Kara-Moon Forum

For a long while now,I have been tormented by the decimation of audio quality when converting WAV to MP3 & regardless of all of the upgrades in gear & painstaking attention to detail in my mixing,I am still left with that horrible warble sound! >:(
At first I thought the problem was with the FreeRip program I had used(uploading from a CD from my old recorder which was 16 bit)....so then I learned to transfer my music directly from my old Zoom 8HD recorder in WAV format directly to my PC,but no improvement.
So I took another step up & utilized the Audacity program,along with upgrading to the 24 bit Tascam 2488neo multitracker.....but no improvment...just can't shake that terrible warbling sound.
Now I am at a point where I have learned to get the most of my Korg keyboards,by fully utilizing all of their sound mastering potential and also brightening & adding more depth to my recordings by upgrading from my Tascam's built in effects,to the Lexicon processor(which improved the stereo field greatly,but did not eliminate that warbling sound that's exclusively inherent in MP3 recordings.
At this point in time,I suspect that the dither switch in the Tascam is to blame,so perhaps I should just transfer my Tascam recordings to my PC without dithering and then do the dithering in Audacity?

Any advice on this matter would be greatly appreciated!

Thanks,

-Thom

I remember Oren had this problem with one of out collaborations. I can't remeber what the final answer was. However I would suggest that conversion to mp3 should always be done in the very last piece of kit the music sees - in your case, Audacity. What bit rate are you using? My experience is that variable bit rate (VBR)  gives the best results. I doubt there are any players these days that can't do VBR, but there are still lots of commercial programs that can't generate it!

I also doubt it's a dither problem. I've attached a copy of a very interesting article I found on dithering.

Hope this helps.

Will,

Thanks very much for the article,as it's quite fascinating.In relation to my issue,I think the cause of the warbling sound may be a result of me putting my music through the dithering process twice?Meaning that if I don't dither it in the Tascam & leave the dithering job to Audacity,then maybe I'll nail down the solution.
I've also decided to change the kbps rate to 256,instead of 128,so I'll see what happens next time I run a song through my Tascam.

-Thom

I don't know if this will have bearing on your problem, but twice I had deal-killer problems with mp3 conversions.

I had two trombones playing a duet, and three quarters of the way through the song it got really ragged..even after several attempts at conversion. The .wav file was perfectly fine.

Increasing from 128 kbs to 160 kbs solved the problem on one, and I had to go up to 192 on the second one.

Yeah....it's funny how the MP3 conversion only effects certain sounds and/or songs........for me,the conversion usually targets my bass & string tracks.
Perhaps changing the kbps rate won't change much of anything....but any improvement at all is worth a go at it.
Thanks for your input Wyatt.

-Thom

If changing to 256 kbs doesn't solve the problem, then it must be something else.

With VBR I find the hihats, cymbals etc get so warbled, I never use it
And since I got a new hifi -amp and speakers I dont listen to 128k files anymore, the difference with 256 and up is just mindboggling to me, never heard that difference before on my old system

Thom,

First of all, dither is something I never do. Dithering is a term some people use interchangeably with sample-rate conversion, but dithering is actually a method of introducing noise into a signal to hide the results of a sloppy conversion. So avoid dithering - turn it off. Plan your work so that sample-rate conversions are kept to a minimum.

Secondly, do your "lossy" compression as the last step, and don't be tempted to expand back to wave format for further processing, nor waste time processing the .mp3 or .ogg file.

Constant bit-rate .mp3 performance in Lame has been upgraded recently, so using their variable bit-rate method is no longer necessary (although the VBR performance is still superior). Ogg Vorbis is, by definition, a variable bit-rate compression codec, and it is my choice of "lossy" formats.

You got me to thinking again, amigo..   ::)

Currently, when I work in Reason it is @ 24/96, and since Audition does not work with 24 bit, but only 16 bit or 32 bit float, I have been exporting audio from Reason @ 16/44.1, since that is the destination bit depth/sample rate of all my projects.

