Music spectra and speaker IMD

[hardisj]

Smaller driver (SOTA Purifi 4-inch with long excursion):

https://www.erinsaudiocorner.com/driveunits/purifi_4/

Test 1: Bass Tone at Fs, Voice Tones from 200-6000Hz

Test 2: Bass Tone at 80Hz, Voice Tones from 200-6000Hz

 

[hardisj]

Interesting to note is the doppler distortion (blue). When you have a smaller driver trying to play lower frequencies, the doppler distortion is going to be higher and it shows in this data. Hard to do straight apples-to-apples but there is enough there to get the picture.

Check my links for the reviews for more distortion data (HD, IMD and multitone).

 

[KSTR]

@hardisj , excellent data.
Testing with lower tone at Fs is a very good idea, it drives the LF tone almost as good as an additional "exiter" speaker mounted to the rear volume of the driver under test. By this the motor distortion applied to the higher tone, mostly from the slowly moving BL(x) operation point, is better isolated. OTOH, testing at Fs of a reflex box does the opposite, it maximizes current but reduces excursion and hence this is more test of BL(i).
Driving at a higher frequency for the low tone in a area where the impedance is moderate probably at good mix of both and pretty representative.

As for Doppler Distortion, this a natural and linear distortion and can be 100% undone with a phase modulator in the upstream signal, before the amp. Up to now, to my knowledge, Doppler Correction has only been applied to active MFB speakers (as their HD/IMD is low enough to let Doppler really shine through) but with woofers as good as the Purifi I think it starts making sense to get applied to non-MFB speakers as well.

 

[vavan]

Care to share your IMD?

https://www.audiosciencereview.com/...ely-is-a-sinad-of-96db-poor.12519/post-471084

 

[Joseph Crowe]

I’m sharing my test of the Dayton RSS315HO subwoofer where I look at IMD at 90, 95,100, 105dB SPL levels. The IMD starts out at 45dB dynamic range against the test signal losing -5dB of dynamic range for every +5dB increase in overal SPL. This creates strong rationale for multiple subs for even normal listening levels (85dB at the listening position?) This assumes of course that these levels of IMD are even audible, which is a separate topic.

https://cdn.shopify.com/s/files/1/0142/2848/8292/files/1592.pdf?v=1638647637&fbclid=IwAR0VFVFBS8rM23pFePyGnf1uYZsf6TuvOmi6B5bOdohkMghUvZcvKLa8tw0

Also the plans for the subwoofer in question…

Subwoofer No. 1592 Plans

By purchasing these plans you will automatically receive the assembly and detail drawings via email. These plans allow you to build the No.1592 subwoofer. These plans are for personal use only. Specification Sheet Click HERE. Details: Funds in CAD currency Assembly and detail drawings (.PDF...

josephcrowe.com

 

[Jukka]

I measured some IMD of the sound system at hand .

Description of system and conditions:
Loudspeaker: 2-way passive, closed box, DIY loudspeaker, 8" woofer and 1" horn loaded compression driver, crossover at 1500 Hz.
Preamps: miniDSP SHD. Dirac and some other dsp enabled.
Amps: Hypex NC400 monoblocks.
Mic: UMIK-1, 12 cm from baffle, midway between horn and woofer centers.
All RTA measurements were done with same positioning and volume settings, sweep may have had different volume.
Background noise sources: Computer, fan of Raspberry Pi, ticking clock. Untreated living room. Subjectively pretty quiet at the moment.
Software: REW.
Sweep: 20-20000 Hz, 256k.
RTA: FFT 64k, Blackman-Harris-7, 4 averages.

The conditions were not ideal, but thats irrelevant to the scale of findings.

Some basics, SPL and phase:View attachment 131289

Normalized distortion:
View attachment 131290
This graph is important to compare against IMD measurements below. This speaker is quite good in frequencies from 100 Hz and up distortion being around -50 dB and less.

