Topic: 24 bit > 16 bit

[illconditioned]

As far as the R09 goes, there is no benefit to going 24bit.  The noise floor (if you short the inputs) is about -90dB, so it is good to only about 15 bits anyway.  (Recall, one bit gives you 6dB of dynamic range, so 16 bits is 16x6 = 96dB of dynamic range.)

Do I understand correctly that this noise floor is inherent to the analog parts of the ADC, and independent of the preamp/gain settings?  Therefore meaning (I think) that with the Edirol it is NOT a good idea to run 24 bit with conservative levels (say peaks at -12 dB) and boot in post / dither to 16 bit later, as doing so will result in an effective dynaimc range of ~ 14 bits?

I'm not sure if it is ADC or a combination of the premp/ADC, but that chip (all-in-one Burr-Brown part) is rated for a max SNR of 90dB, ie., approx. 15 bits.  So, yes, you should run as hot as possible for a good recording.  But even at -12dB or -6dB it is still great.  I suspect many people are not getting even that.  My experience is most of the detail seems to be coming from the mics first, then from the preamp.  The ADC seems to be pretty good.

Any advantage to recording in 24 bit (but with hotter levels) for stuff that is going to undergo significant postprocessing, eg mixing down a mid-side recording?

Yes, I think one should *process* in 24 bit, if you're going to.  Then dither back down to 16 bits.

Also, could it be that random/analog noise is somehow less objectionable that quantization noise, and so there can still be some advantage to running 24 bit even with gear that has a noise floor close to but slightly above the 16 bit dynamic range limit?

I'm not sure about that.  But as I said above *some* high end gear could give you 18 or if really high end, I suppose 20 bits, but I wouldn't count on more than 18 bits really.  So, run your gear 12dB down, dither to 16 and it should be fine.

This is only my experience.  If anyone can report gear capable of really using more bits all through the signal chain, I'd love to hear about it...

  Richard

[boojum]

no, its not
if so 24 bit files would be  256  times larger than a 16-bit recording

i hope that makes sense.

To understand why there is better detail and dynamic range, understand that 16 bit is 2 to the 16th power; 24 bit is 2 to the 24th power. 

Jerry -  I think you are wrong here.  The word length is longer, 16 bit vs 24 bit.  The word itself has the ability to describe that many more degrees because it now has 24 bit switches instead of 16 bit switches.  I will have to check my reference texts on this.  Sampling rate  will make for huge file increases. 

OK, Google is our friend.  I found this:

"Bit Depth refers to the number of bits you have to capture audio.  The easiest way to envision this is as a series of levels, that audio energy can be sliced at any given moment in time.  With 16 bit audio, there are 65,536 possible levels.  With every bit of greater resolution, the number of levels double.  By the time we get to 24 bit, we actually have 16,777,216 levels.  Remember we are talking about a slice of audio frozen in a single moment of time." 

I edited a garbled reference in the 16 bit resolution.  It is beginning to look as if I had it right.  2 to the 16th power is 65,536.  Likewise the 2 to the 24th power is the 16,777,216 which is not used to describe the size of the file.  It is the possible combinations within a 24 bit word as opposed to a 16 bit word.  The link is here:  http://www.tweakheadz.com/16_vs_24_bit_audio.htm.  See also this citation in the Wikipedia:  http://en.wikipedia.org/wiki/Audio_bit_depth

It is the same with digital photos.  300 bit resolution is much better than 72.  Fineness of grain and spectrum of color are vastly improved.  OK??  Let me know if this does not jibe with what you think is right.

Cheers

[Petrus]

More sample depth means ONLY more dynamic range, nothing else. No more "resolution" or such. More dynamic range is a good thing. Yes. Makes level setting easier, keeps artefacts at bay when editing.

