Speaker Impedance measurement ,circuit and sound card output

[John Mulcahy]

If you look at the plot of simulated impedance versus measured impedance for the added mass measurements you will see whether the fit is good, with accurate masses and good measurement the simulation is a very close match to the measurement. If there are significant differences in the position and height of the peak that suggests the mass figures are incorrect. A more accurate scale is a good idea.

 

[icbcodc]

If you look at the plot of simulated impedance versus measured impedance for the added mass measurements you will see whether the fit is good, with accurate masses and good measurement the simulation is a very close match to the measurement. If there are significant differences in the position and height of the peak that suggests the mass figures are incorrect. A more accurate scale is a good idea.

I notice there are two peaks in the no mass measurement plot.

there are two small peaks in the 7.1g mass measurement plot.

two tiny peaks in 14.1g measurement plot.

I will buy a new scale with 0.01g acuracy,and run the measurement again.

 

[John Mulcahy]

Those are typical minor cone resonances, nothing to worry about. I was meaning differences like this for the 14g measurement previously posted with a dual added mass fit:

Doing single-added mass provides a good fit, but the fit parameters are different for the 7g measurement, which suggests it is the mass values which are not correct.

 

[icbcodc]

Those are typical minor cone resonances, nothing to worry about. I was meaning differences like this for the 14g measurement previously posted with a dual added mass fit:

View attachment 32372

Doing single-added mass provides a good fit, but the fit parameters are different for the 7g measurement, which suggests it is the mass values which are not correct.

View attachment 32373

Yes， the value of the mass on my sacle is always changing in around 0.2g range each time I put the mass on the scale。

I just use the value displayed most frequently, so I have to try many times.

I have just bought 3 scales with 0.01g acuracy half an hour ago online, price from 7 dollars to 50 dollars, I will choose the best one or refund for all in 7 days.

 

[icbcodc]

Those are typical minor cone resonances, nothing to worry about. I was meaning differences like this for the 14g measurement previously posted with a dual added mass fit:

View attachment 32372

Doing single-added mass provides a good fit, but the fit parameters are different for the 7g measurement, which suggests it is the mass values which are not correct.

View attachment 32373

Hello, John

First new scale is received.

I realize the simulation and measurement last time didn't provide a good fit.

T/S parameters last time:

This time, I use this new scale, 0.01g accuracy (around 8 dollors).

Simulation in the plot seems better.

BTW, Can I use this way to confirm whether the scale is accurate enough?

New T/S parameters:

 

[John Mulcahy]

The accuracy of the fit is probably a good indicator. You should check the fit for the added mass measurements, particularly around the peak.

 

[icbcodc]

The accuracy of the fit is probably a good indicator. You should check the fit for the added mass measurements, particularly around the peak.

With more mass, the peaks fit worse.

No mass

with 7.00g mass, not fit clearly at the peak.

with 14.02g mass, NOT FIT.

 

[John Mulcahy]

Those look better than the original measurements, but worth seeing if the other scales give better results.

 

[icbcodc]

Those look better than the original measurements, but worth seeing if the other scales give better results.

Hello John,
I used my new scale for a few hours and found it's a bit more accurate scale than the cheaper one,but this more expensive scale (around 50 dollars) did not cause a more accurate simulation.

I'm disappointed by this scale.

T-S parameters:

BTW, is the simulation plot calculated only at the first time when I'm calculating the T-S parameters?

 

[Breeman]

Not to distract from the essence of this topic. I would suggest to try simulate the box responses based on both sets of T/S parameters to see if the differencies are noticeable. It may be that the differencies are not really worth all the effort to improve the accuracy of the T/S parameters in case they are in the ball-park already. Just my 2 cents.

 

[John Mulcahy]

It is probably best to make one set of measurements, noting which masses were used, then measure those masses on the different scales and see which set of mass figures gives the best results for that measurement. There will be variations between measurements, so sticking with one measurement reduces the variables. Breeman's advice is good though, make sure the variations in the resulting parameters are sufficient to be worth the effort of trying to increase the accuracy.

 

[icbcodc]

Not to distract from the essence of this topic. I would suggest to try simulate the box responses based on both sets of T/S parameters to see if the differencies are noticeable. It may be that the differencies are not really worth all the effort to improve the accuracy of the T/S parameters in case they are in the ball-park already. Just my 2 cents.

