New Cintel Scanner from BMD

mike burton Apr 7, 2014

  1. Thanks for sharing that story Marc. I love hearing about the way things were done in the past because it gives a great context to the present state of things. We sure do have it easy today with our unlimited windows and defocus ;).
  2. Yep. There was a time when each defocus window was about $50K, as I recall. Now, it's not even 50ยข.
    Robert Houllahan likes this.
  3. Hi all,

    Nice to be a part of this forum, you always come across here when googling for the tasty bits of grading knowledge. I've had the pleasure of working with this machine with edited camera negative, inter positive and print positive material. Some of it was strongly discoloured. Having worked with the Ursa (and a Renaissance 888), Muller and Arri on restoration, it was a little disappointing. I also believe it can (and will) be better with this hardware.

    I'm not entirely sure, but I think the optics of this thing are virtually identical to the production camera 4k. Mainly because of one big issue: Pattern noise. It's there, and it's bad. But it's bad because of the next point, which might save the day.

    The scanning interface is pretty rough and my biggest issue is that it has no reliable min-max output. As you may know it has something it calls "CRI" files, which I presume stands for Cintel Raw Image. Yet you set the LED diffuse light with a LUT, being positive, inter negative, inter positive or negative. The LUT does a lot with the levels and you end up setting the lights very much too dark if you adhere to the scopes during setup. Sounds to me like something BMD will fix, I've alerted them of the problem. If I'm going to work in RAW I obviously want to see RAW levels.

    Something more problematic is that the stabilisation isn't great, it's still shaking. It places a window automatically, but for some reason it can still shift. Not sure why that happens. It's also unwound (gently however) during frame forward and backward operation or slow shuttling, which works by setting the slider to the lowest point and pushing fast forward or rewind, then using a slider to change the speed. Something easy to change, just not optimal yet.

    Maybe my biggest beef after the levels is that it captures the audio after scanning (and manually telling it to) at a very slow rate. Each image is 22 megs and it shovels through it to stitch the audio. It's very inefficient. A thought would be rotating the saved image 90 degrees so the audio strip is on top and you could theoretically stop loading after getting the audio strip. I also wouldn't mind if they would figure out lossless RAW.

    So no, it doesn't do DNG, it doesn't do ProRes and it has faily poor level setup. Knowing BMD these things might come (to those who wait). I'm not sure if and how much processing power they put in the scanner. These guys love FPGA and they could add functionality if they put in time so on-board ProRes, maybe an on-screen level check on the HDMI out isn't out of the question and maybe, just maybe, real time audio extraction without the need for an optical head.

    I've got some material I'd like to show you but the fact is there's just been a fire in my building and my apartment isn't exactly peachy. My office is a darker shade of grey and I'm not powering up my workstation until I've thoroughly cleaned it. I'm just glad my family and I weren't home and my computers weren't rendering anything but were turned off for a change.
  4. Wow - this is the first review I've seen of this scanner. Sounds like it's still missing a lot of finishing touches!

    Sorry to hear about the fire in your place. I'm looking forward to seeing some samples when you can upload them, though!

  5. Interesting, I figured it was the same poor sensor from the 4K BM Cameras, I think the advertising that the machine would do realtime 30fps scanning was a stretch. I know the fast GPUs on other machines which do machine vision perf stabilization sort of max out around 15-20fps due to GPU processing power and I/O.

    I am not sure what you mean by setting the light with a LUT, on most newer LED Illumination based scanners you can set the RGB mix and intensity. On our Xena machines we can set the RGB illumination duration and the current on each RGB channel.

    All CMOS based sensor scanners can have fixed pattern noise issues but the ability to have fine control over the LED lamp plus the ability to do secondary LUT or CC curves allows for keeping the density of the film out of the exposure range that the sensor works worst in. Although the BMD 4K sensor may be much worse than the sensors used in other scanners.
  6. Jason Myres

    Jason Myres Moderator

    Very cool to finally hear a report from someone who's actually used the BMD scanner.
  7. Sorry to hear about your fire Roel. Glad to hear you are safe.

  8. I don't think it's that much of a stretch, actually. While we don't currently have the speed upgrade for our ScanStation, with it you can scan 4k at 30fps and 2k at 60fps. In the case of the ScanStation the raw sensor data is converted in the GPU and written to whatever format you want, so your speed limitation becomes I/O, if you're writing to formats like DPX. However, the Cintel scanner doesn't do that conversion on the fly, it just dumps the raw sensor data to the hard drive, and you'd need to make DPX from that in an export pass from Resolve. At least, that's how they presented it at NAB when it first came out, so I assume it still works that way. That raw data is relatively lightweight compared to DPX, so you shouldn't need anything special in terms of disk arrays, at least for the actual capture. If the sensor can process at those speeds, I don't see any reason why it would be a problem for it to run at 30fps.

