When reading Tom's frequency response measurements done in magazines,
the frequency response specification is not always given in (what I've
always thought was) the typical +/- 3 dB fashion. Here are some
examples from his subwoofer tests, as pulled from the Web:

48-115Hz +/-2.1 dB
25-117Hz +/-3.5 dB
28-117Hz +/-2.4 dB
25-82Hz +/-3.2 dB
27-90Hz +/-1.7 dB

Does anyone have an explanation for why the tolerances are different
for each speaker? It seems to me that if the speakers were measured
and the results all shown as x-y Hz +/- 3 dB, the numbers would be
more meaningful. As it is, it's not clear which of the above really
has the flattest graph in any particular frequency band.

Thanks for any insight.

Without the context it is hard to tell. One guess is that these represent
the freq range provided by the manufacture as the bottom and high based on
roll off and the low pass xover. Using that range the +/- db figure is
then an empirical observation. Another possibility is that these are the
results measured at a given spl level, where the low will often vary as a
function of that level. In any case he should soon read and reply here.


When reading Tom's frequency response measurements done in magazines,
the frequency response specification is not always given in (what I've
always thought was) the typical +/- 3 dB fashion. Here are some
examples from his subwoofer tests, as pulled from the Web:

48-115Hz +/-2.1 dB
25-117Hz +/-3.5 dB
28-117Hz +/-2.4 dB
25-82Hz +/-3.2 dB
27-90Hz +/-1.7 dB

Does anyone have an explanation for why the tolerances are different
for each speaker? It seems to me that if the speakers were measured
and the results all shown as x-y Hz +/- 3 dB, the numbers would be
more meaningful. As it is, it's not clear which of the above really
has the flattest graph in any particular frequency band.

Thanks for any insight.

tk wrote:

When reading Tom's frequency response measurements done in magazines,
the frequency response specification is not always given in (what I've
always thought was) the typical +/- 3 dB fashion. Here are some
examples from his subwoofer tests, as pulled from the Web:

48-115Hz +/-2.1 dB
25-117Hz +/-3.5 dB
28-117Hz +/-2.4 dB
25-82Hz +/-3.2 dB
27-90Hz +/-1.7 dB

Does anyone have an explanation for why the tolerances are different
for each speaker? It seems to me that if the speakers were measured
and the results all shown as x-y Hz +/- 3 dB, the numbers would be
more meaningful. As it is, it's not clear which of the above really
has the flattest graph in any particular frequency band.

Thanks for any insight.

Seems abundantly clear to me! The only thing we don't know is what
happens beyond the end frequencies.

Gary Eickmeier

Gary Eickmeier

On 22 Oct 2004 18:26:23 GMT, (tk) wrote:

When reading Tom's frequency response measurements done in magazines,
the frequency response specification is not always given in (what I've
always thought was) the typical +/- 3 dB fashion. Here are some
examples from his subwoofer tests, as pulled from the Web:

48-115Hz +/-2.1 dB
25-117Hz +/-3.5 dB
28-117Hz +/-2.4 dB
25-82Hz +/-3.2 dB
27-90Hz +/-1.7 dB

Does anyone have an explanation for why the tolerances are different
for each speaker? It seems to me that if the speakers were measured
and the results all shown as x-y Hz +/- 3 dB, the numbers would be
more meaningful. As it is, it's not clear which of the above really
has the flattest graph in any particular frequency band.

While it would be nice if they were all normalised to the usual +/-
1dB and/or +/- 3dB measures, it is in fact quite clear that the
flattest over the subwoofer range is the last mentioned.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

When reading Tom's frequency response measurements done in magazines,
the frequency response specification is not always given in (what I've
always thought was) the typical +/- 3 dB fashion. Here are some
examples from his subwoofer tests, as pulled from the Web:

48-115Hz +/-2.1 dB
25-117Hz +/-3.5 dB
28-117Hz +/-2.4 dB
25-82Hz +/-3.2 dB
27-90Hz +/-1.7 dB

Does anyone have an explanation for why the tolerances are different
for each speaker? It seems to me that if the speakers were measured
and the results all shown as x-y Hz +/- 3 dB, the numbers would be
more meaningful. As it is, it's not clear which of the above really
has the flattest graph in any particular frequency band.

Thanks for any insight.

+/- 3db is actually quite generous. Today, a good speaker can
easily reach +/- 2db or better.

(tk) wrote:



When reading Tom's frequency response measurements done in magazines,
the frequency response specification is not always given in (what I've
always thought was) the typical +/- 3 dB fashion. Here are some
examples from his subwoofer tests, as pulled from the Web:

48-115Hz +/-2.1 dB
25-117Hz +/-3.5 dB
28-117Hz +/-2.4 dB
25-82Hz +/-3.2 dB
27-90Hz +/-1.7 dB

Does anyone have an explanation for why the tolerances are different
for each speaker? It seems to me that if the speakers were measured
and the results all shown as x-y Hz +/- 3 dB, the numbers would be
more meaningful. As it is, it's not clear which of the above really
has the flattest graph in any particular frequency band.

