Sign in with
Sign up | Sign in
Your question

Arny or other techies - Frequency response graphs?

Last response: in Home Audio
Share
Anonymous
June 7, 2005 3:25:44 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Long story short, I'm trying to test the frequency response of my newly
built LS3/5A's so I can verify that the crossovers are setup properly
(ie. cap & inductor values)....the crossover frequency is 3k and the HF
output is adjustable....the LS3/5A cabinets are built completely to spec
and the speakers are the 15 ohm version using B110-A's and T27-A's

I used RMAA (Right Mark Audio Analyzer) to run some tests through the
LS3/5A's into a single Behringer ECM 8000 test mic. The speakers and mic
are set up in an approximate equilateral triangle with the speakers and
mic about a yard (3 feet) from each other and the mic positioned at a
height about halfway between the B110 woofer and T27 tweeter.

Anyhow, I have no idea what to make of the results of this test. The
freq. response graphs seem to be all over the place even using a variety
of different tests.

The first test was RMAA's generic "Frequency Response" test. Here is a
screen capture of the results:

http://www.geocities.com/jdurangoproductions/freqrespon...

The second test is an IMD+N (I assume that means IMD plus noise)
frequency response test. Here is a capture of the results:

http://www.geocities.com/jdurangoproductions/imdnfreqre...

For the last test I recorded about 30 seconds of pink noise at around
-2dB in Cubase at 16-bits and used RMAA's spectrum analyzer to average
out the freq. response. Here is the screen capture:

http://www.geocities.com/jdurangoproductions/spectruman...

As you can see, in all these graphs, there are large spikes and dips all
over the place and the overall difference between highest and lowest
readings between 20 and 20k is about 30dB (+10 and -20).

I realize that since I am testing in a real room and not an anechoic
chamber I'll get slightly skewed results, but these results seem very
heavily skewed. What do you think?

Anyway, hopefully one of the techies in here can help me make sense of
these readings. Should I try different software or a different testing
method. Basically all I need to know is if the speakers are relatively
balanced or if I should raise or lower the HF output by a couple dB or
more. Thanks a ton!

Jonny Durango
Anonymous
June 7, 2005 8:40:06 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Jonny Durango wrote:
>
> I used RMAA (Right Mark Audio Analyzer) to run some tests through the
> LS3/5A's into a single Behringer ECM 8000 test mic. The speakers and mic
> are set up in an approximate equilateral triangle with the speakers and
> mic about a yard (3 feet) from each other and the mic positioned at a
> height about halfway between the B110 woofer and T27 tweeter.
>
> Anyhow, I have no idea what to make of the results of this test. The
> freq. response graphs seem to be all over the place even using a variety
> of different tests.
> The first test was RMAA's generic "Frequency Response" test. Here is a
> screen capture of the results:
>
> http://www.geocities.com/jdurangoproductions/freqrespon...
>
> For the last test I recorded about 30 seconds of pink noise at around
> -2dB in Cubase at 16-bits and used RMAA's spectrum analyzer to average
> out the freq. response. Here is the screen capture:
>
> http://www.geocities.com/jdurangoproductions/spectruman...
>
> As you can see, in all these graphs, there are large spikes and dips all
> over the place and the overall difference between highest and lowest
> readings between 20 and 20k is about 30dB (+10 and -20).

That's precisely what I would expect given how you've decided to
measure things. The spikces and dips are the direct result of
comb filtering due to the fact that you have two real sources
which are interfering constructively and destructively with one
another.

> I realize that since I am testing in a real room and not an anechoic
> chamber I'll get slightly skewed results, but these results seem very
> heavily skewed. What do you think?

I think your expectations are WAY off. Measuring TWO sources in a
live room will not lead to "slightly skewed" results, it will give
you exactly what you got.

Look for example at your first graph, above 1 kHz. You have dips
at about 2100, 2800, 3500, 4200, and just about 5000 Hz. What to
they all have in common? They're all separated by about 700 Hz.
That corresponds to a delay of about .8 mS delay that causes the
cancellation. That's a distance of around 11 inches. From this
we can guess that there are two paths between the source(s)
and the mic, differing by around 11 inches. This could be the
difference in the distance to each speaker, it could be a floor
or wall reflection, etc.

Basically your measuring as much your improper setup as you are
your speakers.

> Anyway, hopefully one of the techies in here can help me make sense of
> these readings. Should I try different software or a different testing
> method.

Measuring loudspeakers correctly is a TOUGH business: most people
who alledgedly know what they're doing (or at least claim so) don't
get it even remotely right. It requires very careful setup under
well controlled conditions and consistent, well defined procedures.

>Basically all I need to know is if the speakers are relatively
> balanced or if I should raise or lower the HF output by a couple dB or
> more.

