Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[jason levy]

Hello,

For a video project where the original source files are  16 bit, (at 48 ) does it make more sense to edit and mix in 16 bit (via software) or to upconvert everything to 24 bit and dither it back down to 16 after bouncing to disk?

Thanks

Jason

[Level]

All my DVD's are 24 bit in audio. For that, I see no reason to go back to 16 bit...however, if it started at 16 bit to end up at 16 bit, audio editing advantages of going to 24 bit may not outweigh the disadvantages of going back to 16 and adding dither.

To each his own. Try both ways and see for yourself. Your software may provide some surprises.

[henchman]

If it's a video project, it will likely end up on Digibeta which is 16 bit anyway.

So, yes, just stick with 16bit.

[bobkatz]

jason levy wrote on Sun, 03 April 2005 08:14

Hello, For a video project where the original source files are  16 bit, (at 48 ) does it make more sense to edit and mix in 16 bit (via software) or to upconvert everything to 24 bit and dither it back down to 16 after bouncing to disk? Thanks Jason

As soon as you mix, changing gain or other processing in the mixer, the output of the mixer will be the wordlength of the DSP, it will be LONGER than 16 bits. If it's a native system, it is likely to be 32 bit float. You should not have to convert your source files to longer wordlength, but yes you will have to save the output as 24 bit or dither it down prior to saving. In fact, if it's a native system, you really should dither it DOWN to 24 before saving it as 24 bits, to avoid the small quantization distortion from the 32 to 24 bit conversion.

BK

[henchman]

But if it is a native system, and the final is 16 bit, won't bouncing the final to 16 bit already apply dithering?

[bobkatz]

henchman wrote on Sun, 03 April 2005 10:26

If it's a video project, it will likely end up on Digibeta which is 16 bit anyway. So, yes, just stick with 16bit.

Unfortunately, both statements are incorrect.

#1. Digibeta (Digital Betacam) contains 4 channels of 20 bit audio, at 48 kHz sampling.

#2. Even if you start with 16 bit files, as soon as you put them through the DSP/mixer, your wordlength will grow. So, if you are going to put them down to a Digibeta, after mixing, they will have to be dithered DOWN to 20 bit.

Ironic, start with 16 bit, and you end up having to dither DOWN to 20... all part of the way digital audio works.

BK

[henchman]

Yes, this is true that the Digibeta's have 20 bit cahnenls. Howvere most post-transfer house require DA-88's for laybacks.

I think the big problem with the original question is that there is not enough info.

Just saying it is a video project is not enough.

What's the delivery format, and where will it end up.
What kind of video project is it?

[wwittman]

Quote:

all part of the way digital audio works.

If one can say digital recording ever "works" {g}

one learns to lessen one's expectations.

[jason levy]

Thanks everyone for the input.

The original audio was recorded on Panasonic DVC pro 50 vtrs. It is then captured to to 16 bit 48k files in a Final Cut Pro system via SDI with embedded audio using a Decklink card. (www.decklink.com). All that is out of my hands. (We also get material that originates on Betacam SP analog, also captured on a digibeta deck (reads the SP as well) via SDI with embedded audio)

I receive the audio as an OMF file. 16 bits 48k. I edit and mix and must deliver an SDII (or AIFF) file that is laid back to Digibeta via Final Cut Pro and a Decklink card, via SDI embedded audio.

I notice that final cut will accept 24 bit files but I don't know if it plays them natively or does some kind of conversion on the fly. We do need to make DA88s of the stems as well as a mix.

The system we work on is based on Digital Performer software, running on a macintosh computer. Digital Performer will run in either 16 or 24 bit mode.

Again thank you everyone for your comments. They are appreciated. If there is anything I have left out I would be happy to supply that information.

Bob thanks for the comments. Just got your book on mastering and am trying to plow through it.

Jason Levy

[Level]

What is the delivery (final) format?

If it is DVD, go to 24 bits and stay there.

[jason levy]

Final Deliver format - Digital Betacam.

Jason

[henchman]

jason levy wrote on Sun, 03 April 2005 13:41

We do need to make DA88s of the stems as well as a mix. Jason Levy

Save yourself the hassle. keep it 16 bit.
There's absolutely no reason to waste time fileconverting back and forth.

By the time this hits the airwaves, the broadcaster will have screwed with it enough that you could do it at 8-bit, and most people couldn't tell the difference.

