Jump to content

Custom DIY Telecine


Stephen Hargreaves

Recommended Posts

Updated Reversal scan...

 

 

Working on frame steadiness at the moment and have ordered a new camera that will capture 2K up to 15fps although we'll do some head scratching to figure out everything that needs to happen to get it to capture at 15fps :) We'd love to manufacture something that costs the price of a professional DSLR ($4K - $6K) that sits on your desk and transfers 8/S8 to 2K... even 4K... We're currently working on software to write DPX, TIF & JPG.

 

Think this thing will have a place in the market for film lovers?

Link to comment
Share on other sites

  • 4 weeks later...
  • Replies 87
  • Created
  • Last Reply

Top Posters In This Topic

So what is the best method to do frame grab telecine? I might be getting a unit to build up on and would like to build myself a unit. I get that the grain is bigger with a higher res scan but doesn't it mean you can process an even better image in the end?

 

For a camera, has anybody considered a machine vision unit?

 

As for capturing each frame under a different light, isnt it pointless? I remember reading the need to taken braketed shots in digital and blending them because digital images have less lattitude. I thought film couldn't really be improved on by scanning this way.

 

I've always wondered if a 35mm film scanner could be adapted to cine formats. The software is largely there to process the image well. The rest os complex ofcourse but thought it might be a decent avenue.

 

anybody called Kodak and talked to them about the reversal tranfer and the orange to find out what the best way to reverse the image would be? The lab guys there have always been ultra helpfull.

I'm just asking the dumb questions to get some of the basics out of the way.

Link to comment
Share on other sites

Kevin, I think everybody I know that wants to build a solid system has considered the Machine Vision cameras :) We've been looking at the Guppy Pro but the sensor is very very small. Not the pixel size but the physical size. What we're doing right now is projecting the image directly onto a 5D full frame sensor and are loving the results. Obviously we can't continue doing this because we are wasting the shutter.

 

We'd like to purchase the higher end camera with the Kodak full frame sensor but they're $10K. We really want to build an "in home" transfer device that can be bought for the price of a professional DSLR, somewhere around $6K.

 

Yes, you would have to use some sort of HDR capture to preserve the latitude and density of the film. We've considered that as well. We're basically doing what HDDSLR shooters are doing to gain latitude in their images. We peg the saturation and contrast as low as the camera will allow.

 

As far as negative, I've had fairly good luck using the same bright white LED and drastically playing with the white balance until the orange is white(ish). Not a fix, but for transferring home movies it looks pretty good... but always really dirty as the dust is white not black.

 

Good luck, let us know how you do!!!

 

Here's our latest transfer:

 

Justin

Link to comment
Share on other sites

  • 4 weeks later...

Kevin, I think everybody I know that wants to build a solid system has considered the Machine Vision cameras :)

 

I'm using machine vision cameras for 5 years now.

 

1974 Super-8, capturing with my machine vision camera:

 

Rudy_S8_002.jpg

 

 

1973, Rome, 16mm capturing( not by me, but also with a machine vision camera)

 

Rudy_16mm_001.jpg

 

Both pictures are post processed with my special Avisynth-film restoring script.

This script removes all dirt spots etc...

 

Fred.

Link to comment
Share on other sites

  • Premium Member

As usual, Fred, it's very nice stuff! Is this a monochrome camera and RGB exposures? (Edit: Yes, it is. Read Fred's site.)

 

I don't think it's necessary to worry too much about a full frame sensor. If you wanted more dynamic range, you could always do HDR exposures for each channel (for a total of nine exposures per frame, which might slow things down a bit!)

 

P

Link to comment
Share on other sites

As usual, Fred, it's very nice stuff! Is this a monochrome camera and RGB exposures?

 

Hello Phil,

 

Well.. actualy I'm using an 8 bits 400 Euro low-end 1024x768 Imaging Source machine vision camera.

No, it's not monochrome but the debayering is done on the fly with the capturing software.

In the future, I will upgrade to a better IMI camera. The price will be around 2000 Euro.

 

Like you say, RGB exposures would be the max of cource. :)

Frank Vine is working on this. He has developed a Led RGB lightsource.

It is custom made for the use with machine vision cams with trigger.

