[ChrisCoaster]

Softwa Freeware versions of Audacity and MP3Gain.

I recorded approximately 15sec of white and pink noise in Audacity. I
recorded the white noise to approx -12dB in Audacity, left all my
levels ALONE, and recorded a separate mp3 file of pink noise. The
pink noise, understandably due to its weighting, approximated -6dB in
Audacity. *Audacity's recording scale is -infinity 0dB(digital),
just like a digital meter is supposed to appear.*

(As a side note, I did a freqency spectrum analysis of both using
Audacity's tool for that purpose, and White indeed came back ruler
flat, Pink, flat but slanted up toward the bottom end, down towards
the top - indicating Pink's weighting).

However, when I checked the levels of the finished mp3s in MP3Gain,
the White came in at "83db" according to Mp3Gain's scale(75 - 105dB),
and the Pink - at about "90". This is with MP3 gain's default 89dB
gain setting. I did not apply the gain - this was just to check their
levels on a separate application.

To summarize:
(recording levels in Audacity)
White: -12dB
Pink: -6dB

(levels as analyzed by mp3gain)
White: 83dB ~or~ "-22dB"(105 - 83)
Pink: -90dB ~or! "-15dB"(105-90)

My question is: Should I or should I NOT regain the tones to
89dB(White) and 95dB(Pink) in Mp3Gain's app, or leave them ALONE - at
the level I recorded them acc. to Audacity's record meters?

IOW: Are they at reference level as recorded in Audacity, or do they
need correction in mp3Gain?

-ChrisCoaster

[ChrisCoaster]

On Dec 26, 2:04*pm, ChrisCoaster wrote:
Softwa Freeware versions of Audacity and MP3Gain.

I recorded approximately 15sec of white and pink noise in Audacity. *I
recorded the white noise to approx -12dB in Audacity, left all my
levels ALONE, and recorded a separate mp3 file of pink noise. *The
pink noise, understandably due to its weighting, approximated -6dB in
Audacity. **Audacity's recording scale is -infinity 0dB(digital),
just like a digital meter is supposed to appear.*

(As a side note, I did a freqency spectrum analysis of both using
Audacity's tool for that purpose, and White indeed came back ruler
flat, Pink, flat but slanted up toward the bottom end, down towards
the top - indicating Pink's weighting).

However, when I checked the levels of the finished mp3s in MP3Gain,
the White came in at "83db" according to Mp3Gain's scale(75 - 105dB),
and the Pink - at about "90". *This is with MP3 gain's default 89dB
gain setting. *I did not apply the gain - this was just to check their
levels on a separate application.

To summarize:
(recording levels in Audacity)
White: -12dB
Pink: -6dB

(levels as analyzed by mp3gain)
White: 83dB *~or~ "-22dB"(105 - 83)
Pink: -90dB ~or! "-15dB"(105-90)

My question is: *Should I or should I NOT regain the tones to
89dB(White) and 95dB(Pink) in Mp3Gain's app, or leave them ALONE - at
the level I recorded them acc. to Audacity's record meters?

IOW: Are they at reference level as recorded in Audacity, or do they
need correction in mp3Gain?

-ChrisCoaster
_____________
SEE CORRECTION ABOVE - (levels as analyzed by mp3gain) "Pink -90dB"
should read Pink 90dB!! Sorry!

-CC
_____________

[ChrisCoaster]

On Dec 26, 2:04*pm, ChrisCoaster wrote:
Softwa Freeware versions of Audacity and MP3Gain.

I recorded approximately 15sec of white and pink noise in Audacity. *I
recorded the white noise to approx -12dB in Audacity, left all my
levels ALONE, and recorded a separate mp3 file of pink noise. *The
pink noise, understandably due to its weighting, approximated -6dB in
Audacity. **Audacity's recording scale is -infinity 0dB(digital),
just like a digital meter is supposed to appear.*

(As a side note, I did a freqency spectrum analysis of both using
Audacity's tool for that purpose, and White indeed came back ruler
flat, Pink, flat but slanted up toward the bottom end, down towards
the top - indicating Pink's weighting).

However, when I checked the levels of the finished mp3s in MP3Gain,
the White came in at "83db" according to Mp3Gain's scale(75 - 105dB),
and the Pink - at about "90". *This is with MP3 gain's default 89dB
gain setting. *I did not apply the gain - this was just to check their
levels on a separate application.

