how & why to use white and pink noise?? (1 Viewer)

[Manuel Delaflor]

I have the SpectraLab software, I can do my own pink and white noises and record them on a CDR. I also have a RS SPL Meter, how can I use all this to test speakers?
Thanks
------------------

 

[Patrick Sun]

Can you generate sine waves and warble tones of the audible frequency spectrum 10Hz-20KHz)? They would be more useful for measuring speakers.
Or...you can just order a test tone CD, like the one from Stryke which has 99 tracks, most of them are useful for both subs (sine waves) and normal speakers (warble tones).
------------------
PatCave; HT Pix; Gear; DIY Mains; DIY CC; Sunosub I + II + III; DVDs; Link Removed

 

[Manuel Delaflor]

Yes, I can generate sines but apparently not warbles (at least the help doesn't mention it) on all audible spectrum. I already made a subwoofer test CDR and in fact that way I bought a litle sub to use on my computer!
Now, why does everybody use the white and pink noises?
------------------

 

[Patrick Sun]

The better question is: What is your objective?
Have you done a websearch for your questions? Here's a Link Removed and another Link Removed here and Link Removed that answers your main question on pink/white noise.
Pink noise is what you want for speaker testing.
------------------
PatCave ; HT Pix ; Gear ; DIY Mains ; DIY CC ; Sunosub I + II + III ; DVDs ; Link Removed

 

[Manuel Delaflor]

Thanks Patric.
What I'm looking is to understand methodologies to test speakers. I have a software to test and an SPL meter but still don't understand really well what is happening nor how to do some tests.
For example, in the magazines the graph response curve from all speakers is presented in a "linear horizontal" manner. So far the only graph I'm able to produce is "curve horizontal" meaning that HF is much lower in dB than LF.
(I'm not native English speaker so perhaps I'm not explaining myself).
Thats why I'm asking all this.
------------------

 

[Patrick Sun]

The generally accepted industry practice for presenting a speaker's FR (frequency response) is to use a signal with a voltage level 2.83V measured with a suitable SPL meter/microphone from 1 meter away, and measure its output in SPL (Sound Pressure Level).
The best FR will be measured on-axis from the front of the speaker baffle (front side). Sometimes you'll see FR for 30 degrees off axis and 60 degrees off-axis - or some variation of off-axis. This allows XO designers to factor in off-axis response to minimize nulls in the overall FR.
Sometimes you'll see the power level of 1W instead of 2.83V (has to do with sensitivity vs. efficiency of the speaker), but both will yield similar result, usually within 1dB of each other.
So, keeping the volume level constant (and set to 2.83V - how you determine this, I don't know, use a AC DVM perhaps, I just pick a volume level that gives me 80dB measured from 1 meter away, and then leave the volume level as-is for the rest of the measurement session), pump in different frequencies starting from the lowest to the highest, and measure the output from the SPL meter for each frequency (at a minimum, you should use 1/3 octave frequency tones from 20Hz to 20,000Hz).
Once you have all the SPL measurements, just plot them on a graph with the frequency along the horizontal axis, and the SPL on the vertical axis. Usually you'll see logarithmic graphing division on the frequency axis because sound is logarithmic, but you can also plot them as-is in a spreadsheet. I do this.
Some people have nice measurement systems that allow the tester to input the frequencies into the speaker, and the mic is able to record the SPL and that info is sent back to the PC where the software write a file that includes the frequency and the SPL output recorded, sometime phase is also being measured. With a properly set up system, you can generate the data in a matter of seconds.
If you use the Radio Shack SPL meter, there's some corrections mainly for the low frequencies that need to be added or subtracted (depending on the frequency) to smooth out the FR. Factor this into your graph, or plot both curves and shot for something in the middle.
If you see bumps or dips in the response, those can be problem spots. If your tweeter is shelved down in its response, the woofer is overpowering the tweeter, and may need to be turned down (unless the tweeter has been over-padded/attentuated).
------------------
PatCave; HT Pix; Gear; DIY Mains; DIY CC; Sunosub I + II + III; DVDs; Link Removed

 

[Andrew Beacom]

Pat,
>>most of them are useful for both subs (sine waves) and normal speakers (warble tones).

 

[Greg Monfort]

MD,
They use computer programs that take energy density into account like an RTA does.
The reason yours is curved is because the acoustic power of pink noise, sinewaves, etc., falls at 3dB/octave, just like real music, sounds, etc.. When you measure with an SPL meter, then if the FR is flat, the meter will actually give you a plot that falls at 3dB/octave because energy density falls with increasing frequency, or 1/f. If you do the math, you'll see that 20kHz is -30.1dB Vs 20Hz.
GM
------------------
Loud is beautiful, if it's clean

 

[Manuel Delaflor]

Patrick, thanks for the informative post, I wil test some speakers that are far away from my computer using only the SPL meter.
Greg,
"The reason yours is curved is because the acoustic power of pink noise, sinewaves, etc., falls at 3dB/octave, just like real music, sounds, etc.. When you measure with an SPL meter, then if the FR is flat, the meter will actually give you a plot that falls at 3dB/octave because energy density falls with increasing frequency, or 1/f. If you do the math, you'll see that 20kHz is -30.1dB Vs 20Hz."
That was exactly what I was looking for! I never understood why I wasn't able to produce such graph! Now, I think that the software Im using is a RTA (SpectraLab), but my results show exactly what you are describing.
So Im like "trapped" and I dont know how to "get out"! Is there a way to "normalize" the results that I don't know?

