BL distortion prediction / analysis in subs

flowwolf

Member
2011-10-30 2:26 am
Hi guys, i'm planning to build my own speaker drivers for a 4-way open-baffle application.
The details are not so important as it is a typical open-baffle configuration like Linkwitz's and John K's, but
i'm struggling with the subwoofer's excursion since one subwoofer will be used per/baffle. the only important
thing, that it must handle around 14mm excursion as linearly as possible.
Well, i know that the best answer would be 'double the number of drivers', but have taken a look at a typical BL curve
and wasn't satisfied with the results.

Playing with FEMM for a while, checking the distortion levels was a must.
therefore i've made a simple application that evaluates the distortion and THD for a given BL curve generated by FEMM.
Let's take a look at it:

http://s5.postimage.org/jxxv44f47/app01.png

The app eats the BL curve and shapes a given wave, then the distortion is evaluated.
Here is (the already) optimized motor structure:

http://s5.postimage.org/vx9dbfkp3/motor01.png

Since evaluating the BL curve is a pain in the @ss, and FEMM is unable to simulate AC frequencies with permanent
magnetic fields, the BL curve is not entirely precise. I mean, shorting rings will be used, and by changing the
direction of the current in FEMM, the BL curve also changes. btw. the BL curve itself (mathematically) is an integral
of the flux density in regarding to the coil.

So, here is another app. that interprets the 'air gap vs flux density' curve generated by FEMM:

http://s5.postimage.org/xdkvtknlz/motor01_flux.png
http://s5.postimage.org/u9a7ws6tj/app02.png

http://s5.postimage.org/n7caal37r/app03.png

and generates the given BL curve for it. With this, basic distortion checking is simplified.

it is a simplified model, it doesn't take lorentz forces and so on into account so it is not a FEA.

there is more to this story, but i'm waiting for replies.
 
Hi guys, i'm planning to build my own speaker drivers for a 4-way open-baffle application.
the only important thing, that it must handle around 14mm excursion as linearly as possible.
Well, i know that the best answer would be 'double the number of drivers', but have taken a look at a typical BL curve
and wasn't satisfied with the results.
If you actually want to build speakers, I'd suggest getting a Klippel Analyzer like the manufacturers of speakers with excellent BL curves at high Xmax values use to help in the design process.

Why do you want to build your own drivers?
 
I am more interested in HOW you are planning to build your own drivers. I think your approach is interesting, as far as I can see you are trying to reliably simulate distortion/behaviour of a design before you make it, so it would be an interesting tool in adittion to a Klippel analyzer. Before/after.

I assume you will also take into consideration weight of movable parts and physical resistance.
Surface area of gap? Will you be able to adjust gap design in simulation?

Sorry if this seems trivial.
 

flowwolf

Member
2011-10-30 2:26 am
There is a lot more to building drivers than the design side.

Well, then which subwoofers would you recommend me for low BL distortion @ hi xmax? prices around 300$.

maybe you have custom build subs. i would take a look at them then.

KaffiMann, what do you mean by reliably simulating distortion? It's not a FEA, so a simplified model won't give realistic results as it's doesn't take AC fields and so on. into account.
 
Well, then which subwoofers would you recommend me for low BL distortion @ hi xmax? prices around 300$.

maybe you have custom build subs. i would take a look at them then.

KaffiMann, what do you mean by reliably simulating distortion? It's not a FEA, so a simplified model won't give realistic results as it's doesn't take AC fields and so on. into account.

... What? You mean besides these?
AE Speakers Online Store

Yeah, I get the point about not being 100% realistic, but it seems like a good start :)
 

flowwolf

Member
2011-10-30 2:26 am
I know about AE speakers, one of my favorites :) (well, I never heard them, but i'm sure in the results)

Let's take it longer.

You agree that this is just a simplified model. FEMM takes the permanent field of the coil into account, but - in my opinion - it does not matter, because it also changes with the direction of the current, and if you change it to an AC waveform it can be reduced / or almost completely eliminated by a shorting ring. This is what it's called flux modulation as described by others many years before, and also influences the field of the permanent magnet itself.

FEMM flux for DC coil zero positive & negative currents:
attachment.php

attachment.php

attachment.php



Let's ignore them now.
From our perspective, the distortion caused by other things like the spider, surround, or intermodulation distortions are irrelevant now.

Here is the sample BL curve which is generated for the test subwoofer above for a 38mm voice coil:
attachment.php

attachment.php

attachment.php



The prediction shows that it has a BL deviation of about 4% for 10mm and around 10% for 13-14mm. It follows the typical rule that BL deviation of 30% results in around 10% THD but this is an oversimplification. First the motor structure itself is somewhat optimized since it is an undercut & extended polepiece.

