Thiele-Small Parameters measurement using dual added mass method.

Vladimir_1953

New Member
Joined
Jul 5, 2018
Messages
39
My AV System  
Preamp, Processor or Receiver
ROTEL RC-990
Main Amp
ROTEL RB-991
Universal / Blu-ray / CD Player
DENON DCD-2000AE, ROTEL RCD-991AE
Front Height Speakers
Dynaudio Crafft
What is the final ratio of the mass of the additional mass to the size of the speaker?
10 grams per 1 inch?
 

Bernard

Member
Thread Starter
Joined
May 22, 2017
Messages
108
Very good. Thanks for using Speakerbench, and thank alot to John for facilitating it.
Some measurements have been made since then. The main ones are gathered in the attached table.
I continue to measure with REW and send the ZMA curves to SpeakerBench.
The "Fit quality" tab is very handy because it allows to determine the right weight value.
However, this does not prevent you from asking questions.
Should "Error (fail > 1)" and "Zero-mass" be as low as possible?
Should "Bℓ" be as high as possible in the parameters?
Which are the main criteria ?
Fit quality table.png


5 Json files attached : 34g, 34.8g, 35g, 35,2g, and 36g.
35 g seemed the best M2 value to me, but it might be 34 g. Not easy to determine when you're not an expert.
When the weather will be less hot, I will try around 34 g, not 34°C.

27/06/2020 :
I added the corresponding mdat files. You can see an anomaly between 5 and 15 Hz. The anomaly disappears if you measure with the single mass method, M1 or M2.
The "damm" gives good results but it is more tedious than the other methods. However... You have to respect the M2 -> M1 -> M0 measurement procedure and remove the weights very carefully (which I may not always have done...).
You might want to try this new way of measuring... or not!
 

Attachments

Last edited:

sm52

Member
Joined
Mar 14, 2019
Messages
462
speakerbench.com is very slow in generating json files. So always or there is needed browser settings to speed up?
 

Bernard

Member
Thread Starter
Joined
May 22, 2017
Messages
108
speakerbench.com is very slow in generating json files. So always or there is needed browser settings to speed up?
Speakerbench is not slow. Your files are too big as mine were.
Please download my files from message #27.
You'll get information. ;-)
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
5,434
The model fit looks poor in REW, with a big difference between the measured and modelled peak heights. That suggests there is something about the measurements that is not consistent. Is the model fit better in SpeakerBench? There is a graph that shows the measured and simulated curves.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
5,434
They also show big differences in peak heights though. How accurate are your scales? Even small errors in the mass figures have a big influence on the results.
 

sm52

Member
Joined
Mar 14, 2019
Messages
462
The balance is accurate to two decimal places. I first put all the weights on the scales - 8 pieces, looked at the readings of the scales and divided by 8. So I got the weight of one weight. Attached to the membrane by two. So the error in the weight of the weights is no more than 0.05 g. More likely no more than 0.025 g. Calibration and measurements in REW were done at -12 dB. Laptop battery power + external sound card Audient iD14. No network interference. Maybe try to measure at -6 dB? Or 0 dB? Or add an amplifier to the circuit? At what levels are measured in professional laboratories?
 

Bernard

Member
Thread Starter
Joined
May 22, 2017
Messages
108
What is the position of the loudspeaker during measurement?
Attached is my measuring stand.
The stand is in the middle of the room.
The centre is 125 cm (4.1 ft) from the floor and ceiling.
There is no draught or vibration in the room.
A monastic atmosphere! ;-)
I also noticed that you have to remove the masses very delicately with fairy fingers...

Portique de mesure hp.png
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
5,434
The balance is accurate to two decimal places. I first put all the weights on the scales - 8 pieces, looked at the readings of the scales and divided by 8. So I got the weight of one weight.
Why do you need the figure for one weight? The weights are unlikely to be identical. Follow the advice on SpeakerBench and measure the full set of weights as used for the m2 measurement, then the half set you use for the m1 measurement, and check the weights again when you remove them. Make the measurements in the order m2, m1, free air.

