Topic: Just Curious . . . Why Re-Sample

[SmokinJoe]

yeah, ONE single play counts as a full download on the 16 bit. you can't stream 24 bit so unless people are really into getting 24 bit, the majority will just download mp3s.

I'm not exactly saying you can stream 24bit... I'm saying 24 bit files are used to derive MP3s which are streamable if you allow them to be derived.  I assume if people stream this it generates hits on the download counter for the 24 bit source, and if more people stream this source than the 16bit source, it might make the 24bit source look more popular. Here is an example:
http://www.archive.org/details/rmb2010-02-26.akg414-sbd-mix.flac24f  has 101 hits at this moment
http://www.archive.org/details/rmb2010-02-26.akg414-sbd-mix.flac16f  has 150 hits at this moment.

I put the 24bit source there for those who want it, but I don't think 101 people really download 24 bit flac files for high end audiophile systems.  I think most of them streamed it, or grabbed MP3s for an iPod, and most don't know the difference between 16 bit and 24 bit.  The 16bit source has more hits primarily because it went up a few days earlier.  After that it's 5050 at random.

[disco]

so if sample size doesn't really affect much if you record at 48 and convert to 44, does bit depth conversion yield anything better....since the 48>44 doesn't offer anything better, does going from a 24 bit tape to a 16 tape yield any better results?
 I record everything in 24/48 but usually just archive my 24 bit wavs, seed the 16bit flacs and archive a set of the flacs as well. Am I wasting my time/space by recording in 24/48 or is
it just a good practice to record at the highest bit rate/sample available so that you have the best possible pull?

[page]

so if sample size doesn't really affect much if you record at 48 and convert to 44, does bit depth conversion yield anything better....since the 48>44 doesn't offer anything better, does going from a 24 bit tape to a 16 tape yield any better results?
 I record everything in 24/48 but usually just archive my 24 bit wavs, seed the 16bit flacs and archive a set of the flacs as well. Am I wasting my time/space by recording in 24/48 or is
it just a good practice to record at the highest bit rate/sample available so that you have the best possible pull?

for recording, 24bit over 16bit yeilds much better bang/buck then higher sample rates (baring some special thing you are recording like bats). 24 bit puts the bottom bit around -144db (realistically somewhere around -120db) while 16bit's hard maximum is -96db. So what this means is that if you are recording with 16bit and amplify it, you are bringing up the noise floor from 96 while your (mechanical) noise floor ** is probably lower on the 24bit file. As such, running your levels lower and amplifying in post production doesn't have nearly the downside with 24bit recording then it does with 16bit recording.

http://en.wikipedia.org/wiki/Audio_bit_depth

** This doesn't include the noise floor of whatever soundstage you are trying to capture. A crowded bar's "noise floor" is most likely well above -96db, and you microphones noise floor is probably above the -96db threshold as well. What this does is allows you to back off of your gain to prevent clipping without taking a hit on the noise when amplifying it later.

[disco]

so if sample size doesn't really affect much if you record at 48 and convert to 44, does bit depth conversion yield anything better....since the 48>44 doesn't offer anything better, does going from a 24 bit tape to a 16 tape yield any better results?
 I record everything in 24/48 but usually just archive my 24 bit wavs, seed the 16bit flacs and archive a set of the flacs as well. Am I wasting my time/space by recording in 24/48 or is
it just a good practice to record at the highest bit rate/sample available so that you have the best possible pull?

for recording, 24bit over 16bit yeilds much better bang/buck then higher sample rates (baring some special thing you are recording like bats). 24 bit puts the bottom bit around -144db (realistically somewhere around -120db) while 16bit's hard maximum is -96db. So what this means is that if you are recording with 16bit and amplify it, you are bringing up the noise floor from 96 while your (mechanical) noise floor ** is probably lower on the 24bit file. As such, running your levels lower and amplifying in post production doesn't have nearly the downside with 24bit recording then it does with 16bit recording.

http://en.wikipedia.org/wiki/Audio_bit_depth

** This doesn't include the noise floor of whatever soundstage you are trying to capture. A crowded bar's "noise floor" is most likely well above -96db, and you microphones noise floor is probably above the -96db threshold as well. What this does is allows you to back off of your gain to prevent clipping without taking a hit on the noise when amplifying it later.

thanks for the info & linky

[Shadow_7]

Some things I've noticed.

The MP3 encoder will resample your audio to 16/44.1 before it does it's things.  Who do you trust to do a better job of it, the MP3 software or some other software?  A higher quality input almost always gives a higher quality output.  Even when it comes to MP3.

When I do a high quality (slow) resample from a higher bitrate, it seems to have a significant impact on the noise floor of the recording.  But I'm also resampling to sync up two unrelated devices.

Some effects are only available at 16/44.1.  Depending on the software being used.  sox + earwax

Otherwise always keep the highest quality originals.  If only for archival purposes.

[page]

Some things I've noticed.

The MP3 encoder will resample your audio to 16/44.1 before it does it's things.  Who do you trust to do a better job of it, the MP3 software or some other software?  A higher quality input almost always gives a higher quality output.  Even when it comes to MP3.

You bring up a good point, do you trust your MP3 software to do the resample/dither or the outboard software elsewhere? Doesn't change the retention question, but does give pause for considering your workflow.

[piperedworm]

That is kind of funny - Taperj and I were just talking about this last night.  I am going to do the "Pepsi Challenge" (yes I am showing my age here) tonight.

Re-sample one track then convert to 320kbs Mp3 and listen to the same track that I coverted to Mp3 straight from the 24/48 source.