Next, I increase the length of the VC to 45mm:
The distortion figures are improved, 2nd and 3rd mainly. This is a common behaviour/solution but IMO not the best one.
attachment.php

attachment.php


(note here that the evaluation also depends on the quality, but i wasn't able to recognize any improvement above a certain point)

...
 

Attachments

  • motor01_flux.png
    motor01_flux.png
    8.4 KB · Views: 262
  • 2_blcurve3 45 14mm.png
    2_blcurve3 45 14mm.png
    43.5 KB · Views: 255
  • 2_blcurve3 45 10mm.png
    2_blcurve3 45 10mm.png
    42.2 KB · Views: 253
  • 2_blcurve2 38 14mm.png
    2_blcurve2 38 14mm.png
    47.6 KB · Views: 269
  • 2_blcurve 38 10mm.png
    2_blcurve 38 10mm.png
    43.8 KB · Views: 266
  • 2_flux 38.png
    2_flux 38.png
    41.9 KB · Views: 274
  • motor01_fluxneg.png
    motor01_fluxneg.png
    8.4 KB · Views: 263
  • motor01_fluxpos.png
    motor01_fluxpos.png
    8.4 KB · Views: 268
Well, then which subwoofers would you recommend me for low BL distortion @ hi xmax? prices around 300$.

maybe you have custom build subs. i would take a look at them then.
One of the best 18" in regard to low BL distortion and hi Xmax is the B&C18SW115, but it's price has gone up to more than double what you want to pay due to cost of the neodymium magnet structure.
A close second is the ceramic magnet BC18TBW100 at around $349.

You can see distortion results using the B&C18SW115-4 in bass reflex and tapped horn enclosures here:

http://www.diyaudio.com/forums/subwoofers/184992-tapped-horn-vs-bass-reflex-case-study.html

I have never pursued open baffle cabinets, but I can say that a pair of B&C18SW115 run open air driven to 15 mm excursion at 20 Hz have very little distortion and enough output to make me feel ill for quite a while after only a short exposure during break in.
The suspension did loosen up a small amount (perhaps 2 mm) after several hours running around Xmax.

Art
 
Last edited:

flowwolf

Member
2011-10-30 2:26 am
Nice drivers! with good sensitivity.
(It would be great if manufacturers could share the details of copper wires, diameter etc..)

Thanks for the info.
Never used tapped horns, have to do some research before i can comment on it. Pretty backwards mounting :)

What kind of software is that you used there? If there are distortion measurements for pure sine waves, they are so interesting :)

p.s. this is the first time I saw h0 as "EtaZero" in a datasheet
 

flowwolf

Member
2011-10-30 2:26 am
I continue..

One would try to use an underhung design, but have it's own deficiencies for extreme huge top plates, let's see a flux curve given by FEMM for a 45mm top plate design (average steels, magnets):

attachment.php


attachment.php


attachment.php


attachment.php


attachment.php


very good indeed! but can you see that it's not entirely linear? what if you want a design with more than +-14mm?
It won't succeed unless you use high quality materials like pure iron. It start to saturate and you have to increase vcid this results in higher cost, but IMO still the best solution.

...
 

Attachments

  • 2_uh_blcurve3 14mm.png
    2_uh_blcurve3 14mm.png
    44.4 KB · Views: 200
  • 2_uh_blcurve2 10mm.png
    2_uh_blcurve2 10mm.png
    43.1 KB · Views: 197
  • 2_uh_blcurve.png
    2_uh_blcurve.png
    38 KB · Views: 200
  • 2_uh_flux.png
    2_uh_flux.png
    8.3 KB · Views: 202
  • 2_uh.png
    2_uh.png
    86.3 KB · Views: 197

flowwolf

Member
2011-10-30 2:26 am
Next an XBL design:

attachment.php


attachment.php


attachment.php


attachment.php


attachment.php


not bad, better than overhung but not as good as underhung. however, it is possible that it generates more high order harmonics than UH/OH. this is why the THD is not representative. It reveals that something has a certain degree of harmonic distortion, but what harmonics? 2nd 3rd? Here high harmonics are generated (in theory) but (i'm skeptical) i would be curious in real-world examples.
btw. i think the xbl is a clever one, but let's see what happens when the voice coil is 'mis-wound' :

attachment.php


see the peak in the BL curve?
it requires careful engineering. IMO top plates should not be made of laminations like this:

attachment.php


attachment.php


This might worsen the reluctance and more flux will flow on the low reluctance path.


The problematic part in a finished subwoofer that you won't be able to authentically measure BL distortion because of the suspension. It requires special machines. Therefore i'm trying to predict BL distortion as much as possible.

...
 

