I want to give my clients and myself the best possible results every time. DSC Labs has helped me to do just that--shoot after shoot, year after year.
Modern t.v. cameras employ complex digital circuits--a virtual "digital emulsion". In order to optimize this emulsion, its imperative that I can examine the performance of these circuits during different "real world" conditions. DSC's test products allow me to do just that. In fact, the DSC product is, in my opinion, the only product on the market with enough built in technology and precision to match the needs of complex digital cameras. When critical adjustment of HD cameras are necessary, the DSC product is the only choice.
I use both the back lit Ambi, and the front lit Camalign for everything from flare compensation, gamma crossover setting, accurate white shading, color matrix (SMPTE, EBU, 709) installation, detail circuit balancing, etc... In particular, the neutral characteristics of the DSC charts allow for reliable carrier leakage elimination resulting in perfect scene reproduction from the blackest black, to the whitest white. Try doing that with an old paper chart. Tired of crushed shadows? Try the DSC black--everything else is gray by comparison.
For multi-cam shoots there is no comparison. If you can accurately measure and record various camera parameters you can then just as easily install these parameters in another camera. Kind of like taking the emulsion from one camera and duplicating it in another. Perfect, simple, and reliable every time.
Inspite of 12bit A/D conversion and effective compression/ expansion circuits for the knee and toe, is the contrast range of the f900's emulsion much lower than Sony's claim of 10-11 stops? Even if the block is able to output this much information, it's the output to the recording device which is key.
Testing reveals 7-71/2 stops safely. This figure takes into consideration the non-linear regions of the durve. With white appearing at 100IRE, clipping occuring at 105-108IRE, pedestal at 2-4IRE, then the linear portion of the curve will safely contain 5 stops of scene contrast.
With compression circuits such as pre-knee and knee, and expansion circuits like black gamma dealing with the non-linear parts of the curve, an extra 2- 21/2 stops (approx) are injected into the emulsion (~1-11/2 in shadow & ~1 in highlight). This gives us 7-71/2 stops of contrast total.
Keep in mind that the test scene will contain an even distribution of information throughout the contrast range and the gamma crossover point will alllow for an even distribution of signal information. In addition, the pre-knee, knee, and black gamma characteristics will be optimized for the scene. Also, specifications are limited to shadows and highlights having visible detail--not crushed or blown out.
Is Sony wrong? No, not exactly. Are they overly optimistic? Certainly.
Please consider that these results are "real time"--this is what you can expect to see on set via your picture or signal monitor. Also, if the scene is not average--if it's high or low key--then does the camera still have the ability to provide 7-71/2 stops of effective contrast capability?
So, the question becomes, how can a DP using a f900 effectively deal with scene contast in excess of 71/2 stops--10-11 stops?
Some have said that the scene must be changed. The scene contrast must be altered to match the capability of the f900's emulsion. If the scene is changed then it becomes a different scene altogether. If the shadows become more dense, or if the highlights are more blown, than the feeling and emotions the scene evokes in the viewer are different. It's no longer the same picture.
Finishing the picture in camera, on set, has its merits. However, if maximum quailty is to be extracted from the f900's digital emulsion a different approach must be taken.
Tape-to-tape colour correction is the solution.
The practice of capturing the pictures on set with minimal emulsion manipulation, yet at the same time being careful to target precisely the emulsions "sweet spot" and then subjecting the recorded images to a tape-to-tape colour correction in a safe environment via an advanced and powerful colour corrector will accomplish much more than any method of finishing in camera. Unfortunately, colour correction--the practice of finishing in post--while manditory in film production, has never caught on in digital acquisition. Many HDcam DP's are still relying on the traditional television/'Betacam approach--if it looks good on the monitor it's time to roll tape. This is a good approach-- to a point. However, those same images will most likely never benefit from a finish in post.
Subjecting digitaly recorded images (digibeta, and especially HDcam) to a finishing step in a controlled post environment via digital component colour correction has many advantages.
The viewing conditions can be adjusted to ensure the colourist has a proper, and constant, visual reference. Ideally, no ambient light whatsoever will fall on the monitor. In addition, a neutral middle grey (18%ref) will surround the monitor and offer to the colouist a neutral visual reference. Alternatively, as used by the CBC in their transfer suites in Toronto, Canada, a "light surround" will be employed. A translucent plexiglass frame (5-6" wide) surrounds the monitor. The frame is back lit by a source of constant luminance, hue, and chroma. The variables of the light source can be set to match the characteristics of the real world. These conditions are quite specific and will very rarely, if ever, be found on location. Making decisions on set viewing a non reference grade monitor in less than ideal viewing conditions could lead to irrepairable damage--even if the material is subjected to a tape-to-tape correction.
