[Gary Eickmeier]

Peter -

I have gone back to your post with the proposed paper on the Orange Curve
(Orange Noise Characteristic),

https://groups.google.com/forum/#!searchin/alt.audio.pro.live-sound/orange$20noise/alt.audio.pro.live-sound/cMkHUOVP20M/6vqOhGu1Ov4J

in which you state two laws, the idea being that most naturally occuring
sounds, including recorded music, will display the Orange Curve on your
frequency analysis window in Cool Edit or Audition. The Orange Curve peaks
at 320 Hz, falls off above and below that by 3 dB/octave. I have looked at
enough analyses of my recordings to confirm that for the vast majority of
the times and sections of most any track. Most fascinating. In fact, it
helped me correct my "thrumming bass guitar" problem one fine time. I simply
EQed it to bring the curve under the Orange Curve.

I'm thinking it is kind of like publishing a "typical histogram" of a good
photograph. You might propose that it would have a full range of values from
black to white. Not ALWAYS true, but it is usually true that if everything
is skewed to one side, or goes above clipping, the photo will look too
light, dark, or have no detail in the highlights.

In photography, a good camera will accurately record a dynamic range of 8
bits/ channel, all values from black to white and all colors within a
certain range. But the values and the colors that actually appear in the
real world just happen to fall in the curve of the typical histogram unless
you are on a snow bank or in a coal mine. In audio, a good recording or
microphone will record all frequencies with a flat response, but in the real
world of music the actual pattern of the appearance of the various
frequencies will fall under the Orange Curve and hopefully will fill the
area under the curve.

In audio, you say that the curve will indicate the perceived linearity of
the sound in a good recording, which seems to hold true. You might add that
spatially, we expect to hear a soundstage with instruments that are balanced
at the center, with a full spread across the whole, unless it is a solo
guitar or voice.

SO - one of the most fascinating ideas I have read. So how did it do? Did
they publish it? This all happened about 15 years ago, so what has become of
it?

Thanks,
Gary Eickmeier

[[email protected]]

in which you state two laws, the idea being that most naturally occuring

sounds, including recorded music, will display the Orange Curve on your

frequency analysis window in Cool Edit or Audition. The Orange Curve peaks

at 320 Hz, falls off above and below that by 3 dB/octave.

Another common curve is the FM radio de-emphasis curve.
Flat at to about 2.2 kHz, and 6 dB / octave down above that.

I find many recordings follow that curve, and if they don't they sound bright or dull. The ones that follow that curve seem well balanced.

[Ron C[_2_]]

On 5/28/2014 10:43 PM, wrote:


in which you state two laws, the idea being that most naturally occuring

sounds, including recorded music, will display the Orange Curve on your

frequency analysis window in Cool Edit or Audition. The Orange Curve peaks

at 320 Hz, falls off above and below that by 3 dB/octave.

Another common curve is the FM radio de-emphasis curve.
Flat at to about 2.2 kHz, and 6 dB / octave down above that.

I find many recordings follow that curve, and if they don't they sound bright or dull. The ones that follow that curve seem well balanced.



There's another curve out there called green noise; noise natural to
[an] environment. However, it seems said environment is ill defined.

It may (or may not) be related to the orange noise concept.

==
Later....
Ron Capik
--

[Trevor]

"Gary Eickmeier" wrote in message
...
In photography, a good camera will accurately record a dynamic range of 8
bits/ channel,

In fact only crap camera's record that little until unknowing photographers
squash the data into an 8 bit lossy Jpeg. (like recording to a 64kb/s mp3 to
stick with your metaphor) Good DSLR's can capture 14 bits these days, with
even low end ones managing 11-12 bits, as do many P&S cameras now.


all values from black to white and all colors within a certain range. But
the values and the colors that actually appear in the real world just
happen to fall in the curve of the typical histogram unless you are on a
snow bank or in a coal mine.

And quite often something a little less extreme, but still not "average"
enough for the built in light meters which do exactly that, by expecting
everything to be 18% neutral grey.


