Audio issue downconverting to SD
 

Now here’s a weird thing that I wonder if someone can help me with.

I’m required to deliver my footage in Standard Defintion DVCPro50. Using MxM cards in my EX3 was easy. Copy EX3 files to hard disk via MBP 15 laptop, log and transfer into FCP to convert to .mov files, then use MPEG Streamclip to downconvert to DVCPro50.

Having recently acquired a Nanoflash should have made it all even easier, by eliminating the need to Log and transfer. It was for a while. But then I discovered a problem with the audio tracks.

I find that Tracks one and two are both mixed together so that I end up with two identical audio tracks. I can tell you that this was a real shock when I discovered the problem. It never happened with files recorded with MxM cards.

Can anyone enlighten me with what’s going on? I emailed Convergent Design. They say the only difference between their audio and the EX3 is that theirs is 24bit while the ex3 is 16bit and maybe it’s something to do with the way Mpeg Streamclip handles QT files. I’ve emailed Squared 5 who make Mpeg Streamclip but I haven’t yet had a reply from them.

Meanwhile, my only way out is a tedious process of ingesting the Nanoflash files into FCP, and on the timeline, panning ch1 audio to the left, and ch2 audio to the right, then re-exporting. I tried downconverting to DVCPro50 at the same time but the results aren’t anything like as good as MPEG Streamclip, which is a wonderful product, allows me to join up blocks of files into 30minute clips which my editors like, and batch process any number of files overnight. And Mpeg streamclip is, unbelievably, free.

I want to keep using the Nanoflash, not just because of the increased colour space (the final edit is colour graded before broadcast), but I like the pre-roll and what I thought was a simpler work flow. As I often have to deal with all this file processing at night in a hotel room after the day’s shoot I’m looking again at just MxM cards as the simplest way to go.

I'm a regular reader of DVinfo, and i've gleaned a lot of good info from the posts, but haven't posted anything for a long time. Any ideas would be appreciated.

Dear Richard,

As I understand your post, you are saying that we mix Channel 1 + Channel 2, then put the sum of both into Channel 1 and Channel 2, thus giving a mono mix.

In my testing, I have not found that to be the case.

We will specifically test this on Monday.

What happens when you upload our QT files directly into Final Cut Pro?

I hope some users of the nanoFlash and FCP will share their experience.

It seems unlikely that we would have a major issue like this go undetected.
But, we will test it as soon as possible. My tests, in Sony Vegas, this week, did not show the two channels being combined.

Audio issue downconverting to SD
 

Thanks, Dan for replying so promptly.

Your summary of what appears to be happening to my audio tracks is right, though I’m not saying YOU are doing it, just that something strange appears to be going on.

I’ve just run a test again to make sure my observations are consistent. I’m using a Macbook Pro, 15inch, with FCP 7.01.

I’ve recorded two files with distinctly different audio on the two tracks on CF cards in the Nanoflash and MxM cards in the EX3, simultaneously.

Listening to the files on camera, MxM through the camera’s headphone jack, and the nano through its headphone jack, the two audio tracks are clearly distinct and seperate

I now copy the CF files to a hard disc, then import them to FCP and put them on the timeline. I can quite clearly see the two audio tracks with their distinctly different content, as expected. But if I now listen with earphones, the distinction disappears. I’m hearing mono, mixed.

Exactly the same thing happens with the MxM files, imported to FCP via Log and Transfer. I can see the distinction in the time line but the difference disappears when I listen.

Playing the files out of FCP, in QuickTime, is the same. I can hear no difference between the right and left audio tracks with either the CF files or the MxM’s.

Now process through Mpeg Streamclip. The nanoflash files annoyingly come out with two identical mono tracks, but the MxMs actually retain the two audio tracks as two distinct left and right channels, as originally recorded.

It’s bizarre, isn’t it? The Macbook pro is new, and I haven’t fiddled with it. It has to be something to do with the way mpeg streamclip handles quicktime files. However there must be something about the EX3's files and the Nano's to make Mpeg streamclip react differnetly. This wouldn’t be a problem if my editors could work directly with the nanoflash files - they use FCP - and panning track1 to the left and two to the right. But they don’t. I have to deliver in DVCPro50 and that’s that. So probably not many shooters are in my position, but I wonder all the same if anyone else has come across this strange little anomaly, or is it just me?

Richard Corfield,
Sydney Australia

I'll try to do some more specific testing in the next day or so, if you like, but here's what I found when importing some nano .mov files into FCP. The two audio tracks are imported as two mono tracks, not a stereo track. If one doesn't specifically pan channel one to the left channel, and channel two to the right channel, on output from FCP the two channels will both be a mono mix of the original two (distinctly different) input channels.

Importing nano files was the first and only time that the two (stereo) channels had to be panned left and right; with every other format I've ever imported into FCP, they came in already panned properly.

But this isn't something specific to FCP (unless it's specific to its CODEC which Quicktime needs to read your files), and you don't need FCP to check this out. All you need is the pre 10.6 version of Quicktime player (which Snow Leopard will leave installed for you, renamed to Quicktime Player 7). The Snow Leopard version of the Quicktime Player doesn't have much functionality (yet), but the 10.5 version of Quicktime Player does. Open a raw nano .mov file in Quicktime Player 7 (the older version), play it to verify you are hearing it in mono, even though you should be hearing it in stereo.