Spectrum of background noise. The fan of Raspberry shows quite well:
View attachment 131291

Now something more interesting. First 1 kHz sine wave to give perpective:

View attachment 131292

2nd harmonic is close to the Raspberrys fan, but third rises above it. H3 is some >60 dB less than fundamental. There is also mains related 50 Hz spike, although I wonder how it gets there. It's possible to be sourced elsewehere than the actual speaker. Next I used REW dual tone generator and selected the predefined "SMPTE 60 Hz & 7 kHz 4:1", results:

View attachment 131293

Fundamentals tones are as they should, IMD barely above noise floor (remember the raspberry). Lower tone comes from woofer and upper tone from the tweeter. Next "DIN 250 Hz & 8 kHz 4:1":

View attachment 131294

Yup, rising IMD. This is two tones, image what real music looks like. Again lower tone from woofer and higher from tweeter. The next graph is last preset "CCIF 19 kHz & 20 kHz 1:1":

View attachment 131295

Well now, if you wonder why manufacturers don't post their IMD results, it's because they look awful and would scare anyone away. I can't hear the fundamentals with my hearing, but I do hear the distortion loud and clear! Should real music have this much hf content, it would smear the whole response band and probably sound terrible. Lowest distortion product is the only spike below crossover frequency.

If I move the mic up to tweeter center, hf distortion above crossover frequency gets louder and the spike below it gets quieter. When I move it down to woofer center, exact opposite happens. This implies that the distortion is in fact generated before the speaker and the speakers passive crossover splits it to drivers. The electronics of this system are very high quality, each has been measured on this site and are within the top rank of tested devices and have fantastic measured performance. Switching Dirac dsp on/off with the miniDSP remote made no real difference, results were as ugly as ever.

So what gives? That CCIF must be a real torture test, since it generated so much intermodulation distortion and the other test results were a lot more comfortable. Care to share your IMD?

EDIT: A few typos and mistakes. Should proofread first -_-'

I later learned that the tweeter in one of those speakers was faulty, but I'm not sure if it was the one measure. I have now acquired and tested a new set of DIY speakers. The tweeter in my new speakers is the Eighteen sound NSD1095N compression driver on a XT1086 horn. The horn is treated with a butyl mat to dampen ringing of the horn (that's quite essential for aluminum horns). Here are some measurements of the tweeter alone. The equipment is the same as above, just the driver and horn are different and running solo. Output was LR48 high-passed at 150 Hz to protect the driver from damage, but it probably would have handled the measurements just fine. The driver was detached from cabinet and driven on soft platform covered in pillows to reduce echo. Detaching from cabinet can cause the horn to ring despite of damping, so some of the anomalies can be result of that. I was wearing hearing protection throughout the testing.

To put things into perspective, this 18sound compression driver is on a different league than the previously tested. This is nearfield measurement into calibrated mic, at 114 dB at typical crossover frequency, has max THD of 1.1 % over intended frequency range. It's dominated by 2nd harmonic, which is typical for compression drivers and the 3rd harmonic doesn't rise above 0.1 %.

Next IMD. Running the bass notes on a tweeter makes no sense, so I skipped those IMD measurements. Let's go straight to the REW's CCIF signal.

It's as if the test was totally different, this driver shows almost no IMD products when compared to the first speaker that I measured. And this is over 90 dB fundamental, that's more than 10 dB more than with the first speaker. REW has now more dual tone signals. Again I'm skipping those that have low notes. Before criticizing high SPL levels, they are actually needed to emphasize IMD, which is the purpose of these tests. HD and IMD will rise non-linearly with SPL. Besides, less than 100 dB is for wimps (these tests are without EQ).

Spoiler: More Dual Tone Graphs

Now some triple tones.

Spoiler: Triple Tone Graphs

I should have just run J-test signal though it...

I also put it on 112 Ohms of series resistance to mimic current drive. The driver setup was not touch, resistors were added in between cabling. Levels would need to be equal to have comparable results, but I was too lazy to adjust closer than 0.5 dB. Current drive reduces 3rd and higher order harmonics a good deal. If there are such harmonics. I made tests on different resistances and SPLs, but I don't have graphs for all of them now. I learned that the 3rd harmonic is lower with current drive below 2 kHz, but that's almost useless for this driver, because it's so near to my crossover frequency. Also one could argue the audible gain of 3rd harmonic maxing at 0.02 % instead of 0.03% at 110 dB near crossover frequency. Comments below are for comparison against the same measurement above.

Spoiler: Current Drive Graphs

One spike missing at 2 kHz.

Spikes are lower at 1 and 2 kHz, and missing from 7.5 kHz

All spikes between 100 Hz and 10 kHz are noticeably lower, except for d2L.

The other graphs were mostly the same or even smaller changes. I may test the woofer later (also Eighteen sound, 10NMB420), but I already know that I need more analog crossover parts to get useful results. I might have to verify at least some of these test with the old speaker. I still have them.