Thanks for any insight.

These numbers are the tolerance of the speakers response relative to the half
power points (-3 dB.) For example, 27-90Hz +/-1.7 dB, tells us that the half
power points were 27 Hz on the low end and 90Hz on the high end. However it's
possible to have wider variations in between.

For example if this woofer had a 6-dB peak at 40 Hz and a 9-dB notch at 50 Hz
the tolerance would be +/- 7.5 dB. Another way of specifying this would be
(27-90 Hz (+6/-9 dB) but the managing editor sometimes saves a few spaces of
copy with our current copy style.

Cool, an answer from the man himself! Thanks!

(Nousaine) wrote in message ...
(tk) wrote:

27-90Hz +/-1.7 dB

These numbers are the tolerance of the speakers response relative to the half
power points (-3 dB.) For example, 27-90Hz +/-1.7 dB, tells us that the half
power points were 27 Hz on the low end and 90Hz on the high end. However it's
possible to have wider variations in between.

For example if this woofer had a 6-dB peak at 40 Hz and a 9-dB notch at 50 Hz
the tolerance would be +/- 7.5 dB. Another way of specifying this would be
(27-90 Hz (+6/-9 dB) but the managing editor sometimes saves a few spaces of
copy with our current copy style.

Thanks for the response! So the half power points, if I understand
correctly, are the low and high frequencies beyond which (in either
direction) the frequency reponse is always at least 3 dB lower than
the 0 dB reference point, when fed a constant input amplitude. Is
that right?

I'm still not clear as to how to 0 dB reference point is established,
though. In the case of the second speaker above, the +/- 1.7 dB
tolerance indicates a total range of 3.4 dB between the end points. I
assume that the 0 dB reference level is actually attained at some
point between 27 Hz and 90 Hz. Thus:

1. To accomodate a 3.4 dB total range, the response must peak to 0.4
dB above reference at some point, or

2. Maybe at some point between 27 Hz and 90 Hz the response drops to
3.4 dB below reference before climbing up again, and so the max level
is 0 dB, or

3. Maybe the peak level is between 0 and 0.4 dB above 0 dB reference,
which agains leaves the question of how the 0 dB point is chosen.

(I think the answer does have some interesting implications... Let's
say that a given speaker has a +/- 3 dB range between the end points.
If the answer is (1) above, then the frequency response never drops
below that of the end points while within the range (but it does peak
at +3 dB above "0 dB"). If the answer is (2) above, the response
actually is lower (at 6 dB below reference) than that of the end
points at some point(s) within the range, but never climbs above "0
dB". If the answer is (3), then we're in the dark about the nature of
the frequency response within the end points, relative to the levels
of the end points.)

I'm sorry if this has been a very ignorant or pointless question...
Am I missing something really obvious? Thanks for your patience!

"tk" wrote in message
...
Cool, an answer from the man himself! Thanks!

(Nousaine) wrote in message
...
(tk) wrote:

27-90Hz +/-1.7 dB

These numbers are the tolerance of the speakers response relative to the
half
power points (-3 dB.) For example, 27-90Hz +/-1.7 dB, tells us that the
half
power points were 27 Hz on the low end and 90Hz on the high end. However
it's
possible to have wider variations in between.

For example if this woofer had a 6-dB peak at 40 Hz and a 9-dB notch at
50 Hz
the tolerance would be +/- 7.5 dB. Another way of specifying this would
be
(27-90 Hz (+6/-9 dB) but the managing editor sometimes saves a few
spaces of
copy with our current copy style.

Thanks for the response! So the half power points, if I understand
correctly, are the low and high frequencies beyond which (in either
direction) the frequency reponse is always at least 3 dB lower than
the 0 dB reference point, when fed a constant input amplitude. Is
that right?

I'm still not clear as to how to 0 dB reference point is established,
though. In the case of the second speaker above, the +/- 1.7 dB
tolerance indicates a total range of 3.4 dB between the end points. I
assume that the 0 dB reference level is actually attained at some
point between 27 Hz and 90 Hz. Thus:

1. To accomodate a 3.4 dB total range, the response must peak to 0.4
dB above reference at some point, or

2. Maybe at some point between 27 Hz and 90 Hz the response drops to
3.4 dB below reference before climbing up again, and so the max level
is 0 dB, or

3. Maybe the peak level is between 0 and 0.4 dB above 0 dB reference,
which agains leaves the question of how the 0 dB point is chosen.