The ability of modern, state-of -the-art, colour correctors are truly astonishing. The top machines are very powerful and offer everything a camera does--plus more. Top systems employ signal conversion and processing which is more advanced than even the best in-camera circuits.
Specifically, and due to the impressive S/N specs of modern digital formats (especially HDcam), shadow information can be recovered with ease--provided the shadows aren't crushed to begin with. What was once a lifeless shadow is now full of information--clean information. Most impressive is how highlights can benefit from post processing. Finishing the pictures "in-camera" will force the DP to rely on the very limited overexposure capability of the digital emulsion. As discussed earlier, the f900 will allow for approximately 1 stop of overexposure as a result of the cameras compression circuits. If, however, the DP knows that a full tape-to-tape has been budgeted for, then he or she can take full advantage. By underexposing by 1/4-11/2 stop, the DP has now installed a safety factor into their HDcam images. The highlights will now, most likely, avoid the clipper--and in turn avoid being blown-out. The luminance value required by the scene can easily be achieved in post. In addition, and because of the lack of noise in the HDcam signal, the restored scene is noise free. The scene now contains detailed highlight information in a much greater quantity then ever before.
Let's talk numbers. Shadows and highlights can each gain 1 additional stop. This adds 2 stops to the f900's contrast capability, giving a total range of 9-91/2 stops. Impressive!. Additionally, an extra 1 stop can probably be squeezed from the system when required. 1/4 -1/2 stop from the highlights, and 1/2-3/4 stop from the shadows. This brings the count to somewhere between 10-101/2 stops. Very impressive. Please be aware that the shadows still have more flexibility than the highlights. In otherwords, and unlike film emulsion, digital emulsion appreciates underexposure and suffers to tolerate overexposure. It is the noise free nature of the system which allows for manipulation in post without bad side effects. So effectively, you can now have images with a very wide contrast range. The results are similar to film, however, the approach is quite different--even different than conventional television.
While finishing in camera can produce some very fine images, the full potential of HDcam can only be achieved via an unconventional method.
The f900, itself, offers good contrast characteristics. The HDcam system, HDcam acquisition and HDcam colour correction, will offer much more--the ability to safely handle a very wide range of contrast through impressive exposure lattitude.
The topic of exposure lattitude is confusing. Traditionally, exposure lattitude refers to the emulsions ability to be over or under exposed and still produce shadows and highlights which are noise free and exhibit good detail. However, this process relies on the fact that film negatives are always put through a transfer process which includes a colour correction step. This has never been the case with electronic acquisition.
Therefore, film emulsion exhibits excellent exposure lattitude because it relies heavily on a two step process. Without this process is the film negative still as impressive? Well, yes.
Film emulsion has the wonderful ability to gently, seamlessly, fit shadow and highlight information into a small area--the toe and knee. The f900 tries to emulate this behavior by using impressive black gamma and pre-knee/knee circuits. However, the effect is not the same. Even in the 12bit f900 camera, with complex support circuits, the transitions in the toe and knee are not always gentle. If a sawtooth is fed through the system, the result, as displayed on a picture or signal monitor, will be a somewhat coarsely stepped curve. This curve is much smoother than it has been in the past (10 bit), but not as refined as film emulsion. Sony is in the process of developing variable shaped gamma transfer curves to help help solve this problem. By using a multi point look-up table to "smoothly" define the shape of the curve, the resulting shadow and highlight characteristics can more closely resemble film emulsion. These gamma curves will most likely be used in conjunction with the existing shadow and highlight circuits to help fine tune the shape of the curve.
This is especially helpful when transferring HDcam images to film--the DP can now match more closely the gamma curves of the digital emulsion and the transfer stock.
Can it be said that exposure lattitude doesn't realy exist or that it is a latent characteristic? Yes. Exposure lattitiude only becomes fully developed via the colour correction process. In otherwords, for digital emulsion in particular, exposure lattitude by itself is non existent--or a non issue. It is the end result that matters, right? So, if the f900 and the HDcam system can, via a two step process, emulate film emulsions grey scale reproduction characteristics and offer the DP digital images which contain a very wide range of contrast, then who will object? Many people.
