(Page 1 of 1 pages for this article )
Friday, May 09, 2008
Random RED Revelations
Stuff we learned during testing and Art’s spec spot shoot
Tim Blackmore helps John Chater tweak back-focus on a RED ONE
The more we work with RED ONEs, the more we learn about what they do and how to work with them, thus the following grab-bag of observations and experiences.
Flange Focal Depth Adjustment
When we arrived at Chater Camera for our testing a couple of weeks ago, John Chater was busily tweaking the flange focal depth on one of his RED ONEs. John has a Zeiss SharpMax (tm) Back Focus Alignment Device, basically a self-centering tube containing optics that present the image of a Siemens Star focusing target at infinity distance. John had found that the flange focal depth was wrong on this camera; whether it shipped that way from the factory or whether a renter had “tweaked” it was unclear (probably the latter).
During my tests, I found that the 12x Optimo focused on the chart at a very different distance (as reported by the witness marks on the focusing scale) depending on whether I was zoomed in all the way to 290mm, or out to my shooting setting of 40mm or 24mm: it was accurate at full tele, but as we zoomed wider, the visual focal point and the lens scale disagreed more and more. More tweaking followed…
- Final calibration was done using a 16mm Zeiss Ultra Prime. Telephoto lenses have shallow depth of field out in front of the lens, and deep depth of focus back behind the lens. Wide-angles are the reverse: deep depth of field, but shallow depth of focus. Using a wide lens gave us the most critical adjustment, which we then verified was correct for longer primes and for the Optimo zoom.
- Both REDs wound up with a different flange focal depth than they started out with. Once they were recalibrated, the Optimo’s reported focusing distance didn’t vary nearly as much with focal length changes, and the lens scale matched visual focus within the limits of the real-time monitoring outputs.
- We were not able to get consistent results when eyeballing the LCD or monitor outputs; they’re simply too coarse to show the level of fine detail needed to make this adjustment. We wound up making a set of witness marks on a bit of tape on the adjustment ring, based on the range of settings that looked right in real-time monitoring; shooting a few frames at each setting; and viewing them in REDCINE at 1:1, full-res high quality. That seemed to be the only way to figure out which setting was really the right one.
- The back-focus adjustment has some backlash: Turning it in one direction to hit a mark results in a different flange focal depth than turning it the other way. If you always vow to turn it the same way to hit a mark, you’ll have a much easier time getting consistent results.
- Once we had collimated the camera, and verified the setting with both Ultra Primes and the Optimo, the RED 18-50 could no longer be focused at far distances when zoomed wide: it looked like I needed to turn the focus ring another five or so degrees past its end-stop to get things in focus. John left the cameras set up for his other lenses; the 18-50 doesn’t get much use at Chater.
- Properly adjusting flange-back (and verifying it with both near and far tests using at least a couple of lenses) takes a lot of time. Don’t embark on it as a last-minute thing (if you dare embark on it at all), and make sure that it’s OK for you to be mucking about with it in the first place if it’s not your camera!
- If you’re renting your RED out, or in some other way sending it out with strangers, you may want to make a permanent mark of some sort (John Chater used a Sharpie) indicating where proper back-focus should be. That makes it a lot easier to see if somebody has been mucking about with it when it comes back from a gig, and gives you an aiming point for the ensuing re-collimation.
Focusing Frustrations
As I mentioned above and in my test report, neither the RED LCD nor the 720p monitoring output showed fine enough detail to focus critically, even with image magnification engaged. Shooting with RED monitoring is like shooting with most under-$10,000 HD camcorders, using coarse LCDs and EVFs that only hint at what you’re actually shooting. Unfortunately, the 35mm-sized Mysterium sensor offers much shallower depth of field than a 1/3” or 1/2” handycam, so critical focus is much more, um, critical.
In our tests and on Art’s spec spot, our subjects were stationary, or far enough away that minor distance changes weren’t an issue. There was plenty of time to rack focus back and forth, using both the LCD and the 720p output displayed on a 17” monitor, and most of the time that worked fine.
But I’d hate to have to focus the RED by eye on a dynamic, moving target. There just isn’t enough detail in the low-res outputs to be sure of focus.
Focusing by tape measure and the lens witness marks is how the Hollywood folks usually do it, and that seems to be the rule for RED ONE, too. Just be sure that you’ve set the flange focal depth properly before you try this, or you may be disappointed, especially with wider lenses or wider zoom settings.
