[Tig] DI workflow, capabilities and requirements

Christopher Noellert cnoellert
Tue Mar 29 00:32:35 BST 2005

On 3/28/05 8:14 PM, "Steve Shaw" <digital.praxis at virgin.net> wrote:

> Hi Chris,
> The stability of a 'tube' system is not inherently more unstable when
> compared to a CCD as its the 'bulb' in the CCD system that drifts. I've seen
> just as accurate longform scans from tubes as with CCDs, and just as
> inaccurate from CCD as with tubes. Today's calibration techniques have all
> but dealt with this issue.

Hi Steve,

Together again...

My experience I would say has been quite different than yours then.  I've
seen tube machines drift over the course of a session... Even over the
course of a roll.  Stopping to calibrate your telecine constantly because of
drift is not so ideal.  There's a good reason that the bulk of working TCS
boxes I've seen have been on CCD machines.

I have a Mill2 and it wouldn't be the machine I would throw at that kind of
work, even if I had to calibrate it after every roll I would still be
worried about drift.

I'm also very concerned about the stability of the Mill2's S16 gate which as
I understand it is the same gate used for S16 on the DataMill.  Again the
Spirit/Shadow gets my vote for stability.

I know it sounds like I'm relatively negative about the DataMill but in
truth I'm actually not.  We can say that I'm cautiously hopeful.  But so far
I have seen little that makes me believe that the machine is what is
required to do the work that I would need it to do.  Not everyone has the
luxury of S35 for every job.

I am doing 2 perf right now though... That's kind of fun.

> The bigger requirement is for the scanning system to be able to generate
> calibrated LOG output into any format of image, from data, to HD and SD
> 'video'. The Spirit and Shadow can't do this is the image data is
> 'linearised' too quickly within the image path. The dataMill and DSX can
> maintain their high bit count front end until the data is mapped into LOG
> space. This is a far greater requirement. And 444/4K are not options for the
> dataMill as I understand it. SD and HD LOG outputs are, but the price
> differential isn't very large. And as you now don't need a da Vinci/Pandora
> as the transfer is self calibrating you save a whole load that way too.

Someone with greater knowledge of the Spirit can possibly answer this
question better than I can but it's my understanding that everything remains
in Log space internally on the scanner until rack 3 at which point it's
dithered/linearized for output in 422/444 in HD/SD or Data.  Extended ranges
for the Spirit will give you what you're after in terms of DR... Almost.  I
assume in the Spirit4k DR is no longer the issue it was on the Spirit1.

4k comes standard on the DataMill actually... That much I understood, but
you are correct in saying that HD and SD are options.
> And for commercials and music videos you still want to conform first, then
> make the final grade in context - why would you do anything different? I am
> really beginning to see the demise of colourists on film transfer devices.
> DI workflow will be the way all post is done.

Because some people still shoot 80 labrolls and don't cut neg.  Yes.  You
could rush 80 labrolls to SR or data, do the whole online all the effect
shots and all the graphics and then go back and do your color work but
that's not so cost effective for a commercial production that's going to air
in SD only...at least until/if Europa goes HD for broadcast.  I'm not saying
you can't do it, I'm just saying that for the time being the workflow you
describe does not fit every model.  It fits very nicely for feature work,
but in smaller European economies your workflow is budget destroying and
somewhat overkill.

TK Color correction is not dead.  In fact it's far from it.  Your workflow
only suits a job where the project will be completed within the same house,
you're not accounting for the commercial/promo market where you grade with
one guy and edit with another guy and online with another guy and fx by some
other house.

Your workflow would assume that all 80 rolls are going to all these
different houses on NAS drives.  Who's going to pay for that?  It's not that
I don't think things will eventually go that way but touting this workflow
to a commercial client would be financial suicide.  No one wants to pick p
the check.

A TK and a colour corrector have a future for quite some time.
> I do use SR decks for DI workflow, and am working with a facility in the USA
> who are doing a full DI this way. Not sure I can talk about this yet
> though...

I can.  It works great.
> However, I  find NAS a lot cheaper than tape as its a real IT comodity. 1TB
> costs about 1K Euro, if you don't go to 'media' manufacturing companies. And
> obviously we re-use the NAS on a job by job basis. This is for online
> though, not dailies work. For dailies I would suggest HD D5 is more than
> acceptable, and have done a lot with HDcam too. A LOG image in these formats
> looks amazingly good when displayed via non distructive LUTs.

NAS cheaper than tape?  On what planet?

Let's do the math.  Let's say we're talking 444 1080 25p.  You get maybe
about 80 minutes per tb at 1k ?. 2hr SR tapes are running ?329.  So we could
say that I get 6 hours of the same storage for 989? for which you get

The thing you're kind of glancing over here is that time of course is money
as well.  Tape runs in realtime can be graded tape to tape in 444 with
calibration in realtime which means that you "render" in realtime.
While I agree that NAS has it's place, (or san I should say) and I wouldn't
trade our infrastructure for the world, I think it's a big miss to count out
tape just yet.

Everything has a place.

> All good fun. But I think I've said enough for now. Up to others to keep the
> debate going.

Don't leave yet... The party's just starting.


Chris Noellert
Frithiof Film to Video. AB
Sturegatan 58 
114 36  Stockholm

Tel: +46 8 545 678 78
Fax: +46 8 545 678 79
Mob: +46 735 32 00 03
AIM: cmnoellert

More information about the Tig mailing list