Re: [Audacity-quality] ERB scale

[Steve t. F. <ste...@gm...>]

On 31 August 2015 at 18:26, Paul Licameli <pau...@gm...> wrote:
>
>
> On Mon, Aug 31, 2015 at 1:17 PM, Steve the Fiddle <ste...@gm...>
> wrote:
>>
>> On 31 August 2015 at 16:18, James Crook <cr...@in...> wrote:
>> > On 31/08/2015 16:00, Peter Sampson wrote:
>> >
>> > Hi Paul,
>> >
>> > I've been doing my best to update the page in the 2.1.2 Manual
>> > for Spectrograms Preferences:
>> > http://manual.audacityteam.org/man/Spectrograms_Preferences
>> >
>> > I got fairly stuck on the various Scale and Algorithm settings, but
>> > ploughed
>> > ahead with Wikipedia as "my friend" - so it *will* need a fair bit of
>> > editing
>> > in those two sections.
>> >
>> > In particular I note that Wikipedia refers to:
>> >  "equivalent rectangular bandwidth or ERB"
>> >
>> > On this basis it seems to me that perhaps your nomenclature for ERBS
>> > is tautologous in its use of the final S for scale - i.e. a scale which
>> > is
>> > ERB
>> > Scale. And thus the  name for this scale might be better as simply
>> > "ERB".
>> >
>> > And this is an allowably fixable item under semi-freddo.
>> >
>> > Not just allowable.  If these kind of changes are wanted/valid/correct,
>> > positively encouraged.
>> >
>> > Later, in Frozen, it's more locked down.  P1 fixes in Frozen are always
>> > automatically OK.  P2s are generally good, but ask first.
>> >
>> > The RM.
>> >
>>
>> I've been looking at this to see if I can help Peter with it, and
>> there is something very confusing here:
>>
>> The Spectrogram "preferences" and track "settings" refer to "Scale",
>> which I think suggests that it is referring to the scale of the
>> vertical track ruler. However, the units of the vertical ruler are not
>> the units of the scale.
>>
>> Example:
>> This image from Wikipedia indicates that 3500 Hz is approximately 2000
>> mel, https://upload.wikimedia.org/wikipedia/commons/a/aa/Mel-Hz_plot.svg
>> yet in the track Spectrogram view, a frequency of 3500 Hz is shown
>> with a value of "3500" (Hz) on the vertical ruler, not "2000" (mel).
>
>
> It has also been true that log(f) was a choice, and the height the a
> frequency plots to is proportional to its logarithm, but the vertical ruler
> is labelled with Hz or kHz.
>
> The relabelling of the ruler is a future project.
>
>>
>>
>> "1/f" is also confusing because the inverse of a large number is a
>> small number, but it is not shown that way.
>
>
> I said I wanted to change that to "Period" and I have done so.  In this case
> the distance of a frequency from the top of the fully zoomed-out scale is
> inversely proportional to frequency, or directly proportional to period.  It
> is true that periods decrease rather than increasing then as you ascend the
> scale.  But I can't think of a better short name than "period."

What's it for? Why would people use the "period" scale? For those use
cases, wouldn't it make more sense for the smallest period to be at
the bottom and the largest period at the top?

Steve


>
>>
>>
>> Steve
>>