Calibration and measurements in REW were done at -12 dB. Laptop battery power + external sound card Audient iD14. No network interference. Maybe try to measure at -6 dB? Or 0 dB? Or add an amplifier to the circuit? At what levels are measured in professional laboratories?
TS measurements should be done with a small enough signal at the driver to stay in the linear range but large enough to get good signal to noise ratios. Typically the voltage at the terminals should be 100 mV. You could check that by running the sine wave generator at 200 Hz or so and measuring the voltage across the driver terminals with a multimeter.
 

sm52

Member
Joined
Mar 14, 2019
Messages
462
Why do you need the figure for one weight?
This is how I explained that the weight of the weights was measured accurately. For the measurement, 4 pairs of weights were used. Same as 4 weights. I performed the measurements according to the SpeakerBench instructions. But the question is not how the measurements were taken. The question is why REW and SpeakerBench calculated Thiele-Small parameters differently. The measurement files were the same.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
5,434
The question is why REW and SpeakerBench calculated Thiele-Small parameters differently. The measurement files were the same.
Different choices in the curve fits. SpeakerBench compromises the electrical fit a little to optimise the mechanical fit, REW doesn't. Neither manages a good fit to the data though as you can see from the differences between the measured and simulated responses. That is because your measurements are not self-consistent, as you can see if you try to calculate using single added mass for free air and 5.4 g, which gives Mms 11.84 g, then free air and 10.8 g, which gives Mms 12.61 g.
 

sm52

Member
Joined
Mar 14, 2019
Messages
462
I uploaded driver impedance measurements to speakerbench.com to check the quality of the measurement. All data has been loaded. A .json file has been created. But when I try to download it, the browser crashes. Tried many times. I haven't been able to download. Can someone check the download of the .json file?
 

Bernard

Member
Thread Starter
Joined
May 22, 2017
Messages
108
I uploaded driver impedance measurements to speakerbench.com to check the quality of the measurement. All data has been loaded. A .json file has been created. But when I try to download it, the browser crashes. Tried many times. I haven't been able to download. Can someone check the download of the .json file?
Hello,
No problem in France. It takes longer than usual to upoad but then downloading the json files is no problem.
What country are you in?
 

andyc56

Member
Joined
Jun 2, 2017
Messages
45
On a related topic, people may be interested in the Speakerbench youtube channel, which has some helpful videos about Speakerbench and measurement techniques.

In addition, there's a really interesting paper by Claus Futtrup and Jeff Candy (in PDF format) called Physical Accuracy and Modeling Robustness of Motional Impedance Models, comparing the accuracy of various advanced T/S models, including the one supported by Speakerbench. It's well worth a read for those interested in the topic.
 

trobbins

Member
VIP Supporter
Joined
Nov 26, 2017
Messages
96
Location
Australia
Do you have reference weights that you know the mass of? Are they part of the scales system? Are you using a balancing scale, or ?
 

sm52

Member
Joined
Mar 14, 2019
Messages
462
Do you have reference weights that you know the mass of?
Yes. I have reference weights. The measurements are normal. REW calculated the Thiel-Small parameters. I wanted to compare them with the parameters that Speakerbench will calculate.
 

trobbins

Member
VIP Supporter
Joined
Nov 26, 2017
Messages
96
Location
Australia
I note that your parameter window in post #30 showed masses of 10.8g and 5.4g - does that mean you had measured those weights to an accuracy of +/- 0.05g. or +/-0.1g or ? and included the weight of any adhesive product used to affix each weight to the cone ?
 

sm52

Member
Joined
Mar 14, 2019
Messages
462
Yes. My scales are accurate to two decimal places. And I measured their weight along with the scotch tape they are attached to, and with which they are attached to the driver membrane. But now we are not talking about the tasks described in post #30. Now I am making new measurements. Are you able to get the .json file on the speakerbench after uploading the zma files and reference weight values?
 

trobbins

Member
VIP Supporter
Joined
Nov 26, 2017
Messages
96
Location
Australia
My interest and comment was just about that aspect of adding known weights and the likely tolerance of your final thiele small parameters due to uncertainty of the mass of the weights you used. Not everyone has a recently calibrated scale to 0.01g accuracy, or a reference weight that has been calibrated to that level of uncertainty.
 

sm52

Member
Joined
Mar 14, 2019
Messages
462
Java is not installed on my system. Could this be causing firefox to crash while using the speakerbench site?
 

Bernard

Member
Thread Starter
Joined
May 22, 2017
Messages
108
But when I try to download it, the browser crashes.
Would you be so kind as to send us a screenshot with the error message(s) and the three files M0, M1 and M2 so that we can carry out the same operations as you on the speakerbench site ?
 
Top Bottom