I’ve been trying to create a user weighting filter which corrects pink noise input to read as “flat” on the spectrum analyzer (I know white noise already does). This is so I can make the QA403 read as an audio RTA when using pink noise as a source signal. I’ve had no luck in creating this user filter. Has anybody created a filter such as this, and if so could they share it?
Hi @Rammis, instead of doing it with user-weighting, I use the 1/3 octave graph. Then the FFT screen shows the “true” readings and the graph is the “interpreted”. Also, I’m using a hardware pink-noise generator and it doesn’t have much value at finer resolution than 1/3 or 1/6 octave.
1/3rd octave is not nearly enough resolution for frequency response measurements using pink noise as a source. That’s why I was trying to create a user weighting filter. In theory it seemed simple to create such an inverse pink noise filter but I always ran into a boost in the 1kHz region of about 5dB.
What are you using as the pink-noise source?
My hardware generator is not very “smooth” at such a fine resolution, there are peaks and valleys from the generator that are irrelevant for the measurement. I put the pink noise through an amplifier, out a speaker, and picked up from a calibrated measurement microphone, into a QA472 pre-amp, into the QA403 analyzer, and displayed on the 1/3 octave graph. This gives a rough picture of perceived frequency response, room acoustics, etc, much like the analog analyzers of old.
If you are measuring a voltage source, like the output of an amplifier directly, I recommend sticking with white-noise and the standard FFT. This is because the QA403 is a digital device and operates with even-width FFT bins instead of 1/3 octave analog filters.
I don’t need an alternate procedure or solution. I was asking if somebody has created a user weighting filter for pink noise input to read as flat on the FFT. What I’m using as a pink noise source is pretty irrelevant. I’m kind of surprised that a pink noise filter doesn’t come with the original software package as pink noise is so widely used in pro audio in acoustic measurements and things like lining up analog tape machines. Can we please stick to my original question?
Well, I don’t have such a file. Perhaps someone else here may be able to share.
It sounds like you’ve started one but are seeing a slight error in the reading. Are you using the exact octave-interval frequencies or the “rounded” preferred frequencies? For example, the 8kHz band center should actually be 7,943 Hz when you are computing your +3dB/octave.
Hi @Rammis
Take a look at the notes on 1.183 HERE and see if that helps at all. There’s an option to specify LOGINTERP that would should let you specify a pink inverse such as:
LOGINTERP
100,0
10000, -20
Does that solve your problem?
Thanks. It shouldn’t matter what the frequencies are but I’ll give it a try using the precise frequencies. The supplied user weighting filters work extremely well for the RIAA curves, so I see no reason that a curve couldn’t be crafted for something as simple as converting a pink noise spectra to read out as if it were as white noise one. For some reason the stumbling block is only in the vicinity of 1kHz where it deviates by about 5dB.
I’ll give that a try. Thanks.
Hi @Rammis you should be able to fudge the response at 1k however you wish by modifying the endpoints.
I’ve actually resorted to doing that with no luck. The range around 1kHz is always boosted by around 5dB no matter what numbers I enter into the txt file. Very maddening since it should be a dead simple filter to make.
I’ve gotten the user weighting filter to yield a “flat” response when being fed pink noise with the following:
LOGINTERP
10,40
1000,20
10000,10
100000,0
However the overall level is very low at around -70dB. Any ideas how to boost the readout to more closely match what I’m feeding in (about 1 volt)?
This also works, but the voltage display is way off. This doesn’t really matter in my case but it would be nice to correct for the level.
LOGINTERP
10,-20
1000,-40
10000,-50
100000,-60
You mean the level at 1kHz or the RMS?