Repository navigation
Measurement results. Speakerbench results assessment. How to determine the correct measurements. #6
Description
Hello. I measured the Sonido SCW 175-8 driver with different sets of added masses. In one case, I received an error of 0.59 at 200 mV. In the second case, the error is 0.43 at 200 mV. But it seems to me that both did not achieve the goal. The goal was to get the parameters as close to the factory as possible. A little later I will publish measurements.
Activity
TimTom52 commentedon Feb 16, 2025
In the time since my post, I have already made 5 sessions with five different sets of added masses. And only on the fifth set it was possible to get as close as possible to the factory parameters. Why am I trying to do it? The mass of the moving system when assembling the driver can be measured on an accurate scale.. Plus the exact weight of glue. Thus, we have a parameter that is not calculated, does not change much from time, temperature and humidity. So, the coincidence of MMS means an accurate measurement. Usually, with such a coincidence, the rest of the parameters become very close to the factory ones.
I am interested in your opinion on this matter.
First, a comparison. Then the files for this measurement. Error 0.79 for 400 mV. But if you take the best measurement, with an error of 0.61 for 600 mV, then it seems to me worse in terms of the coincidence of the models and Bl.
Z320250216_193639_400_mV_err_0,79.json
ADV20250216_193707_400_mV_err_0,79..json
SBD20250216_194315_400_mV_err_0,79.json
Z320250216_192704_600_mV_err_0.61.json
TimTom52 commentedon Feb 23, 2025
At one of the sets of added masses, I received an estimate excellent from Speakerbench with an error of 0.31 per 100 mV, and 0.43 per 200 mV. But I do not consider these measurements the best.
ADV20250210_164206_200_mV_err_0.43.json
ADV20250210_164835_100_mV_err_0.31.json
Z320250210_164121_200_mV_err_0.43.json
Z320250210_164810_100_mV_err_0.31.json
In the graphs of BL and Model Test, a pale pink band is drawn. It can be wider and narrower. What does it mean?
jcandy commentedon Feb 24, 2025
"So, the coincidence of MMS means an accurate measurement. Usually, with such a coincidence, the rest of the parameters become very close to the factory ones."
Generally speaking, it is problematic to expect precise agreement with factory parameters.
Voltage dependence
As you yourself have shown, the compliance (and so the resonant frequency) changes with applied voltage, and manufacturers do not generally state what this voltage is.
Mms
There is a distinction here between the actual moving mass and the effective moving mass. An oscillating piston in air experiences a so-called radiation impedance. For the typical low frequencies of interest, the effect of radiation impedance looks much like an additional mass added to the diaphragm. However, this mass depends on the baffle geometry in a mathematically complicated way. To a good approximation, when you measure the driver unbaffled (in free air), the extra radiation (air) mass is equal to the mass on one side of a baffled piston. So, the Speakerbench Mms is the moving mass plus the air mass on the front of the cone. In Speakerbench, the box model will add to this an internal effective mass that depends on the box geometry (you may notice the various box geometry inputs).
Creep compliance
The standard Thiele-Small parameters give an approximate description of the driver. The advanced parameter model provides a more precise description of the driver via the addition of extra parameters. The point of Speakerbench is to compute the Advanced model parameters using three impedance sweeps (you are now an expert at this!). FOr convenience, Speakerbench also provides effective values for the classic Thiele-Small parameters (Cms, Rms) which are derived from the advanced parameters (C0,R0, beta), but these effective parameters are not unique.
jcandy commentedon Feb 24, 2025
"In the graphs of BL and Model Test, a pale pink band is drawn. It can be wider and narrower. What does it mean?"
The pink band indicates the frequency range for the motional impedance fit (Mms, C0, R0, beta). The electrical parameters are fit over a different (higher-frequency) range. This region is chosen iteratively to minimize error.****
TimTom52 commentedon Feb 24, 2025
Thanks for the answer. The manufacturer does not indicate the measurement voltage, this is not good. I assume that the manufacturer assumes that the driver will be used on average at an average volume or slightly lower. I would count on it. And this is 1.5-0.9 volts. Judging by several of my measurements, and looking at the factory parameters, the resonance frequency of the driver in factory measurements is 1-2 hertz lower than in my measurements by 0.2 volts. In the maximum, I tried 0.8 volts (except for the earliest measurements, before our conversations), and then the resonant frequency differed by 0.5-1 hertz. So two drivers having factory parameters were measured by about 1 volt. But in my conditions, it is not possible to get the correct beta at 1 volt. I understood your advice, do not try to follow the factory parameters. But if once I can make good measurements on 1 volt, it will mean that I definitely learned this matter.
