Nearfield/Farfield curve splicing

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Some of us have used a technique of splicing a nearfield or dustcap measurement to farfield pseudo anechoic measurements to try and create a realistic wide band look at woofer/box response. In fact every system measurement in Stereophile magazine for several decade has used this approach.

Jeff Bagby has written a white paper on an improved method of doing this which should be more accurate than the current approach. It showed up on a discussion forum on parts express:

New Paper available on How to Get Accurate Measurements Indoors Down to 10 Hz

The actual paper, worth a read, is in PDF form here:

https://app.box.com/s/fefis558wna1d6pd07r3

The thinking has been that a dustcap measurement will give a representative response for low frequencies that is free from the standing waves of your room (or even of an imperfect anechoic chamber). At upper frequencies we can use a gating approach to measure the "anechoic" response. All we have to do is splice the two curves together at somewhere around 200 Hz and we have a wideband and accurate reflection free response curve.

The problem is that the dustcap measurement doesn't properly represent what the woofer looks like in free space because it doesn't include the effects of baffle size and diffraction. Bagby uses a diffraction modeling program to create a 2pi to 4pi correction curve and adds that to the mix. When he adds the "baffle step" correction to the dustcap measurement, it and the farfield measurement show much more agreement over a broader frequency range. Splicing becomes less arbitrary and the proper level of bass vs. midrange is revealed.

Recommended!

Note that every nearfield approach suffers this same issue, as well as microphone in the box measurements, Thiel/Small approaches, P-Spice models, etc. They all assume the driver is non-directional and don't take finite baffle size into account. In that regard they look more like the 2pi response, as infinite baffle mounting will force constant directivity from very low frequencies up to beyond piston band operation.

If you want to play with diffraction modeling try here:

Home of the Edge

Tolvan Data

(Click on “The Edge”)

Don Keele’s classic paper on nearfield measurements is here:

http://xlrtechs.com/dbkeele.com/PDF/Keele (1974-04 AES Published) - Nearfield Paper.pdf

He gives draw away curves for a woofer both mounted in a cabinet and in a large baffle. Draw away curves for the infinite baffle case retain the dustcap response shape while the 4 pi case varies as you move away from the driver.

Regards,
David
 
Just so that the same problem does not occur here as over at the PE Board, Jeff acknowledges that others have identified this approach before but he was unaware of it. Jeff and Charlie have put together software that makes it easier to do and Jeff explains it in a clear fashion that makes this approach more accessible to the masses.
 
Some of us have used a technique of splicing a nearfield or dustcap measurement to farfield pseudo anechoic measurements to try and create a realistic wide band look at woofer/box response. In fact every system measurement in Stereophile magazine for several decade has used this approach.

Jeff Bagby has written a white paper on an improved method of doing this which should be more accurate than the current approach. It showed up on a discussion forum on parts express:

New Paper available on How to Get Accurate Measurements Indoors Down to 10 Hz

The actual paper, worth a read, is in PDF form here:

https://app.box.com/s/fefis558wna1d6pd07r3

The thinking has been that a dustcap measurement will give a representative response for low frequencies that is free from the standing waves of your room (or even of an imperfect anechoic chamber). At upper frequencies we can use a gating approach to measure the "anechoic" response. All we have to do is splice the two curves together at somewhere around 200 Hz and we have a wideband and accurate reflection free response curve.

The problem is that the dustcap measurement doesn't properly represent what the woofer looks like in free space because it doesn't include the effects of baffle size and diffraction. Bagby uses a diffraction modeling program to create a 2pi to 4pi correction curve and adds that to the mix. When he adds the "baffle step" correction to the dustcap measurement, it and the farfield measurement show much more agreement over a broader frequency range. Splicing becomes less arbitrary and the proper level of bass vs. midrange is revealed.

Recommended!

Note that every nearfield approach suffers this same issue, as well as microphone in the box measurements, Thiel/Small approaches, P-Spice models, etc. They all assume the driver is non-directional and don't take finite baffle size into account. In that regard they look more like the 2pi response, as infinite baffle mounting will force constant directivity from very low frequencies up to beyond piston band operation.

