The new Muller HDS scanner, what do you think about that?

[RCBasher]

And anyone that decides to chair an open source project and be the central "go to" person better have a lot of spare time to answer questions and hold peoples' hands because, believe me, it never ends.

Ha, we just simultaneously posted pretty much the same friendly warning!
Frank

[Andreas Wideroe]

I don't agree.

There is a big difference in running a business and supporting your users who have paid money for your services and doing something because you are interested in the subject and therefore investing your time and energy into such a project. Who can demand support? It is open source and free! You could ask for support, but don't expect a rocket scientist to show up on your doorstep. Instead I think the users would group together and help each other out through the channels the project offers.

A good open source project would have a community that supports the users through a forum or similare and of course - many people would participate just out of idealism and pure interest. Not too different from the users of this forum. How many times haven't we explained how a Super8 camera works or how to repair something? I don't mind answering questions and helping out people, but I'm also glad I'm not the only one who can.

/Andreas

[RCBasher]

There is no such thing as “free” for this kind of complex project. I have been supplying my system to other amateurs effectively for free because I only charge pretty much cost plus a couple of beers. I have (and still am) thinking about knocking it on the head because answering tons of emails each day has become quite a chore. You are wrong to imply you don't have to answer because onece you have supplied something, even at cost, you become obliged to held. If I were running a business I would have to charge 10x the price to even start thinking about it being profitable. And don’t say some magic “community” will answer all the problems in a forum. Some will be covered, but for the hard bits it will always go back to the very few real contributors to the project who have the specialist knowledge and created the original designs.

The reality is there are complex parts which very, very few people have the skills or equipment to be able to build. Surface mounting miniature LEDs onto a high-density board is not for the feint-hearted. So who will do it and then ship out to anyone who asks for free. Get real! How may so-called open source projects have fizzled, gone out of date or gone commercial? There are reasons why!

But anyway, I wish you luck and look forward to seeing how you get on.

Frank

[vintagefilm]

Frank and Roger, you guys are right of course. It will be a difficult task to do any support for an open source machine. That's one reason I haven't really started yet. I am really still building the machine, and trying to figure out all the tech involved. It has been a lot to learn, as I am sure you both know. But, can I share a secret? It was the two of you, plus Fred and Andreas, and the projects of several others, that inspired me to actually do this. I am perhaps halfway there on having a machine that others may choose to duplicate. There is still a lot to do. The machine runs and makes nice pictures, but the operation of it is complicated, and the files need post processing to perfect them.

There is no such thing as “free” for this kind of complex project.

Boy, is this accurate! I have probably 15K in this project so far, and a lot of that was used to explore the wrong solutions. Learning Curve! So it is not for the faint of heart. I am not a manufacturer and I don't want to be one. But I am enjoying the quest. For me, mostly it is about the learning. And when I get time, about sharing what I have learned. Which is why I have chosen to be open about the designs. So eventually, when I have a bit of time I will begin to share the details of the scanner here and on my website. In the meantime, if anyone actually cares about that, I am happy to answer questions.

So to return this thread to the original topic, my comment about the Film Fabriek scanner is that it is a real value as it is currently priced, especially since it now has Frank's lighting system installed. The camera choices also add value. I would suggest the highest resolution camera you can. The recent 5K super8 tests of the Lasergraphic illustrate that more pixels are better, even if you think it might be overkill. And the restoration software based on Fred's work adds a lot of value, as well.

Thanks, guys.

[RCBasher]

I agree with most of what you say Grace, but excuse me if I remain more than a little sceptical of the benefit of a 5K scan of Super8. I haven't seen the 5K test results and more importantly any lower resolution scans they were compared with, however I am reminded of similar claims and comparisons in the TV industry of SD/HD1/HD2 where in most tests the lower resolution image was way below what the format was actually capable of, either deliberately or through poor quality testing.

I hope Fred does not mind me quoting him here, but he has been following this thread and just this evening wrote to me:

"Recently, someone has send me a Flashscan HD transfer.
He also has send me the original film, so I could compare.
Guess what? Our system is a LOT better.
OK, less saturation but a LOT more detail.
And saturation is easy to fix in post."

To clarify, "our system" is my lighting and capture system, Fred's PGR Flea2 camera (1384 x 1032) and a modified projector for the film transport. Now the interesting point is his emphasis on a lot more detail. I do not why this should be, but it proves my point that before making resolution comparison claims it is a good idea to be sure what the comparisons actually are!

