How to make quasi-anechoic speaker measurements/spinoramas with REW and VituixCAD

[fluid]

Any work-arounds?

You can do all the scaling, diffraction simulation and merging in Vituix which is designed for the purpose. If you are having trouble with REW you might want to try Vituix.

https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_help_20.html#Merger_tool

 

[TwoEyedJack]

You can do all the scaling, diffraction simulation and merging in Vituix which is designed for the purpose. If you are having trouble with REW you might want to try Vituix.

https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_help_20.html#Merger_tool

I spoke too soon. REW worked fine. VCAD merger tool is great at visualizing and doing all the operations at once as well. Both of them gave exactly the same response.

Thanks for the help.

 

[TheLastGerman]

Why do the peaks and dips in the frequency response vary with the window width? What is the actual frequency response?

 

[Salt]

View attachment 415542

Why do the peaks and dips in the frequency response vary with the window width? What is the actual frequency response?

What setup for measurement was in use?

 

[TheLastGerman]

What setup for measurement was in use?

I forgot to mention: The picture is from here (First page of this thread in the paragraph "3) The On-Axis measurement (sans bass)").

 

[Salt]

The lower the frequency the more the room comes into measurements, and thus time domain.
Klippel recognizes this and counts for it.

 

[fluid]

Why do the peaks and dips in the frequency response vary with the window width? What is the actual frequency response?

The shorter the gate time the less sampling points there are to describe the frequency response. At high frequencies there are still enough points per octave to correctly plot the response as measured. As frequency goes down there is less and less resolution to describe the frequency response and it gets smoothed, until the point where there are not enough samples to even guess what the response should be.

When the gate time is before the first reflection the response is quasi anechoic. Frequencies above the smoothing effect of the gate are generally accurate. When the gate is longer and allows reflections in the frequency response is a combination of the device and the reflections.

 

[mwmkravchenko]

View attachment 415542

Why do the peaks and dips in the frequency response vary with the window width? What is the actual frequency response?

The longer the time the impulse is allowed to propagate between the device under test and the microphone the better the low frequency resolution you have. And also the less smoothing you have. The gentlemen here have given great advice on first reflection. That is the nearest thing to an anechoic test. There are a number of different reasons. Never take any measurement as the final answer, immutable frequency response of a loudspeaker. Change it's position in the room, or change your listening position and the response changes. Happily we can still enjoy the music!

 

[Ageve]

Why do the peaks and dips in the frequency response vary with the window width? What is the actual frequency response?

The actual response is the one without reflections (before the first spike in the impulse response).

When you start to see ripples, you have an inaccurate result (caused by reflections). If it's just small ripples from the mic stand, or the speaker stand, the overall response is often still accurate, but it's always a good idea to get rid of those as well.

As you can see in the gif, the high frequency response remains the same until the resolution is too low to get an accurate result (too short window).

In this example I measured at 1m distance, with a 5 ms window, and then at 2m, with a 3 ms window (to avoid reflections):

Overall response is still the same with the 3 ms window, but resolution below 1.5 kHz is too low to capture the resonance at 1.2 kHz.

 

[TwoEyedJack]

As you can see in the gif, the high frequency response remains the same until the resolution is too low to get an accurate result (too short window).

In this example I measured at 1m distance, with a 5 ms window, and then at 2m, with a 3 ms window (to avoid reflections):

Overall response is still the same with the 3 ms window, but resolution below 1.5 kHz is too low to capture the resonance at 1.2 kHz.


View attachment 415958

Why does the resolution decreases with a smaller window? I though only the minimum frequency decreased because the window time was shorter that the frequency's period ( T ) therefore the soundwave wasn't able to perform a full oscillation in this time window ?

Why did you choose a smaller window for the 2m compared to the 1m? Shouldn't increasing the distance allow for a longer window since reflections would need a longer time to reach the mic as well ?Resulting in a lower captured minimum frequency?

 

[Newman]

Why did you choose a smaller window for the 2m compared to the 1m? Shouldn't increasing the distance allow for a longer window since reflections would need a longer time to reach the mic as well ?Resulting in a lower captured minimum frequency?