You do have enough to make that judgement: smooth over all the peaks
and dips that are on the order of 1/3 octave wide or narrower and you
then begin to see that at least from the viewpoint of overall balance,
your tweeter is running about 3-5 dB too low (assuming your target is
flat axial response). I can also see that the top end of the woofer/
midrange is probably running a wee bit hot as well.
Anonymous
June 7, 2005 10:56:35 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Jonny Durango wrote:
> Long story short, I'm trying to test the frequency
response of my
> newly built LS3/5A's so I can verify that the crossovers
are setup
> properly (ie. cap & inductor values)....the crossover
frequency is 3k
> and the HF output is adjustable....the LS3/5A cabinets are
built
> completely to spec and the speakers are the 15 ohm version
using
> B110-A's and T27-A's
>
> I used RMAA (Right Mark Audio Analyzer) to run some tests
through the
> LS3/5A's into a single Behringer ECM 8000 test mic. The
speakers and
> mic are set up in an approximate equilateral triangle with
the
> speakers and mic about a yard (3 feet) from each other and
the mic
> positioned at a height about halfway between the B110
woofer and T27
> tweeter.
>
> Anyhow, I have no idea what to make of the results of this
test. The
> freq. response graphs seem to be all over the place even
using a
> variety of different tests.
>
> The first test was RMAA's generic "Frequency Response"
test. Here is a
> screen capture of the results:
>
>
http://www.geocities.com/jdurangoproductions/freqrespon...

This actually looks pretty good for a speaker test. Just
guessing but the crossover point in the speaker looks to be
in the 2-3 KHz range, with the tweeter averaging maybe 2 dB
lower than the woofer.

> The second test is an IMD+N (I assume that means IMD plus
noise)
> frequency response test. Here is a capture of the results:
>
>
http://www.geocities.com/jdurangoproductions/imdnfreqre...

IMD+N is not a frequency response test. It is a test of
nonlinear distortion.

> For the last test I recorded about 30 seconds of pink
noise at around
> -2dB in Cubase at 16-bits and used RMAA's spectrum
analyzer to average
> out the freq. response. Here is the screen capture:

>
http://www.geocities.com/jdurangoproductions/spectruman...

The results may show the effects of the normal -3 dB/octave
slope of pink noise.

> As you can see, in all these graphs, there are large
spikes and dips
> all over the place

Welcome to acoustical testing. BTW, normal acoustical
testing as someone else pointed out, involves one speaker
and one microphone.

> and the overall difference between highest and
> lowest readings between 20 and 20k is about 30dB (+10
and -20).

At this point, ignore all of your tests but the first one.

> I realize that since I am testing in a real room and not
an anechoic
> chamber I'll get slightly skewed results, but these
results seem very
> heavily skewed. What do you think?

The only one that seems to be a valid test (the first one)
is really pretty good for an acoustical test.

> Anyway, hopefully one of the techies in here can help me
make sense of
> these readings.

The first test suggests that maybe you want to raise the
level of the signal going to the tweeter by a dB or two.

>Should I try different software or a different testing
method.

See comments, above. Rely on only the first test, and test
one speaker at a time.

>Basically all I need to know is if the speakers are
relatively
> balanced or if I should raise or lower the HF output by a
couple dB or
> more. Thanks a ton!

See comments about. You may want to turn up the level of the
tweeter a tad. Frankly, you already seem to be doing very
well. If they sound good, leave them alone!
Related resources
Can't find your answer ? Ask !
Anonymous
June 7, 2005 10:58:06 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Jonny Durango" <jonnydurango1BUSH_FROM_OFFICE@comcast.net> wrote in message
news:M7mdnV9x8poKoDjfRVn-og@comcast.com...
> Long story short, I'm trying to test the frequency response of my newly
> built LS3/5A's so I can verify that the crossovers are setup properly
> (ie. cap & inductor values)....the crossover frequency is 3k and the HF
> output is adjustable....the LS3/5A cabinets are built completely to spec
> and the speakers are the 15 ohm version using B110-A's and T27-A's
>
> I used RMAA (Right Mark Audio Analyzer) to run some tests through the
> LS3/5A's into a single Behringer ECM 8000 test mic. The speakers and mic
> are set up in an approximate equilateral triangle with the speakers and
> mic about a yard (3 feet) from each other and the mic positioned at a
> height about halfway between the B110 woofer and T27 tweeter.

First problem: you're testing two speakers at the same time, which means
they'll interfere with each other something awful. Second, you're testing
them in a room, probably a pretty live one. The kind of results you show are
about par for the course with those testing methods.

Try testing one speaker at a time, outdoors -- if possible, hoist the
speaker and microphone way up in the air (find someone with a crane :-)}}} )
to lower the frequency of the inevitable ground bounce. Okay, you don't know
anyone with a crane; set the speaker on the top rung of a ladder. Forget
about the swept sine wave which is the generic frequency response; stick
with pink noise.

Peace,
Paul
June 7, 2005 11:25:12 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Jonny,

When you say you are testing two speakers together, do you mean the
LEFT and RIGHT speaker or do you mean the WOOFER and TWEETER. It would
not be correct to test the LEFT and RIGHT together.

If you just want to test the crossover, do an electrical only test.

If you want to test the acoustical response of the speakers, and
eliminate the room reflections and you don't have an anechoic chamber
available, I understand there is a method that uses impulse response
testing where impulses are sent to the speaker and the early impulses
are captured and the room reflections are not. The frequency response
of the speaker alone can then be determined using math (FFT?) on the
captured early impulses.

Sorry I don't have any specifics on this method but someone else here
probably does.

Mark
Anonymous
June 7, 2005 12:26:47 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Jonny Durango <jonnydurango1BUSH_FROM_OFFICE@comcast.net> wrote:
>
>I realize that since I am testing in a real room and not an anechoic
>chamber I'll get slightly skewed results, but these results seem very
>heavily skewed. What do you think?

