[Note: Some of the images in this post may appear truncated when viewed in some browsers (i.e. Safari). Firefox seems to work fine.]

    A little over a month ago there was a discussion here in the forum regarding the proper way to import 48.048-stamped (non-"F mode") audio into a 1080p/23.976 project. Here's a link:

http://community.avid.com/forums/t/71791.aspx

The discussion involved Michael P and Phil Benson and centered around the Audio Import Settings and what check boxes needed to be checked in order to create media that would sync with 23.976 picture.


Figure 1: Audio Import Settings showing the two check boxes in question.

Michael, you tried to explain the functions of Box 1 and Box 2 and made some suggestions regarding which should be checked when importing 48.048 audio into a 23.976 / 48.000 project. But there seemed to be a lot of confusion on what those boxes actually do and what the proper settings should be. Indeed, you both did your own tests and came to different conclusions. Finally, Michael, you advised the following configuration: Box 1 OFF, Box 2 ON.

    It's been my experience, however, that in order to get 48.048 stamped audio to sync with 23.976 picture you don't need EITHER of those check boxes checked. After reading this thread, I got extremely confused as to why this worked. To try to clear up all this confusion, I decided to do some serious testing of my own. Hopefully others will be able to benefit from what I've learned and Michael, perhaps you can confirm with the engineering department if my findings are accurate. 


THE TESTS:

Camera Specs:

    FILM GAUGE / PERF: Super 35mm, 3 perf
    FPS: 24

Production Audio Specs:

    RECORDER: Aaton Cantar
    FILE FORMAT: Broadcast WAV File (BWF), Poly-Rotated
    SAMPLE RATE: 48.048 kHz
    SAMPLE RATE STAMP: 48.048
    BIT DEPTH: 24   
    TC: 30 NDF

Telecine Specs:

    FORMAT: HDCAM
    FPS: 23.98 PsF
    SAMPLE RATE: 48.000 kHz

Avid System & Project Specs:

    OS: Mac OS 10.5.5
    SOFTWARE: Media Composer Adrenaline v3.1.3
    PROJECT: 1080p/23.976 35mm, 3 perf
    PROJECT SAMPLE RATE: 48 kHz
    PROJECT BIT DEPTH: 24
    AUDIO FILE FORMAT: WAVE (OMF)

Procedure:

    I took 4 BWF files recorded and stamped at 48.048 and imported them into a 1080p/23.976 / 48 kHz project. I decided to use more than one test file to make sure the results were consistent across a set and weren't just unique to a single test file. All BWF files were imported with the Project Format flipped to SD 23.976p NTSC for proper handling and preservation of 30 NDF TC on import. Each of the 4 test files was imported with 4 different combinations of Box 1 and Box 2 settings:

        Box 1 OFF / Box 2 OFF
        Box 1 ON / Box 2 OFF
        Box 1 OFF / Box 2 ON
        Box 1 ON / Box 2 ON

    After importing, I measured the durations of all the MediaFiles created by the Avid and compared them to the durations of the original BWF files. To get as accurate readings as possible, I measured their lengths in samples using Soundtrack Pro as well as Duration in frames and min/sec in Media Composer. I also confirmed the stamp of the Avid MediaFiles in both Wave Agent and Text Edit and then checked them for sync against picture.

Here is the bin of files after import:

Figure 2: Avid bin of clips imported with Box 1 and Box 2 OFF and ON.

Here's a brief description of my custom columns shown in Figure 2:

Box 1 = Showing if Box 1 was ON or OFF.
Box 2 = Showing if Box 2 was ON or OFF.
Orig. Duration = Duration of the original BWF file as measured in Quicktime player.
Duration = (Not a Custom Column) Showing the Duration of the file in the Avid in min. sec. and frames.
Duration Frame = Duration of the Avid clip in frames.
Orig Dur Sample = Duration of the original BWF file in samples.
Avid Dur Sample = Duration of the Avid MediaFile in samples.
Fast Import? = If the file imported quickly or slowly.


THE FINDINGS:

Box 1 OFF / Box 2 OFF:

    In this configuration with both boxes OFF, the Avid appears to be doing a straight file-copy of the original WAV data chunk (which is why it's a "fast-import"). This appears to send all recorded samples across un-altered. The Avid then stamps the resulting MediaFile as 48.000 because that's our project sample rate.