[bobkatz]

jason levy wrote on Sun, 03 April 2005 13:41

I receive the audio as an OMF file. 16 bits 48k. I edit and mix and must deliver an SDII (or AIFF) file that is laid back to Digibeta via Final Cut Pro and a Decklink card, via SDI embedded audio.

Assuming that the Digibeta is the final master, then if you dither your AIFF files to 20 bit in a 24 bit file, it will be fine. The Digibeta will truncate the bottom four bits, but you already predithered it.

Final Cut will work with the 24 bit files and if they don't change gain or EQ within Final cut, then your predithering to 20 bit will hold. I'm not sure which plugin that you can get for performer will dither to 20 bits, I suspect the Waves L2 will do it.

However, never turn your back on digital. If you can convince the video house to get a bitscope it would be nice.

BK

[henchman]

But why would you waste the time and effort?

Seriously, converting everything to 24 bits, for Broadcast post is a waste of time.
You just won't hear it. We're not talkign high quality classical recordings here.

You'll usually be dealing with medicore location sound.
We do everything at 16 bit 48k.
99% of the SFX libraries are at 16 bit.

The conversion thing really is a waste of time IMO.

[Joel Silverman]

henchman wrote on Sun, 03 April 2005 15:41

You'll usually be dealing with medicore location sound.

and what about the VO being submitted to you as MP3s

[henchman]

Joel Silverman wrote on Sun, 03 April 2005 17:45

henchman wrote on Sun, 03 April 2005 15:41 You'll usually be dealing with medicore location sound. and what about the VO being submitted to you as MP3s

Exactly. So why bother worrying about 24bit at all, or even the small about of quantization noise.

Again, my advice. Stick with 16 bit.

[Otitis Media]

I agree with Hench on this one.  While I don't have his vast experience with post, I did mix post for 2 years as a 9-5 gig.  I'd routinely receive 64kbps .mp3s for vox, so having even 16 bit linear PCM audio was a luxury!  Working with 24 bit is nice and all, but for broadcast post, it's not a requirement.  

I'd wager that even for features, 16 bit well recorded would be more than adequate.  

[henchman]

Otitis Media wrote on Mon, 04 April 2005 07:52

I'd wager that even for features, 16 bit well recorded would be more than adequate.

We do it all the time.

[bobkatz]

henchman wrote on Mon, 04 April 2005 12:05

Otitis Media wrote on Mon, 04 April 2005 07:52 I'd wager that even for features, 16 bit well recorded would be more than adequate.   We do it all the time.

Sure, as long as you properly deal with all the in between steps...

BK

[jason levy]

Thanks everyone for your input. The more one finds out the less one knows about this complex subject. I'm still trying to understand the issues involved.

Ok  - here's a question. GIven that DIgital Performer's internal processing is 32 bits :

A)  Assuming that the final output will be 16 bits, is it not just as good to work natively with the 16 bit files as opposed to up-converting all the files to 24 bit before (editing and mixing and then) bouncing to 16 bit with dither?

My impression is that given that the original files are 16 bit then there is no difference between working with the 16 bit files or working with files that were originally 16 but then converted to 24.. In both cases you are starting with 16 bits... working in a 32 bit environment, and then dithering down to 16 ( is my conjecture.. am I correct?)

B) What if we bounce to 24 bits from within this 16 bit project? That should preserve more of the quality of the 32 bit internal processing, no? and we can deliver the mixes in 24 bits which FCP can import and use.. ( though I'm still not clear on what FCP is doing with all those bits.. the technical data on the apple website is sketchy).

I assume that in this case i would set the Waves L1 Plug to dither to 24 bits? Or perhaps to 20 bits as someone mentioned given that the final destination is 20 bit Digibeta? I'm just not clear if FCP and the Decklink card is passing the data unaltered to the Digibeta or doing some processing of it's own?? Surely it must be doing something to get the 24 bit file encoded into the SDI stream as 20 bits? or is it 24 bits in that stream and is it the Digibeta that is reducing it to 20 bits?

I have spent hours on the web but can't find the answer to that; how many bits is embeded audio in a SDI stream? is it a standard? can it vary? Downloaded the SMPTE SDI standard ($22 on the Visa card).. no mention of audio.

Sorry, I think I am rambling. That's the way it's been going lately ... all questions and no answers.

Anyhow I am getting an education.. Again thank you for all your input.