Franks site: http://www.cine2digits.co.uk/

 

But we are talking about getting some more detail out of the dark parts here.

My camera gives already pretty good results.

 

PS: my Avisynth script is also removing the vertical jitter pretty good.

So the transfer can be done on any decent (modified) projector, no need for pin registration.

 

Fred.

Link to comment
Share on other sites

Frank Vine is working on this. He has developed a Led RGB lightsource.

It is custom made for the use with machine vision cams with trigger.

Franks site: http://www.cine2digits.co.uk/

Hi there,

As Fred has kind of introduced me, thought I'd sign up....might save a few more direct emails asking me the same questions over again - although it of course it is always nice to hear from other diy'ers and I try to help when I can.

My project site is not fully up to date at the moment as I'm working on a new high-power, high-speed RGB flash system for continuous motion transfer - exposures in the order of just 10 to 40us. My prototype is now achieving around 15us at f5.6 for full CCD exposure with an empty gate whilst maintaining a pretty good purity and evenness across a 16mm frame.

 

But we are talking about getting some more detail out of the dark parts here.

My camera gives already pretty good results.

 

As usual Fred, you do manage to get more out of your camera than your should get by rights ;) but details in the dark areas are perhaps the weakest area. I'm waiting for you to buy that 2/3" IMI or similar :)

 

Frank

Link to comment
Share on other sites

  • Premium Member

The problem is that by the time you're doing RGB exposures plus HDR, you're doing - ulp - nine exposures per frame.

 

By that point, though, you should really, really be seeing absolutely all of it.

 

P

Link to comment
Share on other sites

The problem is that by the time you're doing RGB exposures plus HDR, you're doing - ulp - nine exposures per frame.

 

By that point, though, you should really, really be seeing absolutely all of it.

 

P

Not talking about separate RGB exposures (although this is an option for use with mono cameras) but an RGB adjustable light source which significantly helps optimise the exposure on a CCD colour camera. If you set all camera colour balance controls to mid way then look at a so-called white LED, it will look significantly green to the camera. This surprised me a bit seeing as they mostly use Blue LEDs, but clearly the phosphor coating has the dominant effect. So if we do not somehow filter the light then compensation has to be made in the camera (or worse still, in post) by amplifying the red and blue…which means amplifying the noise too. This can easily be 6dB (often more) and hence loses about one stop of useful information in the blue and red dark areas as well as adding noise in the bright areas, such as in blue sky areas.

 

So the basic principle is to adjust the RGB light source for close to max exposures during capture, which will give very good white balance if there is any white in the image, then this gives the best source for further colour correction in post. Using this method together with a CCD sensor with good sized pixels (like 6.47um for the Sony ICX285 sensor) minimizes the need for HDR and only requires two exposures if HDR is required. Another (more expensive) way to achieve optimum capture is to use a 3-CCD camera with separate control over the R,G & B exposures. I only know of one make which offers this facility, but it falls down in other areas so is not quite suitable IMHO.

 

Frank

Link to comment
Share on other sites

Has anyone had any experience with the Imaging Source All-in-One cameras for DIY telecine?

 

Cheers,

Jean-Louis

Sorry, but what do you mean by "All-in-one"? I don't see anything as such on their website.

In general, Imaging Source cameras are built down to a price. They are good value (Fred is using one at the moment) but have a few serious weaknesses for Telecine, like no gamma correction before conversion to 8 bit output for example.

 

Frank

Link to comment
Share on other sites

For details:

http://www.theimagingsource.com/en_US/products/cameras/usb-cmos-color/

 

Software control of all parameters.

 

Let me know what you think.

Hi Jean-Louis,

 

Ok, you mean their One4all CMOS cameras...I think it will be good for what it was designed for...a security camera ;)

Probably no real use for good quality film transfer, but I haven’t got one to try, so can only comment from CCD camera experience and the data published for the camera and sensor

 

If one could get access to the raw 10 bit data out of the sensor A/D, it might be possible to get something reasonable out of it, perhaps using binning to make up for the tiny pixel size. But as always with Imaging Source IMHO, they go too far with cost cutting and the biggest problem is not having a LUT, or at least some gamma adjustment, available before reduction to 8 bit output.

