I can't read Gearslutz. It makes me genuinely depressed.

VastmanGoddard, one point interests me in particular, as it relates to increasing RECORD BIT DEPTH  to 32, which I am going to try after reading this:
 

 
Setting Record Bit Depth to a value higher than your audio interface is outputting (typically 24 bits) does NOTHING but waste disk space and require more I/O. 
 

I would be very much interested in 32 bit float if it reduces the risk/digital distortion associated with level spikes or accumulations thru combined signals... especially if it doesn't interfere with overall DAW performance too much as I still have margins available as long as I make love to DIVA sparingly!
 

 
Sonar (and many other DAW's) already process using 32bit floating point or higher. Don't assume that the bit depth (or sample rate) input or output to/from a program/plugin/process is what is used internally for processing.
 

Speaking of "love"...I love ambience...stereo imaging, deep effects pushing and pulling at our spirits...I've acquired a lot of them and I plan on using them...so,  lot's of data to be massaged...and will 32bit float handle what I see as magical overload better??? 

 
"lot's of data" is a relative term. What you consider to be "lot's" might be a joke, mathematically speaking.
 

Please, gurus.... answer me that...with real world data or mathamatical extrapolations... as climate scientists are trying to do every frackin' moment we have left...
 
I LOVE open, clean yet emotionally powerful music and while I recognize most of that is achieved via mixing/mastering skills/tools... why not also employ this "edge" or "hedge" as I'm quite sure HZ and every other major producer does... and probably for a reason...
 
better safe than sorry??? What's the cost?  Doesn't it make sense till available cpu overhead is exhausted? Hell, by then I'll have moved to an 8 core haswell! What joys befall a lowely gardener doing eco/save the world music these days!

 
Human hearing does not have infinite resolution. You only need a healthy margin of error around what might conceivably be audible. After that, you gain zero benefit from adding more and more and more.
 
Are you sitting in a chair? The chair was no doubt designed to support a given amount of weight without collapsing (with a healthy margin for error). Should it instead be designed to support 900,000 lbs. just because?
 

 Such things should be measurable... and I am a deep believer in subtle nuances of nature and reality all around us which we can not necessarily SEE but are profoundly influential... 

 
It's not always obvious to people, but anything that makes it past your analog to digital converters has been measured. 
 

Sadly, most childeren, including my own youngling will die of truths being distorted in an emotionally laden fashion...

 
Truth is independent of emotion. That's the beauty of objectivity. So if you and your younglings just learn to ignore subjective rants in favor of objective evidence you can be happy.  
 

and anyone... Does 32bit float even touch on what I've postulated above?  Links, please if u got um...
 

Floating point math has certain advantages and disadvantages compared to integer formats of the same bit depth. A programmer will generally use whatever is appropriate for the task at hand.
 
In terms of audio, it is complex. I would be wary of what you read in audio circles, because there's a lot of bad technical information out there presented authoritatively and often the only way you can tell if something is accurate or not is if you already know the answer. 
 
I would suggest learning about the limitations of human hearing first, so that any mathematical discussions can be put in the context of what might be audible. You can find lots of oddly emotional, supposedly mathematical discussions where audibility is never discussed. But once you get to the point where something is never ever going to be audible, talking about mathematical "improvements" has nothing to do with audio. The argument that, "if you could hear it, it would be better!" is academic if you can't ever hear it.

John
I think Goddard is confused about 1 bit audio recording and the rest of the audio recorders. 
 
This may clear up the confusion http://www.bhphotovideo.c...bit-better-24-bit.html
 
This also may help. http://en.wikipedia.org/wiki/Direct_Stream_Digital

I'd be wary of any source that claims DSD is superior.
 
And I think Goddard might just be pointing out that many converters use oversampling.

Quite possibly. But the article he's complaining about says that, so it's hard to work out what his beef is.

drewfx1

John
I think Goddard is confused about 1 bit audio recording and the rest of the audio recorders. 
 
This may clear up the confusion http://www.bhphotovideo.c...bit-better-24-bit.html
 
This also may help. http://en.wikipedia.org/wiki/Direct_Stream_Digital

I'd be wary of any source that claims DSD is superior.
 
And I think Goddard might just be pointing out that many converters use oversampling.

You're right but the one that is selling stuff is meant only to show the difference. Having never had occasion to hear or use DSD I have no opinion on its fidelity. 

I'm a few days late to the party here but I appreciate the link to the original article/post.
 
In order (as a longtime member but newbie multiple poster here) to go ahead get on the naughty list, I'm going to agree with a lot of the assertions in that piece while I am at it. To me it boils down to 2 words in it, somewhere around them middle of the piece - diminishing returns.
 