Jason

[jason levy]

bobkatz wrote on Sun, 03 April 2005 16:31

Even if you start with 16 bit files, as soon as you put them through the DSP/mixer, your wordlength will grow. So, if you are going to put them down to a Digibeta, after mixing, they will have to be dithered DOWN to 20 bit. Ironic, start with 16 bit, and you end up having to dither DOWN to 20... all part of the way digital audio works. BK

I now realize that what I was just asking was what Bob suggested 2 days ago. I had no idea what he meant then..  You can learn a lot in two days!

Thanks,

jason

[Otitis Media]

Here's what I would do:

use the 16 bit files as they are.  Mix in the DAW - on the master fader insert Wave's IDR or PowR dither set at 16 bit.  Provide deliverables.  

Sure, the DBeta is 20 bit, but the rest of the chain may not have that resolution.  If it sounds good at 16-bit, the extra 4 aren't going to make any real-world difference, IMHO.

[canada]

IDR sounds better than L2 imho.  Good call.

[Otitis Media]

L2 has IDR built in, IIRC.  Either way, it may not be necessary to use an L2 on your master, or you may want to use a different plugin or hardware unit, whatever.  The last insert on your master fader, after anything else, should be dither - even Digidesign dither will do the trick.  I suppose you may not hear anything if you don't dither, either, but it's easy enough to pop a plugin in there and let it do it's thing.

[canada]

If the IDR coding is the same shit built into the L2, then /golfclap myself.  I figured even though it had the big red "IDR" button built into it, the L2 dither might be different.  Haha.

Probably a good idea to have it separate if you have both plugs (I think they come together most of the time) though, just so you're always aware of it.  Sucks to have to do multiple mixes because you forgot about a setting buried in a random plugin.

[Otitis Media]

What's your issue with the L2's IDR dither?  I use Pow-R most of the time now, but really have no issues with IDR, either.

[jason levy]

Thanks everyone for the information and suggestions.

jason

[Ronny]

henchman wrote on Sun, 03 April 2005 15:26

jason levy wrote on Sun, 03 April 2005 13:41 We do need to make DA88s of the stems as well as a mix. Jason Levy Save yourself the hassle. keep it 16 bit. There's absolutely no reason to waste time fileconverting back and forth. By the time this hits the airwaves, the broadcaster will have screwed with it enough that you could do it at 8-bit, and most people couldn't tell the difference.

Sorry, this is not the best advice, IMHO. He said that he was mixing, that means further processing. All high end digital consoles, eq, fx and dyn processors process at high resolution. Native processors typically process at 32 float. High end digi processors at 48 and 64 bit. The reason why digital consoles with only 20 and 24 bit converters process the signal at 32 float is because by floating the decimal you had more accuracy to the processes. For example the Yamaha DM2000, 24 bit A/D, 32 bit native buss, eq has a 56 bit accumulator. There is absolutly no degradation when you process at higher res and return to lower native rate and it's what every digital processor designer that know his salt does. The way it works when you record a 16 bit signal to a 32 float system is, 16 empty bits are added. These bits remain empty until they are utilized in a further process. If you analyze that raw 32 bit recorded signal, it will give you an actual read of 16 bits, as the bits haven't been activated by a process yet. Let's say you  process with a native 32 bit eq with a 48 bit accumulator, the processor will utilize the empty bits for many more possible calculations increasing accuracy, they will be filled and the processor will output back to the 32 bit native depth. You analyze again and the upper 16 bits are now filled and you get an actual 32 bit read.

My suggestion, record or load in, the 16 bit signal at the highest resolution that your system supports, perform your eq, dyn's, fx and leveling processes, after you are finished mixing, dither and reduce word to 20 bit and transfer the stereo bus to the final 20 bit destination. Someone mentioned L2 will be the final process and by the way the L2 dither is IDR, just set the L2 IDR to 20 bit and you can transfer straight to the destination. The L2 has a good bit converter. If you'll read the quality DAW manuals, you'll see that almost all of them recommend raising 16 bit depth material to 32 float or 24 native if that's the max, before performing any processing, including minor processing like changing track gain levels, which will also utilize the upper empty bits, as a gain change is a process unto itself, although it doesn't benefit as much as raising depth to perform an eq process, which requires more processing power.

[jason levy]

Thanks for that informative post. Actually we are using L1 but I was considering investing in L2. Do you think there is a quality gain to be realized by that purchase?

We are mixing in Digital Performer which I am told is uses 32 bits internally. It does not allow mixed format projects so one must either select 16 or 24 bits for the material. The material we receive is all 16 bit. I am not 100% on this, but have more or less come to the conclusion that there is not  any advantage to converting the files to 24 bit to open them in a 24 bit project as even the 16 bit projects are 32 bits internally (unless I am misunderstanding something) and thus (as Bob mentioned) must be dithered DOWN to what ever the output is 16 or 24. I appreciate any corrections if I've got any of that wrong.

