But CMRR is handled by the input analog OP-Amp circuit, before it sees the A/D, unless I’m misunderstanding the circuit.
If different levels, Left to Right is an issue, a 40 dB pad could be added to the speaker voltage input to bring them to similar levels, regardless of which side the sense resistor is on.
Hi @Kravchenko_Audio, release 1.192 is up HERE
This still needs some more polishing, but I think it would be helpful for feedback from you and others in case something needs to change. In short, the plug-in will perform two sweeps if you select the “Do Added Mass Measurement,” and you’ll be prompted to add the mass before the second sweep.
The checkbox for Display Speaker Parameters has been fixed and is now displaying the previously missing parameters. There is a new checkbox called “Open Report” (see below). The plug-in will always generate an HTML report named MyDocs\QuantAsylum\QA40x\Logs\SpeakerImpedance.html. If you check the “Open Report” checkbox, that report will open in your default browser. The HTML has the plot embedded, so you don’t need other files. But, you can easily used the “Microsoft Print to PDF” to take the HTML file in your browser and print to a nice PDF for future reference.
I’ve attached a PDF of the HTML file and it makes for a nice summary (the PDF is a Dayton Audio 8"). The wiki for the plug-in needs to be updated, that will come next week along with other issues that are found. For example, the L(e) calculation isn’t matching up right now. Additionally, most other speaker software will apply some tweaks to the piston effective area. That is, you specify a diameter, and the SW converts to an area. I’m not sure which is preferred.
DaytonAudioDC200-8.pdf (293.9 KB)
Looking promising.
I’m near Ottawa so it is kind of late. I just tested a a FaitalPro 6FE100. Looks reasonable. I’ll do a bunch of SpeakerTester Versus QuantAsylum and see the differences.
I take it that you cannot do DCR with the analyzer? I didn’t even think about this.
Personally I don’t mind calculating the Sd. Many don’t like doing this.
Believe it or not closed box method gets used for tweeters, and planars. That will need to be added. The user specifies the closed box volume.
In the report some of the quantities are non-standard. But that will need me being less tired to take proper notice of them all.
I don’t know how to attach a text file or a pdf file. So I have posted images of these. The calculations are a bit different I’d say.
Hi @Kravchenko_Audio, looking at your data you have specified the Sense R as 0.1 ohms. Are you using an external sense R? Or are you using the QA461 Current Sense output? The QA461 should have a value of 1 (it has a 0.02 ohm sense resistor with a 50X gain, which gives an effective sense resistor of 1 ohm).
If the sense R was specified incorrectly, that would bring your Z from 11.7 to 117 ohms. That is still different from the speaker lab value of 91.7 ohms. But, we can dig into that. Note that your resonant freq was 67.8 Hz or 68.12 Hz, depending on the SW used. What is important is that at the resonant frequency, your phase is 0. So, you can set the QA40x to generate a test tone at the desired frequency, and then run it in idle and probe the amplitudes with a trusted DVM. And since the phase is zero, that means you can do a simple division. For example, on a DC200 on my bench, DATS reports resonant freq of 22.88 Hz, and 35.47 ohms. If I set the QA403 to generate a tone at -20 dBV and 22.8 Hz, then I measure (with a DVM) 992 mVrms across the speaker, and a current of 40.1 mA (I actually measure 40.1 mV at the QA461 current sense port, but the sense R is 1 ohm so 40.1mV means 40.1mA. And 0.992/0.041=24.19 ohms. So, you can track down errors that way.
Hi Matt. I totally missed that setting. I left it as it was because of simple unfamiliarity. And that would seriously mess up all the calculations. Give me a few minutes and I will re-run the measurements. Perhaps a simple check box QA461 or no QA461?
Operator error it is I think. Give me a few minutes.
Mark
So this is the closest yet. I played around with the drive level. Minus 45db seems to get me the closest measurements. A few hertz on Fs even 10% on most of the numbers is not changing your simulation that much. But not everything is lining up.
I should add that the SpeakerTester Pro uses a series of sine wave steps. Not a sweep. So the numbers will be a little different due to the amount of energy put into the loudspeaker. But not too different. I can do T/S via MLS noise as well in a continuous real time measurement method.
Vas is usually listed in Liters. But most people can do the calculation.
Generally you want about 25% of Mms as an added mass. I’ll add one more magnet and see if that changes anything greatly. I use magnets on either side of the cone as an added mass. If the added mass vibrates at all it seriously messes up the measurement.
Mark
P.S. Added mass did not change much.
So your math is solid. Some of the calculations are close. It appears to deviate where you are calculating the maximum and minimum F1 F2 points. And from there the other calculations will deviate.
Hi @Kravchenko_Audio, noted on the various points (eg liters vs m3). The new released HERE has made adjustments to the plug-in to capture most of these. I did some more comparisons of DATS on an 8" woofer and the largest difference I saw was related to drive level. If the drive level was set the same as DATS, the Fs and Zmax matched.
Thanks very much for the feedback thus far.
Ok. I will download and try testing again.
Thanks for the work to get this running. The more capable your product the more potential customers!
Mark
The Black line is the results from the speaker tester, Grey is from the QA403/461 combination.
Just a wee bit of difference.
I have never found the WT3 to be reliable. And I have had clients that used the DATS tester and come up with wild numbers. It seems to be a power supply issue. As in testing and then not waiting long enough to allow the power supply time to be able to re-test and give you a consistent measurement.
The comparisons were made via Hornresp. One of the most comprehensive loudspeaker simulators I have ever seen. Hornresp is rock solid. It is one of the few programs that will actually simulate the resonances within your enclosure. Tested against a lot of very fine equipment by more than a few Phd candidates. Some of whom I know.