Hi Breeman
Thank you for your advice, it is a good way.

I 'm trying to use WINISD(latest version can be downloaded here) cause it's simple, do you recommend any other enclosure simulating software?

I will start the test from the T-S parameters in my POST 63 using cheapest scale.

This is the T-S parameter I post before in POST 63, FYI.

WinISD will caculate some values in colar blue,this assistant function is recommended by it. I checked the blue values, their corresponding values in REW T/S parameters are almost same.

Some values are not available, I just leave them blank.

Not measured speaker, just leave it blank.

I saved the parameters.

When I selected it for box simuation I found an error message. Does it matter?

Simulated enclosure VOLUME 11.75, QTC 0.971

 

[icbcodc]

Not to distract from the essence of this topic. I would suggest to try simulate the box responses based on both sets of T/S parameters to see if the differencies are noticeable. It may be that the differencies are not really worth all the effort to improve the accuracy of the T/S parameters in case they are in the ball-park already. Just my 2 cents.

Second scale is Kubei 0.01g accuracy, price around 8 dollars.

T-S parameters according to New T/S parameters in POST 80

Creat New Driver with T-S parameters:

Enclosure simulation project creating:
This time, T/S parameter integrity checks ok.

Select closed box. EBP 84.5

Everytime I use "1.000" for different project to compare.

Box volume: 11.29L , QTC 0.971

3 dollars scale and 8 dollars scale in closed box simulation make no difference.

 

[icbcodc]

Not to distract from the essence of this topic. I would suggest to try simulate the box responses based on both sets of T/S parameters to see if the differencies are noticeable. It may be that the differencies are not really worth all the effort to improve the accuracy of the T/S parameters in case they are in the ball-park already. Just my 2 cents.

Now use the T-S parameters created by the disapointed scale (50 dollars,brand Yiheng) to simulate the closed box.

T/S parameters come from POST #84, the lastest impedance measurement yesterday.

T-S parameters input:

Closed box simulation, new project:

Box volume: 36.75L (totally different) , QTC 0.986

3 dollars scale and 8 dollars scale in closed box simulation make no difference. but 50$ scale makes big different.

So which one is correct simulation? maybe the cheap scales are better?

 

[icbcodc]

It is probably best to make one set of measurements, noting which masses were used, then measure those masses on the different scales and see which set of mass figures gives the best results for that measurement. There will be variations between measurements, so sticking with one measurement reduces the variables. Breeman's advice is good though, make sure the variations in the resulting parameters are sufficient to be worth the effort of trying to increase the accuracy.

Hi John

I made the mass into four pieces, and divided them into two groups, group one including all mass for one measurement, group two including only 2 pieces of mass for the other measurement.

First, I did three measurements, and input the mass value get from Ying heng scale (price 50$), it looks like a more accurate scale.

mass one : 7.01g
mass two : 14.01g

No warning message.

But the simulation is bad with 14g mass

it looks not too bad for 7g mass simulation.

It looks perfect no mass added.

group one mass 7.01g

group two mass 7.00 , used with group one together running measurement.

7.01g mass is put left and right, this is also for 7.01g measurement.

7.00g mass is put up and down

7.01g mass get 7.00g number on KUBEI scale(cheaper one).

7.00g mass get 7.01g number on Kubei scale(cheaper one)

the saddest thing is the value keep going down.

I know this KUBEI scale is not accurate like the expensive one.

But the expensive one looks good,but the simulation result is bad.

I'm frustrated,anything wrong? Should I calibrate the rig each time before I measure the impedance?

Is it possible the enviroment for measurement is too noisy?

 

[John Mulcahy]

Looks like you cannot trust the scale which drifts. It may help to run a sine wave at the sweep level for a period before starting each measurement, to make sure the driver is in the same condition for each run and try and remove any effect of shifting the cone when adding/removing masses. Also try starting the measurements at 10 Hz rather than 20 Hz, so there is a little more data for the curve fit to use.

 

[icbcodc]

Looks like you cannot trust the scale which drifts. It may help to run a sine wave at the sweep level for a period before starting each measurement, to make sure the driver is in the same condition for each run and try and remove any effect of shifting the cone when adding/removing masses. Also try starting the measurements at 10 Hz rather than 20 Hz, so there is a little more data for the curve fit to use.