    I checked sync on these files and they sync perfectly to 23.976 picture. Indeed, this is apparently due to the altering of the stamp that is done on import. The stamp is what determines the playback speed (or pull). The stamp tells the Avid to play the 48048 samples-per-second media back @ 48000 samples-per-second, which will pull it down to run in sync with 23.976 fps picture. I didn't have any 47.952 files to test, but it's my suspicion that in this OFF/OFF configuration they too would be stamped at 48.000, and by that very nature pulled up.

    You'll notice in Figure 2 that all the Avid MediaFiles have a slightly longer sample duration. If the Avid is doing a straight copy of the original WAV data chunk, then the MediaFiles it creates should be the same duration in samples -- yet they're longer. These extra samples have apparently been added to round the duration up to the nearest frame (since Media Composer can only edit down to the frame). You can see in the table below that the number of samples added to each Avid MediaFile rounds its duration up to the nearest frame:

Figure 3: Math showing samples added to Avid MediaFiles to round up to nearest frame.

The Avid appears to add the extra samples needed at the end of the file, which is the best approach. Adding samples along the way would change the speed and rounding down or truncating samples would be deleting production sound (albeit a very, very small amount). Adding them at the tail makes for a truly "lossless" import. What's interesting is that you can actually see the added samples in the Avid Timeline:




Figure 4: Avid Timeline showing samples added to the tail of "3/t 1."

These added samples are also visible in Soundtrack Pro when comparing the original BWF file to the Avid MediaFile:





Figure 5: Soundtrack Pro Timeline showing the original BWF file "3/t 1."





Figure 6: Soundtrack Pro Timeline showing the Avid MediaFile of "3/t 1."

    The important thing to take away from all this is that in the OFF/OFF configuration the file is not being pulled-down by re-sampling. All original samples are preserved and no conversion is taking place. The only things that change are the stamp -- from 48.048 to 48.000 -- and the samples added to the end of the file to achieve a complete frame. The change in the stamp is where the pull-down is happening, and that is why these files sync to 23.976 picture. And if that's the case, we don't need either Box 1 or Box 2 to be ON to import 48.048 audio.


Box 1 ON / Box 2 OFF:

    Box 1's description says it will "Convert source sample rate to project sample rate on import." But as you can see in Figure 2, files imported with Box 1 ON and OFF are identical: their durations are the same, they both "fast import," and they both get a new 48.000 stamp. Box 1 apparently has no effect. But why? My source files are 48.048 and my project sample rate is 48.000. So, why isn't it converting the "source sample rate" to my "project sample rate?"

    The reason is the Avid doesn't appear to consider 48.048 and 48.000 to be different sample rates. It apparently considers 48.048 to just be 48.000 "pulled up," and likewise 47.952 to be 48.000 "pulled down." Obviously, this is not incorrect. 48.048 is 48.000 pulled up -- but it's also a different sampling frequency. Neither definition is wrong. The problem is it's completely unclear which definition the Avid is using when it says, "Convert source sample rate to project sample rate on import." Is it considering pull-up and pull-down varieties to be different sample rates? Is it going to sample rate convert a 48.048 file to 48.000 by maintaining the "wall clock" duration of the original file? Apparently not because checking Box 1 has no effect. The Avid considers 48.048 and 47.952 to be the pull-up and pull-down members of the 48 kHz "family" of sample rates. Likewise, it considers 44.144 and 44.056 to be the pull-up and pull-down members of the 44.1 kHz family. So, for us in a 48 kHz project, turning Box 1 ON or OFF is going to have no effect on these 48.048 files because as far as the Avid is concerned they're all the same sample rate.

    What Box 1 is really intended to do, is convert source sample rates from a different "family" than the project sample rate. For instance, to convert a 44.1 kHz file to 48 kHz. In this respect, Box 1 is intended to work in conjunction with another setting in the Audio Project Settings: "Convert Sample Rates When Playing." With these two settings, the Avid is presenting us with an option: convert source sample rates on import to the project sample rate OR preserve the source sample rate and convert it to the project rate on playback.