Film DP's, hungry to learn, and with little knowledge of digital acquisition, will do well in the world of HDcam. Conversely, television and video DP's who have established set methods over the years will find that, without a doubt, their HDcam images will not be as impressive. Bad habits are hard to break. Taking a Betacam approach to HDcam is false economy. If a production, determined to use HDcam, has not budgeted for a tape-to-tape transfer, they are fooling themselves. Producers and DP's who traditionally work with film should not cut corners by adopting a"video" method to production. They should continue to finish their images in post. Their video counterparts will gain much by following in their footsteps.
Thanks for the information. I was wondering though, when I look at images made from medium format digital still cameras and camera backs such as the Better Light 6000, Phase One, Foveon, etc. they seem much more impressive than HDcam, especially in the colorimetry and lattitude (this is without post-image manipulation. An example of this can be seen at www.luminous-landscape.com in the comparison between the Canon EOS-D30 and Fuji Provia100-F). I was wondering if this is a result of the YUV 4:2:2 color space of HDcam versus the RGB 4:4:4 color space of digital still cameras. Are there any plans from Sony or whoever to make a camera that captures and records in 14-bit or 16-bit YUV or RGB 4:4:4 colorspace and not throw away half of the color information like YUV 4:2:2 does?
Post Production Artist
Virginia Beach, VA
At the recent SMPTE meeting Lockheed Martin and Arriflex discussed the development of their electronic camera technology. They are currently nearly two years into a seven year project.
Some of the issues discussed were the failure of CMOS technology when it involved motion capture for anything else than still images. Apparently for motion the problem was with aliasing artifacts.
The presenter did state that the recording medium will not be videotape but rather hard drives which are currently in development.
The presentation was an inside look at what the future will hold for the next generation of Electronic Cinematography. However it will be awhile before the camera is ready for mass use.
I believe Sony HD is typically a 12 bit processed signal rescaled to a 10bit linear output Y'IQ colour space with 4:2:2 sampling. The 4:2:2 definitely contributes to a reduction in the effective Colour Space of any signal. The difference between Y'UV and Y'IQ Colour Spaces is subtle in the form of a 33 degree rotation and an exchange of colour difference axes. That is how I have come to understand it and will be happy to be corrected by the Truth Police :)
National Manager, Digital Systems
According to my information HDCAM uses 4:2:0
Hold on to your seat: The Sony camera can - with an optionally available adapter - indeed output a 1920*1080 10 bit component digital 4:2:2 HDSDI signal.
However, the HDCAM format recorder reduces this not only to 1440*1080 (75%) for the Y signal, but also reduces color info to 4:2:0 before compression. (Color data are recorded only for one line out of two, data rate comparable to 4:1:1) This reduces the resolution of the color difference signals to approx. 720*540 pixels per color channel (Less than a fifth from what the camera produces, 18.75%). Only after those reductions, (-25% for luminance, -81.25% for chrominance) the (supposedly 4 times) lossy compression starts.
The eye is amazingly forgiving for this lack of color resolution, but this is important to be aware of it for two reasons: Post production and saturated colors
- If you use very strong colored light like a strong red overcast, or very saturated subjects, resolution will probably visibly reduce.
- Don't expect to get high quality blue or green screen results from HDCAM recordings (same is true for DV and DVCAM, also a 4:2:0 / 4:1:1 format)
- Doing strong color changes in post can significantly reduce detail. By changing color balance, part of the color difference signals end up in the luminance channel en vice-versa, reducing resolution in the Y channel.
Actually with formats like DV and HDCAM, we get confrontred again with some of the problems most of us were familiar with in NTSC and PAL recordings.
Those also have very strong reduced bandwidth for the color signals. (Don't get me wrong: They are still much better than NTSC/PAL but it goes in that direction)
D5HD uses 4:2:2 (color resolution is only reduced horizontally, not vertically)
Seen like this, I look at HDCAM as the DVCAM format in the HD world,
I look at D5HD as if it were the DigiBeta equivalent in the HD world, and see D6 as if it were the D1 from the HD world. D5/D6 records the full signal, but then the signal itself is still 4:2:2.
All 4:2:2 systems were designed for TV broadcasts. Not for hi-quality film work.
All digital film scanners, and most equipment used in high end digital film post production (scanners, recorders, workstations) work in the full 4:4:4 and non-compressed domain. Including the digital animation camera I participated in developing in 1996. We actually started off with a modified high quality professional digital still camera, like you suggest.
class="Body" >Are there any plans from Sony or whoever to make a camera that captures and >records in 14-bit or 16-bit YUV or RGB 4:4:4 colorspace and not throw away half of >the color information like YUV 4:2:2 does?
