Re: [Audacity-devel] [Audacity-quality] Noise generators

[Martyn S. <mar...@gm...>]

On 05/02/2013 21:13, Gale Andrews wrote:
>
> | From Steve the Fiddle <ste...@gm...>
> | Tue, 5 Feb 2013 19:09:54 +0000
> | Subject: [Audacity-quality] [Audacity-devel] Noise generators
>> On 5 February 2013 17:05, Gale Andrews <ga...@au...> wrote:
>>>
>>> | From Steve the Fiddle <ste...@gm...>
>>> | Tue, 5 Feb 2013 15:59:24 +0000
>>> | Subject: [Audacity-quality] [Audacity-devel] Noise generators

<snip>

>>>>> | From Steve the Fiddle <ste...@gm...>
>>>>> | Tue, 5 Feb 2013 04:26:04 +0000
>>>>> | Subject: [Audacity-quality] [Audacity-devel] Noise generators
>>>>>> This patch is against svn r12211 and supersedes the previous patches.
>>>>>>
>>>>>> The Pink noise generator is the same as previous patches, based on
>>>>>> Paul Kellet "instrumentation grade" algorithm. (which I hope will be
>>>>>> nice for Richard Ash when he gets his SPL meter working :=)
>>>>>>
>>>>>> The improvements in this version are to the Brownian noise. The
>>>>>> waveform is less "constrained" than the previous version, so it
>>>>>> doesn't "look clipped" even with 10 million samples duration, but
>>>>>> durations as short as 1 second should still be very close (and not
>>>>>> more than) the user selected peak level.
>
> Six minutes of brownian noise at 0.5 amplitude no longer looks
> very "flat  topped" and the peak is -6 dB as expected.
>
> I got six individual clipped samples with six minutes of brownian
> at amplitude 1, which is less clipping than with the other noise
> types.
>
> Two seconds of brownian at 0.5 always seems to have a peak of
> - 6 dB.
>
> One second of brownian at 0.5 usually seems to have a peak of
> - 6 dB but the peak was -6.8 dB on the first try out of six.
>
>
>>>>>> The parameters for Brownian noise now take into account the sample
>>>>>> rate to provide 6dB/octave from (close to) 20 Hz to (close to) the
>>>>>> Nyquist frequency for all Audacity project rates (8000 to 384000 Hz).
>
> I'm looking in Plot Spectrum at size 256, log, Hanning window
> and in that view the curve seems to be steeper for lower
> frequencies (for pink as well as brown). For example at:
>
>     352800 Hz
>     amplitude 0.5,
>     stereo track,
>     2 minutes 1 second duration
>
> I see 2000 Hz is -24 dB and 4000 Hz is -36 dB, so twice the
> expected drop in level (unless I'm misunderstanding).
>
> At 44100 Hz (otherwise as above), 200 Hz is at -14 dB and
> 400 Hz at -24 dB.
>
> You can begin to see the steeper slope at lower frequencies
> even at default size 512/log.
>
> Although HEAD isn't correct either at low sizes/log, I don't
> see the steepening of the slope there, so it's closer to -6 dB
> per octave at lower frequencies.
>
> What is the "optimal" view to judge this in Plot Spectrum?

I go for a much longer length, like 16k.

HTH
Martyn

>
> Gale
>
>
>
>
>>>>>> The "ToDo" is also fixed by this patch, as is the 2^18 sample glitch bug.
>
>
> ------------------------------------------------------------------------------
> Free Next-Gen Firewall Hardware Offer
> Buy your Sophos next-gen firewall before the end March 2013
> and get the hardware for free! Learn more.
> http://p.sf.net/sfu/sophos-d2d-feb
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>