Re: [Audacity-devel] Renaming the "1/f" frequency scale

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

On 20 August 2015 at 15:27, Roger Dannenberg <rb...@cs...> wrote:
> On 8/20/15 9:42 AM, Federico Miyara wrote:
>
>
> Roger,
>
> Finally, a minor point is that dB has no units, it's just a ratio (well
> actually the log of the ratio) with respect to some  amplitude we
> arbitrarily call "1.0". Indicating "dB" is important to communicate that
> it's a log scale. Similarly, the autocorrelation plots show numbers that
> measure "degree of autocorrelation," but the units are arbitrary. Since the
> scale is linear, I think just putting numbers alone makes sense.
>
>
> The decibel (dB) *is* a unit. Being a unit doesn't mean that it has a
> physical dimension. The radian is another dimensionless unit. See on page
> 127 of the International System of Units brochure:
>
> http://www.bipm.org/utils/common/pdf/si_brochure_8.pdf
>
> "The decibel *is* a unit" seems a little strong. Did you read page 127?
> "some scientists consider that they should not even be called units" and
> "The units neper, bel, and decibel ... are not considered as SI units." But
> OK, I'll agree that *my* statement, "dB has no units" was questionable. I
> stand corrected.
>
>
> However, there are two kinds of dB, absolute and relative ones. Absolute
> dB's refer to some reference, such as sound pressure level (reference = 20
> micropascal), or the digital audio dB (reference = maximum unclipped value);
> relative dB's refer to quantities such as gain or signal to noise ratio.
>
> My point was that *in Audacity track displays*, "dB" does not mean time or
> micropascals or frequency or any other particular dimension, and the
> reference level is unknown. You might argue that there is at least some
> unknown sound pressure level corresponding to 0dB in an Audacity track
> display, but remember that signals are not necessarily audio signals. I've
> loaded all kinds of data into Audacity, e.g. frequency estimates from a
> monophonic pitch detector. "dB" then becomes something akin to semitones.

We have recently had someone asking about how to load market share
prices into Audacity.
I guess if they were displayed as a waveform, the vertical scale
should be dollars ;-)

Steve


>
> -Roger
>
>
>
> Regards,
>
> Federico
>
>
>
>
> -Roger
>
> On 8/19/15 11:43 PM, Paul Licameli wrote:
>
> I think you are speaking of Audacity's Plot Spectrum window, but I am
> thinking of the Pitch(EAC) view of tracks.  Your horizontal is my vertical,
> and your vertical is my color scale.  My horizontal is the center time of a
> sliding analysis window.
>
> I take away that you agree that "period" is a natural enough name for my
> vertical.
>
> In the Plot Spectrum window, the Spectrum algorithm labels the vertical
> scale in dB, but I don't know what units make sense for the vertical of the
> other algorithms.  You tell me there is no particular unit.
>
> PRL
>
>
> On Wed, Aug 19, 2015 at 11:28 PM, Roger Dannenberg <rb...@cs...> wrote:
>>
>> Whether you call it "1/f" or "period", the name applies to the horizontal
>> axis, which is measured in seconds. In fact, you could also call it "time"
>> or "duration" but since the objective is to get a sense of pitch, it's more
>> intuitive to use "1/f" or "period". In fact, the position on the horizontal
>> axis is the lag (measured in units of time) used in the autocorrelation.
>> E.g. at the 10ms point on the horizontal axis, the height is based on the
>> correlation between the signal and a copy of the signal delayed by 10ms. If
>> the signal is periodic with a 10ms period, the correlation will be perfect
>> and the value of the EAC function will be high at 10ms. The correlation is
>> less good and the value of the EAC function will be lower at other points on
>> the horizontal axis, which correspond to different lags or periods.
>>     Your description refers to "height on the scale", but height is a
>> different axis with different units. Just as a spectrum plot is typically
>> magnitude (in no particular units) as a function of frequency (= 1/time),
>> EAC is magnitude (in no particular units) as a function of lag or period
>> (time = 1/frequency).
>>     You can plot any function of frequency f, F(f), as a function of
>> period p = 1/f simply by plotting F(1/p), so we should be careful to
>> distinguish the computation of EAC from the way it is displayed. In the case
>> of EAC, the "natural" parameter is lag or period, i.e. EAC(p), so if you
>> wanted to make a function of frequency, you'd compute EAC(1/f).
>>
>>     -Roger
>>
>>
>> On 8/19/15 10:42 PM, Paul Licameli wrote:
>>
>> I don't like the name "1/f" I gave to the last of the new frequency scales
>> -- what at first I called "undertone" which Roger didn't like -- so how
>> about "period" instead?
>>
>> Why is it there? Historical reasons, because it is the scale that old
>> Pitch(EAC) view was plotted on (though I never saw where the documentation
>> explained that).  If you really liked the old Pitch view (and figured out
>> that you needed to fully zoom out the spectrogram view scale first), you can
>> still get the same displays.
>>
>> 1 / f does not really describe it -- really, pixel height as a function of
>> f is more like a - b / f for certain positive constants a and b.
>>
>> "Period" is easier to explain, period varies linearly with height on that
>> scale, though decreasing as height increases.
>>
>> PRL
>>
>>
>>
>>
>> ------------------------------------------------------------------------------
>>
>>
>>
>> _______________________________________________
>> audacity-devel mailing list
>> aud...@li...
>> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>>
>>
>>
>> ------------------------------------------------------------------------------
>>
>> _______________________________________________
>> audacity-devel mailing list
>> aud...@li...
>> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>>
>
>
>
> ------------------------------------------------------------------------------
>
>
>
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>
>
>
> ------------------------------------------------------------------------------
>
>
>
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>
>
>
>
> ------------------------------------------------------------------------------
>
>
>
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>
>
> ------------------------------------------------------------------------------
>
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>