merging ff and nf

[[email protected]]

after i measure farfield, i average and then bring in the mike in and adjust the volume until the upper portion matches on both. then i average a couple of times and save. dose this give me accurate results? with this method will the phase info be useful or do i need to extract min phase later?
if not, how can i accomplish this?

[dcibel]

Re: merging ff and nf

This presentation is for SoundEasy, but will give you a good idea of the steps to merge near and far measurements. I've done roughly the same using Frequency Response Combiner spreadsheet from the FRD consortium.

http://chadgray.info/soundeasy/Measure_a_Speaker_With_SE.htm

[bwaslo]

Re: merging ff and nf

The magnitude will be accurate if the measurement environment allows for enough common overlap between the nearfield (not good at too high freqs) and farther field (not good at lower freqs). Though the overlap can be pretty poor with indoor measurements. You have to adjust gains (vertical positions) to match there. Phase will NOT be correct from a min phase extraction, as just about no multidriver speaker is minimum phase. Correct phase should be obtainable via comparing summed measurements of the drivers measured separately with their crossovers, a feature that will be in the next OM version release (i don't know when that will be yet, though ).

[[email protected]]

Re: merging ff and nf

Thanks. That is how I'm doing it. What seems logical to me is not always correct.

[[email protected]]

Re: merging ff and nf

Being the bottom of a threeway, is the phase alignment critical??

[ErinH]

Re: merging ff and nf

The magnitude will be accurate if the measurement environment allows for enough common overlap between the nearfield (not good at too high freqs) and farther field (not good at lower freqs). Though the overlap can be pretty poor with indoor measurements. You have to adjust gains (vertical positions) to match there. Phase will NOT be correct from a min phase extraction, as just about no multidriver speaker is minimum phase. Correct phase should be obtainable via comparing summed measurements of the drivers measured separately with their crossovers, a feature that will be in the next OM version release (i don't know when that will be yet, though ).

Bill, I know I'm beating a dead horse here but I really wanted to repost my suggestion from a while back regarding this topic just to maybe keep it fresh, as I believe a lot of us could benefit from some sort of automated or easier way, in general to merge NF/FF plots and save them off as a single file. Especially those of us who present the data on the forum. I like posting on/off axis data and as it stands, I have to do this all in the FF and make a note that baffle and room influences limit the legitimacy of the data below a certain frequency. If that could be mitigated it would be a huge stress reliever.

Nearfield and Farfield Stiching, made easier.
As it stands, if I want to take measurements at 0, 15, and 30 degrees in the farfield and merge them with the nearfield, I have a few options but they're a bit of a process to go through:

• One is that I simply move the NF response down so that it lines up with the FF measurements. Problem here is that the plot looks a bit nasty with 4 lines on it and it's just not a great way to present data.
• Second is to adjust the X-Axis and measure up to/within your predetermined Fmax/Fmin. For example, I know that I'm measuring a woofer in the NF and it's Fmax is 700hz. I have been able to tame reflections in the room and the first real problem comes at 5ms (200hz). That means I have about 500hz overlap that I can try to align the 4 graphs to. So, I'll set the mic up for NF measurements and make the max X-axis value 700hz. Measure. Next, I'll move the mic to FF and set the min X-axis value to 500hz. Measure. This gives all measured data hard limits set by the X-axis. Now I have 200hz overlap and that's manageable. (I realize the problem here is that you're just ballparking where yo u'll stitch without seeing both responses fully... again, another reason I've started this thread).
• Another option I've toyed with is adjusting the .frd file itself. I'll take full measurements of both NF & FF and figure out the best place to stitch them. I then open up the .frd file as .txt, chop off the frequencies I don't want for the respetive response file and then save them. Load them back in to OM, bring down the NF plot to match the beginning of the FF plots, and I've got a much cleaner data set to present.

[bwaslo]

Re: merging ff and nf

Erin,

The next version will allow you to specify the displayed frequency range of "Added" curves. That's about the best I can do for splicing I think.

