Noise Weighting Curves
 

My thanks to all those, both on and off list, who
provided useful info on the noise weighting curves.

The two ITU curves are similar but
ITU-R ARM is a later Dolby Labs
proposal which moves the whole curve
1kHz to the right.

Thanks also to my pal Richard in the UK,
I now have a chart in Excel showing all three
IEC curves, A,B and C, plus the two ITU curves.

http://www.kolumbus.fi/iain.churches...se/ABC+ITU.jpg

Comparison is interesting. One can also see why
the old IEC "A" weighted curve is still popular:-)


Iain

Noise Weighting Curves
 

On Thu, 4 Oct 2007 10:00:34 +0300, "Iain Churches"
wrote:

My thanks to all those, both on and off list, who
provided useful info on the noise weighting curves.

The two ITU curves are similar but
ITU-R ARM is a later Dolby Labs
proposal which moves the whole curve
1kHz to the right.

Thanks also to my pal Richard in the UK,
I now have a chart in Excel showing all three
IEC curves, A,B and C, plus the two ITU curves.

http://www.kolumbus.fi/iain.churches...se/ABC+ITU.jpg

Comparison is interesting. One can also see why
the old IEC "A" weighted curve is still popular:-)


Iain



Iain, I've just looked at the curves, and I don't see the ARM curve
being 1kHz to the right of the 468 - it looks like identical
frequencies but 6dB lower to me. Hard to see from the graphs, but they
appear to peak at the same frequency.

What are the numbers?

d

--
Pearce Consulting
http://www.pearce.uk.com

Noise Weighting Curves
 

"Don Pearce" wrote in message
...

Iain, I've just looked at the curves, and I don't see the ARM curve
being 1kHz to the right of the 468 - it looks like identical
frequencies but 6dB lower to me. Hard to see from the graphs, but they
appear to peak at the same frequency.

What are the numbers?

They both peak at 6276 Hz.
The difference between them is 5.6dB
At 6.276kHz the 468 is +12.2db and ARM is 6.6dB

It looks as if the Dolby recommendation for
the shift upwards of 1kHz was not implemented
I still have a lot of reading to do.

Comparison of noise levels using the different weighting
options is interesting.

Iain

Noise Weighting Curves
 

On Thu, 4 Oct 2007 10:33:24 +0300, "Iain Churches"
wrote:


"Don Pearce" wrote in message
...

Iain, I've just looked at the curves, and I don't see the ARM curve
being 1kHz to the right of the 468 - it looks like identical
frequencies but 6dB lower to me. Hard to see from the graphs, but they
appear to peak at the same frequency.

What are the numbers?

They both peak at 6276 Hz.
The difference between them is 5.6dB
At 6.276kHz the 468 is +12.2db and ARM is 6.6dB

It looks as if the Dolby recommendation for
the shift upwards of 1kHz was not implemented
I still have a lot of reading to do.

Comparison of noise levels using the different weighting
options is interesting.

Iain


Does anything you have say why the curves go on rising all the way to
beyond 6kHz? That is a long way past the peak of the human hearing
response, and seems a strange thing to do for a noise weighting
system.

Not only that, but it all goes a very long way above zero and I'm not
too sure I buy that either.

d

--
Pearce Consulting
http://www.pearce.uk.com

Noise Weighting Curves
 

"Don Pearce" wrote in message
...
On Thu, 4 Oct 2007 10:33:24 +0300, "Iain Churches"
wrote:

"Don Pearce" wrote in message
...

What are the numbers?

They both peak at 6276 Hz.
The difference between them is 5.6dB
At 6.276kHz the 468 is +12.2db and ARM is 6.6dB

Does anything you have say why the curves go on rising all the way to
beyond 6kHz? That is a long way past the peak of the human hearing
response, and seems a strange thing to do for a noise weighting
system.

Don. In a nut-shell, The ITU-R 468 noise weighting curve was
developed by the BBC specifically for noise measurements in
audio equipment. It is said to be better suited in this role than
"A" weighting which is more valid in the measurement of tones
as opposed to noise.


