How to test an EQ

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

Bonsaimaster

Well-known member
Joined
Apr 17, 2013
Messages
216
Dear all,

Sorry for the stupid question if this is an easy one. I have built a number of eq diy projects. I have seen that they have frequencies that they effect. But is there anyway of testing the eq for their performance in a freq range? I see all these eq graphs for eq but unsure how they are generated. Are they theoretical or tested? If tested how does one do that?

Thanks in advance,

Bonsaimaster
 

Attachments

  • UV_EQ_HMF_wide_Q.jpg
    UV_EQ_HMF_wide_Q.jpg
    71 KB · Views: 14
Last edited:
I've used Room EQ Wizard (REW) for this. Once you calibrate it for your soundcard, it's a really handy way to check frequency response of outboard, including EQ sweeps at different settings (as well as using it to measure acoustic response of a room, of course).
 
A sweep is simply the most common way of obtaining a frequency response plot. It is fine if you just want a static plot. However, I like to be able to see the plot in 'real time' so I can see the effect of tone controls as they are varied. And you can do this too with REW. Instead of a sweep you use the real time analyzer (RTA) to view the output spectrum. If you feed a white noise source (RTA has one built in) through whatever you are testing, the RTA will give you a plot of the frequency response directly. And as you change EQ you can see the response change on the screen. Cool eh?

Cheers

Ian
 
A sweep is simply the most common way of obtaining a frequency response plot. It is fine if you just want a static plot. However, I like to be able to see the plot in 'real time' so I can see the effect of tone controls as they are varied. And you can do this too with REW. Instead of a sweep you use the real time analyzer (RTA) to view the output spectrum. If you feed a white noise source (RTA has one built in) through whatever you are testing, the RTA will give you a plot of the frequency response directly. And as you change EQ you can see the response change on the screen. Cool eh?

Cheers

Ian
Good tip, thanks Ian. I've tried using Voxengo Span and a white noise generator in my DAW for this, but found it was a little imprecise (probably had Span set up wrong). I'll try in REW next time I'm working on an EQ.
 
I've tried using Voxengo Span and a white noise generator in my DAW for this, but found it was a little imprecise
Using a spectrum analyzer and pink or white noise is always less accurate than sinewave sweeps, because the width of the filters in the analyzer add to the EQ width.
Let's say you use a 1/3rd octave analyzer to measure a 1/3 octave filter, the resulting graph will be 2/3 octave wide.
Of course, this is a gross example, but it is something to keep in mind.
Most software analyzers use FFT, which uses a linear frequency scale, meaning there is as much measurement between 20 and 40Hz than between 15980 and 16000Hz.
For example, a 1024-point FFT has a BW of about 20Hz, which is half an octave at 40 Hz. At higher frequencies, FFT has a relative BW that goes narrower and narrower.
Many RTA's are FFT-based, meaning that their measurements are derived from an FFT, so their precision at low frequencies may be dubious.
This is easily overcome by using much more points (resolution) for the FFT. It is quite common to use 512k or 1Meg point resolution.
Of course, in that case, the effect on BW is negligible.
However, one remaining factor that may explain your disappointment is that the noise generators are, by essence, unstable. As a consequence, the measurement accuracy depends on the length of observation of the measurement, and its possible integration over time.
RTA analyzers and FFT have their uses and merits, but they are not the best tools for EQ assessment.
 
Using a spectrum analyzer and pink or white noise is always less accurate than sinewave sweeps, because the width of the filters in the analyzer add to the EQ width.
Let's say you use a 1/3rd octave analyzer to measure a 1/3 octave filter, the resulting graph will be 2/3 octave wide.
Of course, this is a gross example, but it is something to keep in mind.
Most software analyzers use FFT, which uses a linear frequency scale, meaning there is as much measurement between 20 and 40Hz than between 15980 and 16000Hz.
For example, a 1024-point FFT has a BW of about 20Hz, which is half an octave at 40 Hz. At higher frequencies, FFT has a relative BW that goes narrower and narrower.
Many RTA's are FFT-based, meaning that their measurements are derived from an FFT, so their precision at low frequencies may be dubious.
This is easily overcome by using much more points (resolution) for the FFT. It is quite common to use 512k or 1Meg point resolution.
Of course, in that case, the effect on BW is negligible.
However, one remaining factor that may explain your disappointment is that the noise generators are, by essence, unstable. As a consequence, the measurement accuracy depends on the length of observation of the measurement, and its possible integration over time.
RTA analyzers and FFT have their uses and merits, but they are not the best tools for EQ assessment.
Thank you for that excellent and very helpful explanation.
 
I use Right Mark Audio Analyzer (RMAA) since years. Combined with a good converter it tells me everything I need.
 
..depends on what I'm after. For detailed analysis like for design work, the AudioTester does really fine - but for at-a-glance and still precise verification, looking at a squarewave on the 'scope is the best

/Jakob E.
 
..depends on what I'm after. For detailed analysis like for design work, the AudioTester does really fine - but for at-a-glance and still precise verification, looking at a squarewave on the 'scope is the best

/Jakob E.
I know you're a big fan of square-wave testing, Jakob, but can you really evaluate a typical 4-band EQ with them?
 
..yes, when the important topic is detecting variation in response between channels, not an absolute numeric target, then the squarewave is far superior imo..
 
Square waves are quick and easy to confirm a path is "working" if you know what to look for.

Swept sine waves are the old school trusted way to measure frequency response.

Back in the day I used graph paper with a log frequency scale for manual frequency response plots.

I do not miss the old days...

JR
 
a lot of companies do this with an audio precision machine. It's how you get the chart/graph in the manual. It's nothing more the a sine wave sweep from 20-20K minimum at fixed level. The graph itself is frequency(x) vs Level(y). So as you sweep up from low to high with the eq settings boosted you will see where it starts to bell, notch or shelf, and how much it will go based on where the eq knobs( frequency, bandwidth and gain) are set.
 
Back
Top