In audio, a good recording or microphone will record all frequencies with a
flat response, but in the real world of music the actual pattern of the
appearance of the various frequencies will fall under the Orange Curve and
hopefully will fill the area under the curve.

Just as with your black cat in a coal mine, there are always exceptions to
an average, and it pays to know when to expect that. Not all music is the
equivalent of "18% neutral grey", but I'm surpised there aren't switchable
filters in every recorder to try and make it so, just as there are in every
camera. Even pro ones.

Trevor.

[Scott Dorsey]

In article ,
wrote:
in which you state two laws, the idea being that most naturally occuring
sounds, including recorded music, will display the Orange Curve on your
frequency analysis window in Cool Edit or Audition. The Orange Curve peaks
at 320 Hz, falls off above and below that by 3 dB/octave.

Another common curve is the FM radio de-emphasis curve.
Flat at to about 2.2 kHz, and 6 dB / octave down above that.

Yes, but that's a de-emphasis curve. Just like CD de-emphasis or RIAA
de-emphasis, it is countered by an inverse filter on the source side, so
the end response remains flat.

People use emphasis in cases where the dynamic range in the channel is
different at some frequencies than others... by equalizing before and
after the channel you can optimize the available dynamic range for the
frequency content of the signal while still having flat response.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Peter Larsen[_3_]]

On 28-05-2014 14:04, Gary Eickmeier wrote:

Peter -

I have gone back to your post with the proposed paper on the Orange Curve
(Orange Noise Characteristic),

https://groups.google.com/forum/#!searchin/alt.audio.pro.live-sound/orange$20noise/alt.audio.pro.live-sound/cMkHUOVP20M/6vqOhGu1Ov4J

in which you state two laws, the idea being that most naturally occuring
sounds, including recorded music, will display the Orange Curve on your
frequency analysis window in Cool Edit or Audition. The Orange Curve peaks
at 320 Hz, falls off above and below that by 3 dB/octave. I have looked at
enough analyses of my recordings to confirm that for the vast majority of
the times and sections of most any track. Most fascinating. In fact, it
helped me correct my "thrumming bass guitar" problem one fine time. I simply
EQed it to bring the curve under the Orange Curve.

It is kinda simple that it should end up like that with sound from
objects. The parameters are object size (acoustic impedance match),
directionality and air absorption. But you need to understand that it is
what the brain expects the reality to be and not the reality per se. So
real world audio may not fit.

A wonderful example is applause, it is by definition an unlinear sound
source because it is horn loaded by the shape of the hands, it is also a
wonderful example because the unlinearity could in theory be defined.

I'm thinking it is kind of like publishing a "typical histogram" of a good
photograph. You might propose that it would have a full range of values from
black to white. Not ALWAYS true, but it is usually true that if everything
is skewed to one side, or goes above clipping, the photo will look too
light, dark, or have no detail in the highlights.

Exactly, some of the time histogram eq - as offered by at least a very
early paint shop pro version - will improve image information
availability and some of the time it will be a distaster than needs an
instant undo.

In photography, a good camera will accurately record a dynamic range of 8
bits/ channel, all values from black to white and all colors within a
certain range. But the values and the colors that actually appear in the
real world just happen to fall in the curve of the typical histogram unless
you are on a snow bank or in a coal mine. In audio, a good recording or
microphone will record all frequencies with a flat response, but in the real
world of music the actual pattern of the appearance of the various
frequencies will fall under the Orange Curve and hopefully will fill the
area under the curve.

Yes, except that it is not so simple. I may write about it some day in a
paper that does not attempt to cover too many topics. In hindsight it
was quite correct that it didn't make it to the printing press.

In audio, you say that the curve will indicate the perceived linearity of
the sound in a good recording, which seems to hold true. You might add that
spatially, we expect to hear a soundstage with instruments that are balanced
at the center, with a full spread across the whole, unless it is a solo
guitar or voice.

It tells you whether what you try to do is probable, not whether it is
true. That's the value as well as the trap.

SO - one of the most fascinating ideas I have read. So how did it do? Did
they publish it? This all happened about 15 years ago, so what has become of
it?