But how much do you need? No need to get overexited here, how many of you can listen to even 16 bit 96 dB clean dynamic range? Typical living/listening room has at least 40-50 dB noise floor, add 96 dB to that and you would have to have 136-146 dB peak volumes to enjoy the full dynamic range... Even if we can pick out detail mixed with the ambient noise you would have to play the recordings at well over 110 dB peak levels to hear the 16 bit limitations. For me and my quite good stereo system 105 dB(A) is the most I can stand for breef periods.

The noise you hear on you recordings does not come from 16 bit limitations, it is the mic & electronics which let you down like some have already pointed out. It might say 24 bit on that switch, but the mic preamps might give only 14 bit equivalent resolution. Very few mics themselves have more than 70 dB dynamic range (13 bits...) etc. Keeping that in mind there is no real benefit in going from 16 to 24 bits (exept safer headrooms when recording).
-------
About the theory behind the bit depth/dynamic range: Simple: adding one more bit to the sample the maximum size of the sample can be twice as big (this is with binary numbers, with our normal numbers adding one digit the numer can be ten times as big). As the sample represents the signal voltage, the loudness level represented can be 6 dB more for each added bit. The loudness levels are not more finely gradiated or anything like that, just the loudness difference between noise floor and maximum level is bigger with more sample depth.

[libfab]

What programs do you guys use to burn DVD-A's with your 24 bit masters?  I have Nero and I don't think it has DVD-A capability (?).  Must be some open source programs out there?

I just got the R-09 and with only a 2 gig card I was worried about running out of space when recording 24 bit, so my first master was 16 bit.  I now have a 4 gig card and want to run 24 bit.

I use DVD-Audiofile or Lplex. Both programs are free and pretty simple to use. These programs will create an image file.



For Nero: In the "Backup" tab select "Burn Image to Disc." You'll love the sound of your 24-bit recordings. Enjoy! 

Do DVD-Audiofile and/or Lplex work on Mac's? If not, what programs will?

DVD-Audiofile has a Mac version, check here http://www.versiontracker.com/dyn/moreinfo/macosx/28674

Lplex is an open-source project, so you can download the source and compile it to make a binary for Mac. A windows binary is distributed on sourceforge here http://sourceforge.net/project/showfiles.php?group_id=171628
I know a Linux binary has been compiled by udovdh (send him a PM) although not yet officially distributed on sourceforge.

[Gollum]

What programs do you guys use to burn DVD-A's with your 24 bit masters?  I have Nero and I don't think it has DVD-A capability (?).  Must be some open source programs out there?

I just got the R-09 and with only a 2 gig card I was worried about running out of space when recording 24 bit, so my first master was 16 bit.  I now have a 4 gig card and want to run 24 bit.

I use DVD-Audiofile or Lplex. Both programs are free and pretty simple to use. These programs will create an image file.



For Nero: In the "Backup" tab select "Burn Image to Disc." You'll love the sound of your 24-bit recordings. Enjoy! 

Do DVD-Audiofile and/or Lplex work on Mac's? If not, what programs will?

DVD-Audiofile has a Mac version, check here http://www.versiontracker.com/dyn/moreinfo/macosx/28674

Lplex is an open-source project, so you can download the source and compile it to make a binary for Mac. A windows binary is distributed on sourceforge here http://sourceforge.net/project/showfiles.php?group_id=171628
I know a Linux binary has been compiled by udovdh (send him a PM) although not yet officially distributed on sourceforge.

This is a little off-topic since the Mac program I'm using doesn't write DVD-A discs. Since I don't have a DVD-A player at home, I've used Toast Titanium 7 to write several "Music DVDs" with my 24-bit files. Encoding options are 24/96 PCM, 16/48 PCM and Dolby Digital. I can fit 2+ hours on one disc using 24/96 (24/48 isn't an available option.)

The discs will play on any DVD player or PC, unfortunately there are short skips between tracks. But at least I can play my 24-bit files for now.

[Brian Skalinder]

More sample depth means ONLY more dynamic range, nothing else. No more "resolution" or such.