Figure 7: Audio Project Settings and the "Convert Sample Rates When Playing" setting.

    So, if your project sample rate is 48 kHz, and you have a 44.100 mp3 you want to import - if you leave Box 1 OFF it will import it preserving the source sample rate of 44.100 and make 44.100 media. THEN, in your Audio Project Settings if you have the "Convert Sample Rates When Playing" set to "Always" the Avid will play that 44.100 file back doing an on-the-fly sample rate conversion to 48.000 (your project sample rate). If you have "Convert Sample Rates When Playing" set to "Never" you will hear nothing when you play the file and the Avid will not display a waveform.

    Now, if you import that same 44.100 mp3 with Box 1 ON, it will do the sample rate conversion on import to your project rate of 48.000. This is how I usually work. I leave Box 1 ON and convert all sources to my project sample rate and set "Convert Sample Rates When Playing" to "Never." This ensures there are NO mixed sample rates in my project, so Digital Cuts and OMFs can be made without any problems and media turned over to the sound department can be used in a single Pro Tools session.


Box 1 OFF / Box 2 ON:

    When Box 2 is ON, the Avid appears to be re-sampling the original BWF file to 48.000 with a 48.000 stamp and maintaining it's original "wall clock" duration. This conversion explains why it takes longer to import these files -- additional processing is taking place. These files DO NOT sync to 23.976 picture - sound runs too fast and ends early.

    Given this, a more accurate description for Box 2 might be: "Convert source sample rate to project sample rate and maintain the "wall clock" duration of the source file."

Box 1 ON / Box 2 ON:

    Files imported in this configuration were identical to those imported with Box 1 OFF / Box 2 ON. Remember, Box 1 has no effect on these files.


CONCLUSION:

    You'll notice that in all of the above configurations that the resulting Avid MediaFile was always stamped as "48.000." This is important because that stamp is what really determined the pull-down for us. Understanding this, and how the Avid treats pulled up and pulled down sample rate varieties is key. Now that we know the Avid thinks of sample rates in "families," we know that when we import 48.048 or 47.952 audio that the re-stamping to 48.000 alone is going to ensure that it plays back in sync with 23.976 picture.

    Based on this understanding and the findings of these tests, I would say the proper way to import 48.048 stamped audio into a 23.976p / 48 kHz project (either HD or SD) is the following:


Figure 8: Recommended settings with Box 1 ON / Box 2 OFF / Box 3 ON.

Box 1 = ON (Leave this on - it won't affect your 48.048 or 47.952 sources, but will convert any 44.100 sources to 48.000 and ensure you don't have any mixed sample rates in your project; In addition, make sure "Convert Sample Rates When Playing" in the Audio Project Settings is set to "Never").

Box 2 = OFF (Leave this off - turning it on will re-sample your audio to 48.000 while preserving the original "wall clock" duration of the file which will not sync with 23.976 picture).

Box 3 = ON (Leave this on as well to "Convert source sample bit depth to project bit depth on import;" Like Box 1, it won't affect your 24 bit sources, but will ensure that other bit depths are converted on import and that you don't have any mixed bit depths in your project).


A note on "F Mode" audio:



Figure 9: Description of "F mode" from the Sound Devices 744T manual.

    "F mode" audio is audio recorded at 48.048, but stamped as 48.000. Based on the findings of these tests, the settings recommended above should work equally as well for them. With Box 1 ON / Box 2 OFF / Box 3 ON the Avid should do a straight "fast-import" of the original WAV data chunk and preserve the 48.000 stamp. That 48.000 stamp would determine the playback speed which would run in sync with 23.976 picture. I had some F mode files to test with (recorded on a Fostex PD-6 in the "48.000 +1 %" mode). I ran all the same tests on the files, and not only did the above settings work, but every combination of Box 1 and 2 settings yielded the same results. Even with Box 2 ON, no re-sampling was done on import. I checked sync on all these files and they all sunk perfectly to 23.976 picture.

Thanks Mike for the detailed work. I will check this against my tests and internal doc I gave to engineering.

Michael

Thank you for such an in depth investigation.

I have printed your work out and stapled into the manual.

