I mentioned in one of the more recent posts that Klippel recently updated the build version and with that came the new versions of their ISC & MTON modules. You can read more about what these are above or watch the video. The net effect, however, is thanks to these two modules I am now able to gather accurate distortion data in the farfield as an anechoic measurement.
There are numerous templates available in the MTON module (you can read about them if you have access to the IEC standard referenced on their product page; I do not have this nor the $350 to buy it
). I played with each yesterday (at least, the ones that made sense for a passive loudspeaker) and prefer the "Long Term max SPL" test. This was the one shown in the above video. I haven't decided *for sure* that this is how the tests will all be conducted going forward but, for now, I am going to share the results of the testing because it truly is next level analysis.
The DUT is the Neumi BS5. Here's the link to the review:
https://www.audiosciencereview.com/forum/index.php?threads/neumi-bs5-bookshelf-speaker-review.14404/
I ran this test
twice:
First from 20Hz to 20kHz.
Second from 80Hz to 20kHz.
Voltage stepped in 1V increments set up from 1V to 30V.
In this "long term spl max" testing the compression limit was set to 3dB and the maximum distortion level was set to -20dB (relative to the fundamental), equaling about 10% distortion.
First Test: Stimulus from 20Hz to 20kHz.
I'll provide the graphics with a short explanation of what each means.
This table shows the max distortion test results. The top line tells me that the DUT hit max SPL due to hitting maximum compression.
The second block tells me what the maximum SPL value is (relative to 1 meter). In this case, the max SPL ~ 82dB.
This block also tells you the compression at this SPL (4.68dB) and the max multitone distortion (-29.62dB; ~3%).
View attachment 73710
Knowing there are two limits set for Multitone distortion failure (compression and/or distortion, mentioned above), let's break the results down in to those sections.
1) First, compression-oriented results:
The below graphic shows the stepped voltages. The 1V line (solid blue) is the reference to which all other measurements are compared. The dashed blue line is the compression limit relative to the 1V output; which is -3dB from the reference. The red line is the case that exceeded whatever threshold (compression and/or distortion as described above). This will always be the case. Red will always be the final measurement. As you will see below, if multiple voltages had passed there would be additional lines besides the reference (solid blue) and the failed (red). In this case you can see the measurement exceeded the -3dB compression limit at approximately 20Hz (noted in orange below).
View attachment 73711
Now, this is the same thing as above, but simply "normalized" to the 1V reference measurement. I prefer this version as it's easier to identify the failure modes more quickly.
View attachment 73712
And, finally, this graphic just gives you the voltage vs compression. I don't find it particularly useful. Single points of data tell me nothing I want to know. So don't be surprised if you never see this from me again:
View attachment 73714
2) Now let's look at the distortion-based results of this multitone testing...
Below are the results of the various voltages' distortion, relative to the fundamental. There will be n-1 gray lines (where n=# of voltages ran in the test; in this case, 2-1 = 1). The green will always be the last measurement. The dashed blue line at -20dB represents the multitone distortion threshold. This is set up in the template so apparently is the IEC standard. I can change it if I want to. Not sure I will. TBD.
As you can see, none of the voltages increased above the -20dB MTON threshold. Which we expected, as the table I provided first told you that compression as the cause for max SPL being reached. Noteworthy here is the 1V measurement shows a higher distortion value below about 100Hz. This is likely an effect of noise floor. I imagine most every case I run will have a similar result with very low voltage (where the signal isn't strong enough to drown out any external noise because testing is done in my garage), however, the noise will almost always be below the -20dB threshold and increasing the voltage lowers the noise floor. But I need to check with the folks at Klippel to see if that logic flows. Anyway...
View attachment 73715
This graphic provides voltage vs relative distortion. I'll probably never post this again; single points of data are useless to me.
View attachment 73716
Alright. Those are the two components for determining max SPL. Again, you can see that the -3dB set for compression threshold failure is exceeded at 20Hz and therefore the (at least in this case) cause for SPL maximum of 81.93dB.
But, given that I think it's hardly fair to expect a small bookshelf like this Neumi BS5 to realistically play down to 20Hz, I ran a separate test with the stimulus spectrum set only from 80Hz to 20kHz to see what the maximum SPL might be closer to in a real-world situation. Note: as of this moment, I do not know a way to apply a shaped stimulus to emulate a nth-order filter. So for now, this "brickwall" of 80Hz will have to suffice.
