I'm quite new to digital video, so this question may look stupid, anyway is it possible to use a SI or Sumix camera to setup a telecine device? Had anybody tried this before? with what results?

thanks.

(sorry for my english, I'm italian/english but the italian half is writing now...)

It is high possible to setup such devecie with SI or similar camera, and the problem will be the dynamic range of the sensor in such camera is too small. for some profession film scanner, the dynamic range is more than 70dB.


I think to get better result, you need the camera with larger sensor ( generally means better dynamic range).

for light source, some LED back light designed for machine vision maybe a good choice.

thank you for the reply, maybe I have to add some details to clear what I meant. I intend to use a projector in front of a cmos camera, speed it down in order to get the biggest resolution, uncompressed video. (SI makes a device that goes up to 6.6 mp - 6 to 500 fps - 61db dynamic range, and there are also 3.2 mp sensors with the same dynamic)

I thought this process would lead to an acceptable mb/s to my hard disk, and I guess it will be much better than putting a dv or hdv camera in front of the projector.
Then I will speed it up to 24fps again in the editing process. Will it work?

Also, can I sync the camera with the projector or the projector with the camera?

Anyway do you think there is any sensor better suited for this purpose? (I do mostly b&w shots, if that helps).

Thanks

david

forgot to say the film to be telecined is super16.

David,
Please have a look at this place:
http://www.moviestuff.tv/16mm_telecine.html
Combined with a capture program from www.bensoftware.com would make a very interesting setup.
I don't know if the software from that site will work with files higher than SD DV, but the projectors seems to be quite good (check with them about the conversion to a super 16mm gate).
Just for short takes I was wondering why not capture with a DSLR camera (good macro lens) at a resolution of your choice (only problem being a sustained rate of 6 fps and too many shutter actuations would harm the camera in the long run).

There are there plenty of programs used by stopmotion animators that can also be used.

The structure of this product
http://www.moviestuff.tv/sniper16_pro.html
is what I really want!

But it seems even with their most expensive model, you can only get the video with PAL or NTSC resolution. it may be enough for broadcast in TV system, but is really useless for editting

So I think, the best way is to buy one projector from Http://moviestuff.tv, then buy a suitable camera and lens with suitable resolution,you can build your own system much faster.

Almost all the camera for Machine Vision has trigger input. So I think the easier method for sync is to use the projector's output ot trigger the camera.

To my knowledge, the life of the shutter in Canon's DSLR is around 150K time, equal to about 2 hour of 24fps. :-(

Sorry for my Chinese style English (Chinglish!). I spend alomst 20 years to study such language, but still feel clumsy on it.

thank you both, valeriu & larry, this is exactly what I meant!

moviestuff show it works (even though they don't say much about their camera sensor, only "pre-owned, broadcast grade 3 chip camera heads". 1/3"? 2/3"? 1/2"? dv? hdv?....).

the only thing now is to check the maximum mb/s that a good computer can handle and choose the right sensor.

I think the device will be quite expensive at the end (I think at least 5k pc not included, of course) but consider telecine rates and the fact that you can do transfers for others and get some money back.

I believe this is worth the effort.

If the camera is a 3-CCD HD camera, they must tell you that, I think.

Normally, 3-CCD camera has better performance in color reproduction than single CCD camera. But I think since most CMOS sensor can provide an output of 10 bit, there will be engough space for post process to get better color performance.

Don't worry about the problem of data rate. The fps of both the projector and the camera is adjustable. a pixel rate 30-40M/s ( 8 bit) can be achieved without too much trouble in most computer.

Rob Lohman on "alternatives to SI and Sumix" thread says:

1920 x 1080 x 12 bits @ 48 fps in RGB = 569.53 MB/s (1 sec. for 1 GB)

so 24fps should be 284.77 mb/s, that means I guess 1 tb (!) per hour of footage.

assume I have 1920x1080x8BITS @ 24fps in GREYSCALE (I'm not sure you can do it, but in still photography with photoshop this means smaller files). How much that will be?

This is because even if I change the projector speed, at the end the amount of data will be the same, and I'd like to know what size the hard disk(s) should be.

The point of building my own telecine is to get the biggest images but still being able to handle it with my computer.

