Hi Paul,

> I might use a higher sample rate to get more resolution with the -T Argument.

Is there a specific reason for rounding the start offset to an integer number of samples? I would guess that allowing arbitrary offsets would allow a slightly better allocation of samples to their nearest symbol intervals.  

Best wishes,
Markus

-----Ursprüngliche Mitteilung-----
Von: Paul Nicholson <[email protected]>
An: rsgb_lf_group <[email protected]>
Verschickt: Fr, 12. Apr. 2019 14:36
Betreff: Re: LF: ebnaut results depend on sample rate

Jacek wrote:
> the results i get from ebnaut are different depending on
> signal bandwidth.

This is possible and quite common.

> 10Hz:
> found rank 6348 ber 4.2147e-01 Eb/N0 -2.0
> carrier Eb/N0: -0.3 dB

> 240Hz:
> found rank 16679 ber 4.3029e-01 Eb/N0 -3.1
> carrier Eb/N0: -1.6 dB

For the carrier Eb/N0, the value of N0 is measured using the full
bandwidth of the input signal.  When the band is wider it may
present a higher N0 because of things like mains sidebands and
harmonics, and other interference.  That's probably the case

For the decode there should not be much difference, and indeed we
see a similar BER. 42.1% and 43.0%.

The Eb/N0 in the 'found' log message is estimated from the BER and
the slope is quite steep, accuracy is poor for weak signals.
The carrier Eb/N0 is usually more realistic, unless you have
interference in the input band.

I usually resample down to about 10 or 20 times the symbol rate,
but at least 10Hz so as to get a reasonable N0 estimate.  If I
expect to have to sweep for the time stamp I might use a higher
sample rate to get more resolution with the -T argument.

--
Paul Nicholson