Nothing. But it is a previously well known audio restoration concept to
compare a type of recording made with modern equipment with the sonic
object to restore, say a piano quartet and match frequency content.

AND: it is in audio restoration/forensics that it really gets fun,
because once you learn the modifiers of the curve to apply for various
contexts you can start using the "back then" recorded audio's probable
spectrum to equalize the total chain and either make it probable or goof
dramatically.

As with statistics, remember what the "aussagegrenz", the limit of
expression validity, is. It would however be extremely helpful for those
that use analyzers to have a selection of bias curves - but that alone
doesn't do it, one also has to know which to select.

Gary Eickmeier

Kind regards

Peter Larsen

[Gary Eickmeier]

Well thanks for that Peter - glad you got your computer fixed!

I don't understand all of that, but I have thought about the restoration
aspect. I have heard some pretty good restorations of recordings from the
early days and wondered how they did that. To again compare it to
photography, if I want to restore an old photo, I might first give it a full
range of values from black to white using the Levels function of the
histogram. Then I might diddle with the contrast and, if it is color, check
the white balance and saturation, all of these moves to make it conform more
closely to a good photo from a modern camera. In audio, I can imagine taking
an old 78 and boosting the bass, adjusting the midrange and treble to
conform to the range of values that we might find in the orange curve of a
good modern recording of the same kind of group.

It is just another way of looking at EQ (?). I mean, check me on this -
suppose I want to use my new DEQ2496 and its amazing variety of functions to
EQ my speakers and room. Traditional method, play some pink noise and EQ to
flat or room curve. Orange alternative, take a variety of well-recorded
music tracks, play throught your system and record that, then put it into
your frequency analysis program and EQ so that it conforms to the orange
curve.

Just thinking out loud, but is that anywhere near worth studying? What uses
do you imagine for your orange curve idea?

Gary Eickmeier


"Peter Larsen" wrote in message
k...
On 28-05-2014 14:04, Gary Eickmeier wrote:

Peter -

I have gone back to your post with the proposed paper on the Orange Curve
(Orange Noise Characteristic),

https://groups.google.com/forum/#!searchin/alt.audio.pro.live-sound/orange$20noise/alt.audio.pro.live-sound/cMkHUOVP20M/6vqOhGu1Ov4J

in which you state two laws, the idea being that most naturally occuring
sounds, including recorded music, will display the Orange Curve on your
frequency analysis window in Cool Edit or Audition. The Orange Curve
peaks
at 320 Hz, falls off above and below that by 3 dB/octave. I have looked
at
enough analyses of my recordings to confirm that for the vast majority of
the times and sections of most any track. Most fascinating. In fact, it
helped me correct my "thrumming bass guitar" problem one fine time. I
simply
EQed it to bring the curve under the Orange Curve.

It is kinda simple that it should end up like that with sound from
objects. The parameters are object size (acoustic impedance match),
directionality and air absorption. But you need to understand that it is
what the brain expects the reality to be and not the reality per se. So
real world audio may not fit.

A wonderful example is applause, it is by definition an unlinear sound
source because it is horn loaded by the shape of the hands, it is also a
wonderful example because the unlinearity could in theory be defined.

I'm thinking it is kind of like publishing a "typical histogram" of a
good
photograph. You might propose that it would have a full range of values
from
black to white. Not ALWAYS true, but it is usually true that if
everything
is skewed to one side, or goes above clipping, the photo will look too
light, dark, or have no detail in the highlights.

Exactly, some of the time histogram eq - as offered by at least a very
early paint shop pro version - will improve image information availability
and some of the time it will be a distaster than needs an instant undo.

In photography, a good camera will accurately record a dynamic range of 8
bits/ channel, all values from black to white and all colors within a
certain range. But the values and the colors that actually appear in the
real world just happen to fall in the curve of the typical histogram
unless
you are on a snow bank or in a coal mine. In audio, a good recording or
microphone will record all frequencies with a flat response, but in the
real
world of music the actual pattern of the appearance of the various
frequencies will fall under the Orange Curve and hopefully will fill the
area under the curve.

