Telecine specs

[Jeff Wexler]

Marc Wielage has kindly provided a .pdf of the telecine specifications and capabilities he presented at our union seminar awhile ago. There may be some updating required but for the meantime this is a valuable and instructive document regarding our evolving workflow.

Regards,  Jeff Wexler

[Bartek]

Jeff,

Its wonderfull that information such as this can be shared freely.

Thank You!

Bartek

[Jeff Wexler]

this PDF is dated almost a year ago with many of the current popular systems untested at that time, have they updated the test equipment list?

Marc informed me that this document woud need a current update but I posted it anyway. I hope to have a current update (things change quickly, as you know, with firmware updates and so forth) and I will post that when I have it.

I am also trying to compile from all the various reports and personal experiences posted here, a general document outlining procedures, deliverables, compatibility issues and so on. This is a moving target, as you know, so it may not be possible (or even useful once compiled). We can only try and make the effort.

Regards,  Jeff Wexler

[larry long]

From WB in February 07.

Warner Brothers Post Production Sound

Digital Dailies Specifications

Version 1.31 – last updated 03/06/2006

File Format: BWF – Broadcast Wave File (Polyphonic / Interleaved files only)

Audio Ref. Level: -20 dBFS

Sampling Frequency: Sample Rate: 48.048 KHz

Bit Resolution: 16 Bit

Note to Deva Users: Do not use the Pull-Up option on the Mirroring Mode screen to achieve pull-up

Timecode Rate: 30 fps

Media:  DVD-RAM Type 1&2 (UDF), DVD+/-R (UDF), CD-ROM (ISO 9600), FTP/AFP Server

For more information please feel free to call our Editorial department at (818) 954-5134 or our Engineering department at (818) 954-2906

[Jeff Wexler]

Thanks for that Larry. This must be a specificatinon for TV shows, right? I am glad to see that they no longer put "Fostex PD6 preferred" in the spec (particularly since that is a machine that is no longer being produced). I understand the 48.048 sampe rate but cautioning only Deva users on these settings does not address the potential problems from other users/machines since they all seem to handle 48.048 differently. The point is, as Courtney so often points out, is that pull down and pull up issues need to be figured out and then stick with it. I also take exception to the insistence on 16-bit and UDF format. I know why they are doing this but it isn't right.

Warner Bros. was quite late coming to many of these technological changes, refusing to allow ANYTHING different than what they wanted to do. I had huge battles with Warner Bros. over my use of DAT and then later my use of the Deva --- I tried to help move the studio to what I knew would be the way we will do our work but they wouldn't budge. My efforts were rewarded by their transfer department screwing up our dailies on "The Last Samurai" even to the extent that they actually deleted takes on my disks and did not tell me they had done this. In typical Warner Bros. fashion they later purchased about 40 PD-6's and a bunch of DV-40's and then "standardized" on that, even suggesting that if you used a different recording "system" the Studio could not guarentee that your dailies would be handled properly. Reminds me of the good old days when Warner Bros. sabotaged Ron Cogswell's Nagra dailies because he was using his own equipment, not the studios. This was a case where Ron had secretly run 2 machines, 2 Nagras, so he had a backup --- Ron had a feeling about what Warner Bros. was up to and so this backup saved his job (and the movie he was working on).

Regards,  Jeff Wexler

[Andrew Bellware]

A question I've had for a long time is: what is it that people really mean by "Audio Reference Level" in a digital format? I've seen this in broadcast specs too. I've always thought that was a legacy specification from analog days, maybe it isn't?

I understand it from an analog tape standpoint, like + or - so many dB's over so many nanowebers/meter.

But from a digital standpoint the "reference level" is... well, 0dBFS at whatever bit-depth one's working in.

So obviously they mean something else. Do they mean where dialog should average if using Vu meters? Or perhaps the RMS level while dialog is present?

Andrew Bellware

From WB in February 07.

Warner Brothers Post Production Sound

Digital Dailies Specifications

Version 1.31 – last updated 03/06/2006

[snip]

Audio Ref. Level: -20 dBFS

[curleysound]

It appears that they expect the reference tone at -20

[Bartek]

The question of reffrence tone in NL is a good one.

I deliver a 1K refference file to post on my "Audio Settings" disc that I provide to post prior to principal photography, its just me talking for a few minutes describing the settings, microphones, timecode, and any special scenarions that I can forsee, recorded at the specified settings. A test dics of sorts.

Now unless otherwise specified I do not repeat the 1k file on every roll.