I too used some of the lovely Alesis ADATs back in the day - I had 2 capable of both 44.1 and 48k. Did 48k "sound better" to me? No, primarily, I guess, because I cannot hear anything at 24khz - and never could actually hear any benefit of the added overhead (but will also add that I never heard anything detrimental when I did try out 48k at times). And from my days with my ADAT decks, I've been one of these 44.1/48k is plenty for me. Always will be. Nothing I've ever produced has ever been critiqued as not hi-fidelity enough. They end up on 16bit/44.1k CDs anyway, so...
 
Nevertheless, I thought it was a great article and more or less just wanted to thank OP for posting it. Really though, I just play a little trumpet, a little piano, a little bass, and a little guitar, and like to record it on occasion...I don't need 192k to make me sound any better or worse, since even 44.1k CD's handily won the argument over "Is it Live, or is It Memorex?"

I have to apologize for what this turned into! OMG, I "thought" I was asking a simple practical question that would be useful to myself (and others) and it turned into rehash of the last one. Ugh...
 
Just out of curiosity (and also because I didn't want to get caught up in it)... I pulled open a few of the 32BFP samples I had made and figured "if those extra 8 bits are empty (an entire byte/character), then the data portion of the 32BPF wav file should have repeating characters (every fourth character) even if opened with NotePad"... so I popped a few open, and basically scrolled to the middle, and found this to be true... the 4th character would repeat for a chunk of samples (no, I didn't count how many), and then shift to a new character that had the same pattern. I scrolled around a bit, then looked at couple other files, and saw the same thing. My "logic" was that even if those characters were not aligned properly to the samples themselves, the 4th character (8 bits) should repeat if "empty." (More layman logic, but was enough to confirm to me what John responded to my initial question... essentially that 32BFP files on disk were not buying me anything (other than 25% wasted space). Why Adobe even has a 64BFP file format available makes even less sense to me now.)
 
For me, this circles back to where it all started, and John's initial reply is still "my" answer, which was that anything beyond 24bit on disk is not gaining me anything, but for processing/rendering higher bit depths should be used (which is pretty standard in all DAWs anyway, so not an issue).

Once insulting buzzwords like "misinformed" and other terms that are direct questions to another's mental capacity are thrown out, it is inevitable a thread will degenerate to personal war and info will become secondary. It's not like this is a topic that there is one, undisputed opinion about.  If so, this thread would probably not exist.

mettelus
Just out of curiosity (and also because I didn't want to get caught up in it)... I pulled open a few of the 32BFP samples I had made and figured "if those extra 8 bits are empty (an entire byte/character), then the data portion of the 32BPF wav file should have repeating characters (every fourth character) even if opened with NotePad"... so I popped a few open, and basically scrolled to the middle, and found this to be true... the 4th character would repeat for a chunk of samples (no, I didn't count how many), and then shift to a new character that had the same pattern. I scrolled around a bit, then looked at couple other files, and saw the same thing. My "logic" was that even if those characters were not aligned properly to the samples themselves, the 4th character (8 bits) should repeat if "empty." (More layman logic, but was enough to confirm to me what John responded to my initial question... essentially that 32BFP files on disk were not buying me anything (other than 25% wasted space). Why Adobe even has a 64BFP file format available makes even less sense to me now.)

It's not as simple as you assume - floating point consists of a sign bit, a series of exponent bits that scale the data up or down, and the actual number ("fraction") as described here (there is also an implied leading digit for non-subnormals that I mention mostly so the technically minded folks don't correct me):
 
http://en.wikipedia.org/wiki/Single_precision_floating-point_format
 
 
 
What ends up happening is your 24bit audio goes into the sign bit, the implied leading bit, and the fraction. The exponent then scales it up or down appropriately. IOW, the extra 8 bits are not just padded zeros because it's not just more bits; floating point means the bits represent things differently than in fixed point (aka integer).
 

For me, this circles back to where it all started, and John's initial reply is still "my" answer, which was that anything beyond 24bit on disk is not gaining me anything, but for processing/rendering higher bit depths should be used (which is pretty standard in all DAWs anyway, so not an issue).

If you are going to be processing further, 32bit floating point can be better because if you happen to have any samples that go over 0dBFS they will not be clipped, and low level precision is preserved if you subsequently raise the volume of the signal. I recommend it for interim formats (i.e. transferring to another programs for further processing or to external mp3 encoders) and for Render Bit Depth - even though might not really be necessary in most cases.

Thanks Drew. I was thinking after I posted that I shouldn't have since it might re-kindle the fire! The samples I had saved had no clipping (they were all rendered to a -0.1 dB normalization prior to save), and I was not shooting for anything scientific to contribute to a white paper but more a litmus test to see how often those additional 8 bits rolled that 4th character (i.e. more just to see with the simplest layman tool I could use). I want to be clear that I do/did not expect a straight-up 8 bits of 0-bit padding. Sorry about that.
 
From the practical stand-point, I walked away from this thread with:
    24-bit files on disk = fine (for me)
    32-bit (or higher) processing = fine (for me)
    44.1/48 kbps sample rates = fine (for me)