Not only that, but it all goes a very long way above zero and I'm not
too sure I buy that either.

Then the old "A" curve is the one for you:-)
It is described as "benign"

During the 1960s, it became evident that the DIN A-weighting
curve did not give results reflecting a real-world situation.
The BBC's report EL-7 in which the results from various
listening experiments using clicks and tone bursts were
published. Those being tested were then asked to compare
these with a tone of 1kHz. The scores were then compared
with noise levels measured using various weighting filters.

Based upon the work done principally it seems by the BBC,
CCIR recommendation 468-1 was published. There were
further revisions made from time to time, up to revision 4.0.
The curve remained basically unchanged, but tolerances were
specified more accurately. CCIR 468 was subsequently
adopted by the IEC (International Electrotechnical Commission)
and the ITU (International Telecommunications Union) and
became the accepted method for measuring noise in broadcast,
professional audio applications and high-fidelity.
When the CCIR was disbanded in March 1993, the
ITU took over the existing standard.

Iain

Noise Weighting Curves
 

On Thu, 4 Oct 2007 11:11:23 +0300, "Iain Churches"
wrote:


"Don Pearce" wrote in message
...
On Thu, 4 Oct 2007 10:33:24 +0300, "Iain Churches"
wrote:

"Don Pearce" wrote in message
...

What are the numbers?

They both peak at 6276 Hz.
The difference between them is 5.6dB
At 6.276kHz the 468 is +12.2db and ARM is 6.6dB

Does anything you have say why the curves go on rising all the way to
beyond 6kHz? That is a long way past the peak of the human hearing
response, and seems a strange thing to do for a noise weighting
system.

Don. In a nut-shell, The ITU-R 468 noise weighting curve was
developed by the BBC specifically for noise measurements in
audio equipment. It is said to be better suited in this role than
"A" weighting which is more valid in the measurement of tones
as opposed to noise.


Not only that, but it all goes a very long way above zero and I'm not
too sure I buy that either.

Then the old "A" curve is the one for you:-)
It is described as "benign"

During the 1960s, it became evident that the DIN A-weighting
curve did not give results reflecting a real-world situation.
The BBC's report EL-7 in which the results from various
listening experiments using clicks and tone bursts were
published. Those being tested were then asked to compare
these with a tone of 1kHz. The scores were then compared
with noise levels measured using various weighting filters.

Based upon the work done principally it seems by the BBC,
CCIR recommendation 468-1 was published. There were
further revisions made from time to time, up to revision 4.0.
The curve remained basically unchanged, but tolerances were
specified more accurately. CCIR 468 was subsequently
adopted by the IEC (International Electrotechnical Commission)
and the ITU (International Telecommunications Union) and
became the accepted method for measuring noise in broadcast,
professional audio applications and high-fidelity.
When the CCIR was disbanded in March 1993, the
ITU took over the existing standard.

Iain



OK, thanks for that. It certainly makes for some interesting changes
to the S/N ratios one can claim for audio gear. The numbers won't look
anything like as pretty under CCIR 468.

d

--
Pearce Consulting
http://www.pearce.uk.com

Noise Weighting Curves
 

"Don Pearce" wrote in message
...
On Thu, 4 Oct 2007 11:11:23 +0300, "Iain Churches"
wrote:

Don. In a nut-shell, The ITU-R 468 noise weighting curve was
developed by the BBC specifically for noise measurements in
audio equipment. It is said to be better suited in this role than
"A" weighting which is more valid in the measurement of tones
as opposed to noise.


Not only that, but it all goes a very long way above zero and I'm not
too sure I buy that either.

Then the old "A" curve is the one for you:-)
It is described as "benign"

During the 1960s, it became evident that the DIN A-weighting
curve did not give results reflecting a real-world situation.
The BBC's report EL-7 in which the results from various
listening experiments using clicks and tone bursts were
published. Those being tested were then asked to compare
these with a tone of 1kHz. The scores were then compared
with noise levels measured using various weighting filters.