Should I?

Bartek

[Jeff Wexler]

The question of reffrence tone in NL is a good one.

Now unless otherwise specified I do not repeat the 1k file on every roll.

Should I?

Bartek

I am not an expert on this and I am also coming from the ancient historical perspective where even though we were recording digitally, everything we recorded was at some point transferred to some analog medium (even if by analog we mean "played out" analog to a mixing panel or a loudspeaker). So, I still put a reference tone at -20 along with the VERBAL "sound roll" ID as the first segment or first recording of the day. Although I cannot give a specific reason why this tone is needed or not, it always seems to me that some repeatable reference is valuable. Sure, I know that it is the nature  of digital audio in all its forms that LEVEL is also part of the whole deal, if we are talking about ANY useage of our sound beyond just a simple file copy, the standard and predictable reference tone still seems to be a good idea. I think it continues to be a good idea, and a good practice, to put this tone at the head of the "roll" (I know, it is just another file that maybe no one ever references), but it is valuable for completely different reasons than we had when everything was analog.

Maybe someone else who really knows the real answer to your question might chime in:  Marc Wielage perhaps?

[Eric Toline]

"This is also why it is important for you to "ride" the level while mixing your sound, so that your peaks never exceed your referenced level.  Especially important now in the digital world we mix in, because digital sound doesn't go into distortion, the term is called "digital tearing" when an overloaded signal is attempted to be recorded."

I question the "peaks never exceed the reference level" statement. If your reference level is -20dbfs why would you want to keep everything at that level? You're wasting tons of headroom and dynamic range. Network audio standards call for -20dbfs reference level with peaks no higher than -10dbfs. Surely production sound for features can handle -10dbfs peaks.

Eric 

[cmgoodin]

...The reference tone, whether it be -20db, or a -8db from the Nagra was "referenced" to 0db on a VU meter.  It is always a good practice to start each "roll" with a head i.d. and tone for reference purposes. 

------

RVD

Typo - I'm sure Richard meant -18db for the Nagra as a reference. -8 would make for pretty hot dailies.

I always treated the Test tone (-18 on Nagra or -20db on Dat) to be the average peak level.  I would have some peaks above that level and some peaks below but the average is -20 for 0 VU.

(Warning! Educational Content follows)

Digital distortion or as RVD described it, "Digital Tearing" which many fear is caused by wrap-around in poorly coded A to D converters.  This is where when the analog signal is converted to digital and it exceeds the maximum level that the bit depth allows it can cause the number that represents the signal level to go from maximum positive to maximum negative from one sample to the next. This happens because in 2's compliment binary math the top bit in the word is used to represent the sign or whether the number is positive or negative.  If the A to D converter does not boundary check on its sampling algorithm, it can cause this overflow which causes the square wave tearing distortion created by causing those peaks to go from full scale positive to full scale negative for every sample that exceeds the peak value.

This kind of distortion was prevalent in early Digital Recorders because in their effort to maximize the performance of the A to D they would not boundary check which allowed the DSP to handle more channels of sound.  However newer DSPs have enough horsepower to boundary check so the peaks that go over are just flat-topped (similar to what happens in analog over-modulation) Peaks are flattened but because wrap-around is prevented by setting any peak that exceeds the bit level of the converter  to the maximum (or minimum on negative peaks) until the voltage drops back to the level within the scope of the digital word. This is sometimes referred to as a "Brick Wall Limiter" if done in the digital converter. 

The presence of the Digital tearing in early converters caused many recordists to become gun-shy of peaks causing a quite conservative use of the dynamic range. In other words they kept an unusually high amount of headroom to avoid these tearing peaks which were quite audible and nasty sounding.   With modern A to D and Brick Wall Limiters, we can run the levels a bit hotter without the fear of an occasional over-mod peak ruining the sound.  This is important especially if the sound starts out or will end up as 16 bit sound.  An occasional flat-topped peak will not be audible so you can run the average level a bit hotter using the full dynamic range of the bit depth.

You should always check the recorder (or Mixer if it is doing the A to D) and record some over-modulated sound and look at the waveform in a digital workstation to see if it has the wrap-around problem.  If it does create the "digital tearing" you have to be more conservative with peak levels. 

----Courtney

[Eric Toline]

Typo - I'm sure Richard meant -18db for the Nagra as a reference. -8 would make for pretty hot dailies.

No. Richard is correct. He's referring to analog Nagras not digital. The built in Nagra tone generator set the reference level at -8db on the meter.