  9. I think it can do 30FPS, but the frames are quite big, about as big as an 8 bit dpx (i know, it's not a fair comparison). So you definately do need special storage. I've tested it along with a G-Speed Studio XL over thunderbolt and found it to be good enough. I mean it should be, at roughly 7000 euros, but one assumes you're not skimping out on the workstation when you get something like this. You can probably get something for less, but frankly, I wouldn't go too low. You don't want to run into speed troubles and it's not just the capturing that takes time with this machine. At least the Ursa goes straight to tape or whatever capture machine you use. You can capture faster or slower though, so if you really want to capture to your USB drive, I'm pretty sure you can set it to 3 or 4 fps. Just be very freaking patient.

    There are workflows thinkable where you capture everything to an internal 1TB drive on a mac pro and then render out to something more tasty, but that means you're stopping your scanner for every export. Sometimes you want to scan a lot and then render a batch, sometimes you have the time to do a take by take. This thing makes you think about workflow way ahead of time.

    Thanks for the support by the way, we're still getting everything cleaned up. Everything is coated with ash and soot, including my workstation. But it boots, so once the gruntwork is done, I'll get the upload going. I think I might have lost a videocard though, I had to disable the secondary card to get it going, so anyone willing to donate a Titan X or Fury X will be my bestest friend.
    Jason Myres likes this.
  10. I may have overquoted the G-Speed, that price is for the 64TB model, it starts at about $3600 for 24TB.
  11. Very true, but even CCD cameras have FPN if you push them far enough. On the Scanity, if you are dealing with very dense prints like Vision Premier, you will notice artifacts in the form of vertical stripes in the blacks. When this happens, I do a two-pass transfer, cranking the LED illumination to the maximum then combining the passes in the Nucoda.

  12. That is not fixed pattern noise, a linear CCD can't have fixed pattern noise. What you are seeing, most likely, is a glaring inherent flaw in how line scan scanners work where any blooming, dirt or optical flaws cause linear streaks in the scan. The Scanity probably has to apply front end gain at the CCD in order to get the speed (25fps for 2k and 15fps for 4K) this is an inherent trade off of system noise vs. scan speed.

    Scanners like our 5K Pin-Registered Xena and the Lasergraphics Director and the Arriscan (although the sensor in the Arriscan is poor by today's standards) allow for multi flash HDR in the scan. That you have to scan the film twice on the Scannity and then combine those passes in Nucoda is a very poor way to overcome the inherent flaw-limitation in the scanner.

    Our 5K Pin-Registered Xena has a 5K 7.4u 14-bit Area CCD which is commonly used in Astro-photography and when run in single tap at 0 gain it is noiseless for film scanning, the exposure times would have to be in minutes in order to approach the noise floor of the sensor. It is slow at about 2sec/frame which is the trade off for flawless scans.
    Jeff Kreines likes this.
  13. Then I stand corrected with regards to FPN on CCD; I guess I haven't done my homework. As to how the Scanity reaches its speed/sensitivity, this is not with analog gain but rather with a technique called time delay integration, whereby the line is actually 96 lines and the charge is moved along them as the film moves continuously, multiplying the exposure time (roughly) by that amount.

    I will be the first to acknowledge the downsides of line array scanners, but I would argue that in the Scanity the signal-to-noise has been implemented very elegantly.
    Robert Houllahan likes this.
  14. The Scannity is an impressive piece of kit, it's just got some of the tradeoffs that are inherent in scanning film at high speed. That you have to scan film twice to do a HDR scan and fix noise is a big drawback IMO.
  15. Sure, that's not very convenient, however the Cintel doesn't do that at all. Although if you're very good at stabilizing i'm sure you could try the HDR trick. Might really make a big difference. Another drawback of the Scanity in comparison to the BMD is sheer size. It's huge! Maybe not for historical telecine standards, but still. I'm not trying to put down the Scanity however. It's a nice machine and at this point I'd much rather work on that than the Cintel. But I'm not entirely sure if the Cintel can't get up there with proper updates. High framerates don't get me warm, most of the time is lost in processing the material anyway and you can do that on seperate workstations with both machines, if you're so inclined.
  16. The BMD Cintel machine has a cheap Bayer CMOS sensor with known fixed pattern noise issues and sprocket drive, it also seems that they do not have stabilization working at this time. The Cintel will never be a true RGB scan and IMO you would never want to do a high end DI with it, it is a scanner made for TV shows and is more like a new style Telecine than it is a real data scanner.

    I don't think any real comparison can be made to the Scanity which is a very expensive and advanced machine.

  17. Because so many TV shows are shot on film these days?

    It seems to me that any new scanner being made today is being made primarily for archival purposes, not for new production. And certainly not for TV, unless you're talking about existing material.
  18. This is anything but an archival scanner. For one thing, sprockets...

  19. Archival doesn't always mean transferring from ancient paper prints from the 1920s. I was referring to existing material rather than newly shot material, because the use of film in television production today is practically nil worldwide. Shows that were shot and finished in the 1980s, 1990s, and 2000s likely have their sprocket holes perfectly intact, and benefit from better registration during the scanning process.
  20. We've scanned film from the 80s with 2+% shrinkage. That's a lot of shrinkage. It probably wouldn't make it through the Cintel.

    It's not just age, it's storage conditions, and a lot of film was not stored properly. BMD is positioning this scanner as archival in their marketing material, which I think is disingenuous, at best.

Share This Page