Robert Harrison reported using the Accuscene viewfinder’s false-color “exposure meter” setting as a focusing aid: the posterized false-color image made it much easier for him to snap things into focus. Since RED acquired Accuscene, that false-color mode is now part of RED’s monitoring; I’ll try using it as a focusing aid the next time I have a chance to work with the camera. Try it for yourself, too, and see if you find it helpful.
Some folks also say that the RED EVF is a much better focusing finder than the LCD. I’ll be keen to test it out at some point.
Finally, we can always hope that a future build of RED ONE firmware will include a higher degree of magnification in its image-enlargements mode, for when critical focusing is required.
Menu Annoyances
The menus continue to trip us up: they only display a few options at a time (and only one at a time on the rear panel), so navigation requires a lot more memorization than on cameras that show more menu options at any one time.
Fortunately I internalized the sequence to get to shutter speeds as a result of doing my tests; that came in quite handy when we needed to tweak speeds during Art’s shoot. But just about every other menu operation we needed to do required a halt in production while we poked through the three different top-level menu trees, trying to suss out where something was.
Obviously any owner/operator will have the menu system committed to memory after a time. But for the rental user, menu navigation is painful. Manuel Wenger in Munich has made up a PDF of the menu system that we found quite helpful, although it only covers build 13 of the firmware (build 15 is current as I write this).
(An aside: I went onto reduser.net to look up the URL for the menu PDF’s, but got sidetracked by this thread which illustrates, as well as anything I can say, why so many of us find an unmoderated-for-quality/maturity/civility forum a poor resource for camera support. I did, however, eventually find Manuel’s site—using Google.)
Scuttlebutt has it that the menu system will change for the better in build 16.
Data Wrangling On Location
Rather than just drag ‘n’ drop clip folders from RED media to FireWire drives, I tried using R3DManager, which allows multiple-destination copies and data verification using checksums. It’s worth checking out if you’re a Mac-based RED Wranger; it automates data movement and provides an audit trail for RED clips much like ShotPut EXpress [Flash required] does for EX1 clips or P2Log [Flash required] does for P2 clips.
On Art’s shoot, while we broke for dinner, I dumped the first magazine using R3DManager. It worked as advertised. There’s a bit of delay while checksums are calculated; then the clips are transferred. Then there’s another delay while the checksums are calculated on the copied clips and compared to the source checksums. At the end of the process, I was told that the copies appeared to be fine, and I could send the RED DRIVE back to the camera for formating and re-use. I opened the clips in REDCINE, and visually verified that I could scrub through all of them, and they indeed appeared to be OK.
However, we only had the one RED DRIVE; every time we wanted to “unload the mag” the camera department came to a halt. That was fine during dinner, but a bit of an imposition during shooting hours. Much as I wanted to use R3DManager, I couldn’t afford to keep people waiting; between the checksumming and the slow transfer rate (limited by the FW400 backup drives we were using), this process was too slow. I wound up dragging ‘n’ dropping clips from the RED mag to a local store (my MacBook Pro’s internal drive, to avoid FireWire slowdowns); loading the clips in REDCINE to verify them, and then releasing the mag back to production. I then manually backed up the clips from my internal drive to a FireWire drive; handed the drive to Tim so he could replicate them on his MacBook Pro, and started another backup. Eventually we wound up with at least four copies of all the clips: two FW400 drives, a G-RAID FW800 drive, Tim’s laptop, and (for some of the clips) my laptop. No clips were lost or corrupted as far as I know.
Obviously this process wasn’t optimal, but it worked: we had a reasonably small amount of data to back up, and time was of the essence getting the single RED DRIVE back to the camera dept.
On a more “serious” production—such as one with two RED DRIVEs, grin— R3DManager would have cut down the number of manual steps considerably; on a long and arduous shoot that level of automation reduces operator workload and eliminates several opportunities for stupid mistakes.
And if I’d had a faster storage pathway, such as an eSATA RAID, or an ExpressCard34 FireWire card feeding a FW800 drive , FireWire400 wouldn’t have bottlenecked the transfer rates, and transfers would have been much swifter. The RED DRIVE (as well as most of the speedy CF card readers) uses FW800 as its fastest offload connection, so you want to make sure your backup strategy doesn’t slow things down any more than necessary: hanging FW400 drives off the bus slows the entire built-in FireWire bus to FW400 speeds.
