what do you make if this "24bit" show?

[heath]

weak

[dmonterisi]

i guess you could send your complaints in to the forum software authors.

http://www.lewismedia.com/

[Brian Skalinder]

Any text file is subject to misinterpreted. I am open for suggestions. I try to be as detailed as possible as you can tell by the super lengthy text file included. At the core of the controversy with the text file and the recording is this..."I did not upsample either source to 24 bit. The multitrack mixdown was done in 24 bit and the 'new source' is now 24 bit". Or some crazy variation, thereof. That's what were are looking into. I don't want to confuse or mislead anyone and I don't think I am.

Thanks for stopping in, Dan.  FWIW, I don't feel you were intentionally confusing or misleading, but I do believe the info nonetheless confuses a lot of people.  The "no upsampling" comment in particular.  I've traditionally seen - mostly in the playback world - the term "upsampling" used to apply to both increasing the sample rate and/or bit-depth.  Thanks to luvean's post, it seems the proper term for changing the bit-depth, either up or down, is dither.  I think we all knew about down, but didn't know dither could also apply to upping the bit-depth (though it makes sense now that I think about it).  At any rate, my point:  the confusion / discussion over converting two 16-bit files to 24-bit has more to do with the general level of knowledge (or lack thereof) around converting 16-bit to 24-bit.  End result:  good discussion all around!

I've put out all info I have on this and we are trying to see if it is worth it. That what was done and I think we all know that. Better description I can include on the text file anyone??

In hindsight - given how often the playback industry uses "upsampling" to apply to both sample rate and bit-depth - I find the "no upsampling" comment a bit misleading, though obviously not intentionally.

I think that is at the heart of what we are trying to get at, no? That's all I am interested in. At the end of the technobabble rainbow, which sounds better. Right?

It's not gonna hurt my feeling if the debate lends itself to the common consensus that saving, distributing and most importantly listening to this 24 bit version is a waste, or possible worse than the same mixdown at 16 bit. I am very concerned that the extra bit are just noise even if ambient agreeable noise. Are the extra bits desirable, useful, the opposite? I feel like we still don't know. I don't trust my "real world" listening comparison enough and hope someone can do their own. I think we still don't know the answer here and curiosity is tweaked. One thing is true. I can save allot of time, money and bandwidth by staying with 16 bit.

You may be interested in Nick's personal experiences on true 24-bit sources (not upsampled/dithered) v. playback upsampling from 16- to 24-bit:   

http://www.taperssection.com/yabbse/index.php?topic=21376.0

Unfortunately, I've not yet done any testing myself - either similar to Nick's or comparing your 16-bit v. up-dithered 24-bit - as I haven't entered the 24-bit playback world.  One of these days...

[dklein]

I haven't downloaded the files yet but what Dan has done makes total sense to me.  

Changing the word length / bit depth is easy when you're making it bigger.  You just add zeros (not noise or dither).  Money analogy: $12.59 can be expressed as 12.59 or 12.590000000000
- nothing destructive going on here
- making it smaller is where the tricks come in (how would you round/truncate this in whole dollars).  This will come later.

Dan is combining two sources and probably performing at least one other step like normalization, maybe even some eq.  Most audio editors will perform all the calculations at higher bit depths (like 32 fp) and dither back down after each operation.  
- as soon as you start doing multiple steps on a value, all of those extra decimal places in the example above could really come in handy but are only available if you're working in a higher bit depth..

Back to the money analogy - it's like doing a bunch of calculations where you had to round to the nearest dollar after each step (16 bit mode).  The alternative is to run CEP in 32 bit fp mode, allowing you to keep these longer, more accurate working values while you edit and then do your reduction at the end.  This produces a more 'accurate' result.

So if you're in 16 bit mode and you do multiple steps (mixdown, eq, amplitude change) then you're better off doing the whole thing in at a higher bit depth and doing one dither at the end.  And in this context, I would think that the 24 bit version could very well sound better when compared to the 16 bit version if all the destructive editing was done at a higher bit depth and dithered down at the end.

Note that this applies to single source recordings as well (not just matrixes).  If you are going to perform more than 1 step of destructive processing (including 'normalization') then in theory, operating at a higher bit depth and dithering at the end produces a more mathematically accurate version that has only been subjected to one dither stage instead of one per editing operation.

