How do I measure speaker frequency response?
ys...@casbah.acns.nwu.edu
I have access to a two phase lockin and a frequency source. What I plan
to do is take the signal from the frequency source and split it into a
reference for the lockin and an input for the stereo (using tape input).
I would then use the lockin with a microphone to measure the speakers output.
This should give the output power as a function of frequency.
Does this seem reasonable? Are there better/easier ways to do it?
How linear is the frequency response of a cheap microphone? Is there
a way to measure this and to correct for its error?
What about using white noise? Is that better? Can you get white noise
by off tuning an AM receiver? I also have access to a computer
with an A-D and FFT, is that useful?
jan Kycia
Richard D Pierce
Why?
>I have access to a two phase lockin and a frequency source. What I plan
>to do is take the signal from the frequency source and split it into a
>reference for the lockin and an input for the stereo (using tape input).
>I would then use the lockin with a microphone to measure the speakers output.
>This should give the output power as a function of frequency.
No, AT BEST, it will give sound pressure level at a single point in
space. And that will include myriads of dependencies on position, speaker
placement, the room, who's standing near the microphone.
>Does this seem reasonable? Are there better/easier ways to do it?
There are FAR better ways of doing it, but none of them are easy, and
none of them are chepa, and all of them are desitgned to give you
SPECIFIC answers to SPECIFIC questions, which leads to my first
quaestion: why do YOU want to do it? What do you hope to learn? The
answer to that very much constrains HOW you do it.
>How linear is the frequency response of a cheap microphone?
If you're talking about cheap microphones that you can buy at Radio Shcak
or that might come as part of a soundcard, the answer is: they're
absolutely awful.
>Is there
>a way to measure this and to correct for its error?
The best way to correct for it is to not have the problem to begin with,
the cheapest, ready-made dependable measurement grqade microphone is
going to cost you about $300, from somebody like Josephson Engineering.
If you're clever and are willing to sacrifice some accuracy, noise margin
and absolute calibration, you can get one of the Panasonic electret
capsules from DigitKet for about $5.00 and build a precision supply and
buffer amplifier for it.
Or, if you don't want to worry at all about ANY problems with a
measurement mike, and just want one that you know is dead accurate, be
prepared to spend up to $5,000.
>What about using white noise? Is that better? Can you get white noise
>by off tuning an AM receiver?
Not even close. Of an FM tuner, not even close.
>I also have access to a computer
>with an A-D and FFT, is that useful?
Sure, if you have the time to do a proper data acquisition, windowing,
scaling, display, etc., software package for it, go for it. BUt be
prepared to resolve ALL the other problems that people who have done it
before you have. Think about aliasing problems, think about windowing
artifacts, think about how your stimulus is going to affect your systems,
and on, and on, and on.
It still comes down to: what are you tryiny to do and why?
--
| Dick Pierce |
| Loudspeaker and Software Consulting |
| 17 Sartelle Street Pepperell, MA 01463 |
| (508) 433-9183 (Voice and FAX) |
ys...@casbah.acns.nwu.edu
Pierce) wrote:
> In article <ysong-12039...@spindry.phys.nwu.edu>,
> <ys...@casbah.acns.nwu.edu> wrote:
> >I would like to measure my stereo's frequency response.
>
> Why?
I want to do it because it would be fun.
> >I have access to a two phase lockin and a frequency source. What I plan
> >to do is take the signal from the frequency source and split it into a
> >reference for the lockin and an input for the stereo (using tape input).
> >I would then use the lockin with a microphone to measure the speakers output.
> >This should give the output power as a function of frequency.
>
> No, AT BEST, it will give sound pressure level at a single point in
> space. And that will include myriads of dependencies on position, speaker
> placement, the room, who's standing near the microphone.
>
Well, That is what I want to know. I would like to fiddle with my
room arrangement and speaker positioning. I would like to fiddle with my
bass and treble knobs. Yes I know I should be able to hear this and
not have to quantify it but it would be fun.
>
> >How linear is the frequency response of a cheap microphone?
>
> If you're talking about cheap microphones that you can buy at Radio Shcak
> or that might come as part of a soundcard, the answer is: they're
> absolutely awful.
>
Thanks for the microphone info.
>
> >What about using white noise? Is that better? Can you get white noise
> >by off tuning an AM receiver?
>
> Not even close. Of an FM tuner, not even close.
>
> >I also have access to a computer
> >with an A-D and FFT, is that useful?
>
> Sure, if you have the time to do a proper data acquisition, windowing,
> scaling, display, etc., software package for it, go for it. BUt be
> prepared to resolve ALL the other problems that people who have done it
> before you have. Think about aliasing problems, think about windowing
> artifacts, think about how your stimulus is going to affect your systems,
> and on, and on, and on.
This seems like a fun project as well ... and I agree a involved one.
>
jan kycia
p.s. I still would like to hear about the better ways of measuring the
frequency response. What do speaker makers do? Do they just listen
by ear? That seems really unlikely.