If you want to play with diffraction modeling try here:

Home of the Edge

Tolvan Data

(Click on “The Edge”)

Don Keele’s classic paper on nearfield measurements is here:

http://xlrtechs.com/dbkeele.com/PDF/Keele (1974-04 AES Published) - Nearfield Paper.pdf

He gives draw away curves for a woofer both mounted in a cabinet and in a large baffle. Draw away curves for the infinite baffle case retain the dustcap response shape while the 4 pi case varies as you move away from the driver.

Regards,
David

InKtoresting, wery inKtoresting

I have been doing it the same way since getting a mic almost year ago. Suppose the engineer in me is trying to be freed. Nagging voice in the back of my head says I shouldn't have stopped building speakers over 20 years ago, but the kid was near death sick. Wadya suppose to do? He's a very healthy 21 year old today :)


Pssst, getting settled into your new digs yet?

Cheers,
Mike
 
Interesting, I had this same idea a couple of years ago when I started doing FR measurements. What is great is a new SW that automates the blending process. Thanks Dave for pointing here the document.

But I bring on a problem I have: what is the correct method of a near field measure if the driver has a phase plug instead of a dust cap.

Ralf
 
But I bring on a problem I have: what is the correct method of a near field measure if the driver has a phase plug instead of a dust cap.

Ralf

Not sure about "correct" but I would just try a number of measurements near the center of the cone and see what the variation is. I think even over the phase plug a curve would look like the nearfield average, for the most part.

Nearfield curves are only accurate in the piston band and will also show cancelation nulls below the piston band from the arrival time of various concentric rings of radiation (see the Keele paper for curves on that).

Try a number of measurements and only trust them over the range where they are consistent. (And report back your findings!)

David
 
Taking a nearfield measurement and adding a simulated baffle step is really nothing new. Implemented in MLSSA, ARTA and probably many other software packages too.

http://www.artalabs.hr/AppNotes/AP4_FreeField-Rev03eng.pdf

Thanks for the link. It looks like their 4pi to 2pi conversion is very simplistic, just a ramp between the two levels, so the final result is not as accurate. But yes, the approach is similar.

David
 
The other thing to note is that Jeff suggests that the blended rather than single point splice might be a more accurate representation of the true response. I think in this case, however, we would need to see a true anechoic comparison to both.

No anechoic chamber needed. The typical problem with splicing two non-smooth curves together at a single point is that a local peak or dip can offset the entire section of the curve above or below that point. The blending allows you to overlap the curves and then smoothly transition from one to another across a range of frequencies that you, the user, choose. The result is more forgiving for the user and in most cases results in a more accurate blended SPL.
 
And then there's the old fashioned method used by AR and McIntosh in the old days, where they dig a hole in the dirt ground out back, and place the speaker in it aimed straight up at the sky, and put a measurement mic several feet above it. Drive the speaker with pink noise and look at the RTA. It's easy, and relatively accurate. Less to go wrong.

Yea but surely doing it that way wont give you the baffle diffraction in your measurement?

I was thinking of buying some 20ft flag poles off ebay and hoisting my speakers and mic up off the floor. Not sure what the neighbors would think though.
 
I'm curious what makes the in-room nearfield dustcap measurement more accurate. I've done it and alas the nearfield sample is more believable than the one taken further back. But if there is a room mode at the speaker position why doesn't it show in the measurements?

Is the mode filtered away by the measuring software?
 
Yea but surely doing it that way wont give you the baffle diffraction in your measurement?

I was thinking of buying some 20ft flag poles off ebay and hoisting my speakers and mic up off the floor. Not sure what the neighbors would think though.
You can also just set the speaker cabinet on the ground aimed up, no hole in the ground, or elevate it up from the ground until you simulate your expected distance from the listening room front wall.
 
Doh!

I'm curious what makes the in-room nearfield dustcap measurement more accurate. I've done it and alas the nearfield sample is more believable than the one taken further back. But if there is a room mode at the speaker position why doesn't it show in the measurements?

Is the mode filtered away by the measuring software?

OK, nevermind, I just read the paper.
I still have trouble visualizing this. It sounds like there is a thin layer of air right next to the cone that is relatively free of interference.

https://app.box.com/s/fefis558wna1d6pd07r3
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.