Proofreading—We’ve Heard of It...
<rant>
OK, this is a trivial thing, but look at these screenshots from the process of reformatting RED media on the camera:
Three misspellings of “magazine” out of six instances? One can only hope the person writing the UI messages isn’t the same person responsible for writing the actual data-storage code!
Seriously, this isn’t the end of the world, but it does bring to mind the concerns several people expressed to me at NAB, when Scarlet and Epic were announced: “gee, they haven’t finished their first science project yet; should they be starting new ones?” And it certainly won’t put a nervous director or producer’s mind at ease should he or she see this; it has to make you wonder what sort of code reviews or other quality-control measures are being followed in the firmware-creation process.
...But It’s Not Just RED, Grin...
I don’t want to dump on RED just because of a little illiteracy. Much bigger companies do even sillier things.
On April 11th, I purchased a MacBook Air (yeah, Mike Curtis told me not to, but it’s a great field machine for writing, Aperture image processing, and web work like PVC updates. I’d say more about this bloody gorgeous machine, but I’d be accused of being a fanboi [and yes, I lust after the Lenovo Thinkpad X300, too; I’m a cross-platform gearfreak]).
On May 5th, I received a worrisome email from Apple:
Support for your MacBook Air is about to expire.
The complimentary 90-day telephone technical support for your MacBook Air ends soon…
Now, I did check, and yes, I have 90 days of telephone tech support, according to all the Apple documentation. But I also checked, and as far as I can tell, this email came 24 days after I bought the flippin’ thing; how does the passage of 24 days get me close to the expiration of my 90 days of tech support?
Innumeracy at 1 Infinite Loop? That’s more scary than a few misspelled firmware prompts!
</rant>
(Page 1 of 1 pages for this article )
Art Adams | 11/21- 08:15 AM
Art Adams | 11/14- 01:44 PM
Art Adams | 10/27- 07:34 PM
Art Adams | 10/16- 08:04 PM
Art Adams | 10/15- 04:10 PM
Adam, it’s OK to like the MacBook Air. Mike told us all not to...but I did as well, and for the same reasons as you. I love it! I also bit the bullet with the X300, I might even like that as much (more?).
Posted by Scott Gentry on 05/10 at 05:31 AM
Great read. Its nice to see some more honest reviews of the camera in real world scenarios without the fanboy hype - I’ll be tuning in here next from now on instead of wasting time on the fansite!
Posted by Stephen Murphy on 05/11 at 02:20 AM
Hi Adam,
You should check out rsync. It works from the terminal, does checksums and gives visual feedback of the progress of the copy. It takes a little longer that a finder copy but a lot less than r3d manager. I’m using this version: http://www.versiontracker.com/dyn/moreinfo/macosx/16814 with great success.
Once installed, in the terminal type:
rsync -ac --progress <source> <destination>
(letter ‘a’ means archival copy, ‘c’ means checksum and progress is for visual feedback).
We have a text clipping with the command on the desktop that we drag to the shell, then drag the red mag, the destination volume, press enter and that’s it.
Hope this helps.
Luis
Posted by luis Bustamante on 05/11 at 11:51 AM
“You should check out rsync.”
I use rsync (the version native to OS X) all the time in development work, but I’m always checking “man rsync” to verify whether I should use trailing slashes, and where to use ‘em. Rsync is incredibly powerful, but like many *nix utilities it’s unforgiving. It’s easy to send stuff where you didn’t mean to, copy stuff one folder too deep or too shallow, or to overwrite stuff accidentally. Even though I had been using it multiple times a day (writing Python, bash, and psql scripts, html and css files, and sql schemas on a Mac, then rsyncing them to linux boxes for testing and deployment), I still wound up double-checking rsync syntax at least every other day to prevent unintended results.
In the confusion and pressure of on-set wrangling, I find rsync a bit too hair-trigger for my tastes. I prefer something that does a bit more “did you really mean that?” checking. R3DManager is pretty darned good that way, but Finder copies aren’t too bad either, given their visually obvious nature.
But thanks for your suggestion; I might try wrapping rsync in an AppleScript Studio UI to make it safer for on-set use by the brainless (i.e., me at 2am, grin).