So  I can disable dither in Reason when I export/downsample..ok..that's easy enough, but would I be better off not downsampling out of Reason in the first place, but just opening the 24 bit files in Audition as 32 bit float, and then downsample to CD quality at the end of the trail?

I have heard that you are best off to maintain the highest level of quality for as long as possible, but I was never impressed with the idea of upsampling.

Thanks in advance!





 

Oren,

Okay...lesson well received & applied to good use....just completed the process in accordance with your advice & my MP3 sounds fantastic!!! :D

Thanks! :-

-Thom

Doing ANY processing after dithering is a no-no
It'll very often come back to haunt you.

It CAN be a beneficial process when done correctly and with a decent dithering app.
That said, poor dithering is worse than no dithering IMO.

Just like trying to go back and reprocess an MP3, trying to reprocess a dithered piece of audio will often inject problems.

It's really only necessary when going from 24 bit to 16 bit for final CD.

Most Dither apps will tell you (quite fervently) not to dither until the final step.

A lot of people read that dithering adds noise and immediately think 'bad idea'.
The problem is that truncating 24bit depth down to 16 without dithering can be 'sharp'.

The process softens the edges as it were.

Thanks a lot..I appreciate your advice.

A point worth keeping in mind is that not all MP3 encoders are born equal :)

Indeed, LAME stands for L(ame) Aint Mp3 Encoder. Although it produces files that are playable by any MP3 player, they are actually generated with different (patent and license free) code.

Most MP3 encoders ceased development years ago so will never improve.

If interested in more, here is another article on it -
http://www.izotope.com/products/audio/ozone/OzoneDitheringGuide.pdf

I personally like the dithering in Ozone above some others I own, but  I speak only thru trial and listening; no testing done..

Didn't know about this article; I'll definitely check it out.

I've been very impressed with what I have heard from folks who use Ozone 4, and the Izotope multiband compressor in Audition 3 is very nice..quality stuff.

Thanks for all of the great advice guys-much appreciated.As I found out,my Tascam 2488neo will serve me very well...provided that I bypass the built in effects & of course,do away with dithering & leave that for Audacity.
Audacity has really become the single most important tool in my studio and in relation to processing great sounding MP3's...I changed the default sample format to 32 bit,left the dither setting at their default positions & changed the kbps rate to 256.
With the latest song I converted in this manner,I actually could not hear a noticeable difference between the sonic quality of my MP3 & my WAV files.....makes me really wonder if OGG is really all that much of an improvement over MP3's? ???

-Thom

Wyatt

If I was 'forced' to mix or master any song with only one plugin (be it a mastering job or a multitrack mixdown) Ozone 4 is about the only one I would even consider.  The single best plugin I have ever experienced.
Many excellent tools built in, you don't have to use them all, and and you can route them in any order.
Its not only the amount of tools ( there are a lot), its the quality of the tools that are included.  So much control is available.
One awesome piece of kit.
Free 30 day trial, but then you suffer from withdrawal afterwards (unless you buy it).

Even if you don't try it, it is worth reading the Ozone mastering guide and the dithering guide linked above.
They explain things in an understandable way..

The folks who develop Audacity have my greatest respect. Although I do most of my work in Linux with Ardour and JAMin, when things get weird, Audacity will sort it out, every time.


Ogg Vorbis level 6 has been tested with a group of "average listeners" to be indistinguishable from the original .wav file. MP3 at 256kbits/sec does the same for you. Cool!
Ogg Vorbis is not necessarily an improvement over .mp3, just free software developed by independent programmers with it's source code available to all for tweaking and further development - and designed that way from it's inception.
That makes it important from a "power to the people" perspective. :-

Yeah, when this subject came up again, I was afraid it was going to cost me some money.    :D

I've had my eye on Ozone 4 for a while. I just take forever to actually buy something, because for every thing I get, I have passed up a lot of other possible additions to the studio, and there is nothing worse to me than to buy something that ends up not serving me after all.

Thanks for your recommendation..a helpful one at just the right moment.

As I remember, you got Audition a while back. Have you had a chance to evaluate its mastering suite?  It has many of the same tools.