< snip >

The noise you hear on you recordings does not come from 16 bit limitations, it is the mic & electronics which let you down like some have already pointed out. It might say 24 bit on that switch, but the mic preamps might give only 14 bit equivalent resolution. Very few mics themselves have more than 70 dB dynamic range (13 bits...) etc. Keeping that in mind there is no real benefit in going from 16 to 24 bits (exept safer headrooms when recording).

I haven't had a chance to do a proper A/B test, but plenty of ad hoc recording and listening reveals my 24-bit recordings sound better to my ears.  If longer word length means ONLY greater dynamic range...and not greater resolution...and the mics are limited to 13-14 dB of dynamic range, then there should be no audible difference between a 16-bit and 24-bit recording (assuming the same sample rate).  So why do my 24-bit recordings sound better than my 16-bit recordings?

[boojum]

I agree that the sound should be better.  After all, the vocabulary to describe the music jumps from 65K words to 16G words.  The vocabulary has increased by several orders of magnitude.  Likewise DPI in photos and computer screens.  I do not have the data but would say the the "vocabulary" is a good metaphor.  As always, YMMV 

[svenkid]

how do we burn these 24 bit dvds in nero? I went to he copy image to disc, but then it lists music dvd, OK then a mp3 dvd, wma dvd or a nero audio dvd. I tried nero audio, so converted a wav to both wma and mp4, which are nero audio formats, and both were drastically smaller than the original wavs, Wha am I missing here?

[StuStu]

how do we burn these 24 bit dvds in nero? I went to he copy image to disc, but then it lists music dvd, OK then a mp3 dvd, wma dvd or a nero audio dvd. I tried nero audio, so converted a wav to both wma and mp4, which are nero audio formats, and both were drastically smaller than the original wavs, Wha am I missing here?

I responded to your PM. I hope that helps. You must create an image file using one of the afformentioned programs. That's what you'll burn in Nero. Don't bother with the Nero Audio or WMA options.   

[kindms]

More sample depth means ONLY more dynamic range, nothing else. No more "resolution" or such. More dynamic range is a good thing. Yes. Makes level setting easier, keeps artefacts at bay when editing.

But how much do you need? No need to get overexited here, how many of you can listen to even 16 bit 96 dB clean dynamic range? Typical living/listening room has at least 40-50 dB noise floor, add 96 dB to that and you would have to have 136-146 dB peak volumes to enjoy the full dynamic range... Even if we can pick out detail mixed with the ambient noise you would have to play the recordings at well over 110 dB peak levels to hear the 16 bit limitations. For me and my quite good stereo system 105 dB(A) is the most I can stand for breef periods.

The noise you hear on you recordings does not come from 16 bit limitations, it is the mic & electronics which let you down like some have already pointed out. It might say 24 bit on that switch, but the mic preamps might give only 14 bit equivalent resolution. Very few mics themselves have more than 70 dB dynamic range (13 bits...) etc. Keeping that in mind there is no real benefit in going from 16 to 24 bits (exept safer headrooms when recording).
-------
About the theory behind the bit depth/dynamic range: Simple: adding one more bit to the sample the maximum size of the sample can be twice as big (this is with binary numbers, with our normal numbers adding one digit the numer can be ten times as big). As the sample represents the signal voltage, the loudness level represented can be 6 dB more for each added bit. The loudness levels are not more finely gradiated or anything like that, just the loudness difference between noise floor and maximum level is bigger with more sample depth.

Very few mics ?

AKG414XLS = Dynamic range 134 dB minimum

[Petrus]

AKG414XLS signal to noise ratio = 88 dB (14.7 bits)

(source: AKG spec sheet)
--------------

Lets get the facts straight: Bit depth means the maximum dynamic range, 6 dB for each bit. Sample rate determines the frequency range, max frequency = sampling rate/2. There is NOTHING more to this, no hidden "resolution" things. Photo analogy does not work, we do not enlarge audio.

When the dynamic range of the recording system gets past the weakest link in the system, there is enough DR. When the upper frequency gets past the human hearing, we have enough samples. That's it. Having some more does not hurt, it might come in handy with heavy editing. With normal recordings, not.