[Brian Skalinder]

I haven't downloaded the files yet but what Dan has done makes total sense to me.  [snip]

Great explanation, David - and you beat me to it (good thing, as you explained it better than I would have), I was digging through CEP to verify it performs internal processing at 32-bit fp.  (It does, though I believe can be configured in multitrack mode to process at 16-bits if desired.)  I think the confusion comes, for me, anyway, from the "no upsampling" comment in the text file as I'm used to seeing the term "upsampling" applied to both sample rate and bit-depth.  At any rate, already blathered on about that in my previous post, so that's enough of that.

[Nick's Picks]

As Heath (hey man) as pointed out my setup is less than rudimentary.  However, I'll need to correct him as I do not live on a houseboat but actually a Motoryacht.  Short summary is 4 digital surround systems (from the low end $500 in bedroom to a $2500 in the main salon).  All 4 units speak to each other and you can walk up from the master stateroom to the salon into the aft cabin and then up to the flybridge as 24 speakers and 4 subwoofers fill each room.  The system also incorporates a 12v system that is wired throughout and is also in the loop.  At that point you can have 34 speakers and 6 subwoofers in unison.  Beside the audio capacity she is also equipped with 7 digital flat panel monitors including a 42 inch plasma.  Just a wee shy of a boom box and a dinghy.

A brilliant discussion indeed!
the one thing that stands out as a true realization for me is that Dan needs to through a fucking party!!


And thats a great analogy as well.
Why is it that money always seems to make more cents?

(rimshot)...thank you...I'll be here all night!

[Tim]

Just saw this...

Dan thanks for the explanation. No harm intended, pithy comments are just my way. It just wasn't clear at all from the .txt file or your original response what was going on, they were pretty ambigous. Thanks for the clarification...

[heath]

Just saw this...

Dan thanks for the explanation. No harm intended, pithy comments are just my way. It just wasn't clear at all from the .txt file or your original response what was going on, they were pretty ambigous. Thanks for the clarification...

yeah dan, don't ever take Tim too seriously....he's always a douche 

[rustoleum]

For the record, I haven't downloaded this but I find the conversation very interesting.  Now just to add to the confusion   

Let's suppose that:

   Source A is a 16bit source... therefore it contains up to 96 db of info

   Source B is also a 16bit source and it also therefore contains up to 96 db of info.

So, what happens if we mix down these two sources in a 24 bit environment, but offset each source by say 12 db (2 bits)?

We could end up with a 24 bit recording where 18 bits are actually in use (108 db of dynamic information vs the original 96).   I have no idea if the difference between mixing down 2 16bit recordings into a single 16bit recording vs this 18 bit dealio would be audible, but the concept of offsetting the sbd w/ the aud is akin to making a matrix mix where the sbd is higher in the mix than the aud.

Thoughts?

Edited to create a picture of what I was saying instead of trying to deal with this darn editor and the spacing issues that occur.  The column on the left represents the 24 bit source that is generated from the two 16 bit sources.

[rustoleum]

One more thing... I believe my drawing is a little misleading in that it would be the top 6 bits that would not be used in this example.  Everything would be shifted 'down' instead of having the bottom 6 bits not used as the drawing illustrates.  I could be wrong on that point, though.

[John R]

 

As Heath (hey man) as pointed out my setup is less than rudimentary.  However, I'll need to correct him as I do not live on a houseboat but actually a Motoryacht.  Short summary is 4 digital surround systems (from the low end $500 in bedroom to a $2500 in the main salon).  All 4 units speak to each other and you can walk up from the master stateroom to the salon into the aft cabin and then up to the flybridge as 24 speakers and 4 subwoofers fill each room.  The system also incorporates a 12v system that is wired throughout and is also in the loop.  At that point you can have 34 speakers and 6 subwoofers in unison.  Beside the audio capacity she is also equipped with 7 digital flat panel monitors including a 42 inch plasma.  Just a wee shy of a boom box and a dinghy.

i would have loved wiring this system, and of course, watching and listening.  how about some pics of the headend, dan?