/////////////////////

I have another question, but I had better do some testing first.

Thanks for your advice.

Used to be I wouldn't have bothered being as picky as this, but I have been
trying to improve my sound and my skills lately, so I am having another
run at this subject.

I just made a track in Reason, and exported it three different ways:

1..16 bit [dithered]

2..16 bit [not dithered]

3..24 bit [not dithered]

I cannot say that I can hear the difference, but that is not setting the bar very high.
Perhaps someone else could.

When I loaded these three track into Audition however, I could easily see that there
are *distinct* differences in each one of the waveforms. No Brainer.

Then I made a copy of the 24 bit [no dither] track and converted them both to 16 bit.
On one I used dithering in Audition, and on the other I disabled dithering in Audition.

This time, I did not even try to hear any differences between them, because I could
not see any differences whatsoever in the two waveforms..not even the slightest.
Perhaps because I have always had set Audition's dithering depth to 1 bit, I don't know.

My conclusion so far is that I will at the very least disable dithering in Reason.

It makes sense to me to maintain the higher quality audio until the end of the process,
but  since Audition automatically upsamples the 24 bit to 32 bit, I am not sure if I am net
ahead after that extra processing.

If I export from Reason as 16 bit, no dithering, then the track doesn't get upsampled and
then downsampled, so that idea appeals to me, as my workflow is a bit tidier.

But that still leaves me wondering: Would I be better off upsampling and then downsampling
for having held onto the higher quality longer?

How many angels can dither on the head of a pin?   :P


 

2nd on Ozone. I've not found a mastering tool I like better.

Thanks Fred!

I *may* be wrong, but here's what I think you are misubderstanding ..

Audition is not 'upsampling' the files to 32 bit.
It is likely using 32-bit floating point method of processing the file while inside Audition.  Many many DAWs do this these days (harder to find one that doesn't) since the advent of 32 bit processing power in the OS being used.

It is different than actually upconverting the file itself.  When it saves the file it will be in 24 bit still, correct?

Every DAW software I own does this.  Standard procedure, even if they don't trumpet it as a feature.

As for 'seeing' the difference, if you read the Ozone dithering guide you will see why this is.  I can 'see' it also, but only judge it by what I can hear.

Every software that does perform dithering during export should let you know it is taking place so you know you have 'finalized' the file.  I am not sying that they all do tell you, just saying that in my opinion they 'should' tell you.

One thing you did not define in your previous post, Wyatt, was whether you original file in Reason was 16 or 24 bit ..

"I just made a track in Reason, and exported it three different ways:

1..16 bit [dithered]

2..16 bit [not dithered]

3..24 bit [not dithered]"

If it was originally recorded at 16 bit, then exporting in 24 bit gained nothing except a ton of 'zeroes' placed in the file .

Oops..The original file is 24 bit.

Right..that is what I was misunderstanding. That puts things in a new light. So if it isn't being upsampled, it's worth keeping in 24 bit.



As far as seeing it but not hearing it, my hearing has gotten a bit worse these last few years. I have come to rely to some extent on visual representations, especially of higher frequencies. Because of seeing there was a difference, I inferred that the 24 bit without dithering must be better quality.

Thanks for explaining that Bob; I really appreciate it.

As far as I know there is only one DAW sotware capable of 32bit files, and it is not well liked by professionals ..

to qoute a famous mastering engineer -
"we live in a twenty bit world" 

Out of curiosity, are you using 24/44  24/48  or ?

Most guys seem to agree nothing above 24/48 is worth the extra ...

I use Audition 3..it give the options for recording and also for mixdown of 16 bit or 32 bit.


In Reason I am using 24/44.1..it will go up to 24/9600.

Audacity offers four dither options: none(my favourite ;D), square, triangle, and shaped.
 As I understand it...

If the converted file sounds just as good as the original file, without dithering, then none would be the best choice. Don't mess with perfection.

If dithering is necessary to avoid weird audio phenomena when down-converting, and you may want to further process the file at a later time,  triangle (triangular probability density function) will allow some subsequent processing without audible glitches.