I have not seen or heard about a scientific double blind tests where people have been able to discern between top quality 16/44.1k and 24/96k. I know several tests where they were NOT able to tell live analog from 16/44.1, or 24/96k downsampled to 16 /44.1 from the original signal. 16/44.1k is good enough for me as the final product.

We have the tools to make audio perfect, only if we learned to use the best available mics, acoustics, and set the levels corectly...

[kindms]

AKG414XLS signal to noise ratio = 88 dB (14.7 bits)

(source: AKG spec sheet)
--------------

Lets get the facts straight: Bit depth means the maximum dynamic range, 6 dB for each bit. Sample rate determines the frequency range, max frequency = sampling rate/2. There is NOTHING more to this, no hidden "resolution" things. Photo analogy does not work, we do not enlarge audio.

When the dynamic range of the recording system gets past the weakest link in the system, there is enough DR. When the upper frequency gets past the human hearing, we have enough samples. That's it. Having some more does not hurt, it might come in handy with heavy editing. With normal recordings, not.

I have not seen or heard about a scientific double blind tests where people have been able to discern between top quality 16/44.1k and 24/96k. I know several tests where they were NOT able to tell live analog from 16/44.1, or 24/96k downsampled to 16 /44.1 from the original signal. 16/44.1k is good enough for me as the final product.

We have the tools to make audio perfect, only if we learned to use the best available mics, acoustics, and set the levels corectly...

You said dynamic range in your last post which is why I pointed to the specs. I thought you might have meant signal to noise but thats not what was in your post and will surely confuse the discussion. Just wanted to clarify. I always find these discussions interesting

[Petrus]

There is a difference between dynamic range and S/N ratio, true, but S/N ratio of 88 dB means that the noise floor is never further than 88 dB form the loudest signal the mic records. That much about the 24 bit detail advantage.

Further thoughts on the audio/photo comparasons and analogies:

Dynamic range, determined by sample depth: in audio, the difference between the softest and the loudest possible sound. In photography, the difference between the darkest and the lightest part of the picture. In photography 8 bits have been the norm, now we are moving to 12 or 14 bit processors. The problem is that printed pictures have only about 7 bit DR, improvement is mostly theoretical, or gives some leeway in exposure (analogous to setting levels in audio). With high enough bits there would not be need to set levels (some new digital mics wok this way), 24 bits is almost like that. With enough bits in photography any exposure would give a perfect picture (sensor technology is not there yet). Basically one bit or 6 dB corresponds to one f-stop in photography.

Sample rate: in audio this determines the highest recordable frequency. In photography it determines the resolution. There is one major difference here: in audio we use only one "size"; we (almost) always listen to the recordings at their original size (speed). In photography we can enlarge the picture untill the resolution gives in, in audio the corresponding thing would be slowing down the audio and loosing the highs! Very seldom done. For that reason we do not really need more "resolution" with audio like we want to have with pictures (even there we seldom make door sized prints).

 

[live2496]

So why do my 24-bit recordings sound better than my 16-bit recordings?

I think it's because the noise-floor is much lower.

All converters add dither as a part of the conversion process. With a 16-bit converter the signal is dithered at the 16-bit level. Typicially with 24-bit there is additional detail not masked by the dither.

A 24-bit recording should sound a bit clearer.

[BC]

while people say 'theres more headroom when you record in 24 bit', this headroom DOES NOT translate to 16 bit unless you normalize it while in the 24 bit realm

in other words, take a 24 bit recording that peaks at -24 dB. the dynamic range is 120 dB. if you convert this to 16 bit with no normalization, your new dynamic range is 96-24dB=72 dB (not so good).

If you were to normalize the 24-bit file to 0 dB, youd still have a dynamic range of 120 dB (you amplified the noise too), then when you dither and convert it to 16 bit, you have the full 96 dB of headroom.

Good point, it's important to remember this when preparing your 16 bit dithered versions from the 24 bit masters.