Then open the Movie Properties window (clover J), select either of the two Sound Tracks, click on the Audio Settings tab, and note that the channel assignment is Mono (not Left or Right), and that the balance fader is centered. If you drag the Balance fader all the way to the left for Sound Track 1 and all the way to the right for Sound Track 2, and then play your .mov again, you'll hear it in stereo. However, there already seems to be in place a system for defining what a track's assignment is. Instead of dragging the Balance fader to one side or the other, note that in the little sub window to the right of the fader, there's a pop menu to define the track's Assignment. You'll see that there are 17 choices, including Right and Left. If you leave the Balance (pan) setting in the middle (called Center), but change the track assignment to Left for Sound Track 1 and to Right for Sound Track 2, then you don't have to bother with the Balance control at all, and the .mov will play properly in Stereo.

I don't know if there's flag word in the SDI stream or in the nano CODEC (or both) that specifies track assignment, but I strongly suspect there is, and you just set it to Mono for each audio track, or don't set it at all...

Now if my supposition is correct, you "just" have another configuration item to create. :) For now, a single setting for Stereo assign or Mono assign is all that's necessary, but when you finish up eight channels of audio, you're going to have to take into account eight tracks of audio track assign as part of the record process. The popup menu I mentioned above will give you hint as to what they are...

I still love the nano. (And I still wonder if you're EVER going to implement hot swapping).

Onward Thru the Fog.

Billy

Dear Billy,

Thank you for your testing, experience, and very informative post.

We will look into this. This does get more complicated with 8-Channel Audio.

It does seem like we need an option for "Left, Right" to properly describe the audio. For others, Mono may be appropriate.

Yes, "Hot Swapping" was promised, and will be delivered. We needed to drastically change the intrastructure, in the nanoFlash, in order to support "Hot Swapping".

Dan,

By change in infrastructure, do you mean some hardware modifications will be needed? If so, this doesn't sound optimistic :(

Dear Piotr,

Sorry, but no.

We had to had change the way we handled certain things in the firmware.

To put it differently, we had to rewite some of the basic code (software/firmware routines) in order to offer "Hot Swapping".

As I understand it, these changes have been completed, but we have not finished "Hot Swapping" yet.

No hardware change was, is, or will be required.

Dear Dan,

No need to be sorry - especially that:

This is great - as you know, sending the units for upgrade would be troublesome for EU customers.

Thanks!

Piotr

Billy,

Thanks for your detailed reply to Dan's comments. I followed all the steps you listed - didn't realise Quicktime had all those options - and for the first time i think I understand what's going on ... ie lack of some sort of flag in the nano's .mov file so that quicktime knows where to pan the files. I thought i was going crazy!

I'll wait to see what develops. What a great forum this is.

Richard Corfield
Sydney, Australia

Richard: Great forum indeed. Glad to have been of help; it's nice to give back.

Dan: I knew you had your work cut out when I figured out what the cause was. It'll probably be relatively easy to add the flag code, but a royal pain in the butt to figure out a good user interface to set it. Maybe a miracle will happen and there's already an assign flag in the SDI stream, which you can then just pass on to the CODEC; no interface needed.

:)

If I get a chance later, I'll look in the manual for my SDI scope's audio analyze chapter. It has all the info about the SDI audio secondary data fields. (Though I vaguely remember once setting them up on my scope and needing to manually define the assign of each audio channel. But I don't remember if it was required or just a means to change the default).

Billy

ps Dan: Just read the SDI scope manual. The initial channel assignment settings are L:CH1, R:CH2, SL(S):CH3, SR:CH4, C:CH5, LFE:CH6, RL:CH7, and RR:CH8. Don't know if this is part of a specification or not, and the labels are changeable on the scope display (e.g., I can label channel 1 as SR if I want to).


pps Dan: Two more things. Most multichannel audio sources in Quicktime show up structured as a single audio "track" with multiple channels. The nano .mov files show up with discreet audio tracks, each containing a single channel. I do not know the significance of this (if any), but since you're about to be looking into this part of the nano code, I thought I'd mention it, so you can check it out at the same time. And two, here is a little picture of the popup menu with all the Quicktime audio channel assign choices:

Dear Billy,

Thank you. I find your posts to be very helpful.

We will be discussing this today.

You should check your audio set up in Final Cut before outputting anything. it could be defaulting to the mix to mono output which technically it would do with any sequence in a project. There may be a preference but not sure.

Interesting thread - one thing I always hated about Premiere and FCP was the way it combines mono tracks into a stereo track. As most of the audio I record in the field is dual mono and not stereo I want my NLE to offer me separate tracks for each audio channel. Yes I know you can go in and split your tracks - but that just seems like another completely unnecessary step.

I prefer how Avid works - the default setup is to have CH1 panned left, CH2 panned right CH3 panned left etc. You can change this to have all tracks centered if you want. Each audio channel is discrete - doesn't mean you can't have stereo, or 7.1 channel sound if you want.

It seems it is a Quicktime "feature" that is actually screwing things up. As you are using MPEG streamclip I'd be interested to hear if the same thign happens in compressor if you have a Mac.

I think the newer versions of premiere - CS3 and CS4 - preserve the dual audio tracks in .mxf files.

I'm using a Mac laptop now, and, yes, the same thing happens in Compressor which will output the audio from my nanoflash files but frustratingly mix left and right together into two identical audio tracks. I must add that the video downconversion in Compressor isn't nearly as good as that from mpeg streamclip.

The SD/SDI out of the EX3 is also great, and can be recorded as SD in the Nanoflash. Same excellent quality as Mpeg streamclip, but still suffers from the mixing together of the audio tracks.

I'm waiting in hope that this mini-bug can be fixed.

Richard Corfield, Sydney, Australia

Dear Richard,

We will try to address this as soon as possible.