Transfer function. Poles not where expected

[Raka]

I'm a bit lost regarding this matter. I have measured nearfield, and I obtain a clean curve, useful from 700Hz and downwards. I also measured the impedance and I got a Fs of 54Hz, and the estimation of the Q is 0.57.
I understand that the poles you get from the impedance curve are the poles of the transfer curve of the SPL (if measured without effects of any resonant system, be it a closed box or a reflex system). I measured nearfield (2cm) with the driver mounted in a dipole, which is kind of free air (you should expect a bit higher Qts, and a lower Fs, but that's all).

Well, If I draw the bass response and the transfer curve of Fs=54 and Qts=0.57, they just don't match.

You can see in the picture, that the resulting Q is equal, but not the Fs!!!!
I understand that the factor that can alter the poles evaluation is the inductance, but in this case the value is really small. Another thing could be that the Q is very small, but clearly this is not the case.

Any hint?

[Iain McNeill]

Was the impedance measured in the dipole or one speaker on it's own?

Did you use peak magnitude or zero phase?

[Raka]

The impedance was measured in the same setup as the SPL.

What do you mean with peak magnitude or zero phase??

[Iain McNeill]

sometimes there can be a difference between the peak in the impedance magnitude curve and the 0° crossing of the impedance phase curve. Some say that the zero phase crossing is a more accurate measure of Fs. Do you have enough resolution on your impedance measurement? What are the frequency steps?

Do you see a difference between the frequency of the peak resonance impedance and the zero crossing of the phase response? Does either agree with the measured 75Hz roll off.

If not then something else is at work here. How does the impedance test differ from the SPL test? There must be a difference somewhere.

It's time for the Sherlock Holmes hat

[Raka]

Ok, I didn't know there can be that difference. I look sometimes to the peak, but mainly on the phase as it's easier for the eye.
The impedance measurement is attached. I used different sample sizes and resolution but there were no significant deviations. You can see that the zero and peak are at the same frequency.
With this impedance test, the Q is about 0,57, Fs=54Hz. The transfer curve of that system is plotted in my first post, together with the actual SPL measurement nearfield.
Looking for measurement mistakes, I have calibrated the impedance setup and is getting good results with resistors.

Is correct that a nearfield measurement is showing just the transfer curve of the driver, without any other superimposed like baffle step, or any other?

[Iain McNeill]

If the speaker was mounted and tested the same in both tests then the data should agree.

The impedance plot looks like a curve of a single driver but you say the SPL is a dipole - you mean two speakers?

Drive level can affect Fs too with higher driver levels lowering Fs. What was the voltage across the speaker in both tests?

[sreten]

Hi,

The driver impedance indicates Fs = 54Hz, and Q ~ 0.57.

The measured response (nearfield) seems to follow Fs = 75Hz ?

I've had a look around and no-one seems to measure the bass
response of the driver in a baffle, they simply trust the derived
parameters and then measure the actual response in a box.

Nearfield measurements for a closed box are accurate to DC,
but I cannot find any criteria for open baffles, which must go
wrong at some point, i.e. they cannot measure DC .....

I might be barking up the wrong tree ... how big is the bipole ?

/sreten.

[Eva]

There is probably some cancellation taking place even with the microphone so close to the driver, these wavelengths are very long. Try measuring it in a sealed box or in infinite baffle conditions.

[AndrewJ]

What method are you using to derive the driver parameters? The original method proposed by Small can be very unrliable.
The best method is to fir the measured impedance curve with an equivelent circuit model of the driver. It is crtitical for an unbaffled driver that this model includes a lossy inductor to simulate suspension creep.
With the correct model, I usually would expect to get a match between nearfield measurement and calculated response to within about 0.5dB.
I have not recent data on an unbaffled driver t show you, so I will try and find time tomorrow to do this.
You should bear in mind that drive level will affect your measurements. It is important to match the level when measuring the frequency response to that used when measuring the impedance. If you derive the impedance measurement by putting a large value resistor in series with the driver then you will significantly alter the drive level across the speaker, and this can greatly affect the results.

Andrew

[phase_accurate]

There is probably some cancellation taking place even with the microphone so close to the driver, these wavelengths are very long. Try measuring it in a sealed box or in infinite baffle conditions.

Usually the accuracy is excellent at low frewuencies when measuring that close (apart form the baffle step that is missing in that measurement). At higher frequencies you will usually have cancellations and such with the nearfield measurement.

Another thing that males me wonder: Do you know the accuracy of your microphone, mic preamp etc ?

Regards

Charles