Do you think 8 bit will be enough for B&W? I can't see much difference in photoshop between 8 and 16bit pictures, but maybe this is because of the monitor or the scanner i use.

For any single-CCD/COMS, there is no difference between the datarate of color mode and the data rate of B/W mode.

So for 1 hour of footage, you really need 1 TB hard disk to store the Uncompressed RAW data.

10 bit and 12 bit is really useful for the post gamma correction and color balance, Rai said, such operation will cost 2 bit of depth. but I think, for telecine application, the 10 bit is not necessary absolutly: you can try the different configruation of gamma and color balance, that means, if you find the result is not perfect, you can just drop it, and rescan the film with another configuration. you need not to change the color , contrast and brightness of the finished video. But for the HD camera which is used in film making, the second shooting means much more trouble. that maybe the reason why Obin is sticking to 10 bit system.

With lossless compression, you can only get a little smaller file size. ( for lossless compression of Canon DSLR, the file size for 6.3Mega pixel 12 bit Raw image is about 6-7 mega Byte).

I think some extent lossy compress is also acceptable. for example, for a picture of 1280*1024, it is very hard the see the difference between of the 200KB lossy JPG file and the 1000KB lossless JPG file.

I think I have to find out which is the best way to (gently)compress data, otherwise I won't be able to edit such big files in fcp.

Have a look at this codec for lossless compression:
http://www.bitjazz.com/sheervideo/.
It is also platform independent and can be used for both stills frames or video files.
For serious colour correction and post-processing it's better to stay in an uncompressed environement or 16bit if you have saved your frames as TIFFs especially if you perform secondary colour corrections.
ColorFinesse plug-in for AfterEffects is a great program for color correction. Give it a test.

<<<-- Do you think 8 bit will be enough for B&W? I can't see much difference in photoshop between 8 and 16bit pictures, but maybe this is because of the monitor or the scanner i use. -->>>

Perform some changes to your files (tonal, gamma, curves...) in both 8 bit and 16 bit. Compare the histograms, especially the 16 bit version after is converted back to 8 bit.

Valeriu.

Off-topic. I've received and tested your opalglass as a fixed gg. - Not so good for light transmission or sharpness because the opaque layer is too thick. On the plus side it seems to have nice colour and it does not seem to have grain of any consequence. An email with captioned .jpg images is on its way.

this is brilliant, valeriu! Yesterday I spent some time looking for some kind of lossless compression, and I read about SPIHT
http://www.siliconimaging.com/SPIHT.htm
I think it's quite good, only problem I'm not sure final cut can handle it.

BITJAZZ seems much better, since it is FCP compatible, and they claim image won't be degraded at all. (actually I think it will, but if it's not noticeable that's fine for me)

for the sensor the best I found is the SI 3170 rgb (2056x1544, 12bit, 66db dynamic range), but if someone knows of a better sensor at a reasonable price just let me know.

it seems that the sensor in 3170 in already 3 years old, but the sensor in 3300 is very new, so I belivey 3300 or other camera based on Micron 's sensor will produce better image quality.

Pixlelink have some interesting cameras:
http://www.pixelink.com/products_info.asp?id=48
http://www.pixelink.com/products_info.asp?id=12
The pixel pitch is small (3.5microns), so... more noise...
I like their interface and in software apart from LUTs you can also assign kneee points where the exposure time can be increased/decreased according to a luminace threshold.

David and all,
It seems like the group is answering your questions very well. Ideally, if your projection like is color balanced (full spectrum) then you don't want to do much correction when you digitize assuming that the final film is already shot with the effects that are intended.

The data rate numbers are easy for storage - the image H * W * fps (either 24 or 30) * #bytes. 8 bit data is one byte, 10 or 12 depends on whether there is packing - two bytes with no packing 1.5bytes with packing.

The 3170 is not what you want - the 3300 is a good low cost solution, especially when frame rate is not important. You want low noise. There will be a new version of the SI-6600 out in about a month that will be lower noise - that is great for this application since there is quite a bit of resolution to mess with (2110x3002) although the aspect ratio is a bit unusual.

The scoop from other threads is that if you are going to do gamma correction, you probably want at least 10 bits prior to processing even if you end up at 24 bit color before YUV conversion.

Please ignore SPIHT for this application. Cineform's compression might be very useful - they understand appropriate losses for this application.