Midrange that plays well and clear in low to mid volumes

Jump to Latest
#41
Hi,

yeah all I'm after is context, what you did and why and so on 🙂 Thanks clarifying.

It looks like the other woofer is louder than the other from this image, making it also better sounding (is it?). Also the responses include room reflections so they differ a lot.

You could take nearfield measurement, mic very close to the woofer, in order to prevent effects of room polluting the measurements. Try to maintain equal distance from tip of mic to tip of dust cup for both as it would affect SPL reading if one mic is farther away. Or perhaps align mic to baffle plane or something, perhaps others have good procedure to suggest.

Then, try to equalize other woofer to match the other, or both match some target curve. For very least adjust levels to match within the passband, preferably deal with any of the breakup and so on, equalize well beyond crossover frequency, two octaves at least. Well, actually the high frequency EQ would need to be made with far field measurement I think. Anyway, try get as good match as you can with the methods you come up with, within 0.5db or better, I almost guarantee you hear difference this small, at least I do 🙂

Your measurements extend quite low, is it a two way system? I bet differences in motor and suspension are audible then, excursion on lows will distort the whole bandwidth and better one could stand out.

Example: I used one driver, with and without an inductor between amp and the driver, and then tried to match both with DSP as close as I could, images below. I remember taking recording with the measurement mic but cannot find it currently, argh. If you open one image and click next to cycle through them, you can see got within +-.5db or so on passband which is roughly below 1-2kHz, and there is deviation well above. I remember hearing a difference, the other one sounds "more bright", so better in that sense, has edge over the other in preference when listening back to back (recordings I made of the two).

If I remember correctly the brighter one was the one with more highs and reduced the distortion, so I think it's just the difference in frequency response. Can't hear distortion on either as such, either it's inaudible, or just so similar between the two there was no practical difference and I couldn' t detect it. Difference in distortion above passband is ~>10db, difference in frequency response is ~2db or so, octave above passband something 3-4Hz and beyond. And I think I hear the difference in frequency response, and no difference in distortion, go figure 🙂 Perhaps another driver, EQ:d to exactly same response would again give some audible difference.
in-application-without-coil.pngin-application-with-coil.pngfrequency-responses.png

While my test might be flawed, it's still fun time with hobby which gives a lot of perspective on things, wanted to bring this up that matching of frequency response and SPL is utmost importance, in order to try to hear any other differences.

edit. found where I posted these https://www.diyaudio.com/community/...ities-of-hybrid-crossover.389129/post-7444428 , I see if I can find the audio file.
 
Last edited:
  • Like
Reactions: Arez
#42
Found the files! Too large to attach a ZIP here so you can listen / download from here:
https://www.dropbox.com/scl/fo/aj0efzewgsm45xd6g8hmm/h?rlkey=16rt6fknjus3y0om6qx6mt4k4&dl=0
There is great audible difference between the two.

For more details, these were measured in my living room and I think at elevated but comfortable listening level with my Beyer MM1 measurement mic perhaps 50cm away or something, cannot remember too accurately. Not calibrated. The rig was not touched between the two measurements. Added ~1.5mH inductor to the speaker cable and adjusted DSP to compensate the differences. Physically they are exactly the same rig and speaker and so on, only difference being the inductor and DSP settings and perhaps half an hour between the measurements at max.

edit. listening them now and indeed there is great great difference between the two, especially with the noise files but also on the Song 🙂 Although the difference is there, I do not know whether it's due to distortion or due to difference in frequency response. => must match frequency response to max, in order to hear difference in distortion for example. The series inductor reduces some driver motor distortion emitting in acoustic domain, like hysteresis and Le(x).

Please take listen to the files behind the link. Do these examples represent difference between better and worse drivers you guys have listened and observed? If they do, then the worse drivers you've listened to have more distortion from motor like these my examples, and could be made to sound better with series inductor ("current drive") like I did here 😉 even the better driver might benefit. Unless, it's all about the frequency response...I have to repeat this test one day, with more careful frequency response matching.
 
Last edited:
  • Like
Reactions: Arez
#43
I tested a few midranges for distortions and Barkhausen noise, I played lots of music, and preserved nonlinear residuals. Here, the input was Mozart's 20th concerto, prefiltered with the range that each driver can reproduce, on 80 SPL RMS, Tutti, from only 15.650 to 16.050 is the easiest to demonstrate the distortions.
In freq domain (dotted lines are the microphone input, solid - residuals):
x2.png

in time domain
time-domain.png

loudspeakers:
SB12cac +12 Ohm load, filtered 300-3000Hz - displays Berkhausen avalanches
SB12pfc filtered 300-3000Hz - displays frequent Berkhausen noise
Focal ps130, filtered 500-5000Hz - same
FANE sovereign 10, filtered 250-2500Hz -same
B&G Neo8 clone, aka Sounderlink ATM-920 - 500-5000Hz - presumably, neodymium magnet based. Much less distortions
HIVI D7500 filtered 500-5000Hz - not too bad
HIVI D7500 loaded with24 Ohm, filtered 500-5000Hz - not too bad
Dayton Audio PM180-8, filtered 250-2500Hz. So far the best

I would not even look at ferrite-based drivers anymore.
 
