MLS measurement at low frequencies - will this method work?

[HiFiNutNut]

Hi, an idea has just popped up from my mind. Let me know if it is workable or not.

Due to room reflections we normally limit the gate to a few ms window therefore the measurement is only valid from a few hundred Hertz up.

Let us assume that the impulse starts at 5ms and ends at 10ms and the first reflection arrives at 15ms.

We do the first measurement from 5ms to 102ms. We then do the second measurement from 10ms to 102ms.
We then substract the second measurement from the first.


Provided that the ambient noise is sufficiently lower than the reflections (would they be?) and the latency of the soundcard does not get changed, the second measurement would contain mainly the reflected singals, which should be identical to those in the first measurement. In this way, we obtain an accurate measurement with a 96ms window which would make the frequency response accurate down below 100Hz.

What do you think?

Regards,
Bill

[John Sheerin]

No, it wouldn't work unless I'm not understanding you correctly.

Think of it this way:
You take a measurement and calculate the impulse response. Now you limit it to 5ms-100ms. Now you subtract the values from 10ms-100ms, setting 10ms-100ms to 0. You need that data to see what the actual LF response is. This would remove not only the reflections and noise, but also the LF response.

[HiFiNutNut]

John,

Thanks for your reply. It helps me to start thinking what MLS measurement is.

I have just started using SpeakerWorkshop to do some measurements. The impulse response is used to identify the gated window. I was thinking about that the first frequency response measurement would contain the signal as well as the reflections, and the second measurement, since we moved the gated window after the impulse, would contain only the reflections. If we take out the reflections from the first frequency measurement, we would have the original signal. But the increased gated window would allow sufficient data points at low frequencies.

Regards,
Bill

[SpeakerScott]

Originally posted by HiFiNutNut
John,

Thanks for your reply. It helps me to start thinking what MLS measurement is.

I have just started using SpeakerWorkshop to do some measurements. The impulse response is used to identify the gated window. I was thinking about that the first frequency response measurement would contain the signal as well as the reflections, and the second measurement, since we moved the gated window after the impulse, would contain only the reflections. If we take out the reflections from the first frequency measurement, we would have the original signal. But the increased gated window would allow sufficient data points at low frequencies.

Regards,
Bill

John is correct, it won't work. However, you can...take your speaker outside, place it on the ground and put a microphone 2 meters away. Make sure it isn't windy, and average 10 or so measurements...you'll be able to get much lower in frequency. In my driveway I can get down to ~60-100 Hz...which makes pasting in low frequency data much easier...

[John Sheerin]

Originally posted by HiFiNutNut
John,

window. I was thinking about that the first frequency response measurement would contain the signal as well as the reflections, and the second measurement, since we moved the gated window after the impulse, would contain only the reflections. If we take out the

Regards,
Bill

You can do the same thing taking only 1 measurement. Just measure the impulse response (in SW). Then copy and paste it so you have two of the same impulse. Then do your different windowing. But you would do the subtraction in the time domain - with the impulses. So you'd just set everything after some point in time to zero, but that's not really what the speaker's impulse response is. The problem is that you can't differentiate what is noise, what is the direct signal, and what is a reflection. If you could, you'd already know the response of the speaker. Hope that helps a bit.

If you could do this, anechoic chambers wouldn't be so useful. But they are. And as Scott said, outside in an open space is usually the next best thing.

[HiFiNutNut]

Thanks, guys.

For outdoor measurement, I presume that you use ground plane measurement, i.e. putting the mic on the ground. How big the gated window do you use?

For a sufficient number of data points, I was thinking about a 52ms window. That requires that the closest boundary be 8.5 metre away, or 4.25 metre away if a 26ms window is used.

Because my speakers are dampped / stuffed open back 3 way, near-field measurement can't be used for low frequency measurement. I have to find a method to measure the low frequency response down to 75Hz. The midwoofer is crossed at 150Hz and will have an acoustic third order slope. So it would need at least one octave accurate data down to 75Hz.

