T/S on DVC sub issues

[Caio Ferrari]

Hi
I've been using REW to measure a DVC sub (JBL GTO1014d) and I'm experiencing and issue.
The sub has two 4 Ohm coils. When I measure the coils in parallel I'm getting some differences compared with coils in series.
Fs, Qms are pretty close.

Qes is 0,7 in series and 0,52 in parallel (should be the same, right?)

Qts is 0,65 in series and 0,51 in parallel.

Vas is 18L in series and 14L in parallel.

I built the exactly that test Jig recommend on REW manual (100R resistor) and I'm using a Behringer UCA202 as sound card.

What should I do?

 

[John Mulcahy]

The UCA202 has some issues with results showing rising impedance at low frequencies, below 30 Hz or so. You can see some examples in this thread. If you attach an mdat file with your measurements we can see how it looks.

 

[Caio Ferrari]

But, my driver Fs is about 37 Hz. Should it presenting problem?

As soon as get my PC I'll come back here to attach me measurements.

 

[Caio Ferrari]

Here, the two measurements made with voice coils in series and parallel

Series Re = 7,1R
Parallel Re = 1,7R

 

[John Mulcahy]

The measurements look OK, the impedance rise from the UCA202 doesn't seem to be affecting the model fit much so the results look valid.

 

[Caio Ferrari]

The measurements look OK, the impedance rise from the UCA202 doesn't seem to be affecting the model fit much so the results look valid.

That's great but, why I'm getting difference in Qes/Qts when I'm connecting voice coils in series or parallel. It should be the same, right?

 

[John Mulcahy]

Don't think so. http://users.on.net/~isaacmcn/audio/dvc/dvc.htm

 

[Caio Ferrari]

No, it really should be the same. Parallel or Series must have the same Qes. Only if I measured one voice coil I would have a Qes difference by a factor of two and I never measured this way.

 

[John Mulcahy]

How precisely matched are the voice coils? How similar is the excursion for the two test configurations? Do you have an interface with a little more drive capability? Laptop headphone out/line in can also be used if available.

 

[Caio Ferrari]

How precisely matched are the voice coils? How similar is the excursion for the two test configurations? Do you have an interface with a little more drive capability? Laptop headphone out/line in can also be used if available.

They have the same graph.
Unfortunately I don't have another interface.

I'd like to ask why the results are so different depending of the frequency range of the measurements.

Since I'm stuck with Qe and Qt value, I ran some tests between 15-100Hz. I got Qt of 0,45.
When I ran the test between 15-20000 Hz, Qt rises to 0,6

 

[John Mulcahy]

If you only provide a limited range of data the model fit will be poor, with no data beyond the motional impedance range the fit for the electrical impedance has little to work with.

 

[Caio Ferrari]

If you only provide a limited range of data the model fit will be poor, with no data beyond the motional impedance range the fit for the electrical impedance has little to work with.

Got it.
But the problem is, when I provide a limited range, the results mach the manual measurement and the speaker datasheet.
When I provide the full range, the Qes and Qts rises too much.

 

[John Mulcahy]

For the manual measurement and the fit with limited data the electrical impedance model is largely ignored. The result may be closer to what was expected, but that doesn't mean it is correct. It is common for measured parameters to differ significantly from the datasheet nominal values. You can use the same connection setup with ARTA's LIMP to get another result.

 

[Caio Ferrari]

For the manual measurement and the fit with limited data the electrical impedance model is largely ignored. The result may be closer to what was expected, but that doesn't mean it is correct. It is common for measured parameters to differ significantly from the datasheet nominal values. You can use the same connection setup with ARTA's LIMP to get another result.

I already did it. Limp gave me a Qt 0,4

 

[Caio Ferrari]

For the manual measurement and the fit with limited data the electrical impedance model is largely ignored. The result may be closer to what was expected, but that doesn't mean it is correct. It is common for measured parameters to differ significantly from the datasheet nominal values. You can use the same connection setup with ARTA's LIMP to get another result.

I just did a few measurements here from 20Hz-20kHz:

Coil 1: Qt 0,928 (Calculation for two coils: 0,464)
Coil 1: Qt 0,928 (Calculation for two coils: 0,464)
Parallel: Qt 0,527
Series: Qt 0,603

 

[John Mulcahy]

When you say "here" do you mean with LIMP? What are the full LIMP results?

 

[Caio Ferrari]

When you say "here" do you mean with LIMP? What are the full LIMP results?

The measurements above I did using REW.

With limp I'm getting the same Fs and Qms as REW.
Qt is 0,38 on "regular" mode and 0,45 using LSE minimization.

 

[John Mulcahy]

So the issue is with Qes, and so with the estimate of total RE which includes the DC resistance and a dR value from the blocked impedance model fit. The dR value is quite high in the measurements you posted, likely influenced by the behaviour of the UCA202 at low frequencies. If I manually force a dR value of zero during the model fit your posted measurements give

series: Qms 6.339, Qes 0.588, Qts 0.538
parallel: Qms 5.681, Qes 0.524, Qts 0.479.

Some have modified the UCA202 to improve that, which might help. Otherwise you could plug your own RE figure into QES = omegaS*CMES*RE.

 

[Caio Ferrari]

So the issue is with Qes, and so with the estimate of total RE which includes the DC resistance and a dR value from the blocked impedance model fit. The dR value is quite high in the measurements you posted, likely influenced by the behaviour of the UCA202 at low frequencies. If I manually force a dR value of zero during the model fit your posted measurements give

series: Qms 6.339, Qes 0.588, Qts 0.538
parallel: Qms 5.681, Qes 0.524, Qts 0.479.

Some have modified the UCA202 to improve that, which might help. Otherwise you could plug your own RE figure into QES = omegaS*CMES*RE.

I applied tihs (QES = omegaS*CMES*RE.) and got a Qts 0,55. Closer to the manual method but a little far from Limp.

 

[John Mulcahy]

If you are ultimately interested in VAS then probably best making added mass measurements so the VAS can be calculated from the full model fit.