Yes, except that it is not so simple. I may write about it some day in a
paper that does not attempt to cover too many topics. In hindsight it was
quite correct that it didn't make it to the printing press.

In audio, you say that the curve will indicate the perceived linearity of
the sound in a good recording, which seems to hold true. You might add
that
spatially, we expect to hear a soundstage with instruments that are
balanced
at the center, with a full spread across the whole, unless it is a solo
guitar or voice.

It tells you whether what you try to do is probable, not whether it is
true. That's the value as well as the trap.

SO - one of the most fascinating ideas I have read. So how did it do? Did
they publish it? This all happened about 15 years ago, so what has become
of
it?

Nothing. But it is a previously well known audio restoration concept to
compare a type of recording made with modern equipment with the sonic
object to restore, say a piano quartet and match frequency content.

AND: it is in audio restoration/forensics that it really gets fun, because
once you learn the modifiers of the curve to apply for various contexts
you can start using the "back then" recorded audio's probable spectrum to
equalize the total chain and either make it probable or goof dramatically.

As with statistics, remember what the "aussagegrenz", the limit of
expression validity, is. It would however be extremely helpful for those
that use analyzers to have a selection of bias curves - but that alone
doesn't do it, one also has to know which to select.

Gary Eickmeier

Kind regards

Peter Larsen

[Peter Larsen[_3_]]

On 01-06-2014 15:46, Gary Eickmeier wrote:

Well thanks for that Peter - glad you got your computer fixed!

I didn't, in as much as OE quotefix seems unable to ever work on it
again after uninstalling something that should have no bearing on how OE
behaves. But I can live with Thunderbird for news, if only they had
bothered getting import right ...

I don't understand all of that, but I have thought about the restoration
aspect. I have heard some pretty good restorations of recordings from the
early days and wondered how they did that. To again compare it to
photography, if I want to restore an old photo, I might first give it a full
range of values from black to white using the Levels function of the
histogram. Then I might diddle with the contrast and, if it is color, check
the white balance and saturation, all of these moves to make it conform more
closely to a good photo from a modern camera. In audio, I can imagine taking
an old 78 and boosting the bass, adjusting the midrange and treble to
conform to the range of values that we might find in the orange curve of a
good modern recording of the same kind of group.

You're on the track, there is not a lot of difference between the
concepts of audio processing and image processing. For optimum playback
experience from a 78 rpm record: use a proper grammophone with a quality
needle, a HMV travel grammophone will do, but having a larger horn is good.

It is just another way of looking at EQ (?). I mean, check me on this -
suppose I want to use my new DEQ2496 and its amazing variety of functions to
EQ my speakers and room. Traditional method, play some pink noise and EQ to
flat or room curve.

Room curve, start with - 6 dB at 20 Hz, 0 dB at 200 Hz and -6 to -10 dB
at 20 kHz, the recommendation is for danish (or british) size listening
rooms, for a larger room you may need to start with 40 Hz to 1 kHz flat
and then - 3 dB pr. octave, normally the starting point for a PA in a
500 seater.

Orange alternative, take a variety of well-recorded
music tracks, play throught your system and record that, then put it into
your frequency analysis program and EQ so that it conforms to the orange
curve.

No. Concept broken. Do not fix the audio in the system. Same error with
your loudspeaker concept, fix what is broken where it broken, do not
break it differently somewhere else to compensate.

Just thinking out loud, but is that anywhere near worth studying? What uses
do you imagine for your orange curve idea?

That is just how real world audio generating bodies tend to behave, note
the wording "tend to". Then there are some room size modifications that
remain as yet unpublished ... those that mix with an analyzer as "fact
sheet" would however do well with a suitably modified curve as reference
curve.

It won't tell you whether it is "right", but it will tell you whether it
is "probable", nothing more. That starts to get interesting in case of a
signal that you know is decayed, but you don't really know what happened
to it, you just need to make it listenable. As with cooking - a required
caveat - it is not enough to have the recipe, you also need to know how
to cook.

Gary Eickmeier

Kind regards

Peter Larsen