 

It is also CMOS (most wisdom says not as good as CCD for dynamic range) and a rolling shutter. Rolling shutter is ok for static frame capture, but no use for a continuous motion capture or for RGB light balance adjustment by changing the relative colour exposures.

 

Frank

Link to comment
Share on other sites

Hello Frank,

 

Ooops! yes, one4All. Didn't the Three Musketeers say: "All 4 one and one 4 all" ;-)

 

Thank you for your insight into these cameras. However, with all these limitations, I still marvel at how Freddy can get such amazing results using one....

 

Is there a camera you can recommend for the DIY telecine that fulfills the criteria you mentioned?

 

Cheers,

Jean-Louis

Link to comment
Share on other sites

I still marvel at how Freddy can get such amazing results using one....

 

1) I'm using a TIS CCD camera, not CMOS.

2) I'm using my Avisynth filmscript afterwards

3) After 5 years of testing, I have some experience by now

 

:)

 

But one day I will upgrade my system, and then I'm going to follow Franks advice.

 

Fred.

Link to comment
Share on other sites

Hi Jean-Louis,

 

However, with all these limitations, I still marvel at how Freddy can get such amazing results using one....

Well that prompts the question: what quality level are you looking for? Without doubt, Fred's results are plenty good enough for most people and for displaying on a TV set. They only start to fall down with close inspection at high resolutions and dark level details with contrasty films.

 

Is there a camera you can recommend for the DIY telecine that fulfills the criteria you mentioned?

Tricky one to answer! The usual rule of increasing quality costs increasing amounts of money applies, but usually the former goes up linearly and the later to a log law! You need to decide which parameters are important for you: like film format size(s) to be scanned, purpose of the scan (archive, show on TV in SD or HD), speed of scanning, how much money you are prepared to outlay, etc.

 

For me, it is as much a technical challenge to get the best out of the equipment I was prepared to buy. My IMI cost around US$2500 but fortunately the exchange rate to the UK£ at the time was significantly more favourable than it is now!

 

There are many other options in between the TIS priced models and my IMI, so "pays your money, takes your choice" ;)

 

Some tips: Resolution is not as important as pixel size as the biggest challenge in film transfer is dynamic range. For example, if you want HD1 from S8, then 1388x1024 2/3" CCD will do the job nicely enough.

 

Most machine camera manufacturers other than TIS usually have have a better set of in-camera features, but build quality can vary...even with the same manufacturer. For example, the Han and Rhein series of models from IMI would appear to be technically the same (on a sensor type basis) but the more compact design of the Han model causes it to have a much worse dark field performance due to internal electrical interference (clock noise I think).

 

Sorry, I've rambled on a bit here and not given you a clear answer...think you need to answer some of the basic questions regarding your own requirements first!

 

Frank

Link to comment
Share on other sites

Thanks for your detailed answer Frank.

 

My reasoning so far has been the following. You can correct me without hesitation if it is faulty.

 

I would like the S8 footage to be transferred to HD 1920x1080.

But I want to reserve until later the decision to display the full 1.33:1 image or to crop it down to 16:9.

So I would like the master to have at least enough horizontal resolution to encompass the full 1920 pixel width and the vertical of a 4:3 ratio which would be 1440 pixels height I believe. Something close to 2K in fact. Does this make any sense?

That's why I was looking at the TIS One4all CMOS cameras with resolutions of 2592x1944 or 2048x1536.

 

Thanks for mentioning IMI. I had not heard of this manufacturer until now.

I had been looking at AVT and the Pike camera but the cost is too high for experimentation.

This is why I am trying to gather all the information I can before deciding on a image capture system.

 

Also, thank you Freddy for your clarification.

 

Cheers,

Jean-Louis

Link to comment
Share on other sites

IMI are an actual manufacturer and I believe often OEM to other brands, like Unibrain. Point Grey and Basler are a couple of other good quality brands to consider.

 

I would not get too hung up about about resolution...if you truly want 1920x1080 original pixel data then you will need a sensor in the order of 3840x2160 plus allowance for cropping, because the bayer pattern on a single colour CCD only gives a true resolution of half area for green and quarter area each for red and blue. Ok, I'm being a bit pedantic, as a good debayer algorithm will make up a bit, so you will probably get an effective resolution of about 75% of the quoted sensor resolution.