  • Like
Reactions: lrisbo, Hörnli and ScottG
#44
Hi @mikets42 : Thanks for the measurements, highly informative! Testing with real music signal is an unconventional, but obvious method.

Just for understanding: The "residuals" = (input signal to driver) - (received signal from microphone), somehow equalized that linear distorions of FR are neglected. So what we see in the residuals is all harmonics, IMD and so on during playing that music signal? So thats a great, fresh view to the driver's behaviour.

But I have issues with the interpretation how to identify the Barkhausen noise. e.g. peaks in playback signal of SB12CAC may also be harmonics, when some frequency excites the cone breakup resonance at a point in time (see residual spectrum response at 12kHz of SB12CAC).

EDIT: this was my first interpretation of the SB12CAC behavior:

1710662424077.png


1710662554460.png


What is your pattern to identify the barkhausen noise, and to distinguish from other nonlinear effects?
 
Last edited:
#47
Kwesi said:
Hi @mikets42 : Thanks for the measurements, highly informative! Testing with real music signal is an unconventional, but obvious method.

Just for understanding: The "residuals" = (input signal to driver) - (received signal from microphone), somehow equalized that linear distorions of FR are neglected. So what we see in the residuals is all harmonics, IMD and so on during playing that music signal? So thats a great, fresh view to the driver's behaviour.

But I have issues with the interpretation how to identify the Barkhausen noise. e.g. peaks in playback signal of SB12CAC may also be harmonics, when some frequency excites the cone breakup resonance at a point in time (see residual spectrum response at 12kHz of SB12CAC).

EDIT: this was my first interpretation of the SB12CAC behavior:

View attachment 1287044

View attachment 1287046

What is your pattern to identify the barkhausen noise, and to distinguish from other nonlinear effects?
Click to expand...
The residual is {mic - conv(RIR, spk)} removes the linear-related part of mic signal.

I have even more issues with the interpretation of what I observe. Yesterday I talked to a friend of mine who is a theoretical physicist (on quite a good level) for hours debating if and how quantum mechanics can provide mathematical apparatus for describing the effect.

There are 2 major sources of LTI distortions: magnetic effects and cone resonances. Magnetic-related are split into 2 classes - low frequency, displacement-related, and high frequency, unclear etiology. The first one has been analyzed to death by Klippel and Co. Second one... it seems that the process of demagnetizing/remagnetising happen in steps, as Barkhausen predicted, creating discontinuities of 0th-order (dnf/df^n), whose spectrum shall decay as 1/f. Then shorting coil comes into action as (1st order of astatism) feedback loop which takes a derivative of distortions and creates delta-function out of Barkhausen's discontinuities. These delta functions shoot through the spk cone and become shaped as TF - here you get the 12kHz spike on SB12cac.

BTW, that's why the high-frequency distortions grow 6dB/octave. The magnetic domains can be small and large, and the discharge can happen rarely as an avalanche, or more often, but less spiky.

I am not a physicist, only a mathematician. I would love to discuss the effects I see with a scientist who specializes in this area of physics.
 
  • Like
Reactions: Kwesi and stv
#48
@mikets42 : Are you aware of the work of Purify company? They have spend some efford on deep analyses of such effects:

https://purifi-audio.com/blog/tech-notes-1/this-thing-we-have-about-hysteresis-distortion-3
https://purifi-audio.com/blog/tech-notes-1/combating-hysteresis-distortion-part-1-amplifiers-2
https://purifi-audio.com/blog/tech-notes-1/distortion-the-sound-that-dare-not-speak-its-name-8

I assume Lars Risbo may be one of the experienced discussion partner for this, he is active as @lrisbo here in the forum:
https://purifi-audio.com/about
 
#49
Yes, I am aware of Purify. I contacted them when I ran into 3rd harmonic staying on the same relative level - which is actually very similar to transient distortions in class B amplifiers, and easily simulatable. They first told me that they would get back to me - but they never did. Lars Risbo is obviously not interested in discussing anything with me. I have no idea why but I respect his choice.
 
#50
Hi @mikets42 , I am sorry if I (or my colleagues) have missed a message from me. I love math and discussions. It is a bit different form cross over distoriton since this gives a rising relative distortion as the level goes down. cross over distortion does not have memory. A 3rd order nonlinearity (a la x^3) gives harmonics that scale with amplitude squared (A^2) relative to the fundamental. However, for the magnetic and mechanical hysteresis we see that they drop only A^1 or even close to A^0. I am postulating that this is not possible with a memoryless nonlinearity. It simple resuires memory like we see in hysteresis.

Cheers,

Lars
 
  • Like
Reactions: Arez, Kwesi and analogkid455
#53
Last edited:
You must log in or register to reply here.
Top Bottom