[GreenKeys] ITTY/RTTY Sampling tests

[Harold Hallikainen]

On Thu, September 23, 2021 7:13 pm, ad7i wrote:
> With the 8K sampling, how many bits per sample were used per sample?
>
> The reason I ask is because the 1970s T1 systems did sample at 8000 sps
> and
> transmitted 8 bits per sample, but those 8 bits represented a compressed
> (compandered) value.  Although I've forgotten the specifics, the entry on
> Wikipedia for uLaw coding seems to indicate that the compander transformed
> the data from 14 bits uncompressed to 8 bits compressed, and vice versa at
> the RX end.
>
> Paul, ad7i

Yes, u-Law encoding is just a nonlinear binary code for each sample that
improves the signal to noise ratio but adds some distortion. The SNR is
important with voice transmission since it has a wide dynamic range. Low
level portions of the voice could go below the noise level. With FSK,
however, there are no low level signals. It's all at 0 dB FS. So, u-Law
encoding is not useful here.

I think there's something going on between the ADC sample rate,
compression sample rate, and transmission sample rate. If the frequencies
and bit rate are lower, I suspect there's a buffer filling up somewhere.

I've never done anything with MP3 files, so I'm not familiar with how they
are encoded. A WAV file, however, has the sample rate in the header. Does
the MP3 encoder know that the incoming samples are 8 kHz? Maybe it's
putting the wrong sample rate in the header.

https://soundbridge.io/audio-formats-file-types/ says MP3 can do 8 kHz
sampe rate with a resulting bitstream of 16 to 96 kbps.

It just seems that the audio is being sent more slowly and the buffer will
fill, the bucket will fill with paper tape, etc.


Harold



-- 
http://w6iwi.org