No, that looks about typical. I've seen much worse.

This is why impulse systems like MLSSA are required to give you anything
approaching accurate speaker response without a chamber.

I did say this before, the last time you brought this up.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
Anonymous
June 7, 2005 1:02:03 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Scott Dorsey wrote:
> Jonny Durango <jonnydurango1BUSH_FROM_OFFICE@comcast.net>
wrote:
>>
>> I realize that since I am testing in a real room and not
an anechoic
>> chamber I'll get slightly skewed results, but these
results seem very
>> heavily skewed. What do you think?
>
> No, that looks about typical. I've seen much worse.
>
> This is why impulse systems like MLSSA are required to
give you
> anything approaching accurate speaker response without a
chamber.

Scott, how do you know that RMAA doesn't use similar kinds
of test procedures as MLSSA?
Anonymous
June 7, 2005 1:09:41 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

In article <WLKdnaF6uY9WBzjfRVn-2g@comcast.com>,
Arny Krueger <arnyk@hotpop.com> wrote:
>Scott Dorsey wrote:
>> Jonny Durango <jonnydurango1BUSH_FROM_OFFICE@comcast.net>
>wrote:
>>>
>>> I realize that since I am testing in a real room and not
>an anechoic
>>> chamber I'll get slightly skewed results, but these
>results seem very
>>> heavily skewed. What do you think?
>>
>> No, that looks about typical. I've seen much worse.
>>
>> This is why impulse systems like MLSSA are required to
>give you
>> anything approaching accurate speaker response without a
>chamber.
>
>Scott, how do you know that RMAA doesn't use similar kinds
>of test procedures as MLSSA?

Because Jonny specifically mentioned pink noise for excitation.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
Anonymous
June 7, 2005 1:51:56 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Scott Dorsey wrote:
> In article <WLKdnaF6uY9WBzjfRVn-2g@comcast.com>,
> Arny Krueger <arnyk@hotpop.com> wrote:
>> Scott Dorsey wrote:
>>> Jonny Durango
<jonnydurango1BUSH_FROM_OFFICE@comcast.net> wrote:
>>>>
>>>> I realize that since I am testing in a real room and
not an
>>>> anechoic chamber I'll get slightly skewed results, but
these
>>>> results seem very heavily skewed. What do you think?
>>>
>>> No, that looks about typical. I've seen much worse.
>>>
>>> This is why impulse systems like MLSSA are required to
>> give you
>>> anything approaching accurate speaker response without a
chamber.
>>
>> Scott, how do you know that RMAA doesn't use similar
kinds
>> of test procedures as MLSSA?
>
> Because Jonny specifically mentioned pink noise for
excitation.

The RMAA FR test noise seems to be pink-like to some... ;-)

In fact RMAA use three distinct test signals:

(1) A swept tone that has variable amplitude (decreasing
amplitude at higher frequencies - no doubt tweeter-friendly)
(2) A series of about 20 ca. 1 octave chirps at ascending
frequencies
(3) A pseudo-random white noise signal.

There really does not seem to be a lot of doc about the
details of the RMAA speaker test, but judging by the test
signal and its actual performance, there is a bit more there
than a 1/3 octave RTA.
Anonymous
June 7, 2005 9:34:11 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Jonny Durango"
>
> Long story short, I'm trying to test the frequency response of my newly
> built LS3/5A's so I can verify that the crossovers are setup properly (ie.
> cap & inductor values)....the crossover frequency is 3k and the HF output
> is adjustable....the LS3/5A cabinets are built completely to spec and the
> speakers are the 15 ohm version using B110-A's and T27-A's


** That requires electrical testing - not acoustic. A calibrated,
variable audio oscillator and scope or good AC meter is all you need.

As far as acoustic response tests go - I find the Denon Audio Technical CD
( 38C39-7147) with 29 separate 1/3 octave band pink noise signals is very
useful when used with a good SPL meter set to C weighting.

A pair of late 70s Quad ESL57s I have here at the moment give a result (in
the loungeroom at 1 metre ) that is flat ( +/- 2dB ) from 125 Hz to 16 kHz
tested this way.

A sub woofer carries the bass on down to 32 Hz and because it is placed near
a corner is just a bit lumpier ;-)




............. Phil
Anonymous
June 7, 2005 9:34:12 PM

Archived from groups: rec.audio.pro (More info?)

On Tue, 7 Jun 2005 17:34:11 +1000, in rec.audio.pro "Phil Allison"
<philallison@tpg.com.au> wrote:

>
>"Jonny Durango"
>>snip
>A pair of late 70s Quad ESL57s I have here at the moment give a result (in
>the loungeroom at 1 metre ) that is flat ( +/- 2dB ) from 125 Hz to 16 kHz
>tested this way.
>
>A sub woofer carries the bass on down to 32 Hz and because it is placed near
>a corner is just a bit lumpier ;-)
>

what subwoofer/ crossover are you using?




martin
Anonymous
June 7, 2005 10:35:43 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Phil Allison" <philallison@tpg.com.au> wrote in message
news:3gl0vdFcu9ifU1@individual.net...
> As far as acoustic response tests go - I find the Denon Audio Technical
CD
> ( 38C39-7147) with 29 separate 1/3 octave band pink noise signals is very
> useful when used with a good SPL meter set to C weighting.
>
> A pair of late 70s Quad ESL57s I have here at the moment give a result
(in
> the loungeroom at 1 metre ) that is flat ( +/- 2dB ) from 125 Hz to 16
kHz
> tested this way.