Yes, there are several. But, as of my knowledge, none of them are anywhere close to completion.
Is this the 4K resolution Lockheed/Arri camera that I've been hearing rumors about? I've been trying to dig up some information on this camera on the internet with no avail. Do you know of any place that might have some more information on this development? Gosh though, 5 years left! I guess it may seem like a long time now, but I'm sure the time will pass quickly.
It will be very interesting to see what they come up with. One of the primary reasons that 10 bit depth is used is shear data volume. Each step up is a doubling of the data to be compressed and stored! I can't imagine how much data a 4k x 4k chip would give out at RGB or 4:4:4 at a 16 bit depth! The 10bit SDI out on the Sony Camera is 1.485Gbits a second uncompressed! (except for the 4:2:2 sampling) Then again, at what point do you "see" or not "see" this additional resolution and on what Visual
Medium is it and the full RGB Colour Space fully and honestly presented?
National Manager, Digital Systems
You spoke about HDCAM & chromakey. I tend to agree with your observations. However, what is interesting is Panavisions approach to this issue.
In spite of bandwidth/information reduction at the vtr, Panavision has, I believe, made some impressive improvements which increase blue/green screen performance.
Via custom patented modifications, Panavision HD cameras enjoy having their CCD imagers receive more precise R,G,B light. This will serve to enhance the ability of Panavisions camera to perform very well with blue/green screen.
It's not just the lenses which are different--the whole camera is different
Interesting. Is there anybody from Pana listening who could give us more details about this?
But even if Panavision enhances the color response of the camera, it will be hard to obtain more that 720*540 (roughly equivalent to an SDTV luminance signal) resolution for your key mask, if the VCR does not record more.
I would still highly recommend connecting a D5HD (or better) recorder to the camera and record on that if possible. A majority of blue/green screen shoots happen in studio, and the recorder can be linked to the camera with a simple BNC coax cable. (Even better than the three that go to the monitor now). It would double the resolution of the mask.
Don't forget that the key mask generally ends up for 100% in the luminance channel on the composite shot.
VFX Cinematographer Brussels, Belgium, Europe
Unless this a new development, at NAB this year both SONY and PANAVISION told me that the CCD imagers and block were exactly the same from the PV model to the commercially available Sony F900 camera.
Only the bridge support, the extension viewfinder and the PV modified mount and consequencial displacement of the front electronic package were different from the commercial model.
Obviously the engineering set up menu can be taylored to better handle key situations.
As far as I can tell, the F900 doesn't have any digital connections to the outside world, only the 3-wire analog RGB connectors to a monitor. Neither does the Panavision version. Yet another move by Sony that one can consider either clever or diabolical and self serving, depending upon how you look at it.
Maybe Sean Fairburn can correct me (we were both at Panavision yesterday) but Nolan Murdoch told us that a dockable version (separate deck) would be coming out next year maybe. Maybe I misunderstood him. Sean?
Cinematographer / L.A.
I have seen a HDSDI adapter that clips onto the rear of the F900 which allows outputting to an uncompressed system or another vtr.
This same unit was available for the 700A camera.
The HDSDI unit was on display at Panavision. Panavision showed some film footage that compared green screening compositing from HDCAM versus a uncompressed recording. You should see them yourself to view any difference.
I'd be very interested in seeing that. According to what I've been told, the Star Wars shoot was recorded on the DVCam format, which, if true, is interesting if the alternative was available. BTW, for those interested, there is a day long 24P production symposium taking place this Thursday (11/16) at the L.A. Film School, sponsored by various vendors (Sony, Avid, Panavision, Bexel, et. al.) and Digital Cinema and Videography magazines.
I'll report back anything new or interesting.
Mike I think your were referring to 24psf HDCAM not DVCAM which is Sony's 4:1:1 25mbps SDTV format. I heard that Star Wars was shot using the onboard recorder and an external HDCAM recorder to act as a backup dual record system. If this is correct then the HDSDI adapter could have been used to feed the vtr. I can't confirm this so ask Panavision to verify this story.
Contact Panavision in Woodland hills to see the film footage. It was shown at their open house. It will give you a good insight into what is possible as far as compositing is concerned.
As a side note the Panasonic HD camera's come standard with HDSDI outputs from the camera.
So how did that go?
Also, did anyone see part 1 of An American Tragedy (the OJ Movie, with a brief appearance by CML's own Bob Tur)? My DSS was down and I saw it on really noisy cable (the CBS affiliate's cable channe is the same as their transmitter channel, noise bars abound) but I thought there was something strange about the look (shot by the fine Prince of Darkness Bruce Surtees). Obviously there was material shot on video meant to look like video, but the rest of the show looked like it might not have been film.