Eric

[cmgoodin]

I stand corrected.  Brain Fart. I guess too many years of recording on DAT at -18db. I remember now that the Modulometer on the Nagra had that -8 db level for their tone. It was a completely different type of metering that read out percent modulation with pretty slow ballistics as I recall... It has been a while at least 12 years since I used an analog machine.  Still have them though.

---Courtney

.

[Marc Wielage]

A question I've had for a long time is: what is it that people really mean by "Audio Reference Level" in a digital format? I've seen this in broadcast specs too. I've always thought that was a legacy specification from analog days, maybe it isn't?

I've seen this hotly debated many times. The EBU spec (insisted upon by the BBC, among others) is -18dBFS; the North American spec is generally -20dBFS. But I've also seen everything from -20 to -12 used.

I think as long as the disk(s) and sound reports are labeled, it's OK. We'd prefer to consistently have -20dBFS with digital media for post, but that's just because all of the Panasonic D5 and Sony HDCam/HDCamSR HD machines are usually set up for -20, as are the Tektronix 764 digital audio scopes we use for monitoring in all 11 of our telecine rooms in Hollywood. This is pretty standard for Ascent, WB, Universal, and most of the other LA post houses I know of.

There's a separate question as to average dialog level, and in final mixes, most networks request (some demand) dialog peaks of -10dB. Some distributors (notably Disney and Discovery Channel) will kick back master tapes if any momentary peaks go over about -4dB, but that's a final mix issue, not one for dailies.

Note that I have rarely gotten any complaints from dailies tracks that were too loud, but I have heard complaints about dailies tracks that were too quiet. I believe editors are just looking for consistency more than anything else, so the tracks will cut well. We try to make sure all the dailies sent to clients are an exact replica of the audio tracks given to us, and will not change or "ride" levels unless explictly told to.

--Marc W.

[Marc Wielage]

I remember now that the Modulometer on the Nagra had that -8 db level for their tone. It was a completely different type of metering that read out percent modulation with pretty slow ballistics as I recall...

If it's like the last Nagra IV-S I used, it was PPM. Definitely a very different animal from VU. I've seen Nagra-T's that were calibrated for VU or PPM. Having grown up in radio, I'm used to VU, but I'll adjust to whatever I have to do to get the job done.

By the way, Courtney, BWF-Widget saved us in a commercial dailies job the other day, where a client had given us 50 bad tracks from a 744T. At least with BWF-Widget, we were able to tell the mixer how and why the files were bad, and we were able to save the job by going to the DAT backup. Very frustrating when these things happen.

--Marc W.

[cmgoodin]

By the way, Courtney, BWF-Widget saved us in a commercial dailies job the other day, where a client had given us 50 bad tracks from a 744T. At least with BWF-Widget, we were able to tell the mixer how and why the files were bad, and we were able to save the job by going to the DAT backup. Very frustrating when these things happen.

--Marc W.

Marc,

What was wrong with the 50 tracks?  Just curious.  Wrong settings?  Curious if they were beyond correction.  I guess going to the DAT backup is easier than trying to fix them.

---Courtney

[Andrew Bellware]

I think there are two things going on here. One (reference level) is I believe a "legacy" from all - analog systems. The other (peak levels) is still important when interfacing digital gear with analog gear when the signal chain has "issues" with their gain staging and headroom.

In the digital world it's irrelevant where you decide your "reference" level is. You can say your reference level was -50dBFS and still put your dialog peaks at -4dBFS. How would anyone know? (Well, they'd know that if you had 46dB of headroom you'd have a pretty interesting mixer, but still.)

Unlike with analog tape where one could set up one's recorder at, say, +4 over 350 nanoWebers/meter, or +6 over 350nW/m; in the digital world -20dBFS is always -20dBFS. It's not like someone playing back your tape should be wondering how you'd set up the electronics because the numbers are absolute. -12dBFS on your digital machine can't and won't play back as -18dBFS on their machine.

The real question I think, is where you are putting your program material. Typically, in an all digital system, the only rule you need to know is to never ask the A/D converters to go above 0.

But in some (especially, in my experience, broadcast) facilities, the analog outputs of digital recorders and playback machines were not properly gain-staged taking into account the headroom of the head-end (analog) electronics. What this did was give us a situation where broadcasters are dictating that the signal never go above, say, -10dBFS. (I blame the picture-centric EIC's who installed digital machines in place of analog machines without changing the gain staging.) I imagine this issue might also be true with optical dailies, although one would intuit that it should not arise with digital dailies.

