Preferred Method for Measuring Equalizer Parameters

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Gold

Well-known member
Joined
Jun 23, 2004
Messages
3,670
Location
Brooklyn
When looking at EQ performance graphs I mostly see FFT's with what I assume is White or Pink noise as a test signal. This shows center frequency and bandwidth in a readable form. After talking to a friend and reading some literature it looks like swept sines are less prone to erroneous measurements and generally more accurate. You have to get the windowing right for a particular FFT measurement.

Are there any pros and con's I'm missing. What is recommended?
 
It’s more of a practical problem than a theoretical question. I want to measure bandwidth for a given center frequency of an equalizer. 

I have an AP Portable One. It can do swept sine in the generator. The display is to small to be useful as a graph. I don’t have the print option which can output a tab delineated spreadsheet.

I have an audio interface and REW. Would I be better off using the noise generator in REW or the swept sine generator? Which would display better in REW?

 
I don't know REW specifically so you'ld probably need to try both.  I can say though I've found ARTA with pink noise and a long FFT to be sufficient for the majority of EQ measurements including bandwidth.
 
The problem with FFT is that it will give different results when changing windowing.
Swept sines give less questionable results, as long as the sweep speed is low enough.
The most accurate results are obtained by targetting the desired points with a swept sine, and pinpointing by dialling frequencies (which can be considered some kind of ultra-slow sweep).
After that comes the $64 000 debate of what makes a proper definition of BW in an EQ...
Certainly the -3dB definition that is favoured by many is inadequate.
 
abbey road d enfer said:
The problem with FFT is that it will give different results when changing windowing.

I don’t know the ins and outs of FFT. If you keep the windowing type the same and adjust the display to only cover the target frequency area would that give consistent results across the audio bandwidth? Any easy way to get consistent results for looking at EQ bandwidth?

Swept sines give less questionable results, as long as the sweep speed is low enough.
The most accurate results are obtained by targetting the desired points with a swept sine, and pinpointing by dialling frequencies (which can be considered some kind of ultra-slow sweep).

The way I found the center frequencies was to rock the frequency generator and look for maximum and minimum gain. Usually a couple hundredths of a dB.

If I use swept sine how do I get the display to look like a bell? Do you have to sync the swept sine with the spectrum analyzer? Does the display hold values for the duration of the sweep?

After that comes the $64 000 debate of what makes a proper definition of BW in an EQ...
Certainly the -3dB definition that is favoured by many is inadequate.

This is just so I can see what’s going on. I can sidestep this question.
 
Gold said:
I don’t know the ins and outs of FFT. If you keep the windowing type the same and adjust the display to only cover the target frequency area would that give consistent results across the audio bandwidth?
No. FFT operates on a linear frequency scale. resolution is high at HF, low at LF, so bell curves tend to be a tad tilted.
Any easy way to get consistent results for looking at EQ bandwidth?

If I use swept sine how do I get the display to look like a bell? Do you have to sync the swept sine with the spectrum analyzer? Does the display hold values for the duration of the sweep?
I don't really get it. You want to use a spectrum analyser for display? What spectrum analyser? You should use something that creates a Bode plot. REW can do it. You must use the built-in sweeps and the Measure command.
I'm not sure REW can sync to an external sweep.
 
abbey road d enfer said:
You should use something that creates a Bode plot. REW can do it. You must use the built-in sweeps and the Measure command.
I'm not sure REW can sync to an external sweep.

Thanks. That’s enough to get going on.
 
My preferred method is sine wave generator sweep and level plot.

I have discussed this here at some length and there is some debate about definition of Q/Bandwidth for boost/cut EQ sections.

Over a decade ago, after a few years, I gave up lobbying the AES standards committee to come up with and publish a proper Q/bandwidth definition standard for boost/cut EQ sections.

Q/Bandwidth is well understood for basic bandpass filters, but boost/cut EQ that are based on these band-passed sections using different topology can deliver different Q/bandwidth results for different amounts of boost/cut. In general measuring q/bandwidth at maximum boost/cut will better agree with the underlying bandpass filter section's Q/bandwidth. 

JR
 
john12ax7 said:
Measurements will often come down to a trade-off between speed vs accuracy. It really depends on what you need at the time.

I’m starting to understand this better. With pink noise and FFT I get a real time display but less accuracy than a swept sine measurement.

For EQ bandwidth I want very accurate frequency and level measurements so a swept sine is more appropriate.

If I want to look at the noise floor of a device than an FFT would be more appropriate because I could see things like 60Hz hum more easily.
 
Gold said:
I’m starting to understand this better. With pink noise and FFT I get a real time display but less accuracy than a swept sine measurement.

For an FFT frequency response measurement you need white noise (equal power at all frequencies). Pink noise has equal power per octave.

Cheers

Ian
 
ruffrecords said:
For an FFT frequency response measurement you need white noise (equal power at all frequencies). Pink noise has equal power per octave.

Do you change the settings to target certain frequency areas? If looking at a low frequency range do you increase the number of bins by a certain amount to keep the resolution similar to a higher frequency range? Do you decrease the sampling rate as you look at lower frequencies?

 
Gold said:
Do you change the settings to target certain frequency areas? If looking at a low frequency range do you increase the number of bins by a certain amount to keep the resolution similar to a higher frequency range? Do you decrease the sampling rate as you look at lower frequencies?

REW allows you to take 132K samples which I find more than adequate at all audio frequencies.

Cheers

Ian
 
When building my EQ’s I divided the 10K pot range into five equal positions. For lack of a better method. It sounds like the positions are evenly spaced  and the middle position sounds like the middle position. I’d like to correlate what I’m hearing to a display.

The positions are labeled 1-5 for easy recall and to avoid controversy.
 
Gold said:
When building my EQ’s I divided the 10K pot range into five equal positions. For lack of a better method. It sounds like the positions are evenly spaced  and the middle position sounds like the middle position.
It depends on how wide is the range. Typically, the slope is gentler close to the mid-position and tends to become steeper at the ends. It's generally visible when the amount of B/C exceeds 12-15dB.
Also the apparent BW often varies with the amount of Boost/Cut.
 

Latest posts

Back
Top