Noise added to down sampled audio files

[LMO_Matt]

Hi Audition Community,

I updated Audition a few days ago and now experience noise being added to the files I down sample. This is my first time posting here so I hope I'm following the correct protocol.

I've attached a short segment of the original 48k 16 bit file and two examples of how noise is added when down sampled. I have a ulaw and a 16k,16 bit example. The noise is added across the file but is most pronounced where there should be silence.

The only settings I am using to down sample through the Batch converter is:

Format: libsndfile

Sample Type: 8,000 Hz, Mono, 8 bit

Format Settings: libsndfile: Waveform Audio

u-law 8bit PCM (ITU-T G.711)

Default Byte Order

I've tried dithering with Triangular and Gaussian settings as well as changing the Byte Order to Little-Endian. All outputs contain the same added noise.

I am perplexed! Any help or guidance will really be appreciated.

Thanks,

Matt

[SteveG_AudioMasters_]

All 8-bit files have a very poor signal to noise ratio (SNR) - it's simple to work out; each bit of resolution you use gives you 6.02dB of SNR, so if you have an 8-bit file, the noise floor will be at about -48db, which is worse than a lot of cheap cassette recorders manage. And this is dictated by the Laws of Physics. So if you downsample a file from 16 bits to 8 bits, you've halved the noise floor. And that's exactly what I can see on your files - except it's slightly worse on the u-law one, as  it's dithered. The lowest 'real' sample value you have is at -42dB and the sample below that is at  -infinity. That's a lot of noise!

Whenever you save files for use on low-SNR systems, you have to use compression judiciously to make sure that all of the message can be heard above the noise. Generally this means limiting the signal dynamic range to about 10dB at a a maximum, and making sure that your final file is normalized to 0dB, otherwise you're giving SNR away for nothing.

[LMO_Matt]

Hi SteveG,

Thanks for getting back to me. I truly appreciate you reading and responding to this thread.

While your explanation makes sense I still don't quite understand how the down sampling Audition is interpreting the SNR. If the original file had a decent amount of noise I could understand the 8 bit file coming out with much more noise. 

On  a side note, I thought I had disabled dithering when down sampling the samples yesterday. Also, I thought I could attach more files to this post but it appears I need to link to them so I'm linking to a folder on Box.

I've linked 3 additional samples to show that something has changed with the down sampling function in Audition. There is a 48k 16 bit file (UK_48k.wav), the 8k 8 bit file I converted from UK_48k.wav back in December 2019 (UK_8k_Old.wav), and the 8k 8 bit file I converted today (UK_8k_New.wav).

Link: https://app.box.com/s/av34e4hytxarj1575zfjltaws4f3rzxs

The down sampled file from December is fine. Today's down sample has a tremendous amount of noise.

The settings are the same as mentioned  in my original post except this time I made sure dithering was disabled as I had done in Decmeber.

Thanks again in advance for any response you may have.

Matt