Richard

[videoFred]

to capture the film in at full speed, and yet end up with one video frame for each film frame


Maybe you are talking about 16mm only?
This was made at 24fps...

Regular 8 and Super-8 can be 15-16-18-24 fps and even something between it, and even all these speeds together :) , like Richard says.

But I agree, a frame accurate (1 film frame= 1 avi frame) and progressive transfer is the best way to start.

But then, for DVD use, it must be converted again.
MVFlowfps is the tool I use for this.
It smooths out zooming and panning very accurate.
I use Depan, too.

OK, sometimes MVFlowFps gives you wierd frames, on scenes with fast motion, but you barely see these artefacts on TV, at 25fps. The same should be true for NTSC users.

A really perfect solution would be an electronic synchronisation between the projector and the digital camera. It exists, but I hear it is not working 100%.....

Fred.

[dokworm]

Yeah I am talking 16mm mainly and 24fps Super8.
For DVD I am happy to just use the PAL methodology and speed it up to 25 fps.
Thanks John for the detail, it explains what I was missing, I was ending up with stabilised frames and couldn't work out why (Duh!)
Now it all finally falls into place in my thick skull.

I can zoom in far enough because I have a removable lens setup, (not on the FX1 obviously) so extension tubes and manual focus get me there easily.

[videoFred]

Hey guys,

It's time to show us some results!
A few frames would be fine!

Fred.

[dokworm]

Wouldn't a sync setup just be a matter of stripping out the sync pulse from the output of the camera and driving a stepper motor to advance the projector one frame (using the sync pulse as the trigger).
You could then have the projector being driven by the stepper motor at 25fps (for a PAL camera) and the camera just grabbing the frames when the stepper was stationary.
You could then come in at full speed without the need to discard any frames.