"So, the Speakerbench Mms is the moving mass plus the air mass on the front of the cone." In your experience, how do these masses compare on average? Does it depend on the size of the membrane? A large membrane and a strong motor will be able to push a large mass of air easily. But it can be equal, because for a small membrane, the mass of air in front of it is also small. Are you familiar with the AKABAK program? Does it model the driver's behavior accurately enough?
I asked about the pale pink band on the graphs because I was trying to find some signs that clearly spoke about the quality of the measurement.
jcandy commentedon Mar 10, 2025
Yes, fs will typically drop as measurement voltage increases. The air mass on the front of a cone when mounted on a baffle depends on the baffle geometry. On average, this mass is very roughly 1% of the driver mass (we will print accurate estimates of the various acoustic masses in a soon-to-be released major update of Speakerbench).
Also, SEAS measures at 2Vrms
I doubt AKABAK could provide an alternative tool to model what we are discussing. But perhaps you could clarify why you are unhappy with the present level of agreement. Your earlier scans of voltage were extremely accurate (perhaps more accurate than what manufacturers are doing).
TimTom52 commentedon Mar 11, 2025
"But perhaps you could clarify why you are unhappy with the present level of agreement."
Measurement on low voltages gives accuracy for several parameters, which cannot be obtained at higher voltages. But the resonant frequency at low voltages will differ from the resonant frequency at higher voltages by 2 hertz. Knowledge of the resonant frequency is necessary to calculate the volume of the box and the port of the phase inverter, for large drivers. The difference of 2 hertz is too big so as not to pay attention to her. I am sure that in 99% of cases the measured driver will work at a larger or much larger input voltage. That is, the calculated box for the driver will be bad.
By your link, R&D team measures the Seas driver. Are these the same people who created AKABAK?
jcandy commentedon Mar 11, 2025
The voltage dependence is a fact of life, and probably one of the easier challenges of successful design. In any case you can compare a low-V fit to a high-V fit to see the expected variation in the box SPL. Note that in the upcoming Speakerbenech release, you will be able to save and load a so-called TUNE file, which will store all the box model settings so you can easily load the box setting at the start of a session.
If you plan on designing vented systems, the problem becomes much worse because the physics of the vent is poorly understood . You can also expect level dependence of the vent, and will want to build prototypes to know exactly how the system will behave.
TimTom52 commentedon Mar 11, 2025
Yes. This should help. Thank you.
If you are familiar with VituixCad, can I ask you about the enclosure tool? I am interested in the behavior of some quantities, or rather graphs. In the context, what is worth achieving in order to get a good result. For example, through direct messages on the REW support forum.
TimTom52 commentedon Mar 26, 2025
Hello. I've already seen the Speakerbench update. Congratulations.
Can you take into account the displacement of the resonant frequency depending on the applied voltage in the box simulator?
To be precise, it is possible to measure the impedance without weights at multiple voltages when measuring the driver. Minimum (0.1 V), maximum (3 V or 14.15 V) and several intermediate to understand how the resonance changes for a given driver.
jcandy commentedon Mar 31, 2025
Good question. There is no automated process in Speakerbench for building in the V-dependence of parameters. Actually, I am not aware of any tool that does this. That said, with the Speakerbench advanced model you have a sufficiently accurate framework to precisely calculate the voltage dependence of the creep compliance. You did a great job of carrying out the required measurements to compute the V dependence. How to perfect this process (and without having to add mass as you point out) would be a nice project but is not something we are currently prioritizing.
TimTom52 commentedon Apr 13, 2025
I uploaded some of the previously made measurements to Speakerbench. There is a difference in the assessment of the quality of measurements before and now. For example, one of them. When I first uploaded it (before the Speakerbench update), the error was 0.36. Now it is 0.49. Were there any changes in the way measurements were evaluated during the Speakerbench update?
jcandy commentedon Apr 13, 2025
Can you please attach the Z file so I can check? We improved the data handling for cases with questionable data, but for good data the answer should be unchanged.
TimTom52 commentedon Apr 13, 2025
Z320250210_164512_150_mV_err_0.36.json
jcandy commentedon Apr 14, 2025
Please try again. It should work now.
TimTom52 commentedon Apr 14, 2025
Yes. Loaded about 8 measurements. Everything is as before. Thank you.
I upload the sbd file to Datasheet, then change the aspect ratio of the box. For example, there were 1 and 1. I do 1.5 and 1.5. I click apply. In the info tab, the three box sizes were the same, and remain the same after my change. Is this how it should be?