I tried again from 10hz to 20000hz and added a sweep before each measurement.

But the result is still not very good.

perfect simulation when no mass

Not that good 7.02g mass

Bad result.

I double check the mass after I get it down from the speaker, I believe the weight value is accurate, but the simulation result with more mass is worse.

Something still wrong ?

 

[John Mulcahy]

The measured phase looks odd on the 14g measurement, I wonder if the added mass is still a little too much, or perhaps the masses were not quite symmetric? Worth trying with smaller masses.

 

[icbcodc]

The measured phase looks odd on the 14g measurement, I wonder if the added mass is still a little too much, or perhaps the masses were not quite symmetric? Worth trying with smaller masses.

I think they are symmetric on the speaker cone.

and each piece is around 3.51g。

I'm preparing to use 80% of MMS as first mass, and 64% of MMS as second mass, is that O.K.?

 

[John Mulcahy]

Why not make a big step to 50% and 25% and see what difference that makes?

 

[icbcodc]

Why not make a big step to 50% and 25% and see what difference that makes?

I'm reading the manual made by Jeff Candy and Claus Futtrup, and try their online app using their sample data, because I want to find some clue about the relation between MMS and Added Mass qulity.

I download the sample measurement files from the the MANUAL page, and generated a json file needed later in the COLLECT page and download the json file.

and upload the json file in FIT page, calculating automatically and pass.

There are some errors but still pass.

The MMS of the speaker is around 14.8959g

You can see the first mass is 8.017g, the second one is 16.048g, and 16.048g > 14.8959 (Mms)

Zoom in the impedance plot, at the peaks, the simulation is also bad when added mass is 16.048g.

It's not better than the simulation in REW.

I'm not sure whether this is caused by the mass quality which is larger than Mms.

 

[Bernard]

I'm reading the manual made by Jeff Candy and Claus Futtrup, and try their online app using their sample data, because I want to find some clue about the relation between MMS and Added Mass qulity.

I download the sample measurement files from the the MANUAL page, and generated a json file needed later in the COLLECT page and download the json file.

and upload the json file in FIT page, calculating automatically and pass.

There are some errors but still pass.

View attachment 32623

The MMS of the speaker is around 14.8959g

View attachment 32624

You can see the first mass is 8.017g, the second one is 16.048g, and 16.048g > 14.8959 (Mms)

View attachment 32625

Zoom in the impedance plot, at the peaks, the simulation is also bad when added mass is 16.048g.

It's not better than the simulation in REW.

I'm not sure whether this is caused by the mass quality which is larger than Mms.

View attachment 32626

Something's wrong. Your quality analysis is the same as the Seas driver.
Can you send your 3 impedance measurement mdat files ?

 

[icbcodc]

Something's wrong. Your quality analysis is the same as the Seas driver.
Can you send your 3 impedance measurement mdat files ?

There isn't any mdat file. I didn't do any measurement in POST 96.

Because in POST 96, I JUST downloaded and uploaded the sample measurement files provided by "speakerbench" and then generated online T/S parameters and simulation plot.

 

[Bernard]

There isn't any mdat file. I didn't do any measurement in POST 96.

Because in POST 96, I JUST downloaded and uploaded the sample measurement files provided by "speakerbench" and then generated online T/S parameters and simulation plot.

There's a misunderstanding.

The first step is to create a first ZMA file from your existing MDAT file with the mass of 0 g,
then a second ZMA file from your existing MDAT file with the mass of 7 g,
and finally a third ZMA file from your existing MDAT file with the mass of 14 g.

The second step is to upload your 3 ZMA files into the site Speakerbench using the Collect tab. Speakerbench will create your JSON file from your data, named Zyyyymmdd_hhmmss.JSON .

The third and last step is to use this personal JSON file to get the quality fit report.
If you are unable to create the ZMA files, attach the mdat files with the 3 weights.

 

[John Mulcahy]

He was just trying with the sample data first to see how it behaved.

Although the peak heights with the sample data aren't an exact match the measured and simulated responses are quite well aligned in frequency. I did discuss the fit with Claus and Jeff while implementing the dual added mass method, whilst the Ritter 3PC model does a good job of modelling the frequency dependence of compliance there remain other aspects of the driver's behaviour that aren't reflected in the model, hence the less than perfect fit. It is nonetheless much better than the basic TS model.