So nowhere near flat then if you were using C weighting!

MrT.
Anonymous
June 8, 2005 1:02:10 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Mr.T"
"Phil Allison"

>> As far as acoustic response tests go - I find the Denon Audio Technical
> CD ( 38C39-7147) with 29 separate 1/3 octave band pink noise signals is
> very
>> useful when used with a good SPL meter set to C weighting.
>>
>> A pair of late 70s Quad ESL57s I have here at the moment give a result
> (in the loungeroom at 1 metre ) that is flat ( +/- 2dB ) from 125 Hz to
> 16
> kHz tested this way.
>
>
> So nowhere near flat then if you were using C weighting!
>


** Small omission - the particular ( Rode ) SPL meter used has been
modified ( one styro cap removed IIRC ) so the "C" setting is not subject
to a high frequency roll off above 8 kHz - so is flat ( -1dB ) to 18
kHz.




.............. Phil
Anonymous
June 8, 2005 1:09:54 AM

Archived from groups: rec.audio.pro (More info?)

"martin griffith"
"Phil Allison"

>>A pair of late 70s Quad ESL57s I have here at the moment give a result
>>(in
>>the loungeroom at 1 metre ) that is flat ( +/- 2dB ) from 125 Hz to 16
>>kHz
>>tested this way.
>>
>>A sub woofer carries the bass on down to 32 Hz and because it is placed
>>near
>>a corner is just a bit lumpier ;-)
>>
>
> what subwoofer/ crossover are you using?


** A 250mm sub woofer in a 65 litre ported box, tuned to 32 Hz, cone and
port facing carpeted floor to suppress mid output - box supported on 50 mm
high castors with an 18 dB /oct Butterworth LPF filter feeding a 50 watt
amp with -3dB point at 80 Hz.

Also similar 18db/oct 80 Hz HPFs in each main ( Quad) amp channel.



............ Phil
Anonymous
June 8, 2005 2:50:08 AM

Archived from groups: rec.audio.pro (More info?)

On 7 Jun 2005 07:25:12 -0700, "Mark" <makolber@yahoo.com> wrote:

>If you want to test the acoustical response of the speakers, and
>eliminate the room reflections and you don't have an anechoic chamber
>available, I understand there is a method that uses impulse response
>testing where impulses are sent to the speaker and the early impulses
>are captured and the room reflections are not. The frequency response
>of the speaker alone can then be determined using math (FFT?) on the
>captured early impulses.
>
>Sorry I don't have any specifics on this method but someone else here
>probably does.

I can recommend WinAIRR, available from the Old Colony folks;
works fine; about $50.

But 1/3 octave filtered pink noise is maybe more useful for this
application. I use dedicated hardware, but Phil, I believe,
mentioned the Denon CD, from which I've sadly gotten separated.
Might still be available. *Much* nicer to use than full bandwidth
noise.

Chris Hornbeck
"What, never?"
"No, never."
"What, never?"
"Well, hardly ever."
"HMS Pinafore"
Anonymous
June 8, 2005 4:33:17 PM

Archived from groups: rec.audio.pro (More info?)

"Chris Hornbeck"

> But 1/3 octave filtered pink noise is maybe more useful for this
> application. I use dedicated hardware, but Phil, I believe,
> mentioned the Denon CD, from which I've sadly gotten separated.
> Might still be available. *Much* nicer to use than full bandwidth
> noise.
>


** The Denon CD is still available:

http://www.amazon.com/exec/obidos/tg/detail/-/B0000034M...

Plus - any record dealer should be able to get one for you.

Just getting all those super low THD sine waves from your CD or DVD player
is worth the cost.



............... Phil
Anonymous
June 9, 2005 6:00:31 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Phil Allison" <philallison@tpg.com.au> wrote in message
news:3gld5lFcmcldU1@individual.net...
> ** Small omission - the particular ( Rode ) SPL meter used has been
> modified ( one styro cap removed IIRC ) so the "C" setting is not subject
> to a high frequency roll off above 8 kHz - so is flat ( -1dB ) to 18
> kHz.

The C weighting is now flat ??????????
Why not use the flat response setting in the first place?

MrT.
Anonymous
June 9, 2005 6:15:41 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Arny Krueger" <arnyk@hotpop.com> wrote in message
news:VYudnbFbzIThOzjfRVn-3w@comcast.com...
> >> Scott, how do you know that RMAA doesn't use similar
> kinds
> >> of test procedures as MLSSA?
> >
> > Because Jonny specifically mentioned pink noise for
> excitation.
>
> The RMAA FR test noise seems to be pink-like to some... ;-)
>
> In fact RMAA use three distinct test signals:
>
> (1) A swept tone that has variable amplitude (decreasing
> amplitude at higher frequencies - no doubt tweeter-friendly)
> (2) A series of about 20 ca. 1 octave chirps at ascending
> frequencies
> (3) A pseudo-random white noise signal.
>
> There really does not seem to be a lot of doc about the
> details of the RMAA speaker test, but judging by the test
> signal and its actual performance, there is a bit more there
> than a 1/3 octave RTA.

But from your own description is probably not "MLS" type impulse.
Why would it need to be anyway, it is designed primarily to measure sound
cards in the electrical domain, not speakers in the acoustic domain.