Ralph T. Muehleisen
: If you're clever and are willing to sacrifice some accuracy, noise margin
: and absolute calibration, you can get one of the Panasonic electret
: capsules from DigitKet for about $5.00 and build a precision supply and
: buffer amplifier for it.
I would argue that if you have access to the right equipment, these cheap
little mic's can outperform just about anything around.
I needed may channels of reasonably phase matched microphones for some
measurements I needed for my thesis. Phase Matched B&K's (the defacto
standard of measurement mic's) go for $2500 a pop - without including the
cost of the amps. So, I bought a bunch of the Panasonic WM 63's and found
that most were phase matched within 2 degree's across the audio band and that
I could find 4 (I bought 10) that were matched within +/- 1 degree. They
matched within +/- 200 mdB amplitude across the audio band as well. They were
were nearly perfectly flat from about 40Hz-20kHz.
Using the simple circuit in the Digikey catalog, I had a relatively high S/N
ratio compared to the B&K mic's. I found however, that by upping the supply
voltage to the microphone and adjusting the resistance, that for a certain
resistance value (unique to each mic capsule) the distortion and noise floor
drop way down. I was getting s/n and distortion measurements in the 70-90
dB range!!! I figure it has to be something like matching the bias current
to the idss of the fet or something. I tried biasing with a constant current
source but it didn't work any better than the standard biasing circuit.
(The resistance went from 2k up to the 8k-26k range)
So, in short, with a little bit patience tweaking a pot (and the correct
instrumentation to measure the levels) you can turn those $3 panasonic
mic's into virtual B&K's.
Now some things to note:
The panasonics are not as stable as the B&K's, but having found the correct
load resitance value for the mic's, they seem to drift a bit less.
The WM 63 are not availble anymore. The WM 60's I've tried do not match
anywhere near as well and drift more than the WM 63's. (The WM 63's had
an epoxy back while the WM 60's only have a phenolic back - maybe it's
a temperature thing).
So, those of you who have been using the standard panasonic circuit, start
tweaking those resistors and lowering that distortion.
--
Ralph Muehleisen
Graduate Student in the Graduate Program in Acoustics at Penn State
Just learnin' to quiet down ducts...
It's a bird! It's a plane! It's a (quack) ... duct!
-----------------------------------------------
Ralph Muehleisen (r...@sabine.acs.psu.edu)
phone (814-865-0422)
fax (814-865-3119)
-----------------------------------------------
Richard D Pierce
<ys...@casbah.acns.nwu.edu> wrote:
>In article <D5CKE...@world.std.com>, DPi...@world.std.com (Richard D
>Pierce) wrote:
>> >I would like to measure my stereo's frequency response.
>> Why?
>I want to do it because it would be fun.
Well, I do it for a living, and it ain't that much "fun", I guess because
sometimes I end up doing for 8-10 hours a day!
However, let me warn you that one of the traps you'll fall into, even
assuming that you have the right collection of equipment and adopt a
dependable methodology, is that you will be absolutely overwhelmend with
information. The data you'll get will often be very confusing, often
contradictory. Without a specific path, how do you propose to make the
choices and sorts to get where you want to go? (This could be as much of
a rhetorical question as a practical one, mind you).
>> >This should give the output power as a function of frequency.
>>
>> No, AT BEST, it will give sound pressure level at a single point in
>> space. And that will include myriads of dependencies on position, speaker
>> placement, the room, who's standing near the microphone.
>
>Well, That is what I want to know. I would like to fiddle with my
>room arrangement and speaker positioning. I would like to fiddle with my
>bass and treble knobs. Yes I know I should be able to hear this and
>not have to quantify it but it would be fun.
Well, but moving the microphone position an inch or two might give wildly
different measurements at high frequencies, especially if you're using
sine waves as the stimulus. How do you reconcile what might seem two
totally different curves with the fact that you'll probably be moving
your head similar amounts and more while listening?
>> >I also have access to a computer
>> >with an A-D and FFT, is that useful?
>>
>> Sure, if you have the time to do a proper data acquisition, windowing,
>> scaling, display, etc., software package for it, go for it. BUt be
>> prepared to resolve ALL the other problems that people who have done it
>> before you have. Think about aliasing problems, think about windowing
>> artifacts, think about how your stimulus is going to affect your systems,
>> and on, and on, and on.
>
>This seems like a fun project as well ... and I agree a involved one.
Yeah, try about 4-8 man-years to do it right. That's what's typical for a
minimally acceptable commercial package with limited capability but that
gives repeatable, dependable results. For your purposes, with an ugly but
useable user interface, you might hack something together in 4-6 weeks of
work. But by what criteria do you judge the quality of the results?
Or, you could go out and buy something ranging from a sound-card based
system(which reviews indicate are pretty awful), through something like
and IMP system available from Old COlony for about $500 assembled (which
seems not bad), to a LMS from Linear-X for about $1200 (which works okay,
but has its own set of problems inherent in the measurement technique
that I find unacceptable) to a variety of real-time analyzers for prices
ranging from $800 to $2500, to a DRA MLSSA system for about $3600 (which
has its set of problems that I find acceptable) to Techron systems and up
to the tens of thousands of dollars.
