Down Sampling 44khz 16bit to 8bit

[Shadey]

Hi guys just wanted some opinions

Trying to down sample some wav files from 44khz 16 bit to 44khz 8 bit
but getting a large amount of noise in in the 44khz 8 bit files.

Using programs like Acoustica and so on.

Can someone suggest a good tool to use to down sample with noise reduction etc or other methods of reducing the noise levels i get please.

Regards
Shadey

 

[VIPER_1069]

HOW big are these wavs shadey mate ?! what about windows sound recorder !? simple but effective for short wavs !?

not tested for long wavs but hey its free so cannot hurt to try !?

maby wavlab !? or goldwave !? or cool edit pro 2?!

 

[LTR12101B]

Audacity? - I assume that could save in various formats.

It seems a bizarre format though... 44K sample rate = high quality / 8 bit (miserable quality) - unless something absolutely requires that format, there have to be better ways to shrink the size of an audio file.

 

[rebootjim]

Transcode to mp2 in tmpgenc, it'll be much smaller, and you won't lose the quality.

 

[celtic_druid]

SSRC should be able to do it --bits 8 I would imagine, should be about as higher quality as you can get. As said though better ways to shrink it. I mean 16bit-->8bit = 1/2 size which you could achieve using a lossless compressor like monkey, flac, etc. or go for transparancy with aac, LAME mp3 or mp3 and still get pretty small file sizes.

 

[Shadey]

HOW big are these wavs shadey mate ?! what about windows sound recorder !? simple but effective for short wavs !?

not tested for long wavs but hey its free so cannot hurt to try !?

maby wavlab !? or goldwave !? or cool edit pro 2?!

Wave files sizes range from about 300Kb to 2Mb - They are mainly recordings of spoken phrases.

Thanks for the suggestions - I'll check them out - Got stuck with this acoustica program cos it ws all we had licensing wise. Cheers Viper - apologies for not answering you on ICQ - was asleep

Audacity? - I assume that could save in various formats.

It seems a bizarre format though... 44K sample rate = high quality / 8 bit (miserable quality) - unless something absolutely requires that format, there have to be better ways to shrink the size of an audio file.

Not Audacity - Acoustica

Yeah the recordings (mentioned above) were made in that format - so not much can be done about the source files - the 16 bit to 8 bit conversion is necessary because we're experimenting with digitising these sounds to ICs used in handheld devices equiped with speakers - experimenting with sound samples and quality - because its a new IC we're using. Cheers LTR12101B

Transcode to mp2 in tmpgenc, it'll be much smaller, and you won't lose the quality.

Thanks for the suggestion rebootjim - unfortunately the source files we are working with are wav files and the format cannot be changed - unless the recording is re-recorded into mp2 but that means booking recording studios and hiring the speech talent again - but thanks for the advice - could use it next time round save this hassle

SSRC should be able to do it --bits 8 I would imagine, should be about as higher quality as you can get. As said though better ways to shrink it. I mean 16bit-->8bit = 1/2 size which you could achieve using a lossless compressor like monkey, flac, etc. or go for transparancy with aac, LAME mp3 or mp3 and still get pretty small file sizes.

Cool. Got a good place find these lossless compressor/codecs ? It would be worthwhile to experiment with them - as the noise levels i have at the moment are terrible - deeming these files as near useless. Cheers celtic_druid

= = = =

Thanks for he advice guys - got lots to think about and do. Would appreciate any links for information on using these lossless solutions.

Regards
Shadey

 

[VIPER_1069]

@ shadey no worries mate

hey is that the record for the most replys in one reply lol

 

[rebootjim]

Um...I don't get it.
Is it that 8 bit is a necessity, or filesize is a necessity, or what?
A .wav is a .wav, and just about any audio app will resample/transcode to 8 bit mono/joint stereo/stereo from any .wav, and save as .wav
Using Audacity or Goldwave, or Sounforge, or WaveLab, one could also clean up the audio to remove noise.
I guess I need to know exactly what you need the finished product to be, and what the source really is, before going further.
As I understand it now, you have 16 bit .wav files, and need a simple conversion to 8 bit. What's the problem again?

 

[scarecrow]

I don't get it, either... this "conversion" is totally nonsensical to me.
Can you provide a good reason on WHY you wanna do it, or you just wanna do it?

 

[LTR12101B]

I was suggesting Audacity... http://audacity.sourceforge.net
And thinking that if it's to be 8 bit quality, 44k sample rate is overkill - 22k or even 11k/8 bit would be more in keeping with 8 bit - or ADPCM, which uses fewer bits, but doesn't drop as much as a straight chop to 8 bits (IMA ADPCM uses only 4 bits/sample, but sounds better than 8 bit uncompressed) - to use any compressed format, obviously requires that the playback system supports that format - WAV is merely a wrapper, as you can wrap MP3 data in a WAV (Riff-Wave header).