Posted by Adam Wilt on 05/11 at 03:54 PM
Good unbiased article, but continuing on from the other recent piece you did on Red, isn’t there now a bigger question?
Shooting Red then down-converting to SD, which somebody recently said they did 70% of the time; this all seems like a big pain. If your final output is SD TV or SD DVD what’s the point?
Reminds me of the old quote, “just because you can do it doesn’t mean you should”.
There are also easier ways to achieve shallow depth of field.
Conversely, and here’s where the real fun starts; what about theatre projected Red One projects? If focus is off at all, it’s going to be magnified ten-fold on the big screen; this camera is being touted as a cinema type, film replacement option, in addition to other things.
So that leaves HD for DVD and broadcast, which is full of major manufacturers products.
As I understand it, there are also lens issues related to sensor size and image quality, even from some of the best glass out there. Another headache with highly magnified/projected images to worry about.
We’re hardly in to RedOne 4K yet, and the next one’s a 5k job.
I want to like this camera, I really do, but looking at it practically, where does it fit?
The human eye is really the true benchmark, but we’re seeing, (pardon the pun) equipment here that produces images probably beyond what my brain can comprehend. The final product will output at 2K anyway, at least for a while. Very few high end projects can afford 4K.
I believe this is the way it’s all going, (if you look at the still camera model), but critical focus issues, a difficult if not convoluted workflow, maxed out computers and $47,000 lenses that don’t do the job; you’ve gotta have pause for thought.
When the 5K version hits the market, everyone will want one of those, but where’s the glass to match that?
Have to admit, it all makes for good reading tho’
Cheers!
Posted by on 05/17 at 03:27 PM
“...isn’t there now a bigger question? ... I want to like this camera, I really do, but looking at it practically, where does it fit?”
It fits nicely in the toolchest as an affordable way to shoot raw, 3.2K, 35mm depth of field imagery, digitally, with a tolerable data rate.
Easier ways to achieve shallow DoF for SD? Sure, shoot 35mm, but then there’s the hassle of the photochemical film workflow. Use a 35mm adapter? They add bulk, are a hassle to set up and collimate, often consume a lot of light, and sometimes invert the image--and all of them are unsharp compared to direct imaging without the adapter. (On the plus side, they make any video camera as ungainly and awkward as a handheld RED, grin.)
For cinema release (and HD, as well), focusing errors will be magnified. But film shooters have been dealing with this for decades: collimate the lens and the body, make sure the calibrations are accurate, and have a skilled 1st AC pull focus by the numbers. The trouble with focusing by eye on a theatrical shoot is that by the time you can see a focusing error, it’s too late to fix it, because it’s part of the shot (that’s a slight oversimplification, but only slightly).
For HD, versus major manufacturers’ products? The RED still gives you that shallow 35mm depth of field that requires a 35mm adapter on other cams (or shooting DigiPrimes wide open--but what if you want the shallowness of T2 or T1.3 on 35mm, if you’re shooting 2/3”? There aren’t many T0.25 HD lenses...).
For all of them, RED gives the low-budget producer a more affordable production / acquisition package than either film (where film and processing costs will eat you alive) or tape (no need for expensive decks), and offers more after-the-fact control than non-RAW video formats, and more resolution (and 4:4:4 sampling) for VFX work. The real RED “revolution” isn’t just the camera, it’s the low-data-rate recording and the affordable access to that data.
But it’s a tradeoff: the RED ONE is awkward compared to a shoulder-mount video or film camera; realtime monitoring is compromised compared to video cams or other digital cine cams (since RED has decided not to support the RAW data port at present, there’s currently no opportunity for an S.two or a Codex Digital recorder to take the full-res data out and deBayer it for realtime display, and of course the onboard monitoring is tragically insufficient); you have to be the film lab / telecine / knee circuit all rolled into one; the camera is an ongoing science project as opposed to a proven, highly-evolved design from a vendor who has already learned most of the hard lessons.
[There are other digital cine cams, too, like the 4K Dalsa Origin, the Panavision Genesis, Sony F35, and the Arri D20/D21 (and the Viper and F23 in 2/3"), but they require expensive offboard, high-bandwidth recording. Only the Silicon Image SI-2K offers the same sort of revolutionary raw recording at non-heroic data rates, using Cineform--but it’s a 2K camera with a 2/3"-sized sensor, and without the marketing juggernaut of Jim Jannard and Ted Schilowitz behind it.]