-------------- (break) --------------------
Second Test: Stimulus from 80Hz to 20kHz.
Now, let's see what the results look like using a limited, more realistic for a bookshelf speaker, stimulus of 80Hz to 20kHz.
So, that's it. This test method isn't cemented just yet. I'm going to explore the other modules and shaping stimulus when I have time. But, it gives you an idea of where I am going. Overall, I think this is fantastic news as it gives me a repeatable, highly accurate way to determine the actual maximum SPL based on a set of conditions that make sense for the average user.
With this limited stimulus the result of max SPL testing is:
View attachment 73717
Whoa! Now we have a maximum SPL of 95.65dB with compression at 3.37dB. The Multitone distortion at this output level is -40.13dB (1%).
That's a big,
big difference from the 20-20kHz test results.
What are the corresponding results?...
Well, the -3dB compression limit was reached at about 1.37kHz. But barely.
View attachment 73720
How barely? By about 0.37dB.
View attachment 73721
And here is the graphic I said I wouldn't show again. Maybe I'm softening. At any rate, this one shows the max compression value relative to voltage input. Note how the curve is not linear (isn't that an oxymoron?... meh, you know what I mean).
View attachment 73723
So, how does the distortion look, then? Well, the -20dB limit is never reached. But, if I were to change this -20dB (10%) to 3% (-30dB) it still wouldn't have been exceeded with this limited stimulus at the compression limit voltage of 9V.
View attachment 73722
There you have it. Using a limited stimulus allows a higher SPL. Makes sense. But, most importantly, we have a repeatable, accurate method for determining maximum SPL with the underlying factors for why the SPL is limited to x dB. Pretty awesome.
---------- (break) ------------
Additional stuff:
To me, it's interesting to compare the overlapping voltages' distortion and compression profiles. So, let's do that. Remember, with the full 20-20k spectrum I was only able to provide a 2V input before reaching the max SPL. That's where Test 1 and Test 2 will overlap, then. So, let's directly compare what the MTON distortion and compression curves look like using 2V (I could have used 1V as well but I'm trying to save some time here):
First, this assumes 1V (reference) measurement is the same, so let's look at that by way of looking at the two compression limits (blue for the 20-20k and green for the 80-20k):
View attachment 73726
Hey, they are almost identical. Good enough for me to continue on...
Now, let's keep these compression limits but also add the 2V comparison between the full stimulus (red) vs the limited stimulus (gray). Remember, the former failed at 2V due to exceeding the threshold at 20Hz while the latter failed at ~1.37kHz. Still, how much does the limiting of excursion on the 5-inch woofer drive a change in the compression profile?...
View attachment 73733
Not a lot of difference here. A couple dB here and there. But not much; and they both only seem to deviate where the 1V (reference) for each respective measurement lies. Okay.
Moving on, let's compare the MTON distortion profiles using the same voltages. Green = Full spectrum (20-20kHz) and gray = 80-20kHz).
View attachment 73743
Interesting. The limited signal result is between 5-10dB lower in distortion until you get to about 2kHz. This shows that the woofer playing low is causing additional distortion throughout the mid/woofer's passband.
But, even beyond that, above 5kHz there is a pretty drastic difference as well with some portions reaching 10dB more distortion :
So, what can we learn here? Well, things we either already knew or would have logically assumed: a multitone signal is more dense and therefore more stressing than a single tone stimulus like a harmonic distortion or even a two-tone stimulus like an IMD test. This is a good way to resolve complex distortion drivers that wouldn't otherwise be shown. Notably at points far away from the loudspeaker's drive unit's individual Fs (where, typically, suspension related distortion is prominent). However, I didn't expect the difference in distortion to be so significant as high in frequency (even out to 20kHz) by simply limiting the stimulus above 80Hz. That was really eye-opening.
...... aaaaaaaaaaand... done.
Once I settle in on a test method I'll figure out if I want to use Klippel's default window for providing results or write my own script. In this case, though, I think I like theirs just fine. In other cases it made more sense to write my own. That was a topic of conversation in our hour-long skype session a couple weeks ago. They asked for feedback on their graphics (they noticed I was using the data but writing my own scripts for plotting). It looks like some of my suggestions may have already been incorporated. I'll be beta testing their next major release for next year and I'm excited to see what they have changed with their GUI.