If dithering is necessary, and this is the last bit of conversion and processing the file will undergo, then shaped (absolute threshold of hearing - or - noise shaping) would be the best choice.

The folks at Audacity use shaped dithering as their default setting. This is a reasonable choice, assuming the file is not slated for further processing.

I appreciate your thoughts..

Yes, when I cannot see any difference at all in the wave form in Audition, it makes perfect sense to not be dithering around with it.   :D

Audition has 5 different types of dithering, and when I was using it, triangle was the default. It also defaults to one bit depth.

As for the noise shaping, it has 11 different patterns, and it suits me fine to leave dithering out of the equation, since I haven't a clue, (yet), as to what they are all about.

So as of now, I am turning off dithering everywhere, and leaving it that way unless and until the time comes when I might have a reason to use it.

Thanks for your advice.

As far as sampling rate goes, is there an advantage to using 24/48  over  24/44.1, since my destination is 16/44.1 ?

Excellent!!

..hitchhiker's guide to dithering.

 ;D


Eleven - Woh!!!


Seems reasonable...


There is a tremendous amount of disagreement on this point, even among the experts. Personally, I can't hear the difference between the three. Theoretically, there is some advantage to using a higher bit-rate and sampling rate when processing digital audio, but once again... I can't hear any difference in the results.
For a while I was using 24bit/48kilohertz for all my audio files, but could not realistically perceive any improvement in sound quality. Consequently, I've reverted to using 16/44.1, unless someone gives me an audio file that's mixed down to one of the higher rates.

How much of this trend to push audio processing to the limits of a modern computer's capability actually results in real-world audio improvement?

Interesting..thanks for your advice.

Now maybe I can let this question rest for another year or two!   :D

I use 24 bit when I'm just playing around, especially with sampled pianos, where I do "think" I hear a difference. But for recording and distribution, I just use 16/44.1. What has always bothered me about higher quality is that you eventually end up dithering it down most of the time anyway to CD quality. Then whatever sonic quality improvement achieved is lost. Further, as you are developing your project, you are theoretically listening to something different than the eventual result. So, although you want the best rendering possible, what's the point...really. If you work in 16/44.1, your hearing exactly what the end consumer will be listening to. Looking at it in that way, it actually seems like an advantage to work in CD quality.

I can hear the difference between recording the same thing at 16 and 24 bit.  That I have tested.

It still does make sense to work in 24 bit (IMHO) and dither at the very end.

The difference is in the size of each sample (or chunk) getting grabbed digitally.

24 bit grabs larger samples than 16 bit.  These samples occur at a certain speed (44.1kHz for example) which means they are small digital samples of a sound over time.

Now if you could grab a small sample (16bit) with tweezers, imagine the sample you could bet with a big set of tongs!

How these samples are bigger is because they take a larger bite of time.  This adds more depth to the sound.  During processing every effect has better data to work with.. so gives better results.

The resulting 'depth' gets retained during dithering (as much as possible)

There IS a difference, which is why studios use 24 bit versus 16..

The key is to use a good A/D/A to begin with and a good dithering program at the end, so the depth is retained as much as possible.
I used to really be confused about it all, and then spent some time reading, listening, testing ..
still not an authority but I can hear the difference now.

Moon can probably explain about the exponential difference this size sampling makes.. it sounds like an 8 bit difference, but it isn't.. its much greater

This difference in dithering results is why I said earlier that Ozone is the best one I have used to date.

Makes sense!

I'm with you on the bit depth part..but I still have a question about the sample rate.

I have stuck to 16 bit up until now, because I couldn't hear the difference, (though admittedly, that isn't setting the bar very high). I knew that in my restoration work, noise reduction works better, (with less aliasing), at higher bit depth, but I have developed a number of strategies that have given me very accurate, clean noise reduction @ 16 bit..so I kept it simple.

Now that I am using Reason in conjunction with my 'real' instruments, (ooooooohw..did I just say that?), I am ready to hang onto its native 24 bit without dithering until the bitter end, as per your suggestion.  (Thanks for explaining that 32 bit float isn't actually upsampling, btw).