One tiny point, do excuse my muddled thinking, but does not the box 2 option resample the audio so that it runs at the original "wall clock" speed at a sample rate of the project. (-and so does not sync up with NTSC frame rate)

David Y

Thanks so much for this, Mike.   This confirms what I found as well.

For the record, I raised this issue mostly because a client warned us that the DEFAULT setting for a new 23.96p project had Box #2 SET ON, which put his 48048Hz material out of sync!

It could be he just inherited this setting from the last person that rented the machine.   To check, I just confirmed that my Import Audio Settings for an existing 1080p/23.976 project were correctly inherited into a brand new 1080p/23.976 project.

I'll confirm with the client what version of Media Composer they were using when he was convinced the system set Box #2 to ON.

Thanks again for all the hard (and rigorous) work, Mike.

-Phil

Mike Fay:

Note: Some of the images in this post may appear truncated when viewed in some browsers (i.e. Safari). Firefox seems to work fine.

Or reply to the original post.  The compose window in the reply dialog will show the images uncut.

NubusAvid:

Thank you for such an in depth investigation.

I have printed your work out and stapled into the manual.

One tiny point, do excuse my muddled thinking, but does not the box 2 option resample the audio so that it runs at the original "wall clock" speed at a sample rate of the project. (-and so does not sync up with NTSC frame rate)

David Y

David,

It appears your thinking isn't muddled at all, but rather quite correct. It looks like my math confirming the duration of the Box 2 ON files was a little off. What I neglected to do was measure the Box 2 ON files without their added samples and when divided by 2000 (which would be the number of samples-per-frame for 48.000 audio if it were matched with film running at 24 fps) instead of 2002 (which would be the number of samples-per-frame for 48.000 audio running at 23.976).

If I take the Avid MediaFile of "4/t 1" and open it in Soundtrack Pro, the last sample of sound before the samples of silence added at the tail by the Avid is 5427625. So, that's the duration the Avid has converted it to with Box 2 ON. If I divide that by 2000, I get 2713.8125, which as you can see in Figure 3 matches the wall-clock duration of the original BWF file which was running at 48.048. So, yes, with Box 2 ON it's maintaining wall-clock duration and re-sampling to 48.000.

Thanks for spotting that. I've re-written the Box 1 OFF / Box 2 ON section and the Conclusion with this updated info.

Also, before you print this out or take any of what I've written here as gospel, I would recommend waiting on the feedback Mr. Phillips will hopefully be getting from engineering as well as the cross-checking done by readers like yourself here on the forum. I would hate for someone to base a workflow on this before it's had the chance to be properly vetted.

Thanks,

Mike

MichaelP:

Thanks Mike for the detailed work. I will check this against my tests and internal doc I gave to engineering.

Michael

Thanks very much, Michael. Please be aware that with the help of David Y I've already made one addendum to the Box 1 OFF / Box 2 ON section and the Conclusion.

Phil Benson:

Thanks so much for this, Mike.   This confirms what I found as well.

For the record, I raised this issue mostly because a client warned us that the DEFAULT setting for a new 23.96p project had Box #2 SET ON, which put his 48048Hz material out of sync!

It could be he just inherited this setting from the last person that rented the machine.   To check, I just confirmed that my Import Audio Settings for an existing 1080p/23.976 project were correctly inherited into a brand new 1080p/23.976 project.

I'll confirm with the client what version of Media Composer they were using when he was convinced the system set Box #2 to ON.

Thanks again for all the hard (and rigorous) work, Mike.

-Phil

 

Phil,

It sounds like the state of those check boxes is tied to the user setting, as opposed to the project. I know those check boxes always remember their last state, but when opening a new project I agree that it should always revert to the default - which I think should be the settings I recommend above (ON/OFF/ON) and with "Convert sample rates when playing" set to "Never." That would at least help protect the user against having mixed sample/bit rates in their project. It's so, so easy to import things incorrectly, however, that one needs to be really vigilant about checking the settings each time to make sure they are set correcly for whatever they're importing. Hopefully, your client detected it only after the first day's dailies and didn't have that much to re-import and re-synch.

And thanks to you and Michael for the original discussion that got me going on this in the first place! It was a great discussion. :)

Mike

Yep, that makes sense.  The client confirmed he was using MC 3.5.

pb