Now I see from their website that the LaserGraphics ScanStation has a continuous transport system and an area scanner, just like the Muller. I like facts, so let's explore their 5K 30fps claim in detail. For a 4:3 format film frame, 5K horizontal resolution is 3750 vertical. The film is continuously moving, so to actually register different image data on each line of the sensor, we need to make sure the film does not move much more than a line (let's say 2 lines as probably a Bayer sensor anyway) during the exposure. 30fps means a full frame will pass the sensor each 33333us, but we need to "freeze" the frame with a fast exposure so that the film moves not much more than a line during the exposure. I make that 33333 / 3750 = 8.9us, 18.8us for two lines. Good if you can do it, but I'd love to see the diffused lighting system, lens and camera sensor which can capture a well exposed, high dynamic range film in that time range. As Jusetan has tested the ScanStation, maybe he can add to this data by telling us the actual typical exposure times? I would guess they are somewhat longer and hence the real image data on the film will be spread across a few lines, nullifying the potential resolution available. Horizontal resolution will hold up better of course but only with pure vertical object detail.

My lighting can give an empty gate exposure of around 10us (with f4 lens aperture and 0dB camera gain) but with some films, even 300us is not really enough, imagine how much vertical image blur that will give at 30fps! As far as I know, there are not many solutions to this problem. An intermittent film transport is the obvious, immense lighting is another and a rotating prism to track the film movement onto the stationary senor is another (as in super high speed cameras). Both of the later two options are beyond the amateur and I suspect any commercial cost-effective scanner. But throw enough money and technology at it then things start to become more possible I guess. I would like to know the technical details and numbers behind the ScanStation to back up the claims which justify a 5K camera at 30 fps!

Frank

[Andreas Wideroe]

Frank, let's meet in Amsterdam with VideoFred and the others who are interested. Grace and Roger - you are also invited! How awesome would that be? At IBC in September we can talk with the Lasergraphics engineers, MWA, Filmlight, Digital Vision, Blackmagic and others. I was there last year (and several times before that), I will go in september. Lovely in Holland at that time of the year. Would be awesome to have a "live" talk about all this. We can also hook up with Frank at Super8lab.nl.

Filmfabrik could also meet up. I had a meeting with Mr. Müller at IBC for about 5 (?) years ago when he was building version 1 of his Müller scanner. Mr. Müller unfortunatly passed away a couple of years later. Very nice and kind man btw.

http://www.ibc.org/ (exibition)

PS! About small format resolution, I believe there is more too it than you think Frank. ;)

Here are some quick samples from the Kinetta scanner and it confirms what I think and also hear from my friend and colleague Mike Waldie (UK) who has been engineering scanners since the 1970s.

http://www.kinetta.com/downloads.html
https://dl.dropboxusercontent.com/u/284 ... 202012.pdf

Grace - I admire your project and your scanning business. Keep it up! Same goes for Franks project, Videofred's project and Rogers machines. Great work from all of you - and also all the others who do interesting film/tech projects.

Andreas

[RCBasher]

Frank, let's meet in Amsterdam with VideoFred and the others who are interested.

Interesting thought but cannot commit to such at this stage as will be building a new house!

PS! About small format resolution, I believe there is more too it than you think Frank. ;)

No, I think you were missing my point about how to actually achieve the benefit of a 5K sensor when the film is moving at high speed.

Here are some quick samples from the Kinetta scanner and it confirms what I think and also hear from my friend and colleague Mike Waldie (UK) who has been engineering scanners since the 1970s.

http://www.kinetta.com/downloads.html
https://dl.dropboxusercontent.com/u/284 ... 202012.pdf

I have looked at these before but just looked at the pixel/small format document again as a reminder. There are two issues I see in that document, the scans are low dynamic range - proven by the histogram and suggest a low dynamic range sensor. The second is the lower resolution scans are crap! My scans on a 1920x1388 IXC674 sensor blow those shown in the document away. So I stand by my comment of making sure one is getting the best out of any format before assuming one has to go up in size to do better. But please be sure I am not against higher resolutions, I just want to see some return on the investment and larger file sizes! Finally, one has to consider what the end target is for viewing the scans - I have two 2560x1600 monitors but my TV is only 1920x1080 and to get the scans to the TV they will need to be compressed to MPEG. Compression does not like noise (which is what the grain will look like) so filtering will be required and bang, there goes all that extra detail - if has been truly captured!