Again, these are usually for the car/Ipod or for my friends that have no clue what 24/48 even means let alone what to do with it . . . but none the less now I am curious.

[Todd R]

The MP3 encoder will resample your audio to 16/44.1 before it does it's things. 

Wondering where you got this from?  Is this down-sampling to 16/44.1k before mp3 encoding specific to the encoder you use and is this pre-encoding downsampling discussed on the manual for the encoder or something?

If that is indeed how it works (and just for some encoders or all?), I agree, it sounds like it would be best to do the downsampling yourself first with high quality audio software. I've tried to do a fair amount of investigation into this since I derive my own 320 kb/s from my 24bit files and I want to make sure I'm getting the best results I can.

Here is another discussion on the 24bit>mp3 thing:
http://taperssection.com/index.php?topic=136471.msg1779929#msg1779929

I never kept track of the various studies and technical discussions I found when I was researching, but at least the post above has links to the relevant mp3 standards.  The mp3 standard accepts 24bit inputs and it seems like a really dumb way of encoding to first down-sample to 16bits when it doesn't seem at all necessary.

I'm not trying to bag on you at all, just trying to get to the bottom of things, but I'd like to know if you have solid information that your mp3 encoder or all mp3 encoders downsample to 16bits before encoding or if you are just assuming that that is how it works.  I've never seen that mentioned in the technical stuff I was able to dig up, and it seems totally at odds with the mp3 standard.

[Shadow_7]

The MP3 encoder will resample your audio to 16/44.1 before it does it's things. 

Wondering where you got this from?  Is this down-sampling to 16/44.1k before mp3 encoding specific to the encoder you use and is this pre-encoding downsampling discussed on the manual for the encoder or something?

It might be lame specific.  But I read it in the documentation for that (or of that) encoder which seems popular as the goto MP3 encoder.  It also makes sense in that MP3 is optimized for CD quality.  Not to imply that ALL encoders do the resample first step.  Or that the codec itself is CD quality specific.

lame/USAGE

--- low bitrates ---
lame will automatically convert to 16 bit (as will FhG and Xing)
(slightly paraphrased)

--- CRC error protection ---
Yes this really does work correctly in LAME.  However it takes 16 bits per frame that would otherwise be used for encoding.

--- sampling frequency in kHz ---
LAME will automatically resample the input file to one of the supported MP3 samplerates if necessary.

lots of references to "The input file can be any input type supported by encoding", but no list of encoding supports directly...  Without delving into the actual code anyway.  But it stands to reason that most of the bug reporting and fixes would be related to 16 bit 44.1kHz samples (CD quality).  So even if the other rates are supported, there may be unreported and unresolved issues related to content (if only in theory).  And on the manpage for frequency options the list 8/11.025/12/16/22.05/24/32/44.1/48 is listed.  Which would imply that 24 bit 96kHz is NOT a supported encoding

under "-b n" in the man page.
- For MPEG-1 (sampling frequencies of 32, 44.1 and 48 kHz)
- For MPEG-2 (sampling frequencies of 16, 22.05, and 24 kHz)
- For MPEG-2.5 (sampling frequencies of 8,11.025, and 12 kHz)
(with a list of n values for each)

Mpeg-1 Layer III for what passes as a playable MP3 on my MR-1000 field recorder.  (stereo ONLY, if the latest firmware hasn't addressed that quirk on that device)

[Todd R]

Thanks Shadow for passing that info along.  It'd be great to get a clear answer, maybe more digging will uncover it.  Not sure though from what you posted that it would down-sample 24bit to 16bit.  The passage about automatically converting is stated for low bit-rates, which I'm reading as it will upsample 8bit or 12bit material up to 16bit for the conversion, but not necessarily that it would down-convert 24bit to 16bit.  Which makes sense as the mp3 standard seems to be defined for 16 and 24 bit inputs.

Not sure what it is getting at with the CRC correction, but on the sampling frequencies, from the standard I would expect that re-sampling to 32/44.1/48 would happen as those seem again to be the frequencies that the mp3 standard supports.  And yes, I would definitely expect since 96k (and anything other than 32/44/48) would be re-sampled as the mp3 standard doesn't make allowances for them. 

Still though, even from what you've posted, I'm hopeful that material that is recorded at 16/48, 24/44 and 24/48 would be converted to mp3 directly -- meets the input standard for mp3, and from my reading of what you posted, it doesn't seem that LAME is saying otherwise.

[fingers crossed]

[Shadow_7]

The hint that it specifically resamples to 16/44.1 might be a ffmpeg quirk.  But that's more for video.  I do remember reading specifically that 16/44.1 was the standardized input to a specific MP3 output.  And adjusted most of my routines to preconvert as much as possible.  But I'll have to research that harder.  And that info might have been in error to start with.  Or perhaps related to how things worked two or more years ago.

[lastubbe]

I was brought up as a young tape trader (15/16 years old) in the spirit of sharing and trading.  That's how I got into this, and that's how I've always been.

The spirit of the music, and the concept of taping live shows, is to spread and share.  If someone didn't share with me, I may not have ever had the pleasure of the being introduced to the music I follow and collect today.

It makes no sense to me why tapers would not want to share. (another topic)

With that said, it is still 2010, and 16/44.1 is still the commonly used format, as most listeners use iPod and CD.  Since I use a high resolution recorder, it is necessary for me to resample for people to be able to take advantage of the shows I share.

Now a High Resolution iPod would be cool.  You know that's coming!

HDTV.  HDIPOD.

[newplanet7]

 
Well put sir.