(But remember I was sitting way off axis, 13 feet away, with a really noisy signal.)
While I thought Executioner's Song was an amazing book, and a great collaboration between Larry Schiller and Norman Mailer, I'm not so sure of Mr. Mailer's ear for dialog. (He's buggin'....?)
Jeff "watching television with all the attention it merits*" Kreines
*good days for news junkies like myself, though
Duhhh, sorry, I'm awake now. That was, of course, a typo. I meant HDCam.
I believe I did get that information from Nolan about a month ago when we shot a test with the Panavision camera on "Ally McBeal." My point was that the improvement on the recorder end would only exist if the recorder was something other than an HDCam. Say, D5 or possibly a Voodoo (Philips' uncompressed recorder). I'm assuming, of course, that the HDSDI adapter outputs a full bandwidth HD stream, which, if it is taken directly from the camera, it theoretically should. Needless to say, an external recorder would also require cabling (limiting its use if you're using Steadicam, for instance) and, in all probability, another crew member. And would that crew member be Local 600 or 695? Hmmmmm......
I would venture to say that unless Panasonic can come up with a manufacturing/marketing partner well known to the film industry (Arri, for instance, or possibly Aaton) it is very unlikely that they would gain any real foothold in "Hollywood" based production. Only my opinion, of course.
Please tell us what you saw, since many of us will not be able to see this footage. Was the difference noticable? Were the tests interesting and done in critical situations? What were your impressions on the results?
Thank for your comments,
class="Body" Kommer Kleijn
VFX Cinematographer Brussels, Belgium, Europe
Tony and David are correct the adaptor to the f-900 is the HD-SDI camera adapter HDCA-901 and we used it on Nicolas and is a great device because it not only carries a clean signal but also all the audio and timecode down one cable. and Yes I reccomend if youir on stage taking that signal straight to a D5 Deck or better yet straight to the HDDDR (computer harddrive) like the sierra or any other device that you may be working in, if its nearby as in the case of Stargate Films where the stage and the computers are in the same building. then you bypass any compression of Tape, you can use what you need and store as backup the footage on tape only after it has been put on the computer.
This lil device also opens up the use of channels 3&4 for audio. so ch 1&2 are input at the camera body and 3&4 are input on the Camera adapter. this is also where Panavision will put Metadata information for motion control or CGI work on Channel 3 of Audio if required.
The smaller Camera Head that Nolan Murdock spoke of to David Mullen and I about was in responce to a question about Lipstick cameras or small cameras that could be used with High Def. Nolan said that an effort is being made to take the Head of the 950 and give it a smaller profile that would need to be cabled (10 meter) to a CCU and then to a HD record Deck.
Not a one man operation. But again for possible MOCO shots it is very small.
David Canning (genius) and myself with Matt Morresy put together a rig for Michael Mann and Emmanuel Lubezki DP to see and shoot tests with for upcomming film ALI. Michael wanted a small handholdable camera in HD that he could shoot like a Filmo or like a Pistol at the end of the arm to be able to get in close and tight to the boxers. We found that the Ikegami HDK-37 3ccd camera was the smallest camera that was true HD that we could field at the time. We mounted the Fujinon Prime 8 mm lens from Plus-8-Video then screwed the handle of my Eyemo in had handed it off (No viewfinder) it worked great but there was still the cable that we had to run up to the cieling and over to get out of the way. only downside it was 1080i
The Sony HDK 10's I think are currently being used in Madison Square Garden for Basketball and Hockey supplied by Fletcher Chicago
Call Tom Fletcher for more info on that.
B. Sean Fairburn SOC
Studio City Ca
Although they may be coming out with a separate camera at some point, I don't think Sony will be making any more dockable units (a dockable unit being defined as one where different format tape decks dock directly to the back of the camera). They were pretty clear on this on several occasions.
Personally, I'm looking forward to seeing them make a separate camera. It gets real tiresome to see the cameras talked about and evaluated by the tape formats the decks they are coupled with use.
A separate camera will actually have to be judged on its own merits.
Mike's right on the basic camera - but there is an SDI out option that fits on the back of the camera for an extra 10K - this gives you access to audio tracks 3&4 and AES in. This is what PV uses for Metadata.
I believe there is an add-on unit for the F900 like the Digibeta 700's 701 unit that supplies a digital video signal out as well as 2 additional audio channels in.