Thanks for the suggestion to dither down to 20 bit. We are going to try a test and see how that sounds.

Jason

[henchman]

Ronny wrote on Fri, 15 April 2005 00:17

Sorry, this is not the best advice, IMHO.

That may well be your opinion.
But as soemoen who has mixed hundreds of hours for TV, I can tell you, it'sd not worth the hassle and time converting everythign to 24 bit.

Remeber. We're not taqlkign about a 3 1/2 minute song here.
We're talking an hours worth of stuff. When you add up the time that it would take to convert everything on import, it's just not worth it.

[bobkatz]

jason levy wrote on Fri, 15 April 2005 07:45

Thanks for that informative post. Actually we are using L1 but I was considering investing in L2. Do you think there is a quality gain to be realized by that purchase? We are mixing in Digital Performer which I am told is uses 32 bits internally. It does not allow mixed format projects so one must either select 16 or 24 bits for the material. The material we receive is all 16 bit. I am not 100% on this, but have more or less come to the conclusion that there is not  any advantage to converting the files to 24 bit to open them in a 24 bit project as even the 16 bit projects are 32 bits internally (unless I am misunderstanding something) and thus (as Bob mentioned) must be dithered DOWN to what ever the output is 16 or 24. I appreciate any corrections if I've got any of that wrong.

To summarize: You can make a 16 bit project, no problem; that's just to accomodate your source files.*

For Digital video you should be working at 48 kHz sampling. The output of Performer's mixer will be 32 bit float. Insert a dithering plugin at the end of the mixer to dither down to your destination format. It can't hurt and it will probably help! If the destination is going to be digital video tape, dither it down to 20 bits, this will fit in the 24 bit file length and output 4 zeros at the bottom.

It will be extremely hard, if not impossible, to hear the difference between dithering down to 20 versus 24 bit.

Simple... straightforward!

* However, if you are going to do any internal captures or bounces of processed material and use them within your mix, then you should convert your project to 24 bit ahead of time so as not to add any further degradation. In other words, this is so that your bounces can be 24 bit files. Pity that Performer does not incorporate 32 bit float files, because then you would not have to dither its mix buss to 24 to capture those bounces you might make.

BK

[Ronny]

henchman wrote on Fri, 15 April 2005 09:12

Ronny wrote on Fri, 15 April 2005 00:17 Sorry, this is not the best advice, IMHO. That may well be your opinion. But as soemoen who has mixed hundreds of hours for TV, I can tell you, it'sd not worth the hassle and time converting everythign to 24 bit. Remeber. We're not taqlkign about a 3 1/2 minute song here. We're talking an hours worth of stuff. When you add up the time that it would take to convert everything on import, it's just not worth it.

I see your point if time is a factor. However, the key phrase here is "further processing", it's not just my opinion, but as I mentioned all quality gear designers will increase word length at their processors input to more accurately process the signal. It also depends on the gear, I can print a word increase on an hours worth of stereo material in less than 5 minutes, if I couldn't, I might not think that it's worth it to convert per individual project, but that doesn't change the fact that increasing word allows for better processing of the digital signal. In the case of the poster, he's already having to increase word to 20 bit from the 16 bit original, therefore my advice to load in the material at the highest res that his system supports, perform the eq, dyn, fx and level blending at the higher depth, is advice that most digital gear designers would recommend, time not being an issue of course.

WRT, the question about the differences between the L1+ and the L2. I have L1 and L2 software. I sold my hardware L2 when I bought the L3, but have L1, L2 and L3 software versions now. While the L2 typically can get me a db or two more threshold gain before degradation than the L1+ and it has improved 9th order noise shaping, the L1+ has some advantages. One it can be used as a multi-track "mono" limiter and a perceived level stereo limiter, whereas the L2 is really a last in the chain stereo limiter. My advice is, if you don't have the Waves linear phase processors, to look at the L3 instead of the L2. Basically it's an L2 with the LP multi-band compressor built in, so that you can get a little hotter gain on most material than the L2. It takes a little practice learning to tweak it to perfection, but as far as Waves limiting goes, it's the best perceived gain limiter that they've come out with to date and has made the L2 all but obsolete.

[jason levy]

Thanks to y'all for all that advice. I appreciate everyone's having taken the time to comment on my itsy "bit"sy problems..

Jason Levy