There is an open source program for speaker testing available called
Loudspeaker Workshop. I haven't used it, but it's probably a better tool for
speaker testing than RMAA I imagine.

MrT.
Anonymous
June 9, 2005 6:15:42 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Mr.T wrote:
> "Arny Krueger" <arnyk@hotpop.com> wrote in message

>>There really does not seem to be a lot of doc about the
>>details of the RMAA speaker test, but judging by the test
>>signal and its actual performance, there is a bit more there
>>than a 1/3 octave RTA.
>
>
> But from your own description is probably not "MLS" type impulse.
> Why would it need to be anyway, it is designed primarily to measure sound
> cards in the electrical domain, not speakers in the acoustic domain.

It's the same problem.

Sweep or MLS either one gives you the impulse response when
cross correlated with the matching signal. The sweep method
has the advantage that harmonic distortion is separated out
of the result. That's a big advantage considering how much
speakers typically distort at low frequencies.


Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein
Anonymous
June 9, 2005 6:15:42 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Mr.T <MrT@home> wrote:
>
>
>But from your own description is probably not "MLS" type impulse.
>Why would it need to be anyway, it is designed primarily to measure sound
>cards in the electrical domain, not speakers in the acoustic domain.

No, the whole point of using the impulse response is that you have a
short excitation, and you extrapolate the frequency response from that
short impulse response measurement.

You can make an impulse response measurement in a room, because the
measurement window is closed by the time the first reflection from the
room occurs (if it's a big enough room and you did the math for the
window right). The problem is that if the window is small enough to
do this, you don't get very accurate low end measurements.

If you do a swept-sine measurement, there is no way to avoid room
issues. The whole point of the impulse response stuff is that it
makes it possible to do actual speaker measurements without an anechoic
chamber. The bad news is that it now requires a microphone with a
much better impulse response than you needed for swept sine.
--scott


--
"C'est un Nagra. C'est suisse, et tres, tres precis."
Anonymous
June 9, 2005 6:15:43 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Scott Dorsey" <kludge@panix.com> wrote in message
news:D 89gc8$q8c$1@panix2.panix.com...
> Mr.T <MrT@home> wrote:
> >
> >
> >But from your own description is probably not "MLS" type
impulse.
> >Why would it need to be anyway, it is designed primarily
to measure sound
> >cards in the electrical domain, not speakers in the
acoustic domain.
>
> No, the whole point of using the impulse response is that
you have a
> short excitation, and you extrapolate the frequency
response from that
> short impulse response measurement.
>
> You can make an impulse response measurement in a room,
because the
> measurement window is closed by the time the first
reflection from the
> room occurs (if it's a big enough room and you did the
math for the
> window right). The problem is that if the window is small
enough to
> do this, you don't get very accurate low end measurements.
>
> If you do a swept-sine measurement, there is no way to
avoid room
> issues.

One common dodge is to follow the response with a tracking
filter. The filter may be actually implemented in hardware
or implicit in a calculation. By coordinating the swept
frequency of this filter with the swept stimulus, you can
effectively select the speaker's response, delayed by a
certain amount of time.

>The whole point of the impulse response stuff is that it
> makes it possible to do actual speaker measurements
without an anechoic
> chamber.

Right, but the thing you usually want to avoid is actually
using an impulse. Thus you optimize the dynamic range of the
measurements. The autocorrelation of anything with uniform,
broadband content makes a nice impulse.

http://www.acoustics.salford.ac.uk/student_area/bsc3/ro...

nicely explains why the MLS is often chosen. But, subject to
other constraints its not always the best answer.

> The bad news is that it now requires a microphone with a
> much better impulse response than you needed for swept
sine.

If you want to have good results, you're stuck having to obt
ain at least half-good measurement equipment. ;-) Most of
speaker measurement technology relates to finessing the
environment. It does little for mitigating the need for a
source or receiver with good performance. It trades-off
fancy computation which used to be prohibitively expensive,
but is now relatively dirt cheap.
Anonymous
June 9, 2005 7:15:49 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Mr.T"
>
> "Phil Allison"
> ** Small omission - the particular ( Rode ) SPL meter used has been
>> modified ( one styro cap removed IIRC ) so the "C" setting is not
>> subject
>> to a high frequency roll off above 8 kHz - so is flat ( -1dB ) to 18
>> kHz.
>
> The C weighting is now flat ??????????
> Why not use the flat response setting in the first place?
>


** You assume there is one.




............. Phil
Anonymous
June 9, 2005 7:34:36 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Phil Allison" <philallison@tpg.com.au> wrote in message
news:3gq1k6Fdm3faU1@individual.net...
> > ** Small omission - the particular ( Rode ) SPL meter used has been
> >> modified ( one styro cap removed IIRC ) so the "C" setting is not
> >> subject
> >> to a high frequency roll off above 8 kHz - so is flat ( -1dB ) to
18
> >> kHz.
> >
> > The C weighting is now flat ??????????
> > Why not use the flat response setting in the first place?

> ** You assume there is one.

Fair enough, but why did you mention using C weighting at all?