--

So the issue isn't what you are REFERENCE to in a digital system, just how your digital peaks might be above the headroom of analog electronics further down the signal chain.

If your (analog) recording mixer's headroom is, say, 24dBu, you can set up your digital recorder so that you can go right up to 0dBFS without distorting from a "reference" level of -6dB, -12dB, -18dB, or even -24dB. You can put tone on a digital file at any level you like.

But if your digital PEAKS come out as an analog signal which is above the headroom of the next (analog) device in the chain, then it would be necessary to ask that recorded signal never peaks above that level.

In other words, if you have a DAT machine feeding an analog device, and the DAT's analog outputs produce +24dBu at 0dBFS, but the next analog device can't handle signals hotter than +18dBu, then you better keep the peaks you record onto that DAT at -6dB (although doing so effectively reduces your digital system by a bit of resolution).

So my point is that:

Reference level and tone in a digital recording system? I don't see any reason why post still asks for this.

Maximum digital peaks? In an all-digital system this shouldn't be, but with badly gain-staged analog systems further down the signal chain it might be a necessary evil.

My .02!

Drew

[RPSharman]

There is still a reason for providing a file with tone in the digital world.  I didn't think it was neccessary, but on one job recently the editor was not importing audio properly into FCP.  I don't know what the issue was, but I hadn't given him a tone file with which to check his system.  I went to his house, record tone at -20dB on the 744T and he imported the file.  It came in much lower than -20dB on the FCP meters.  He listened to the recordings played directly from the 744T at normal headphone levels, watched the meter levels during playback, and understood the tracks were recorded properly.  I left, and he must have fixed the issue, because I did not hear back.

[Andrew Bellware]

There is still a reason for providing a file with tone in the digital world.  I didn't think it was neccessary, but on one job recently the editor was not importing audio properly into FCP.  I don't know what the issue was, but I hadn't given him a tone file with which to check his system.  I went to his house, record tone at -20dB on the 744T and he imported the file.  It came in much lower than -20dB on the FCP meters.  He listened to the recordings played directly from the 744T at normal headphone levels, watched the meter levels during playback, and understood the tracks were recorded properly.  I left, and he must have fixed the issue, because I did not hear back.

Of course, in this example, you're helping a picture editor with his technical incompetence. I would certainly hope that Warner Brothers would be able to generate their own tone files and set up their gear on their own! ;-)

[Guest jimg]

I've run across several smaller market telecine facilities that treat DV40's and DV824's just like a tape machine, using the analog audio outputs as the source for the machines they are recording to. So in this case, a roll ID/tone file is pretty essential. I can't recall which houses in particular have transferred spots I've done this way, but they are in Dallas, Atlanta and Chicago if memory serves me right. I've done a couple of spots recently where the transfer facility wanted TC DAT only or wanted a DAT as backup because they are still figuring out file based transfers. And it was easier for the engineers at these facilities to just unplug the Nagra T or DAT playback deck, plug in the DV 40 or DV824 in the same way and work as they have been for the last few years.

In the commercial world, it's back to the same proceedures I had to follow back when the 4S-TC and Nagra T hit the market. Call the transfer house for each job.... Some can deal with a 48.048 file, some only want 48K/16 bit files and a few are still tape-based.

Best to all,

Jim

[Philip Perkins]

I've run across several smaller market telecine facilities that treat DV40's and DV824's just like a tape machine, using the analog audio outputs as the source for the machines they are recording to. So in this case, a roll ID/tone file is pretty essential. I can't recall which houses in particular have transferred spots I've done this way, but they are in Dallas, Atlanta and Chicago if memory serves me right. I've done a couple of spots recently where the transfer facility wanted TC DAT only or wanted a DAT as backup because they are still figuring out file based transfers. And it was easier for the engineers at these facilities to just unplug the Nagra T or DAT playback deck, plug in the DV 40 or DV824 in the same way and work as they have been for the last few years.

In the commercial world, it's back to the same proceedures I had to follow back when the 4S-TC and Nagra T hit the market. Call the transfer house for each job.... Some can deal with a 48.048 file, some only want 48K/16 bit files and a few are still tape-based.

Best to all,

Jim

SF.

Philip Perkins

[Marc Wielage]

What was wrong with the 50 tracks?  Just curious.  Wrong settings?  Curious if they were beyond correction. 