The further the microphone is from the driver in the same room setup, the less is the additional path distance that reflections travel to the microphone (compared to direct sound from the driver to the microphone).

For example, if the mic is 1 cm from the driver and the floor is 100 cm away, the path difference is 100+100-1 = 199 cm.

But if the mic is 100 cm from the driver and the floor is 100 cm away, the path difference is 111+111-100 = 122 cm.

 

[Ageve]

Why does the resolution decreases with a smaller window? I though only the minimum frequency decreased because the window time was shorter that the frequency's period ( T ) therefore the soundwave wasn't able to perform a full oscillation in this time window ?

As the frequency increases, the time needed to capture data with adequate resolution decreases. Lower frequency = longer wavelength = more time needed.

As you can see in the graph, there's no information below ~360 Hz with the shorter window* (not enough data).

* It was actually 2.8 ms.

Here's another example, with different windows.

Resolution below 2 kHz is very low with the 2 ms window, and there's no data < 500 Hz.

This is the speaker measured btw:

Bowers & Wilkins 686 S2 Spinorama measurements (CTA-2034)

Here are measurements of the Bowers & Wilkins 686 S2 bookshelf speaker. It was released in 2014, and discontinued in 2018, and it's two generations behind the 607 S2 measured by Amir. I Think the MSRP was 399.99 USD / pair. My measurements are quasi-anechoic, with near-field port+woofer...

www.audiosciencereview.com

 

[TwoEyedJack]

As the frequency increases, the time needed to capture data with adequate resolution decreases. Lower frequency = longer wavelength = more time needed.

As you can see in the graph, there's no information below ~360 Hz with the shorter window* (not enough data).

* It was actually 2.8 ms.

Here's another example, with different windows.

Resolution below 2 kHz is very low with the 2 ms window, and there's no data < 500 Hz.

View attachment 416404

This is the speaker measured btw:

Bowers & Wilkins 686 S2 Spinorama measurements (CTA-2034)

Here are measurements of the Bowers & Wilkins 686 S2 bookshelf speaker. It was released in 2014, and discontinued in 2018, and it's two generations behind the 607 S2 measured by Amir. I Think the MSRP was 399.99 USD / pair. My measurements are quasi-anechoic, with near-field port+woofer...

www.audiosciencereview.com

I understand why there is no data below a set frequency ( for example for 360Hz the wavelength is 0.952 m and the period is 2.77 ms , your gating frequency).

But why is the measuring resolution smaller at around 1.2kHz in this example and you only capture the dip with lower gating?

Is it because the dip in the response is caused by something that happens later in the time than the initial impulse like an internal reflection and the lower gating doesnt have enough time to gather enough sampling data?

 

[3ll3d00d]

It just means the frequency resolution is too low given the short time window (which is a fundamental tradeoff with an fft) to even see it let alone know what generates it. Having said that,.given the way the response through that section is that little bit lower, I would guess it is present and generated early in the response.

 

[Elgrosso]

How big is the baffle you are measuring? A rule of thumb assuming a baffle barely wider than the largest driver is mic distance at least 3 times that width. So for a nominal 200mm wide baffle for a 2 way 6.5'" woofer design. You can probably get down to 600mm mic distance without masking baffle diffraction or driver summation accuracy. You can see tools like vituixcad to model and optimise at a listening distance e.g 2.5m

The above will gain you some resolution. For a 3 way this doesn't really help if you are crossing below the gate as is often the case

I tried to find some references for larger baffle/speakers and I'm struggling.
My boxes are very large, a bit smaller than jbl 4350 or Superwax: 70x100cm with the horn on top, its center at 85cm.
Since it is large and low, the gating at 2 or 3 meters gets very short!
1.82ms for 2meters and 1.3ms for 3m, so nothing below 500 or 800hz.

(here in an old position)

Taking baby steps for now before finding a plan for outside measurements.
So I will still try but it looks like I can only do a mix of near-field for bass and low midrange to get the whole picture.
Or do you have any other advice?