Hornresp will validate the parameters. Which means your calculations are correct. Perhaps we need either a common test sample of another means to verify the actual values that are of importance. In two software iterations you are closing in on accurate measurements.
Hi @Kravchenko_Audio, thanks for sharing this. There are a few “black box” items that seem to be a source of error amongst the comparisons across packages.
As you’ve pointed out, test amplitude can have a sizable impact on on Fs and Zmax. DATS looks to be testing at -13.76 dBV and I don’t see a way to change that. However, if the QA403/QA461 is set to drive at the same level, the Zmax and Fs are extremely close.
The F1 and F2 frequencies that flank Fs seem to be computed differently across the web. Richard Small’s original paper specified F1 and F2 as the frequencies where the impedance magnitude were SQRT(Re*Zmax). But others may use something else. For example, the DATS manual indicates F1 and F2 are where Z = sqrt(r0) * Re, where r0 is Zmax/Re. That will result in a deviation of Qms.
It’s possible to interpolate Re from AC measurements. But most AC coupled analyzers and sound cards have a lot going on around 1 Hz or so, and the phase due to the HPF is still going nuts around 10 Hz, and so, it seems risky (to me) to try and deduce Re from the sweep. Especially since a $100 Fluke can easily measure a few ohms with far less uncertainty.
Piston diameter will often have some tweaks applied by the various software package, and what these tweaks are aren’t clear (that is, you enter a diameter, and the piston areas is calculated by taking the pi*r^2, and then another factor of 0.9 or 0.95 is applied behind-the-scenes to arrive at true diameter). And since Sd is squared when computing Vas, even small tweaks can result in a sizable change to Vas
Any insight you have to the above would be helpful. I’m pretty sure the differences between the packages will stem from the above.
It is an interesting position to be in. That in which the fundamentals of how to measure the drivers need to be addressed. I remembered some papers by Ivo Mateljan. I will see what else that I can dig up.
If you need any of the AES papers I can get them for you.
https://www.bksv.com/media/doc/bo0384.pdf (nice and detailed, again might spark a few useful thoughts)
"Derivation of Moving-Coil Loudspeaker Parameters Using Plane Wave Tube" by Brian Eric Anderson (the comparison part may have a few nuggets of useful information)
I am pretty sure that ARTA, and other software based measurement systems cannot use your products. How difficult is it to write software for this ensemble? In the background I have contacted a few people that are very capable engineers. But I don’t have any useful answers for them.
Mark
But wait! There’s more!
I have been diving down a rabbit hole over most of today. Many papers that are behind paywalls or on the AES website I have hunted down and now have. If there is anything that you want, I just may have it. There are papers that do directly talk about what doesn’t work in the Small paper. Even Rod Elliots Sound pages has a reasonable discussion on that!
https://www.roomeqwizard.com/help/help_en-GB/html/thielesmall.html
Many sources there, again if you want them. These are more up to date methods of calculating driver parameters.
Mark
The differences in the parameters are still there. But it is getting closer. It is demonstrating how very sensitive this system of measuring a loudspeaker parameters is to accuracy of the added mass. Relatively small errors are not insignificant.
Second thought. Duration of the sweep, or how long the driver is excited over the period of measurement. Could you make a few options? Longer sweep, or if possible individual sine steps? In the Speaker tester Pro you specify the number of steps and the software chases the zero crossing point with as many steps as required to get it accurately.
Claus Futtrup has asked me to compare the results using the Speakerbench website method. That will be next.
D’appolito, in his “Testing Loudspeakers” book, suggests a 25% Fs difference between the 2 measurements for the added mass Vas calculation.
Why above 1k Hz there is all that garbage in the graph?
I agree with your statement regarding 25% of Mms. There are a few places that recommend more.
I have no idea what that garbage is. I am guessing that it may be my setup. But I cannot figure out why. When I do the same measurement via the Speakertester Pro it is nothing at all like this.
Mark
Mark, 25% of Fs, that is you measured an Fs=67.65 Hz, with added mass you should get an Fs of around 51 Hz. Instead you got an added mass Fs=58.23 Hz.
About the garbage in the graph, maybe it could improve using a stepped sweep, like STP and Audiomatica Clio do. I noticed (with Clio) that using a simple sweep brings incorrect impedance graphs.
I agree Claudio that a sweep is the least accurate. And that is proven through many papers and people that are in this business that agree that a stepped sign wave measurement is the most accurate. It takes a few minutes and you can trust what you have.
As for the mass added Speakertester Pro gave us 32.6 grams and I place 6.4 grams as an added mass. Near 20% instead of the 25% goal.
Mark
Why above 1k Hz there is all that garbage in the graph?
The sense resistor used inside the QA461 is 0.02 ohms, followed by a 50X diff amp with killer CMRR (INA199). But I suspect as the impedance rises and the current decreases, the noise of the INA199 is coming into play.
If you used an external resistor, it’d probably clear up. I think a future version of the QA461 needs the ability to switch current sense resistors.
Hi @Kravchenko_Audio, a good sanity check here is comparing a sweep of a resistor versus using a tone. The sweep doesn’t know the the resistor is purely resistive, it’s just grinding through the math as though it were reactive.
For example, using the RLC Automated Test I get the following sweep of a 10 ohm resistor, and you can see at 1009 Hz the impedance was computed at 9.658 ohms.
If I do the same test with a tone, I get voltage peak of 985.1mV, and a current of 101.9 mA. That is 9.667 ohms.
The two techniques should give you nearly identical results. And you can do the same using a reactive load. You just need to know the phase and then grind through the math.
So, have faith in the sweep!