Damndest thing we've ever seen. 55 tracks, and the Fostex couldn't read them. BWF-Widget reported that their starting timecode was 00:00:00:00, and the ending timecode was 00:00:00:00 (!). And there was sound there. My feeling is, the guy rolled but the timecode feeding the 744T (or the 744T's own internal generator) went stupid.

The DAT was fine, so go figure. Oh, and every other take on the DVD-RAM was missing -- literally no file there. It was as if the mixer didn't go into record on X number of takes. It was there on the DAT, but the log from BWF-Widget reported no file there.

[Marc Wielage]

Reference level and tone in a digital recording system? I don't see any reason why post still asks for this.

We ask for head tone and a slate telling us what reference level is being used so that we won't screw it up. All incoming audio to most LA dailies facilities goes through some kind of digital mixer (typically a Yamaha 03D or a DM1000), and it's nice to have the tones there so we can check it.

Some shows ask us to combine 1&2 to a mono mix track for the editor, so it's also a way for us to make sure the levels are identical on each side.

For these same reasons, we also record a minute of digital bars and 1K tone at the heads of all HD videotapes. Ironically, we don't do this for digital files on servers. As tape starts going away, maybe things will change; the crystal ball is cloudy.

[cmgoodin]

Damndest thing we've ever seen. 55 tracks, and the Fostex couldn't read them. BWF-Widget reported that their starting timecode was 00:00:00:00, and the ending timecode was 00:00:00:00 (!). And there was sound there. My feeling is, the guy rolled but the timecode feeding the 744T (or the 744T's own internal generator) went stupid.

The DAT was fine, so go figure. Oh, and every other take on the DVD-RAM was missing -- literally no file there. It was as if the mixer didn't go into record on X number of takes. It was there on the DAT, but the log from BWF-Widget reported no file there.

Very curious. Especially since BW-Pro calculates the End Timecode based on the number of bytes in the Data chunk.  So if there is sound there it should show an end time code equal to the length of the file in HH:MM:SS.  Even for non BWF files BW-Pro will calculate an end time code based on starting at 00 and ending at the duration (based on the default TC rate listed in the Options Menu) I guess if there is a bext chunk, but gibberish in the TC field it could cause BW-Pro to calculate a negative number for the start time or something, that would screw-up the Out TC calculation.

It could be that the data chunk did not get the chunk-length updated at the end of the recording.  This can happen if you remove the CF card or shut off power before the file is closed.  IF a device locates the start of the Data chunk and there is sound there it might play it until the file runs out.  Perhaps something stupid was going on like shutting off the ext power to stop recording. That would really screw things up.

This validates my opinion that the manufacturers' claim that the recorder will warn you if there is a problem with writing the file(verification during recording)  is really a myth.  There is no "Real" confidence during recording on most HD recorders.  Espcially those that record or mirror on multiple media. At least with DAT we could monitor off the recorded take with only a few millisecond delay that would serve as confidence that the data was recorded and playable (at least by the source Dat machine).

---Courtney

[Jeff Wexler]

This validates my opinion that the manufacturers' claim that the recorder will warn you if there is a problem with writing the file(verification during recording)  is really a myth.  There is no "Real" confidence during recording on most HD recorders.  Espcially those that record or mirror on multiple media. At least with DAT we could monitor off the recorded take with only a few millisecond delay that would serve as confidence that the data was recorded and playable (at least by the source Dat machine).

---Courtney

I agree that there is no "real confidence" in the traditional and older sense of the word with the file based recording technologies (all of them including laptop/software based recording systems) and certainly each of the systems in use (whether they are dedicated hardware recorders, Deva, Cantar, SD 744T or software recorders) utilize different routines to "verify" the recording.

I can only really speak for the Deva but I have found that the method of recording to the primary drive utilizing the Deva OS and MARF file format, has been 100% reliable. The Deva also has really good warning indications if there has been a problem with making the recording. So, to that extent, I really don't worry at all --- my confidence comes from experience with the machine. Now, on to what Courtney is talking about in the real world, our recordings having to be put on multiple media, either as a mirror (copy) process or as a simultaneous routine as in many of the recorders, is still fraught with potential problems and ugly consequences. I have had a few difficulties even with the Deva regarding the mirror process to optical disks where there is a certain degree of confidence from the copy process but it is not by any means 100% foolproof. I don't know what can be done about this without further complicating the whole process. I am sure Courtney can explain to us all the things that can happen even when everything is in one perfectly configured full computer setup --- and obviously there is a higher potential for problems when dealing with small batery operated field recorders.

Regards,  Jeff Wexler