To summarize:
(recording levels in Audacity)
White: -12dB
Pink: -6dB

(levels as analyzed by mp3gain)
White: 83dB *~or~ "-22dB"(105 - 83)
Pink: -90dB ~or! "-15dB"(105-90)

My question is: *Should I or should I NOT regain the tones to
89dB(White) and 95dB(Pink) in Mp3Gain's app, or leave them ALONE - at
the level I recorded them acc. to Audacity's record meters?

IOW: Are they at reference level as recorded in Audacity, or do they
need correction in mp3Gain?

-ChrisCoaster
______________________

Hmmm. Folks 'round here must think I'm some sorta TROLL or something.

This is serious business, this levels thing!

-CC

[Scott Dorsey]

ChrisCoaster wrote:
Hmmm. Folks 'round here must think I'm some sorta TROLL or something.

This is serious business, this levels thing!

MP3gain adjusts a scaling factor that basically tells the MP3 decoder
how loud to play the stuff back at. It does not adjust the data going
into the encoder.

When you adjust the data going into the encoder, you are varying the
data that the encoding process throws away. This will affect the output
levels; the incoming and outgoing levels do not have a 1:1 relationship
because of the encoding process. Turn the gain up 3dB going in, and what
comes out is not necessarily 3dB higher.

Datasets like pink noise don't encode cleanly either, and the end result
is not always predictable. MP3 is really intended to be used on music
that has a lot of correlation between different bands and it falls apart
with random noise.

On the other hand, when you adjust the scaling factor with MP3gain, you
ARE getting a 1:1 relationship with the output sound level, because in
fact you are directly adjusting the output sound level.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[ChrisCoaster]

On Dec 27, 1:39*pm, (Scott Dorsey) wrote:
ChrisCoaster wrote:
Hmmm. *Folks 'round here must think I'm some sorta TROLL or something.

This is serious business, this levels thing!

MP3gain adjusts a scaling factor that basically tells the MP3 decoder
how loud to play the stuff back at. *It does not adjust the data going
into the encoder.

When you adjust the data going into the encoder, you are varying the
data that the encoding process throws away. *This will affect the output
levels; the incoming and outgoing levels do not have a 1:1 relationship
because of the encoding process. *Turn the gain up 3dB going in, and what
comes out is not necessarily 3dB higher.

Datasets like pink noise don't encode cleanly either, and the end result
is not always predictable. *MP3 is really intended to be used on music
that has a lot of correlation between different bands and it falls apart
with random noise.

On the other hand, when you adjust the scaling factor with MP3gain, you
ARE getting a 1:1 relationship with the output sound level, because in
fact you are directly adjusting the output sound level.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
_______________________

Thanks Scott! Now that I "sort of" understand MP3gain's level system,
in Audacity, are levels really what Audacity says they are? And how
can I confirm. record 1kHz test tones at certain levels?

-CC

[Scott Dorsey]

ChrisCoaster wrote:
Thanks Scott! Now that I "sort of" understand MP3gain's level system,
in Audacity, are levels really what Audacity says they are? And how
can I confirm. record 1kHz test tones at certain levels?

The peak levels in Audacity are what it says they are, but the levels indicated
are peak and the metering is rather crude.

You can verify by creating a 1 KHz test tone at one level in Audacity,
then saving it as a .wav file and then using od to look at the actual
data in the file as 32-bit signed integeners. Skip past the header,
look for the highest and lowest values and those are the peaks.

However, once you export anything into an MP3 file and go through the
MP3 encoder, all bets are off. What will happen to levels through the
encoder is not easily predictable.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Scott Dorsey]

ChrisCoaster wrote:
Listen, like a complete ASS I hit "report discussion" completely by
accident. Serves GoogleGroups right for having that option right next
to "Collapse All" which I frequently use when the number of posts gets
long in a thread. I open each reply, read it, and close it, similar
to flippng pages in a book. (some habits die hard!)

I HOPE it is recognized by the admins as a MISTAKE and is dismissed.
Apologies to Scott or anyone else if it causes any problems.

I don't think google actually does anything with any of that information.
If they did, they'd probably do something about their spam problem... as it
is, the vast majority of Usenet spam comes from google. So I would not worry.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."