I have the space in my backyard but it is in a slope so I am afraid that I could only do it in a 15-25 degree vertical axis. Tilting the speakers is not practical. If it was not a MTM it would be OK but it is a MTM and I would expect some lobbing issues in the vertical axis.

I have a relatively dampped room and have installed sound damping panels on the first reflection points on both the side wall and the ceiling. After stacking 6 cusions on the floor betweeen the mic and the drivers with a 1.2 metre distance, I was surprised to find that from the impulse the only identifiable reflection comes after 18ms, clearly from the backwall, not the floor, ceiling or side wall. But the data points are still so rare below 300Hz and I would not be able to use the measurement to design the XO.

[SY]

It strikes me that you're trying to do this the hard way. Why not do a nearfield measurement at low frequencies, then splice it onto a 1 or 2m gated measurement for midband and treble? True, you'll miss room effects, but that's also true of ground-plane.

[bwaslo]

Originally posted by SY
Why not do a nearfield measurement at low frequencies, then splice it onto a 1 or 2m gated measurement for midband and treble? True, you'll miss room effects, but that's also true of ground-plane.

Sy is right. The problem with the whole idea is that the room effects are exactly the same as the thing you are attempting to remove. But the room effects are a very real part of the speaker's low frequency response -- while those effects are different for different rooms and for different speaker placements! Sure, you can measure in an anechoic chamber (or outside) or try to use some processing tricks to get the same effect, which might be a response you can compare to other speaker measurements made the same way. But it won't be near what will happen in any real room, it can at best show how low in frequency the speaker can make energy to work with. Probably the most useful measurement for low frequencies is to do the smoothed in-room measurement at the placement the speaker has to be put in (probably put there for other reasons than LF response, such as mid and high frequency imaging). Ignoring the floor and walls is a bad idea.

[HiFiNutNut]

It strikes me that you're trying to do this the hard way. Why not do a nearfield measurement at low frequencies, then splice it onto a 1 or 2m gated measurement for midband and treble? True, you'll miss room effects, but that's also true of ground-plane.

The midwoofer runs from 150Hz to 3kHz in OPEN BACK. Nearfield measurement is good for a close box but will not take the back waves into considerations therefore can't be used.

Sy is right. The problem with the whole idea is that the room effects are exactly the same as the thing you are attempting to remove. But the room effects are a very real part of the speaker's low frequency response -- while those effects are different for different rooms and for different speaker placements! Sure, you can measure in an anechoic chamber (or outside) or try to use some processing tricks to get the same effect, which might be a response you can compare to other speaker measurements made the same way. But it won't be near what will happen in any real room, it can at best show how low in frequency the speaker can make energy to work with. Probably the most useful measurement for low frequencies is to do the smoothed in-room measurement at the placement the speaker has to be put in (probably put there for other reasons than LF response, such as mid and high frequency imaging). Ignoring the floor and walls is a bad idea.

I am building a 3 way with OPEN BACK midwoofers but the woofers are sealed taking care of 150Hz and below. I could use nearfield measurement on the woofers. However, I also thought of the same approach as what you recommend, ie. including the floor bounce and other room reflections.

I thought that since the woofers are placed low just above to the floor and the XO point is fairly low as well at 150Hz, I should be able to measure them gated or farfield and use the result in designing the XO. However, I posted the question a few months ago but I did not get an answer.

One problem may be that the reflected signal is no longer minimum phase.

[SY]

Nearfield measurement is good for a close box but will not take the back waves into considerations therefore can't be used.

That will also be true of ground-plane. As Bill says, you really want to get accurate in-room response. I've found REW to be a very useful piece of software for capturing LF in-room LF response; I average over several mike positions near where my head goes when listening. There's also an excellent freeware program from Liberty Audio (Bill's company), the name of which escapes me, that uses a prerecorded MLS-style signal.