>p.s. I still would like to hear about the better ways of measuring the
>frequency response. What do speaker makers do? Do they just listen
>by ear? That seems really unlikely.
Some of the system above are what are found in many speaker manufacturers
arsenal of equipment and techniques. Personally, I own B&K swept sine
wave measurements, swept filter measurement, 1/4 octave real time
analysis equipment, a DRA MLSSA system (with which I do the bulk of the
acoustic measurement), a pile of custom made stuff for special purpose
measurements (for example, a phase-locked loop system that measures
loudspeaker resonance drift with time/temperature/level), and a
complement of dependable microphones and calibrators.
There are some loudspeaker manufacturers that have NO equipment. Some of
them rely on consultant like myself for their measurements, some try to
wing it with pure simulation software. There are some, even of legendary
status, that get away with nothing. They tune purely by ear and customer
response and, as a class, take MUCH longer to settle on final designs that
are minimally acceptable than those that have more objective means of
attaining information that are not subject to fatigue (a VERY important
consideration) or lack of reference points (equally important).
Len Moskowitz
<ys...@casbah.acns.nwu.edu> wrote:
>I would like to measure my stereo's frequency response.
Dick Pierce has explained that it's not straightforward. The best way
to learn about testing like this is to buy the last few years of Speaker
Builder magazine. Look for articles about the IMP, a low cost tester
that'll do impulse and MLS testing.
Speaker Builder Magazine
Old Colony Sound Labs
PO Box 243
Peterborough, NH 03458 (603) 924-6371 (9-4, M-F)
--
Len Moskowitz
Core Sound
mosk...@panix.com
Robert Matthews
> I would argue that if you have access to the right equipment, these cheap
>little mic's can outperform just about anything around.
Depending upon the application I would agree, especialy for the price.
>...[deletions]...
>The WM 63 are not availble anymore. The WM 60's I've tried do not match
>anywhere near as well and drift more than the WM 63's. (The WM 63's had
>an epoxy back while the WM 60's only have a phenolic back - maybe it's
>a temperature thing).
Which model of the WM-60's did you try? There are 2: WM-60AT, and WM-60AY.
The AT model is the *recomended* substitute. I'm not sure of the differences
except that the AT has a higher sensativity. I haven't tried either one
but I am going to build several mics in the near future and would like to
know. I can still get the WM-063T from Panasonic, but they want $9 for it.
Digi-Key has the WM-60's for about $2.25. If the AT's were the ones that
you tried, I may just bite the bullet and pay Panasonic for the 63's. If
you could respond with some more information, or if ANYBODY knows a better
way to get some 63's please let me know.
Thanks,
Robert Matthews
Robe...@Novell.com
/* These are my views, not my employer's */
Ralph T. Muehleisen
: >The WM 63 are not availble anymore. The WM 60's I've tried do not match
: >anywhere near as well and drift more than the WM 63's. (The WM 63's had
: >an epoxy back while the WM 60's only have a phenolic back - maybe it's
: >a temperature thing).
:Which model of the WM-60's did you try? There are 2: WM-60AT, and WM-60AY.
:The AT model is the *recomended* substitute. I'm not sure of the differences
:except that the AT has a higher sensativity. I haven't tried either one
I have the panasonic mic catalog and the only difference in the 60AT and
the 60AY seems to be the sensitivity. They have the same spec sheet, only
the sensitivity parameter is different. (AT is 44dB+/- 3dB and the AY is
42dB +/- 3dB). As I said before, I'm inclined to think it's the phenolic
back VS the epoxy back. You might be able to cure the problem with a bit
of clay stuck on the back of the AY's. (I'd give it a try but I've got my
63's working fine and I need to finish my measurements to get my thesis done).
Robert W. Igo
clarity and frequency response, and how much would such a procedure cost?
This would have to be in a chart format which could be shown to a customer.
[********************************* Bob Igo ********************************]
"This Emacs is from OCTOBER!" -- John, my office mate
[***************************** gryp...@cmu.edu ***************************]
Richard D Pierce
Robert W. Igo <gryp...@CMU.EDU> wrote:
>How about this question, then: Where can one get speakers professionally analyzed
>for clarity and frequency response, and how much would such a procedure
>cost?
>This would have to be in a chart format which could be shown to a customer.
Well, more for information than for self promotion (since, at the moment,
I'm booked solid and couldn't possibly squeeze something in for several
months), people such as myself do this on a professional basis. The
equipment I own can provide frequency response, phase response, multiple
axis curves, impedance, distortion, etc. etc., all in nice, multicolored
charts. Rates for this kind of work vary from site to site, but I charge
on an hourly basis unless it's a regular client.
However, I know of no such measurement as "clarity" as that is a very
subjective term. One man's clarity is another's brittleness, and all that.