MrT.
Anonymous
June 9, 2005 7:54:18 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Mr.T"
>
> "Phil Allison"

>>> ** Small omission - the particular ( Rode ) SPL meter used has been
>>>> modified ( one styro cap removed IIRC ) so the "C" setting is not
>>>> subject to a high frequency roll off above 8 kHz - so is flat
>>>> ( -1dB ) to
> 18 kHz.
>> >
>> > The C weighting is now flat ??????????
>> > Why not use the flat response setting in the first place?
>
>
>> ** You assume there is one.
>
>
> Fair enough, but why did you mention using C weighting at all?
>


** Cos the "A/C" button gets pressed in on the particular meter for
speaker tests - dickwad.




............. Phil
Anonymous
June 9, 2005 10:12:36 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Phil Allison" <philallison@tpg.com.au> wrote in message
news:3gqbjuFdm2lfU1@individual.net...

> ** They needed to know not to use A weighting

Yes, you should have said flat in the first place.

> as I did not.

What you don't know is endless.
Including the bass reponse of that meter it would seem :-)

MrT.
Anonymous
June 9, 2005 10:26:24 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Mr.T" = Mr Turd


** Trevor = another anonymous, autistic turd !!

You are and have always been a brain dead, trolling, useless,
psychopathic
cunt !!

You leave mere morons spinning in your wake.

Get back to your kiddie porn and public dunny trolling.




.......... Phil
Anonymous
June 9, 2005 10:26:25 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Phil Allison" <philallison@tpg.com.au> wrote in message
news:3gqcphFdprmuU1@individual.net...
>
> "Mr.T" = Mr Turd
>
>
> ** Trevor = another anonymous, autistic turd !!
>
> You are and have always been a brain dead, trolling, useless,
> psychopathic
> cunt !!
>
> You leave mere morons spinning in your wake.
>
> Get back to your kiddie porn and public dunny trolling.

Hey Quad Boi!

Didn't the RodBot say you were the undisputed master of primary school dunny
loitering?
Surely you don't want to give your crown away to the humble Mr. T?




>
>
>
>
> ......... Phil
>
>
>
>
Anonymous
June 9, 2005 11:10:04 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Phil Allison" <philallison@tpg.com.au> wrote in message
news:3gqcphFdprmuU1@individual.net...

As usual when cornered, Phil has nothing to say except to claim his own
attributes apply to someone else.

>You leave mere morons spinning in your wake.

You must be very dizzy then Phil.

MrT.
Anonymous
June 10, 2005 12:45:16 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Mr.T" = Mr Turd


** Trevor = another anonymous, autistic turd !!

You are and have always been a brain dead, trolling, useless,
psychopathic cunt !!

You leave mere morons spinning in your wake.

Get back to your kiddie porn and public dunny trolling.




.......... Phil
Anonymous
June 10, 2005 5:17:02 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Phil Allison" <philallison@tpg.com.au>

> autistic turd !!
>
> brain dead, trolling, useless, psychopathic cunt !!
>
> moron
>
> kiddie porn and public dunny trolling.
Anonymous
June 10, 2005 5:22:42 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Scott Dorsey" <kludge@panix.com> wrote in message
news:D 89gc8$q8c$1@panix2.panix.com...
> No, the whole point of using the impulse response is that you have a
> short excitation, and you extrapolate the frequency response from that
> short impulse response measurement.
>
> You can make an impulse response measurement in a room, because the
> measurement window is closed by the time the first reflection from the
> room occurs (if it's a big enough room and you did the math for the
> window right). The problem is that if the window is small enough to
> do this, you don't get very accurate low end measurements.
>
> If you do a swept-sine measurement, there is no way to avoid room
> issues. The whole point of the impulse response stuff is that it
> makes it possible to do actual speaker measurements without an anechoic
> chamber.

Thanks for filling me in on what I already knew!
The point I made is that RMAA is not an MLS measurement and was never
intended for speaker measurement.
There is other software better suited to the task.


>The bad news is that it now requires a microphone with a
> much better impulse response than you needed for swept sine.

No, but you need a calibrated mic in either case if the measurements are to
have any purpose.

MrT.
Anonymous
June 10, 2005 5:28:31 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Bob Cain" <arcane@arcanemethods.com> wrote in message
news:D 891v2027ak@enews1.newsguy.com...
> Sweep or MLS either one gives you the impulse response when
> cross correlated with the matching signal. The sweep method
> has the advantage that harmonic distortion is separated out
> of the result. That's a big advantage considering how much
> speakers typically distort at low frequencies.

The main purpose of MLS impulse testing is to eliminate room reflections
above the gated cut off frequency.
This is obviously unnecessary for soundcard testing (RMAA), and you do want
separate measurements for FR, THD etc.
One would expect the effect of distortion on FR with any soundcard would be
very small though!!

MrT.
Anonymous
June 10, 2005 5:38:06 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Mr.T wrote:
> "Bob Cain" <arcane@arcanemethods.com> wrote in message
> news:D 891v2027ak@enews1.newsguy.com...
>
>>Sweep or MLS either one gives you the impulse response when
>>cross correlated with the matching signal. The sweep method
>>has the advantage that harmonic distortion is separated out
>>of the result. That's a big advantage considering how much
>>speakers typically distort at low frequencies.
>
>
> The main purpose of MLS impulse testing is to eliminate room reflections
> above the gated cut off frequency.

Why does it matter for this whether the stimulus is a sweep
or a pseudo random white noise sequence?

> This is obviously unnecessary for soundcard testing (RMAA), and you do want
> separate measurements for FR, THD etc.