(I think the answer does have some interesting implications... Let's
say that a given speaker has a +/- 3 dB range between the end points.
If the answer is (1) above, then the frequency response never drops
below that of the end points while within the range (but it does peak
at +3 dB above "0 dB"). If the answer is (2) above, the response
actually is lower (at 6 dB below reference) than that of the end
points at some point(s) within the range, but never climbs above "0
dB". If the answer is (3), then we're in the dark about the nature of
the frequency response within the end points, relative to the levels
of the end points.)

I'm sorry if this has been a very ignorant or pointless question...
Am I missing something really obvious? Thanks for your patience!

I had exactly the same reaction to Tom's explaination. It just doesn't seem
like a very orthodox use of tolerance specs, and leads to the speculations
you have made above.

(tk) wrote:



Cool, an answer from the man himself! Thanks!

(Nousaine) wrote in message
...
(tk) wrote:

27-90Hz +/-1.7 dB

These numbers are the tolerance of the speakers response relative to the
half
power points (-3 dB.) For example, 27-90Hz +/-1.7 dB, tells us that the
half
power points were 27 Hz on the low end and 90Hz on the high end. However
it's
possible to have wider variations in between.

For example if this woofer had a 6-dB peak at 40 Hz and a 9-dB notch at 50
Hz
the tolerance would be +/- 7.5 dB. Another way of specifying this would be
(27-90 Hz (+6/-9 dB) but the managing editor sometimes saves a few spaces
of
copy with our current copy style.

Thanks for the response! So the half power points, if I understand
correctly, are the low and high frequencies beyond which (in either
direction) the frequency reponse is always at least 3 dB lower than
the 0 dB reference point, when fed a constant input amplitude. Is
that right?

I'm still not clear as to how to 0 dB reference point is established,
though. In the case of the second speaker above, the +/- 1.7 dB
tolerance indicates a total range of 3.4 dB between the end points. I
assume that the 0 dB reference level is actually attained at some
point between 27 Hz and 90 Hz. Thus:

1. To accomodate a 3.4 dB total range, the response must peak to 0.4
dB above reference at some point, or

2. Maybe at some point between 27 Hz and 90 Hz the response drops to
3.4 dB below reference before climbing up again, and so the max level
is 0 dB, or

3. Maybe the peak level is between 0 and 0.4 dB above 0 dB reference,
which agains leaves the question of how the 0 dB point is chosen.

(I think the answer does have some interesting implications... Let's
say that a given speaker has a +/- 3 dB range between the end points.
If the answer is (1) above, then the frequency response never drops
below that of the end points while within the range (but it does peak
at +3 dB above "0 dB"). If the answer is (2) above, the response
actually is lower (at 6 dB below reference) than that of the end
points at some point(s) within the range, but never climbs above "0
dB". If the answer is (3), then we're in the dark about the nature of
the frequency response within the end points, relative to the levels
of the end points.)

I'm sorry if this has been a very ignorant or pointless question...
Am I missing something really obvious? Thanks for your patience!

You're working the analysis too hard :-) The transfer function measurement
technique simply assigns the zero value to an average of output over a useful
bandwidth. You could do trhe same thing by using a sound pressure value.

But if the useful bandwidth of a speaker is X-Y Hz there is not reason that
tolerances would necessarily reside within any given interval.

For example it is common for a horizontally arrayed center channel to have
severe lobing off-axis. For example a speaker might be half power (-3 dB) at
100 Hz at the low end of its useful baandwidth and 20 kHz at the upper end. So
how would you deal with a 4 dB hump at 500 Hz and a 20 dB notch near the
crossover frequency at 30-degrees off-axis? (These are not values that I've
never seen, BTW)

Does the speaker have useful output at 100 Hz? At 20 kHz? Are these NOT the -3
dB pointsover it's entire bandwidth? Is the deviation NOT 24 dB? Isn't the
average +/- 12 dB?

This is not, as claimed by another poster, an unusual manner of describing
speaker performance but I agree that it also requires a graphical
representation of the response to gain a full understanding.

If life were to fall my way I'd simply publish a full set of graphs with all
directivity data included. But I'm simply using the space the managing editor
allows for presentation.

There's more to speaker performance than frequency response at a given fixed
level of output. That's why I use a separate technique to illustrate the linear
dynamic capability of subwoofersand other speakers (Bass Limits.)

For subwoofers this is far more interesting than a depiction of the speakers
transfer function at a fixed level. For subwoofers, in general, most
manufacturers have figured out ways to get pretty uniform output under
near-field and anechoic conditions.

Thanks for the followup!

(Nousaine) wrote in message ...
(tk) wrote:
You're working the analysis too hard :-) The transfer function measurement
technique simply assigns the zero value to an average of output over a useful
bandwidth. You could do trhe same thing by using a sound pressure value.