Myself, I go outside when I need anechoic curves to lowish frequencies. That will be the most accurate. Of course off-axis curves are meaningless with near-field mic acqusitions in low midrange and down, so outside is the only real choice for that. The splicing is basically for 'marketing' curves.

[ErinH]

Re: merging ff and nf

Sounds great. Looking forward to the update. Any other treats you can share with us?

Also, Bill, have you guys considered allowing the screen shots to be of higher resolution? The only reason I ask is because I've actually had people complain to me when posting data that the window resolution could be higher. Not a real big issue... just tossing it out there.

I really appreciate you being so receptive to ideas, even if they're silly. A company that at least listens to its customers is pretty rare.

[jcandy]

Re: merging ff and nf

Its mentioned in the Soundeasy tutorial, but nobody seems to be talking about it in the thread. You need to take the nearfield measurement and add the diffraction gain. The maximum validity of this result is restricted by the cone diameter because of complexity in the soundfield in the proximity of the driver. To avoid this complex region (where the nearfield response is not neatly related to the farfield) you must have ka << 1. In real terms, this means f >> 728 Hz / D, where D=d/6in (i.e., D=1 for a 6in effective cone diameter). More precisely when f=728/D, the error is about 16% (if I am not making an algebra mistake). The error drops to about 2% when f=364/D. If you are below this limit you can add diffraction gain to the nearfield to obtain the correct anechoic low-frequency response. This also assumes your diffraction gain calculation is correct, which is another can of worms. ALL diffraction gain algorithms you get with standard software are high-frequency approximations, and can be inaccurate in the 100Hz region.

With regard to farfield, another caveat is related to an impression I have, and that is that people tend to trust a gated FFT well below its region of validity. If you want to use the farfield result at a given frequency, make sure the result is not sensitive to gate size at that frequency. That is, do not put all your trust in the simple "reflection-free window" estimate.

[charlielaub]

Re: merging ff and nf

I just posted an Excel based tool that can be used to merge nearfield and farfield measurements. I call it the "FRD Response Blender and Minimum Phase Extractor". I created this after finding other methods cumbersome or difficult to understand. Please see the thread about it in this forum:

http://techtalk.parts-express.com/showthread.php?234735-FRD-Response-Blender-and-Minimum-Phase-Extractor-is-now-available&p=1869069#post1869069

I am using it to combine box models, nearfield measurements, and diffraction responses with gated frequency response measurements that I make with ARTA. Any program that can generate FRD files can create the input file. I don't own OmniMic, but I assume it can export data as FRD files...

There is a tutorial in the link, above, with a couple of examples.

-Charlie

[dcibel]

Re: merging ff and nf

I'm surprised there hasn't been a greater response to this. Charlie, thanks for the spreadsheet. It's just what everyone with an Omnimic (or any measurement rig) needs. Very straight forward and functional!

[kendomusic]

Re: merging ff and nf

Newbie question: After merging the FF and NF measurements, how is the baffle step taken into consideration?

[charlielaub]

Re: merging ff and nf

Newbie question: After merging the FF and NF measurements, how is the baffle step taken into consideration?

The baffle step should be incorporated into the merging process.

In general the frequency response can be broken down into:

The low frequency regime
The baffle step
The high frequency regime

A measurement of the frequency response of the driver might capture the high frequency regime and part or all of the baffle step. Usually the low frequency regime must be obtained by:

A box model using the driver's Thiele-Small parameters and the loading (sealed, vented, etc.)
A nearfield measurement
Other low frequency measurement (e.g. ground plane)

Commonly you will only need to stitch together the low frequency measurement/model and a high frequency measurement that already contains some or all of the baffle step. But it's not always possibly to capture enough of the baffle step in the high frequency measurement to be very confident in it. In that case, you can use a model the diffraction response (that causes the baffle step) and combine this third "input" with the other responses.

I typically use a box model, plus a diffraction response model, plus a high frequency measurement, and then combine all of these together keeping the SPL level of the high frequency measurement unchanged and scaling the other responses to match it.

These kind of blending tasks are easily handled by the FRD Response Blender...

-Charlie

[[email protected]]

Re: merging ff and nf

I really look foward to giving it a try. Thanks.