However, I still am not entirely comfortable with the sample rate situation. I mean really, did they have to confuse this issue mathematically by having two standards that are in such an odd relationship to each other as 44.1 and 48?

Reason works at a sample rate of 24/96. My sound card's A/D converter works at 24/192. Now if I were downsampling from 192 or 96 to 48, I would be a lot more comfortable with the situation, because 48 is exactly half of 96, but isn't downsampling to 44.1 going to introduce more problems? Even if I export audio from Reason @ 88.2, (twice 44.1), then it has still gone through that odd numbered conversion in Reason.

So in the end I can't avoid this odd conversion, it has to happen somewhere..I have to choose between downsampling from Reason in the beginning, or waiting till later and doing it in Audition.

So should I follow a similar principle with the sample rate conversion, as with the bit depth?  Specifically, should I export audio from Reason in its native sample rate of 96, and then wait until I am ready to burn to cd, and then convert sample rate and bit depth at the same time?

The important bitrate happens as the audio is converted to digital .  This takes place inside the soundcard A/D/A convertors.  The software may tell the card what bitrate to use, but the actual function happens inside the card.

The software, during recording, is simply saying "OK I'm ready to recieve data at such and such a format" (which is why the software used to record isn't really as important as the soundcard - PCM wav audio data 'is what it is' once inside the computer)

Reason,  Audition, Powertracks, Reaper, whatever, the software is receiving what the soundcard sends for  a format.
After that anything done is mathematical. So if the soundcard sends 16/44 then there is no problem for the software, it just adapts.  If its 24/96 then it knows that and adapts... but the adaptations are not going to change the sound;  So adjusting the recording rate to work with software is defeating the real source- soundcard convertors.
That is where the rubber meets the road and the sound you get to work with is determined.
All the recording softwares actually handle the audio as 32 bit floating point for the math they do.  So saying the 'native rate' for a software makes no sense to me; it handles whatever the soundcard gives it.  Maybe Reason could be an exception because it actually 'generates' audio in some instances and this audio may be at a certain rate, dunno..
  Upconverting just adds zeroes that will be removed later. I can't see that making any improvement

I mentioned 24/48 because during tests people can't tell anything beyond that... that is the best rate-for-filesize because beyond that people just don't hear it.
 
There is a bigger difference between 16/44  and 24/44 than there is between 24/44 and 24/48.

The 16 to 24 bit depth is a big change, where as taking a sample 44,000 times compared to 48,000 times per second isn't that big a difference.  Its the fact that every one of those samples grabbed a larger chunk of sound in 24 bit, and THAT adds up.

I have an EMU 1212m..so good convertors..

The software, during recording, is simply saying "OK I'm ready to recieve data at such and such a format" (which is why the software used to record isn't really as important as the soundcard - PCM wav audio data 'is what it is' once inside the computer)


Reason doesn't record any audio..just midi. It can export midi tracks as audio, which is when I can select the bit depth and sample rate. The midi data from my keyboard does go through my soundcard. The .rex files I believe are already 24 bit, but I have no idea what the sample rate is on them.


So now that technology has outstripped human perception, as Oren also alluded to, there are things that really do matter and things that really don't.

Thank you for helping to sort all that out.

.. and thanks for your patience, I appreciate it.

8)

Wyatt

Maybe not quite about dithering but certainly interesting with regard to the 16/24 bit and mp3 quality question, i saw this video some time ago, its a long one though:

http://www.philoctetes.org/Past_Programs/Deep_Listening_Why_Audio_Quality_Matters

This is really long, but I am finding it very interesting. Kind of amazing to me how much better the first vinyl cut sounds than the CD redbook cut that precedes it.. Especially considering that I'm listening to a recording of this discussion on my computer. Must have been even more noticeable being in the room.

Finally somebody is making sense... :;

Over the last three years, I have looked at the information provided by various audio experts on bit-rate, sampling rate, dithering, digital processing, and compressed digital audio.They have provided a lot of numbers, but not much in the way of clarity.  It is astounding how frequently they contradict each other, and even themselves.