At the end of the day, if you can tolerate the camera’s flaws and idiosyncrasies, it gives you a richly detailed, naturalistic image that can look extremely beautiful. It uses the same lenses that 35mm shooters are comfortable with; keys well without annoying edge artifacts; gives you room to push in and reposition in 2K post; even lets you correct (slightly) for exposure errors, or pull up more shadow/highlight detail in post than a “traditional” video camera might give you. And, of course, “the workers can own the means of production”. This has both positive and negative aspects, grin, but it’s one of the key factors behind the hysteria surrounding this camera.
It’s a digital cine camera with a certain “look” to it, as well as an evolving workflow that’s not the traditional film workflow, not the tape workflow, not even the P2 / SxS workflow. Does it obsolete film? No. Does it obsolete video, whether SD or HD? No. So where does it fit? It’s another tool in the toolkit, another tube of paint in the paintbox, nothing more.
When the 5K cam comes, it will offer the opportunity to pull a true 4K image from it--but as you mention, 4K post is a rarity still, and arguably will remain that way for quite a while yet. The 5K camera only gives you 20% more resolution to work with, but costs twice as much; we’re looking at RED ONEs for theatrical work but currently have little interest in the 5K camera for our 2K and HD target markets.
As to the glass, all the lenses I’ve tested show evidence of easily exceeding the resolving power of a 4K sensor, so will likely handle a 5K sensor with aplomb.
Posted by Adam Wilt on 05/18 at 09:16 PM
“Focusing by tape measure and the lens witness marks is how the Hollywood folks usually do it”
“...collimate the lens and the body, make sure the calibrations are accurate, and have a skilled 1st AC pull focus by the numbers.”
Hi Adam,
Thanks, great info about real-life issues with focusing. Could you point me to a good resource where the techniques that you mentioned above are explained in a bit more detail, especially as they relate to focusing on dynamic, moving targets?
thanks a lot,
Hugo
Posted by Hugo on 06/03 at 06:12 PM
“Could you point me to a good resource where the techniques that you mentioned above are explained in a bit more detail...?”
I can recommend Douglas Hart’s “The Camera Assistant - A Complete Professional Handbook”. David Elkins’ “The Camera Assistant’s Manual, Fourth Edition” looks good, though I haven’t read it. Search for “Camera Assistant” on amazon.com and both books will come up at the top of the list.
Posted by Adam Wilt on 06/15 at 06:17 PM
This is about my 18th or 20th time trying to post this: the first 2 days I tried, I didn’t know that one had to be Registered & Signed-in, to be able to post ( Preview got me a blank screen: no html, no nothing ).
The third day, I finally got fed-up trying different Permit-Scripting settings, and tried Submit, without Preview, and it THEN told me I had to be registered / logged-in, to post.
Fire/suspend the person who coded that bit of your web-site, until they learn to test with secure browsers, of several kinds, and to inform us of what is required to work with the system implemented.
So, then I registered, and logged in ( in a different tab ), and came back to this page’s tab, and refreshed to be able to post, and it threw away my posting.
I pray to god I never implement a UI as opposed to informing-me as this one is, to informing-you.
--
The REASON for my bothering with all this, is simply that if you’re using tape for focusing, you’re non-necessarily wasting your entire-crew’s time, and worse, boshing your work’s flow:
There is a brand of laser-rangefinder that is *accurate* ( as-in to +/- 1.5mm, over about 200m ), unlike most.
Leica DISTO A5 is something you would find pays for itself in a couple of days.
It *doesn’t* require one person at each end, as tape does,
it *doesn’t* require more than a single second to do,
and it *doesn’t* obstruct/bosh the work’s flow.
Leica-Geosystems.com DISTO A5
Metric version:
Froogle: Leica Disto A5 740689
American version:
Froogle: Leica Disto A5 740690
I’m recommending the A5 over the other versions, because:
a) it can survive being dropped ( though I don’t recommend that )
b) it covers usable distance-range
c) it’s splash/dust proof ( why rely on something that’s excessively fragile? )
Keep on shootin, eh?
Cheers!
--
PS: “Preview” still produces a blank screen.
Posted by on 08/04 at 12:33 AM
|