 

[Greg Monfort]

If you're using an RTA and pink noise, then all you need is a mic hooked to it and positioned to feed the signal back to the RTA, which should display a "flat" representation of it. If your soundcard won't allow this, then get some log paper and draw your "flat" line along the Y axis starting at 10Hz on out to 20kHz, then measure up the X axis 33dB and draw a line from the X/Y coordinate to it.
Now when you use the SPL meter, plot the results along the rising slope line and you'll get the "flat" version, just don't be surprised if it's anything but.

GM
------------------
Loud is beautiful, if it's clean

 

[StephenMSmith]

You need to set the scaling to 1/3 octave in SpectraLabs when measuring pink noise. Logarithmic scaling will produce the downward sloping result.
However, you should use the Real or Complex Transfer functions in Spectralabs for the most accurate results (and a flat target curve regardless of scaling). See the help file. Basically, you plug both the SPL and a line out from your preamp into the 2 channels on your soundcards line input. Spectralabs then displays the difference b/t the two. This eliminates any FR issues due to your soundcard itself.
With the transfer functions, you can also use the Delay Finder to precisely calculate the exact delay times for each of your speakers. This is much more accurate than just measuring distances.
Steve
------------------

 

[KevinP]

Manuel,
White noise has equal energy at all frequencies. Pink noise has higher energy at lower frequencies at a 3db per octave rate. The choice of which to use depends on what type of spectrum anallyzer you use. One type is the constant bandwidth analyzer which has a passband of fixed widthas it is tune throughout the spectrum, a 5hz bandwidth is common. Whitenoise with its flat energy gets displayed as a flat line. The other type of analyzer is the constant percentage where the bandwidth changes with frequency, 1/3 octave bands are typical, 1/3 of 100 hz is 23 hz, 1/3 at 10khz is 2300 hz so the analyzer gathers more energy at 10khz at a rising rate of 3db per octave, using pink noise with this type of instrument will display the flat line.
Having said all this, if your goal is to get levels between drivers even with an spl meter (as far as I know these are constant bandwidth devices) you will want to use white noise. You will also be assuming that the speakers having flat response within their range, this will make your spl readings less usefull and difficult to interperet.
I do not know Spectralab, but there are many pc based measurement options that work quite well, Spectralab may be one of them.
This is great stuff, have fun first.
Kevin

 

[Manuel Delaflor]

Steve
Thanks for the information. It's good (at least for me

) that you know the program. Did you learned to use it using the help file alone? I'm new to this kind of software, I'm not an engineer and just want to have fun and to learn to make objective tests of speaker systems.
"You need to set the scaling to 1/3 octave in SpectraLabs when measuring pink noise. Logarithmic scaling will produce the downward sloping result."
I already did that, and I don't like the resulting graph because is to "discrete" and I want more information instead of specific bands.
"However, you should use the Real or Complex Transfer functions in Spectralabs for the most accurate results (and a flat target curve regardless of scaling). "
Now, that's really interesting! WOW! I will have to learn to do that. It appears that SpectraLab is the best software out there, but it's expensive. Do you know an alternative?
------------------

 

[Manuel Delaflor]

Greg and Kevin.
Thanks for your comments, it is a nice thing to be part of this forum!

------------------

 

[StephenMSmith]

The help file is pretty good, but is obviously targeted towards people who already understand the concepts. I learned a lot by searching on "Spectralab" right here in the HTF, but mostly just a whole lot of experimenting. It's fun!
You can go higher than 1/3 octave for better octave resolution, but you'll need to up the FFT size to keep the same range, and then things start to slow down. This is where all the experimention comes in to play.
See the help file on Real Transfer Function. It pretty much explains it all. I'm lucky in that my ACT-3 pre/pro lets me output a full-range signal on the sub outs, so that is what I run into the right channel of my soundcard's line in (the left in RS SPL). This is definitely the best way to go.
Steve
------------------

 

[Manuel Delaflor]

Regarding the transfer function I dont have a pre, so I cant do some of the tests.
However Im thinking in a way to do this, perhaps using a cable to feedback my soundcard with a pink noise from spectralab??
Anyway, I played with the settings, but dont understand which to use to get what you mentioned. Averaging? Coherence?
------------------

 

[Greg Monfort]

>I'm lucky in that my ACT-3 pre/pro lets me output a full-range signal on the sub outs, so that is what I run into the right channel of my soundcard's line in (the left in RS SPL). This is definitely the best way to go.
=====
Why is using a known incorrect 20-20kHz reading device (RS SLM) better than a mic feedback?
GM
------------------
Loud is beautiful, if it's clean

 

[StephenMSmith]

There is a known compensation chart for the RS SPL that you convert into a SpectraLabs mic compensation file. This gets the RS SPL remarkably accurate! I don't know the URL offhand, but it's routinely sited in posts on this forum. A search on "mic compensation" will take you there.
Steve
------------------

 

[Julian Data]

I wouldn't soley based the RS compensation chart for all RS Meters as the tolerances vary from one another. The meter isn't really accurate

 

[Greg Monfort]

>There is a known compensation chart for the RS SPL that you convert into a SpectraLabs mic compensation file. This gets the RS SPL remarkably accurate!
====
I have several different RS comp files published over the last 6yrs by folks with the necessary measurement tools, and my RS meter isn't close to any of them based on a comparison with a calibrated B&K unit, so I don't share your opinion. A so-so mike is cheap and gets you closer IMO if real world numbers is the goal.
GM
------------------
Loud is beautiful, if it's clean