Okay, I think I get the point; essentially in these measurements you
are saying "this is the main useful bandwidth of the speaker, and the
speaker has this degree of fluctuation within the bandwidth." At
least, this is the impression I get.

If life were to fall my way I'd simply publish a full set of graphs with all
directivity data included. But I'm simply using the space the managing editor
allows for presentation.

I understand -- it would be cool for the web site to host enhanced
versions of your articles.

There's more to speaker performance than frequency response at a given fixed
level of output. That's why I use a separate technique to illustrate the linear
dynamic capability of subwoofersand other speakers (Bass Limits.)

That's another question I had -- how you decide what frequency to use
for the "useful bass extension" or "bass limit" number, e.g. 75 dB at
20 Hz versus 83 dB at 25 Hz.

The other thing that I sometimes found confusing was the measurement
of max SPL at 10% THD. (E.g. a speaker maxes out at 100 dB at 25 Hz
with 10% THD but only at 85 dB at 20 Hz with 10% THD). I had thought
this was indicative of frequency response, but it's apparent that that
is quite a different issue.

Thanks for the feedback. It's very useful that someone has done so
many measurements of different speakers under the same conditions,
making it possible to compare performance in a meaningful way.

(tk) wrote:

Thanks for the followup!

(Nousaine) wrote in message
...
(tk) wrote:
You're working the analysis too hard :-) The transfer function measurement
technique simply assigns the zero value to an average of output over a
useful
bandwidth. You could do trhe same thing by using a sound pressure value.

Okay, I think I get the point; essentially in these measurements you
are saying "this is the main useful bandwidth of the speaker, and the
speaker has this degree of fluctuation within the bandwidth." At
least, this is the impression I get.

If life were to fall my way I'd simply publish a full set of graphs with
all
directivity data included. But I'm simply using the space the managing
editor
allows for presentation.

I understand -- it would be cool for the web site to host enhanced
versions of your articles.

Well perhaps I might do just that someday. But again 'comparing speakers' on
paper is pretty easy. Learning to implement and live with your choice may be
more problematic. I routinely advise both home and car enthusiasts to optimize
the equipment they already own before deciding that 'throwing money at the
problem' is a useful alternative.



There's more to speaker performance than frequency response at a given
fixed
level of output. That's why I use a separate technique to illustrate the
linear
dynamic capability of subwoofersand other speakers (Bass Limits.)

That's another question I had -- how you decide what frequency to use
for the "useful bass extension" or "bass limit" number, e.g. 75 dB at
20 Hz versus 83 dB at 25 Hz.


The frequency 'falls out' of the experiment. I just keep moving upward in drive
level and downward in frequency to find the lowest frequency where the
subwoofer can generate a sound pressure (= 10% distortion) at an SPL that will
be audible to listeners.

Here's the procedure. Start at 62 Hz (In my estimation the beginning of the
'subwoofer' range .... IOW 100 Hz isn't a subwoofer frequency; arguably 62 Hz
isn't either) and give the subwoofer input (with the volume control set to
maximum and the crossover controlset to maximum bandwidth) a 6.5-cycle ramped
tone burst (aka Linkwitz and Keele) at that preferred 1/3 octave frequency and
keep increasing drive until distortion reaches 10% ( at 10% a woofer sill still
sound 'clean' but will be leaving its linear operating range and distortion
will be increasing exponentially) and then record the Sound Pressure Level.

Now repeat this process for every 1/3 octave frequency at lower frequencies.

Record the Bass Limit as the lowest frequency that a speaker will produce a
clean SPL above the threshold of audibility.

For subwoofers average the SPL figures over the 26-62 hz range (the heart of
subwoofer action.)

I also publish a Bass Uniformity figure (the 25-62 Hz number / maximum SPL at
any frequency) to draw attention to those products that handicap the 25-62 Hz
number with products that have a huge output at the upper-end of their
bandwidth but with rapidly falling lower frequency capability.



The other thing that I sometimes found confusing was the measurement
of max SPL at 10% THD. (E.g. a speaker maxes out at 100 dB at 25 Hz
with 10% THD but only at 85 dB at 20 Hz with 10% THD). I had thought
this was indicative of frequency response, but it's apparent that that
is quite a different issue.


It can be. That's because frequency response measurements are made with a fixed
(and fairly low) drive level. I've evaluated products that claimed an 8-Hz
bandwidth and found that the 8 Hz could only be obtained at very low
frequencies with close-mic measurements and small drive levels.

At typical levels in a real room the actual bandlimit was closer to 40 Hz.

Thanks for the feedback. It's very useful that someone has done so
many measurements of different speakers under the same conditions,
making it possible to compare performance in a meaningful way.

Thank you. I'm so happy that folks are listening.