As with the participants in this in this discussion video, I have to conclude that the only real measure of an audio tool or technique is how positively it affects the listening experience.
      _________________________________________________________________________________________________________________

A group of audio experts I have had my eye on for some time (who actually do seem to make sense) recommend setting your DAW at maximum resolution (32bitfloat/96KHz), and outputting files at 24bits/96KHz. Then, for file sharing and general listening, compress the 24/96 file to a 256Kbit VBR file (either Ogg Vorbis or Lame VBR) to produce a small audio file that is superior in audio quality to a 16/44.1 wave file. I have not taken the time to work with any of these suggestions, but it feels like the time is right... ;D This is one of their sites:
http://jthz.com/mp3/#TOP

They referred me to little article on microphone pre-amps.
 A $5.00 solid-state preamp proved to be the listeners' favourite in an internet test:
http://www.audiomasterclass.com/arc.cfm?a=well-could-you-hear-the-difference-could-you
     ________________________________________________________________________________________________________________

It might interest some of you that in all the time I have been mastering digital audio and encoding files to Ogg Vorbis and Lame, no-one has suggested I ask the folks who provided the original files if they had dithered in the process of producing the audio I was about to manipulate. Given our discussion in this thread, it would seem to be quite a critical bit of information... :o

Well I'll be very honest about this....
I'm a musician, not an audio expert or even an audio engineer.
By this definition, I only take care how it sounds in a live environment.
I also do my best to post the best tracks I can here on the forum, but that's it.
The day I'll decide to make a commercial CD, I wont do it myself, I'll go in a studio and let the guys do there job  ;)
And I'll try to play the best I can....

Does this sound naive ?

k

Sounds like a smart musician.

Another tidbit: When a 16/44.1 file is introduced into a DAW working with 16 bit internal processing, the workstation file measures the same size as the original wave file.
                       When a 16/44.1 file is introduced into a DAW working with 32 bit floating point processing, the workstation file measures twice the size of the original wave file. This suggests to me the file is fundamentally altered (up-converted?) as a track in your DAW.

We may have come full circle, amigo, but it's been a hell of a trip!   :D

Thank you so much for filling in the blanks.

8)

Wyatt

Which workstation and software?

(Curious)

Audacity, in 64 bit Linux.

Curiously, importing a 16/44.1 file into a 44.1Khz sample rate DAW results in an identical workstation file size, but importing the same 16/44.1 file into the same DAW set for a sample rate of 96KHz does not result in a larger workstation file.

My guess is that at import time it does an automatic upsampling that hase (at least) the resolution of it's highest spec so it can then downsample with little or no loss when it needs to.

Curious as to how you check the filesize after importing .. do you save it first, or look at a temp file, or does Audacity tell you somehow while the song is loaded?

This 32 bit file increase may be an Audacity only thing (the filesize when working in 32-bit floating point)

I have other 32 bit floating point (processing) software here that this does not occur in.  Maybe Audacity is actually joining Audition in supporting 32 bit bitdepth ..

As a side dish-
I have a 64 bit system ordered which will be here in a few days!  Bit the bullet again.
dual core AMD (2X 2.1gHz) with 4 gig RAM .. 320gig drive ... lappie
I'm excited!

Yup. Save the project, and then check the project file size. I chose Audacity for two reasons:
- the results are verifiable in Linux, Windows, and Mac.
- Audacity saves the file(s) separately from the workstation settings, so a hoser can examine them individually.

Yup, again. ;D Audacity (in 64 bit Linux) will export in a staggering number of formats, including 32-bit PCM, 32-bit floating point, and 64-bit floating point.


All that in a laptop. Huge! What size display?

Display size??
what difference does that make for music?

It.s 15.6 I believe .. 16:9 layout

It starts to make a big difference when you're lugging it around in a back-pack, or balancing it on your knees while doing some audio editing on a bus or plane... ;)

I see yer point!

I never liked working on a laptop, but I got a wireless mouse and keyboard for it so hopefully it feels more better..