Getting accurate phase data

[5th element]

OK here goes when measuring loudspeakers there are certain phase characteristics that you normall get. A phase starting off high +180, which then slowly falls to -180 then quickly jumps up to +180 again. Then is goes between -180 and +180 a few times.
Now this is what I get when measuring drivers. See pic. this is a merged response near field to get bass and far for upper mids etc. Now there is an option to set the distance the mic is away from the loudspeaker which is necessary to get good phase. BUT what i need to know is how far away do you set the mic. If I have the mic literally a milimeter away from the duscap for the nearfield response I get the -180 to +180 at about 2kHz, setting mic distance to zero. Then if I get the farf response at 80cm, the phase thing happens at about 10k. Then if I try different distances, ie 50cm it happens at about 8k and then 30cm about 6k. This to me doesnt make any sense whatsoever, the phase should stay the same right? its just the near and far is required to get the low and high frequencies. This creates problems when simulating because the simulated combined response is different from the actual combination. Is there anyway to know what distances to use? I am using a panasonic WM61A and a preamp.

[SY]

What you're seeing is the excess phase from the time delay between the mike and the driver zero delay plane. The "jump" is because +190 degrees is the same as -170 degrees. If you want to merge two plots, diddle the distance of one of them until the phases match at the merge point.

[5th element]

Yes I understand that entirely its just how do I know which is the correct plot to use.

The problem is when I change the mic distance and compensate for this by telling the program how far away the mic is, the phase doesnt remain the same. If you keep the mic distance the same and change the distance in the program, then the phase data moves to the left or to the right. If you keep the distance in the program the same and move the mic then the same thing happens which is what you would expect.

But if you alter the distance of the mic and the distance in the program you would expect all the phase data to stay in the same place, or at the same frequency, but it doesnt.

At one distance mic and program the 180/180 swap will happen at 5k, but at another distance and compensated for by the program this swap will happen at say 10k.

[SY]

Which plot you use as the reference phase plot is arbitrary- the point is to have the phases match at the point where you're doing the merge. I'm assuming here that you're not trying to combine two different measurements of two different drivers; this only works if you're trying to merge a near-field and a far-field measurement of a single driver.

The phase unwrapping phenomenon is very useful if you want to determine the location of the driver zero delay plane with respect to the baffle. d'Appolito's book on measuring loudspeakers has a really clear explanation of how to do this. You ought to get that book anyway- it's in paperback and not too expensive.

[5th element]

Hmmm not that its your fault but youve not quite got the point, unless there is something im missing which is quite possible!

Im not asking about merging two plots that was done as an example of a phase response that I can acheive.

The problem is dertermining which point to use in the first place.

I measure the near field response of the driver and the 180/180 phase happens at 2k.

Then I measure the farf where the 180/180 happens at 6k.

Which set of data do i take as the correct one, the near or farfield?

[SY]

That was my point- the phase is arbitrary. If you take the sine of x, then take the sine of x + 360, it's the same thing. As is the sine of x + 720. None of those is more "right" or more privileged than another- that's why we do a wrap of the phase at +/- 180.

If what you want is the minimum phase plot, you can get that by differentiating the magnitude plot with respect to frequency (Hilbert transform). If what you want is the phase data to allow crossover design, mount both drivers on the baffle and take the data at the same distance.

[cocolino]

just a quick note:

accurate acoustical phase measurements are quite sensitive to signal noise ratio.
Keep that in mind when comparing near field and far field phase measurements and for far field crank up the volume of the test signal eventually.

[Svante]

I don't know if I understand the question correctly, but it seems as if you expect the phase to be the same when you move the microphone and tell the program that you have moved it?

If so, I think that you cannot expect this to work if one of the positions is in the near field. When you put the microphone really close to the membrane the mic will "see" sound from the closest point, but also from points further towards the edge of the membrane. All parts of the cone surface will add up to the sound that reaches the microphone, and it is easy to imagine that there are frequency response issues towards higher frequencies as a result of this, but also phase issues, like you experience.

So, looking at only the phase, the effective distance to the speaker does probably *not* vary as the physical distance.

Did that answer the question, or am I way off?

[ucla88]

If you plan to use both measured spl and measured phase then all measurements must use the same window starting point, that is, the same offset distance.

In other words, pick a measurement distance, say 1m. Do all measurements at this distance. If you offset the measurement, use the exact same offset for all drivers. It does not have to be the actual measurement distance-this is totally arbitrary-it just has to be the same for all measurements.

You cannot use merged near and farfield spl and phase in lspCAD.


It is possible to use a merged graph, but you have to disregard the actual phase and use minimum phase data. This is difficult with justMLS and lspCAD-though it can be done. I think Ingmar has a tutorial on this.

Point is, for xover design, forget the merged graph and use the same physical measurement distance for all measurements and the same offset for all measurements. Though, as previously noted, the offset and the true physical distance don't have to be the same.

Hopefully that answers your question.


mark k

[5th element]

Yes that looks to be quite helpful. So for xover design I just measure at a set distance for both drive units and dont change the distance in MLS.

Ill give this a try tomorrow when i get in from uni and report how it goes.

I had been following ingmars ugly duckling tutorial anyway but just got a bit stuck on that one point.

Also I take it that I do still move the mic so its level with the drivers centre?

Cheers again Matt