That's what is really cool about the sweep method. You get
both at once and in far more detail than a simple THD
measurement. You get a separate impulse response at each of
the harmonics.

> One would expect the effect of distortion on FR with any soundcard would be
> very small though!!

Yeah.


Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein
Anonymous
June 10, 2005 5:42:15 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Scott Dorsey wrote:

> No, the whole point of using the impulse response is that you have a
> short excitation, and you extrapolate the frequency response from that
> short impulse response measurement.
>
> You can make an impulse response measurement in a room, because the
> measurement window is closed by the time the first reflection from the
> room occurs (if it's a big enough room and you did the math for the
> window right). The problem is that if the window is small enough to
> do this, you don't get very accurate low end measurements.

And you have very low ambient noise immunity with an impulse
stimulus. Not much energy in an impulse.

>
> If you do a swept-sine measurement, there is no way to avoid room
> issues. The whole point of the impulse response stuff is that it
> makes it possible to do actual speaker measurements without an anechoic
> chamber.

With the swept sin, you just take the calculated impulse
response and truncate it at the time to the first
reflection. Low end suffers in the same way.

> The bad news is that it now requires a microphone with a
> much better impulse response than you needed for swept sine.

Why?


Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein
Anonymous
June 10, 2005 8:16:33 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

> And you have very low ambient noise immunity with an impulse
> stimulus. Not much energy in an impulse.

So you average them. When I was making impulse measurements, I usually ran 8
or 16.
Anonymous
June 10, 2005 11:25:18 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

William Sommerwerck wrote:

>> And you have very low ambient noise immunity with an
impulse
>> stimulus. Not much energy in an impulse.

> So you average them. When I was making impulse
measurements, I
> usually ran 8 or 16.

In the end the SNR of any of these measurements, whether
actual impulse or continuous-tone driven, increases with the
time you spend making measurements. But, you get more SNR
per period of wall clock time with signals that have a lower
peak-to-average ratio.

For a broadband signal with low peak-to-average ratio, I can
think of nothing worse than an impulse, and nothing better
than a continuous swept tone.You can get the same data out
of either one, but you get cleaner data faster with the
swept tone.
Anonymous
June 10, 2005 11:25:19 AM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

> For a broadband signal with low peak-to-average ratio, I can
> think of nothing worse than an impulse, and nothing better
> than a continuous swept tone.You can get the same data out
> of either one, but you get cleaner data faster with the
> swept tone.

Please explain how waterfall displays can be derived from swept tones.
Anonymous
June 10, 2005 12:04:59 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

William Sommerwerck <williams@nwlink.com> wrote:
>> For a broadband signal with low peak-to-average ratio, I can
>> think of nothing worse than an impulse, and nothing better
>> than a continuous swept tone.You can get the same data out
>> of either one, but you get cleaner data faster with the
>> swept tone.
>
>Please explain how waterfall displays can be derived from swept tones.

The same way they are usually derived from impulses: they are drawn
up by commercial artists in the marketing department.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."
Anonymous
June 10, 2005 1:09:16 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

William Sommerwerck wrote:

>> For a broadband signal with low peak-to-average ratio, I
can
>> think of nothing worse than an impulse, and nothing
better
>> than a continuous swept tone.You can get the same data
out
>> of either one, but you get cleaner data faster with the
>> swept tone.

> Please explain how waterfall displays can be derived from
swept tones.

Repeat after me - an impulse response derived directly from
an actual impulse is the same as an impulse response derived
by means of the cross-correlation of broadband signals.
Only, the latter probably has a far better SNR, when the
wall-clock duration of the test is held constant.
Anonymous
June 10, 2005 3:27:47 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Mr.T" = Mr Turd


** Trevor = another anonymous, autistic turd !!

You are and have always been a brain dead, trolling, useless,
psychopathic cunt !!

You leave mere morons spinning in your wake.

Get back to your kiddie porn and public dunny trolling.





.......... Phil
Anonymous
June 10, 2005 3:48:01 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Scott Dorsey"
>
> You can make an impulse response measurement in a room, because the
> measurement window is closed by the time the first reflection from the
> room occurs (if it's a big enough room and you did the math for the
> window right). The problem is that if the window is small enough to
> do this, you don't get very accurate low end measurements.
>
> If you do a swept-sine measurement, there is no way to avoid room
> issues.


** Sure there is - you park the mic nice and close to the driver in the
NEAR FIELD.

Long as the sound pressure from room reverberation arriving back at the mic
position is less than 10 % of the near field SPL (easily achieved) good
accuracy is possible in ordinary rooms.


> The whole point of the impulse response stuff is that it
> makes it possible to do actual speaker measurements without an anechoic
> chamber.


** Always easily possible by going to the great outdoors - remember all
those AR speaker adds from the 70s with the half buried speaker in an open
field ????


> The bad news is that it now requires a microphone with a
> much better impulse response than you needed for swept sine.


** Obtaining such a mic is not a problem - plus any decent omni condenser
has way better response and transient performance that just about any
speaker system ever made.

