Can anyone test "Plot Spectrum" in Audacity 1.3.8

[steve]

I notice that "Plot Spectrum" in Audacity 1.3.8 produces a different output from earlier versions.

I'm using Audacity 1.3.8 on Linux, but I expect this to be the same on all platforms.

Here is an example of the type of thing:

In any version of Audacity 1.3.x
1) Generate 20 seconds of silence at 32 bit 44.1kHz
2) Set Audacity Preferences to "High Quality Conversion" > "Shaped dither" (in "Quality" tab)
3) Export as 16 bit WAV
4) Import this file into Audacity (it will have had shaped dither applied due to the bit rate being dropped from 32bit to 16 bit).

In Audacity 1.3.8 I get this and it appears that no dither has been applied:

Silence downsampled from 32bit to 16bit in Audacity 1.3.8

screenshot_138.png (43.52 KiB) Viewed 1701 times

When using Audacity 1.3.7 the characteristic high frequency noise of the shaped dither is clearly evident:

Silence downsampled from 32bit to 16bit in Audacity 1.3.7

screenshot_137.png (36.04 KiB) Viewed 1707 times

Can anyone test this in other software?
(Koz, do you have your Cool Edit Pro installed anywhere?)

[Gale Andrews]

The spectrum is off the scale. Reduce the Meter/Waveform dB Range on the interface tab of Preferences


Gale

[steve]

The spectrum is off the scale. Reduce the Meter/Waveform dB Range on the interface tab of Preferences


Gale

But why is it off scale in Audacity 1.3.8 and on scale in previous versions of Audacity?
Both images below show a range of -90dB to -80dB, but in Audacity 1.3.8 the peak is at around -110dB rather than -80dB !

[Gale Andrews]

The spectrum is off the scale. Reduce the Meter/Waveform dB Range on the interface tab of Preferences

Gale

But why is it off scale in Audacity 1.3.8 and on scale in previous versions of Audacity?
Both images below show a range of -90dB to -80dB, but in Audacity 1.3.8 the peak is at around -110dB rather than -80dB !

This was discussed on -devel list. I've commented here:
http://n2.nabble.com/Plot-Spectrum-in-1 ... 35287.html

and the thread where it was discussed in detail is here:
http://n2.nabble.com/Re%3A-Question-abo ... 92212.html

From user's point of view (if coming from previous versions) I do think the new scaling is confusing, and needs a full explanation in the manual. I've made my concerns clear on -devel list.


Gale

[steve]

Thanks Gale,
I've not yet read the "Numerical Recipes in C" - can't open it in Evince, and don't currently have Acrobat installed - also probably a bit heavy for bedtime reading - but the "bins" explanation worked for me. I look forward to reading the full explanation in the manual.