Based upon the work done principally it seems by the BBC,
CCIR recommendation 468-1 was published. There were
further revisions made from time to time, up to revision 4.0.
The curve remained basically unchanged, but tolerances were
specified more accurately. CCIR 468 was subsequently
adopted by the IEC (International Electrotechnical Commission)
and the ITU (International Telecommunications Union) and
became the accepted method for measuring noise in broadcast,
professional audio applications and high-fidelity.
When the CCIR was disbanded in March 1993, the
ITU took over the existing standard.

Iain



OK, thanks for that. It certainly makes for some interesting changes
to the S/N ratios one can claim for audio gear. The numbers won't look
anything like as pretty under CCIR 468.


Which explains why many still stubbornly use the long
superceded "A" curve.

Iain

Noise Weighting Curves
 

"Iain Churches" wrote in message
ti.fi...

Which explains why many still stubbornly use the long
superceded "A" curve.


Don. I forgot to mention, I was looking at a British valve amp
yesterday built 1990s. The difference between the noise
floor measured "audio band" and "A" weighted was 20dB !!

A magical improvement at the press of a button.

Iain

Noise Weighting Curves
 

On Thu, 4 Oct 2007 11:32:30 +0300, "Iain Churches"
wrote:


"Don Pearce" wrote in message
...
On Thu, 4 Oct 2007 11:11:23 +0300, "Iain Churches"
wrote:

Don. In a nut-shell, The ITU-R 468 noise weighting curve was
developed by the BBC specifically for noise measurements in
audio equipment. It is said to be better suited in this role than
"A" weighting which is more valid in the measurement of tones
as opposed to noise.


Not only that, but it all goes a very long way above zero and I'm not
too sure I buy that either.

Then the old "A" curve is the one for you:-)
It is described as "benign"

During the 1960s, it became evident that the DIN A-weighting
curve did not give results reflecting a real-world situation.
The BBC's report EL-7 in which the results from various
listening experiments using clicks and tone bursts were
published. Those being tested were then asked to compare
these with a tone of 1kHz. The scores were then compared
with noise levels measured using various weighting filters.

Based upon the work done principally it seems by the BBC,
CCIR recommendation 468-1 was published. There were
further revisions made from time to time, up to revision 4.0.
The curve remained basically unchanged, but tolerances were
specified more accurately. CCIR 468 was subsequently
adopted by the IEC (International Electrotechnical Commission)
and the ITU (International Telecommunications Union) and
became the accepted method for measuring noise in broadcast,
professional audio applications and high-fidelity.
When the CCIR was disbanded in March 1993, the
ITU took over the existing standard.

Iain



OK, thanks for that. It certainly makes for some interesting changes
to the S/N ratios one can claim for audio gear. The numbers won't look
anything like as pretty under CCIR 468.


Which explains why many still stubbornly use the long
superceded "A" curve.

Iain


Just created the curve in my DAW, and it makes white noise look nearly
7dB worse. I haven't tried the A curve, but I imagine that compared to
that, it will be an even bigger change.

d

--
Pearce Consulting
http://www.pearce.uk.com

Noise Weighting Curves
 

On Thu, 4 Oct 2007 11:38:01 +0300, "Iain Churches"
wrote:


"Iain Churches" wrote in message
hti.fi...

Which explains why many still stubbornly use the long
superceded "A" curve.


Don. I forgot to mention, I was looking at a British valve amp
yesterday built 1990s. The difference between the noise
floor measured "audio band" and "A" weighted was 20dB !!

A magical improvement at the press of a button.

Iain


Just done something similar on my DAW. I've also created an A curve.
Made a chunk of white noise and compared the average noise levels.

Flat -19.44dB
A wt -21.7dB
468 -12.59dB

So that is a 19dB difference. I only eyeballed the filters, so I could
be a bit out. This is all a bit odd. I can see there being differences
between two subjective curves purporting to be of the same thing, but
19dB? Something smells nasty.

d

--
Pearce Consulting
http://www.pearce.uk.com