BTW

The REAL reasons impulse testing methods are widely used are:

1. It is the cheapest technology to use.

2. It gives flattering graphs even for atrocious sounding speakers.

3. Graphs are very easily enhanced or plain fudged by speaker makers
working in privacy.




........... Phil
Anonymous
June 10, 2005 4:18:05 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Scott Dorsey wrote:

> < ...snip.. >
>
> The same way they are usually derived from impulses: they are drawn
> up by commercial artists in the marketing department.
> --scott
>
> --
> "C'est un Nagra. C'est suisse, et tres, tres precis."

Damn it Scott, now I have to clean up all the coffee spray!


later...

Ron Capik
--
Anonymous
June 10, 2005 6:36:06 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

William Sommerwerck wrote:
>>For a broadband signal with low peak-to-average ratio, I can
>>think of nothing worse than an impulse, and nothing better
>>than a continuous swept tone.You can get the same data out
>>of either one, but you get cleaner data faster with the
>>swept tone.
>
>
> Please explain how waterfall displays can be derived from swept tones.

Roughly, by doing FFT's of blocks that first contain the
whole IR, then moving the block out in fixed time steps,
dropping more and more of the start of the IR, and finally
plotting the data in three dimensions, time, frequency and
amplitude. Given an IR, no matter how you obtain it, the
procedure is the same.

In a noise free envioronment whether you give the system
under test a real impulse, a random noise sequence, a
sinusoidal sweep, or any full band sequence (along with the
calculated "inverse" for the last three) you get the same
IR. In an environment with noise, the impulse is ruled out
because the amount of signal in the stimulus relative to the
noise of the environment is usually _very_ low.

Yes, averaging can help (if you have a perfectly time
synchronized impulse source.) The noise goes down by
1/sqrt(n) where n is the number of them averaged. It takes
really large n to match what you can do with a sweep or
sequence that lasts 30 seconds or so. For equal noise
immunity you probably have to have an n on the order of the
number of samples in the sweep or sequence. Implicit in the
cross correlation of sequences which calculates the IR is a
very large degree of averaging. Each result sample is a
weighted sum of all the measurement samples, each containing
uncorrelated noise.

I followed the dead end path of actual impulse measurement
to its conclusion using a hefty spark generator I made (and
time synchronization of the data using fractional sample DSP
delays.) Total waste of time as everyone had tried to tell
me it would be. If I'd understood then the averaging
implicit in cross correlation of sequences I wouldn't have
wasted that considerable time.


Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein
Anonymous
June 11, 2005 3:33:10 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Bob Cain" <arcane@arcanemethods.com> wrote in message
news:D 8bjhe01t0a@enews2.newsguy.com...
> Why does it matter for this whether the stimulus is a sweep
> or a pseudo random white noise sequence?

As I said, for speaker testing in a reflective environment, it can matter.
Near field measurements can help too, but may not properly show any
interaction between drivers near the crossover frequencies.

MrT.
Anonymous
June 11, 2005 3:33:11 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

Mr.T wrote:
> "Bob Cain" <arcane@arcanemethods.com> wrote in message
> news:D 8bjhe01t0a@enews2.newsguy.com...
>
>>Why does it matter for this whether the stimulus is a sweep
>>or a pseudo random white noise sequence?
>
>
> As I said, for speaker testing in a reflective environment, it can matter.

I'm asking you to give me more detail on that. It isn't
true to my knowledge. I do a lot of that kind of thing and
if there is a better way that I don't understand I'd like to
hear about it.


Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein
Anonymous
June 11, 2005 3:40:34 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Phil Allison" <philallison@tpg.com.au> wrote in message
news:3gs8kkFdlafjU1@individual.net...
> "Phil Allison" = Repetitive, tiresome, moronic imbecile"
Anonymous
June 11, 2005 3:40:35 PM

Archived from groups: rec.audio.tech (More info?)

On 6/10/05 9:40 PM, in article
42aa411b$0$14817$afc38c87@news.optusnet.com.au, "Mr.T" <MrT@home> wrote:

>
> "Phil Allison" <philallison@tpg.com.au> wrote in message
> news:3gs8kkFdlafjU1@individual.net...
>> "Phil Allison" = Repetitive, tiresome, moronic imbecile"


You kids play with the fresh cow patties WITHOUT the crossposting.. Ok?
Good.
Anonymous
June 11, 2005 5:59:39 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Phil Allison" <philallison@tpg.com.au> wrote in message
news:3gv2n9FefubvU1@individual.net...

<typical Phil shite>

It's not even a full moon Phil. Lost your meds or something?

MrT.
Anonymous
June 11, 2005 6:04:03 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Bob Cain" <arcane@arcanemethods.com> wrote in message
news:D 8dlhf016tq@enews1.newsguy.com...
> > As I said, for speaker testing in a reflective environment, it can
matter.
>
> I'm asking you to give me more detail on that. It isn't
> true to my knowledge. I do a lot of that kind of thing and
> if there is a better way that I don't understand I'd like to
> hear about it.

Maybe the bit you snipped would tell you why near field swept measurements
have limitations too.
The best technique is to use all available methods to give you the best
possible picture, and understand the limitations of each.

MrT.
Anonymous
June 11, 2005 6:16:39 PM

Archived from groups: rec.audio.pro,rec.audio.tech (More info?)

"Mr.T" = Mr Turd


** Trevor = another anonymous, autistic turd !!

You are and have always been a brain dead, trolling, useless,
psychopathic cunt !!

You leave mere morons spinning in your wake.

Get back to your kiddie porn and public dunny trolling.





.......... Phil
!