Frank

[VideoFred]

A picture is worth a thousand words

Super-8 footage from my own private collection, made in 1967 on Agfa film. (yes, I was young once)


My transfer system is different from the Müller, it's just a modified Eumig projector, but I use the the same light source (Franks), the same capturing software (again, Franks) and a similar camera. So the end result is nearly the same.

Please notice that the original might look a litte 'flat' and low saturated, but all information is there and can be easily made visible in post.

Now, about xxxK scans: those files are very huge, taking lots of disk space and they are very slow to render in post.
I'm not against them, but I think we only need them for very special HQ footage, not for average 8mm and even not for 16mm.

Frank wrote to me: "it is not the number of pixels you have, it is the quality of the pixels!" and I agree with that.

Fred.

[RCBasher]

Hey Fred, I bet you would never have guessed at the tender age of 14 you would be digitising the film nearly 40 years later!!

Looks like you transferred at 16bit and applied gamma in C2D before saving out to 8 bit avi. This gives the "flat" look but is the way to do it if not saving to 16bit TIFFs and has allowed you to bring up good detail in the blacks.

Frank

[VideoFred]

Hey Fred, I bet you would never have guessed at the tender age of 14 you would be digitising the film nearly 40 years later!!

Nearly 50 years Frank! ;)
Actualy this film and similar other films have 'triggered' me to start this project, 8 years ago.

Looks like you transferred at 16bit and applied gamma in C2D before saving out to 8 bit avi. This gives the "flat" look but is the way to do it if not saving to 16bit TIFFs and has allowed you to bring up good detail in the blacks.

Frank

Yes, and another reason for the flat look is the following: as you can see at the frame counter, this scene is at the end of the film. Other scenes are less 'flat'. But to avoid constant manual gamma adjustment, I always capture with a bit higher gamma than needed. This way, gamma setting can be the same for the entire film. My latest script corrects this automatically (in RGB) for the entire film, as you can see on the histogram. The other side (255) from the histogram is already adjusted at capturing stage by your exellent auto-levels algorithm. No more blown out whites!

Fred.

[vintagefilm]

excuse me if I remain more than a little skeptical of the benefit of a 5K scan of Super8

It was Justin from Canada-Toronto, I think? who posted a link to a super8 transfer he had done on the LG5K. I expect it was a well shot piece of film, and I thought it looked pretty good, resolution-wise. I hope you will jump in Justin, and clarify what I am about to say. I believe that the 5K resolution is rated for a 35mm film frame, but when you scan 16, S8 or R8 the LG uses an "Area of Interest" to capture these smaller frames, as opposed to an optical zoom, or a different lens. So an S8 frame has resolution down in the HD/2K range at best. And the LG 2K resolution is down around 1600 pixels for 16mm and 800-1000 or less for S8. Maybe. If I am correct. Justin?

I like facts, so let's explore their 5K 30fps claim in detail.

I think you can adjust the LG speed to what ever, so you could acheive more effective resolution by running the film at 18 fro S8 fro example. But that does not change your facts Frank, about the Bayer loss and the motion blur.

PS! About small format resolution, I believe there is more too it than you think Frank.

I think the idea of using a 4K or 5K camera is really about oversampling, and mitigating the losses inherent in the motion blur and bayer losses. So if the optics can deliver the frame to the full resolution of the camera, I think the results will be better. How much? Who knows? 5%? 30%? And right now it probably doesn't matter too much for HD delivery. But I am getting asked if I can deliver 2K or 4K now. So I think the market will demand the pixels eventually, even if it doesn't really add much.

Frank, let's meet in Amsterdam with VideoFred and the others who are interested. Grace and Roger - you are also invited!

IBC? wow, that would be fun! I wish I could. NAB is just 5 hours away by car, and I haven't gone for years. The walking just kills me. But I hope as many of you that can go, will get together and explore what the manufacturers are really up to.

[Gargamel]

Hello everybody !

The thread is very interesting, but after Grace I suppose that "the idea of using a 4K or 5K camera is really about oversampling"...

The film gauge matters more (9.5, 16mm) or less (8, Super8).
Transferring my own old Super 8, through good projector and HDV/DV cam, then post-cleaning, stabilizing etc thanks to Fred or John Meyer scripts... I preferred keeping a DV mode; my tests of HDV gave a very, very minimal improvement in that size of film (with my own reels).
Even not so perfect as Fred's ones, my clips to burn familial DVDs are "honest".
https://vimeo.com/105469092
https://vimeo.com/81387519

Transferring is the first step; but post-treatment is very important too. Friendly, thank you again, Fred !

