MIME-Version: 1.0
In-Reply-To: <CAA8k23RTFF1D1TDNxKN_pRQPF6vc3PC84HYrW-Q5cCeZfM5U=A@mail.gmail.com>
References: <CAA8k23Tpk=77uv=VqenyRV52LDBMoKKfGrE9k32YA3EQ+A1tnA@mail.gmail.com>
	<000901cd302c$671ed2e0$1502a8c0@Clemens04>
	<CAA8k23RTFF1D1TDNxKN_pRQPF6vc3PC84HYrW-Q5cCeZfM5U=A@mail.gmail.com>
Date: Sat, 12 May 2012 20:35:00 +0100
Message-ID: <CAA8k23S7k=nNNWiiRD_Ee6AK_0kjJ2+m2dj0sMonpUvnt7p0tg@mail.gmail.com>
From: Andy Talbot <andy.g4jnt@gmail.com>
To: rsgb_lf_group@blacksheep.org
Subject: Re: LF: Re: Demonstrating audibility of weak signals - utilty to play with.
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
X-AOL-SCOLL-URL_COUNT: 0  
X-AOL-SCOLL-AUTHENTICATION: mail_rly_antispam_dkim-m233.2 ; domain : gmail.com DKIM : pass  
x-aol-sid: 3039ac1d40cb4faebbe37e94
X-AOL-IP: 195.171.43.25
X-AOL-SPF: domain : blacksheep.org SPF : none


Correction,
The 0dB tone reference has an RMS value of 10000 in the generated .wav
file, not 20000 as stated previously.
The peak is therefore 14142 which gives 7.3dB of headroom for a
resonable amount of added noise before output scaling becomes
necessary

Andy
G4JNT


On 12 May 2012 13:22, Andy Talbot <andy.g4jnt@gmail.com> wrote:
> Clipping is where the generated samples of signal plus noise exceed
> the 16 bit range of the .WAV file. =A0 All values are calculated as
> floating point, scaled to keep close to the maximum. =A0ie the tone at
> 0dB amplitude has a peak of 28282 (RMS =3D 20000)
>
> Noise is calculated based on an RMS at 0dB of 20000 in the wav file
> calculated for teh full sampling rate bandwidth, the same as the tone
> 0dB level, in a truely statistical manner. =A0 So you can see that for
> noise levels much above -20dB some spikes will occasionally exceed the
> alllowed maximum. =A0 If noise were 0dB at full bandwidth, and output
> scale factor were also 0dB it would clip something like 63% of the
> time as the RMS is defined as the one-signma probability. =A0The need to
> adjust noise amplitude for defined signal bandwidth also adds to the
> confusion
>
> Noise is scaled depending on the requested S/N ratio, and the ratio of
> defined signal bandwidth ro sampling rate and added to the tone
> waveform. =A0The resulting sample generated, are converted to integers
> by multiplying by the scaling factor (0dB =3D 1) =A0 Any that exceed 3276=
7
> / -32767 are clipped to those values to prevent overflow, and a count
> of the number of occurances made =A0 =A0If just the odd few are present,
> the amount of clipping is shown,, but =A0if the number of clipped
> samples exceeds 1% of the total for the file, this is labelled
> "Severe" =A0 Users can then adjust the output scaling appropriately
>
> Andy
> www.g4jnt.com
>
>
> On 12 May 2012 11:45, Clemens =A0Paul <cpaul@gmx.net> wrote:
>> Hi Andy,
>>
>> thanks for the nice little program.
>> I've been playing a bit with the utility and wonder under which
>> circumstances clipping occurs.
>> See attached screenshot.
>> With the parameters used in this example there is clipping.
>> I guess the level in 1Hz BW, here 33dB is the crucial parameter?
>> Woud you mind to explain what output scaling means?
>>
>> 73
>> Clemens
>> DL4RAJ
>>
>>