And if you start with 44k sample rate, 16 bit WAV, then you can convert them to any format you like.
If you are experimenting with audo, you should join www.hydrogenaudio.org - they torture test various audio codecs, and there is a lot of expertise there.

If you are involved in developing something, it may be worth looking at the integer math version of VORBIS designed for embedded processors - though it still needs a reasonable amount of power.
http://www.xiph.org/ogg/vorbis/

Or Speex http://www.speex.org/ - a low bitrate codec designed for speech

 

[Shadey]

@ shadey no worries mate

hey is that the record for the most replys in one reply lol

LOL It most definately is

Um...I don't get it.
Is it that 8 bit is a necessity, or filesize is a necessity, or what?
A .wav is a .wav, and just about any audio app will resample/transcode to 8 bit mono/joint stereo/stereo from any .wav, and save as .wav
Using Audacity or Goldwave, or Sounforge, or WaveLab, one could also clean up the audio to remove noise.
I guess I need to know exactly what you need the finished product to be, and what the source really is, before going further.
As I understand it now, you have 16 bit .wav files, and need a simple conversion to 8 bit. What's the problem again?

The conversion is simple - like you said - we've done it. The problem is we have noise after the conversion. I do not understand where the noise is coming from. The source files don't have the noise, but the converted files do. We want to eliminate as much of the noise as possible during the conversion.

So noise is my problem - want to make this a one step process if possible as theres a huge number of files to process.

Final product - is a handheld talking dictionary using a 8bit audio chip or IC

I don't get it, either... this "conversion" is totally nonsensical to me.
Can you provide a good reason on WHY you wanna do it, or you just wanna do it?

Nonsensical maybe but is feasible and cheap for production.

Have to use 8 bit because we need to use a specific audio chip - but the source is 16 bit. Hence the conversion is ncessary.

I was suggesting Audacity... http://audacity.sourceforge.net
And thinking that if it's to be 8 bit quality, 44k sample rate is overkill - 22k or even 11k/8 bit would be more in keeping with 8 bit - or ADPCM, which uses fewer bits, but doesn't drop as much as a straight chop to 8 bits (IMA ADPCM uses only 4 bits/sample, but sounds better than 8 bit uncompressed) - to use any compressed format, obviously requires that the playback system supports that format - WAV is merely a wrapper, as you can wrap MP3 data in a WAV (Riff-Wave header).

And if you start with 44k sample rate, 16 bit WAV, then you can convert them to any format you like.
If you are experimenting with audo, you should join www.hydrogenaudio.org - they torture test various audio codecs, and there is a lot of expertise there.

If you are involved in developing something, it may be worth looking at the integer math version of VORBIS designed for embedded processors - though it still needs a reasonable amount of power.
http://www.xiph.org/ogg/vorbis/

Or Speex http://www.speex.org/ - a low bitrate codec designed for speech

The final file format needed is a wav 8 khz 8 bit. I started this thread asking about 44khz because thats the source file sample rate.

Although no matter the sample rate - 16 bit to 8 bit still introduces noise to the converted files. I dont understand why i get this noise.

Thanks for all these resources, although i am really interested in reading them and trying them all - i have a struggle with time. So i think im looking for a quick solution.

- - -

Although I realised i was silly thinking about lossless solutions - not much i can do about not loosing information when i chop 16 to 8 bit ? ? ?

 

[LTR12101B]

Would it be possible (copyrights etc.) to post an example - Original 16 bit, and an 8 bit showing the problem?

Hydrogenaudio never seem to have any problem with encoder test smples of 30 second duration - does that count as "fair use"?

 

[rebootjim]

Yes, even a 30 second sample clip that we can play with.
I have a feeling that there's enough talent around here to find a method that works, without adding any noise.
It may be the hardware at fault even, not the software, or a combination.
Between such apps as cheap as Audacity (free) and Wavelab ($$$), someone will find a way

 

[zver]

In this zip you got a 30 sec sample of a cllasic rock song and same song encoded with foobar to 44.1khz-8b.Sample sound good,so your handheld will sound even better-couse its a talking-which will take just few khz passage.
So its sound ok
But as the guys said..try a speex or vorbis which are good for low bitrate
Soory..attachment too big..
Anyway..ditch the acoustica,play that thing on a foobar or any good decent player which can encode too,and go with diskwriter and youre ok