But OK, the next time I re-numerize all my footages, I'll use a frame by frame scanning ! ;-)

Andre.

[carllooper]

A picture is worth a thousand words

Like.

[supa8]

So how many bits is the muller HDS scanner?
8? 16?
Does the restoration software limit the depth to 8 bits altogether? or is it 8bits per channel?

When supplying files to a postproduction company for grading, surely it would be best to provide them with something higher than 8bit right?
Seems like a wasted to have uncompressed really high quality video, not being able to lose that information for the grading phase.

I can see Muller's suggestion of cameras, the Grasshopper3 has bit depths of 8, 12, 16 and 24bits, so why talk about 8bit?
http://www.filmfabriek.nl/usb3.0-cameras.html
http://www.filmfabriek.nl/cine2digits-software.html
About the software though, which seems like a necessity to use, it says Capture in 16bit (white balance in 16 bit). Output to 8bit
Not really liking the sound of that, as far as future proofing is concerned. Is it camera dependent, meaning if one eventually upgraded the camera on the machine to something new down the line, the possible 8bit constraint would disappear?
Or is it an issue with avisynth/Fred's Script/Film9 limitation to 8bit?
Thanks a lot for your feedback.

[RCBasher]

So how many bits is the muller HDS scanner?
8? 16?
Does the restoration software limit the depth to 8 bits altogether? or is it 8bits per channel?

When supplying files to a postproduction company for grading, surely it would be best to provide them with something higher than 8bit right?
Seems like a wasted to have uncompressed really high quality video, not being able to lose that information for the grading phase.

I can see Muller's suggestion of cameras, the Grasshopper3 has bit depths of 8, 12, 16 and 24bits, so why talk about 8bit?
http://www.filmfabriek.nl/usb3.0-cameras.html
http://www.filmfabriek.nl/cine2digits-software.html
About the software though, which seems like a necessity to use, it says Capture in 16bit (white balance in 16 bit). Output to 8bit
Not really liking the sound of that, as far as future proofing is concerned. Is it camera dependent, meaning if one eventually upgraded the camera on the machine to something new down the line, the possible 8bit constraint would disappear?
Or is it an issue with avisynth/Fred's Script/Film9 limitation to 8bit?
Thanks a lot for your feedback.

There are no limitations with the scanner, it is down to the camera and how the user wishes to take the output. AVI predetermines an 8bit container (unless using special CODECs like HDX from Grass Valley which can do 10 bit) and most users I know of prefer this method as it gets them to their final target easily. The purpose to scan in 16 bit mode even for 8 bit output is that Cine2Digits is able to apply accurate gamma on the fly before dropping to 8 bit, hence minimising banding when pulling details out of the blacks. Gamma can even be set differently for the R, G & B channels although this would normally only be done with badly faded films. It is also possible to apply some colour correction on the fly, invert colour negative and process to sRGB for example, and this is done in full bit depth mode before optionally dropping to 8 bit RGB (3 x 8 bit) output to AVI. This usually results in only a small amount of final grading for most purposes and being 8 bit at this point is less of a problem.

A 16 bit scan will usually be RAW data from a Bayer sensor and will contain image data to whatever bit depth the camera is capable of (and is set to by the user). So this will usually be 10, 12 or 14 bit for most cameras. For those wanting to grade in full bit depth mode, then Cine2Digits will output to 16bit TIFF file sequences which can then be brought into a linear editor and will be about as pure from the sensor as you will get, with just de-Bayer normally applied. This means the saved scans will be 3 x 8 bit RGB if within an AVI, or 3 x 16 bit RGB if saved to TIFF sequences. It is also possible to save in RAW mode if the user wishes to do their own de-Bayer at some later point in the process (or is using a mono camera for example), then the files will be 1 x 8 bit or 1 x 16 bit of course.

Avisynth only works in 8bit mode (3 x 8bit for RGB images) so scanning in full bit depth, applying de-Bayer in full bit depth, applying gamma in full bit depth then dropping to 8 bit for AVI output is a very good option for those wanting to post process with Avisynth.

There are no camera constraints with the software, providing they are USB 3 Vision compliant (or DCam compliant for Firewire) and have a global shutter, so upgrading is no real issue.

Hope the above helps to clarify.

Frank