 

There's also only so much information you will get out of an S8 frame and at the risk of starting yet another useless debate on the subject, HD1 (1280x960) if cropped is probably about it. A 1920x1080 with side borders to maintain the original format is also about right.

 

Take a look at the wedding image on the Results page of my website ( http://www.cine2digits.co.uk ). Click on the image to download a full res version of it. It is 16 bit straight from the camera (only 12bit of actual data) with no tweaking at all. Try up-res'ing it to 1920 width in Photoshop or the like and see what you think.

 

The big problem with higher resolution sensors is in order to get all the pixels in, they have to either make the silicon huge or the pixels smaller. The former makes it a very expensive sensor, the later kills the sensitivity and dynamic range....so unless you want to spend the rest of your life doing HDR transfers, best not to go down in pixel size ;)

 

Frank

Link to comment
Share on other sites

  • 8 months later...

Nice stuff.

Has anyone considered doing an infra red capture as well to get a 'mask track' for digital dust and scratch removal?

Most colour film stocks are 'invisible' to infra red, so the only thing you get on the capture is an image of the dust and scratches. Makes removing the defects easy whilst presercing the detail.

 

Does anyone have an easy to build, functioning wetgate?

Link to comment
Share on other sites

Has anyone had any experience with the Imaging Source All-in-One cameras for DIY telecine?

 

Cheers,

Jean-Louis

 

Unfortunately they are rubbish for telecine work, I have tried one personally and it was awful quality.

 

The DBK41 and the newer model of the DBK21 are okay, a little noisy and lower resolution, but give great results in practice.

 

IS cameras do cut some corners, and I have used the Pt Grey cameras with the same CCD as the DBK21 and the Pt Grey had less noise, and no artifacting at 60fps (the IS camera had edge ringing at 60fps but was okay at 30fps).

 

So they are good cameras, but not best of breed.

 

If I was going to use one, I would take some darks, and stack them and use that as a noise removal mask, it works well, but if you can afford it, there are better models with the same CCDs out there.

 

But for home use, they are very good.

Link to comment
Share on other sites

  • 3 months later...
  • Premium Member

Hi all,

 

This thread has been a good read..

 

Recently i picked up a 2nd hand jk optical printer with 16/35 gates, I've been looking at the vast array of machine cameras on the market, and it's quite a headache, specs are hard to find and prices are staggering... I feel this rabbit hole has no end.

 

I've thought about a continuous light source but noticed all the frame by frame scanners on the market use a strobe style light source..

 

Thoughts?

 

 

Cheers,

Sean

Link to comment
Share on other sites

Hi Sean,

 

If you can stretch to it, then the Point Grey Grasshopper 2 Firewire camera probably has the best performance/value trade-off for telecine....but depends on what you intend to scan from and your intended display platform.

 

It has a few hiccups, but they can be worked around. The biggest is the internal gamma correction is flawed, so need to use 16 bit output mode and apply gamma on the fly in the host if wanting to save to 8 bit avi, or just save as 16 bit and do all the work in post. It will do 26 to 27fps in external trigger mode at 16bit output.

 

Not much else around with that sensor without a big hike in price. There are cheaper and higher res sensors, but these have smaller pixels and hence less dynamic range capability.

 

On the lighting front, flashed makes sense as the exposure times are quite short compared to frame period times, so significantly less power required and no waste heat to get rid of. I have gone a stage further with my RGB system where I independently alter the duration of the R, G and B pulses per frame to adjust colour balance. My latest design has exposure times in the 10's of microseconds with the lens at f5.6, 16mm film and the Sony ICX285 sensor. The flashing means I'm generating the equivalent of a 300W of LED light, but still able to power it from a USB port at 30fps ;)

 

Frank

Edited by Frank Vine
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Forum Sponsors

Metropolis Post

New Pro Video - New and Used Equipment

Gamma Ray Digital Inc

Broadcast Solutions Inc

Visual Products

Film Gears

CINELEASE

BOKEH RENTALS

CineLab

Cinematography Books and Gear



×
×
  • Create New...