Attachments

  • 2_xbl_flux_bad.png
    2_xbl_flux_bad.png
    12.5 KB · Views: 197
  • 2_xbl_bad.png
    2_xbl_bad.png
    80.9 KB · Views: 203
  • 2_xbl_blcurve_bad.png
    2_xbl_blcurve_bad.png
    44.6 KB · Views: 192
  • 2_xbl_blcurve3 14mm.png
    2_xbl_blcurve3 14mm.png
    49.7 KB · Views: 195
  • 2_xbl_blcurve2 10mm.png
    2_xbl_blcurve2 10mm.png
    48.1 KB · Views: 195
  • 2_xbl_blcurve.png
    2_xbl_blcurve.png
    45.5 KB · Views: 201
  • 2_xbl_flux.png
    2_xbl_flux.png
    10.2 KB · Views: 201
  • 2_xbl.png
    2_xbl.png
    84 KB · Views: 195

flowwolf

Member
2011-10-30 2:26 am
p.s. please note that typical xbl BL curves are compared with the average BL curves of overhung designs.
at post #10 there is a BL curve for a well optimized sub:

attachment.php

(i.e. if you compare the xbl with an average overhung, the xbl is winner)
 

Attachments

  • motor01.png
    motor01.png
    73.1 KB · Views: 193
Nice drivers! with good sensitivity.
(It would be great if manufacturers could share the details of copper wires, diameter etc..)

Thanks for the info.
Never used tapped horns, have to do some research before i can comment on it. Pretty backwards mounting :)

What kind of software is that you used there? If there are distortion measurements for pure sine waves, they are so interesting :)

p.s. this is the first time I saw h0 as "EtaZero" in a datasheet
OOps, gave you the wrong link. The distortion figures are in post #12 of this one:
http://www.diyaudio.com/forums/subwoofers/185588-keystone-sub-using-18-15-12-inch-speakers.html

I used the Smaart program to look at the actual distortion speakers have when driven with a sine wave tone , an example showing 35 and 45 Hz at 77.5 volts(1500 watts in to a nominal 4 ohms) is posted below.
Using the distortion calculator harmonic distortion levels can be determined.
Slowly.
 

Attachments

  • Distortion Tests.png
    Distortion Tests.png
    99.1 KB · Views: 37
  • Distortion Calculator.png
    Distortion Calculator.png
    52.1 KB · Views: 32
Last edited:
Super info.

Both 2nd & 3rd harmonic @ -28dB and 4nd & 5th harmonics @ -40dB, 7th @ -51dB, close to inaudible above 400hz.
This is at xmax i suppose?
The tricky part is that you cannot tell the contribution from the suspension.
Does a Klippel reveals that?
I really have no Klippel experience, just have heard others write about it, and how the B&C neo drivers have the best BL charts.
They also use a very interesting coil winding topology, the details of which I have forgot.
I suspect B&C drivers are designed with simulations, then tested to see how they vary from the (proprietary) sim, after many tests the sim can be tweaked to closer match reality.
Fewer test iterations are probably required for each new model as the software is tweaked generation to generation.

The Keystone excursion maxima is right around 45 Hz, at 77.5 volts it would be around Xmax (Xvar), 30 mm peak to peak.
At 35 Hz, the excursion is probably only about 8mm peak to peak.
The suspension really puts the brakes on excursion above Xvar, the BC18SW115-4 would probably burn up before it hit Xlim.
Although I did not measure distortion, the driver sounded OK at 120 volts 60Hz, three times in a row (it was testing mistake), about 1 second per time.
 
Last edited:

flowwolf

Member
2011-10-30 2:26 am
Interesting, Weltersys,
i will read through the Keystone Sub thread you posted before.
Understand. sure they have their proprieraty stuff for sims.

Progressive spiders are extensively used in subs.
According to Klippel's paper

http://www.klippel.de/fileadmin/kli...Diagnosis_and_remedy_of_Nonlinearities_00.pdf ) and Hutt's patent
https://docs.google.com/viewer?a=v&...nxlcXVpdHlzbmRpbnZ8Z3g6N2FhNjY3NTNiY2Y5MDUyMw

a regressive spider is better in linearity.


At worst case the distortion of the suspension could be measured separately. A spider compliance measurement machine may be present at manufacturers. I'm wondering whether it is able to measure distortions too.

I may have an idea about better coil winding topology, i will come up with the details in a few posts.

--------------------------

Here is a split-coil model.

attachment.php


First I tried it on the existing flux curve of the optimized subwoofer and got worse results:

attachment.php


was sceptical and did the Bl curve evaluation in FEMM, the result were the same:

mm BL
5 5.54122
4 5.70404
3 6.03851
2 6.35937
1 6.46567
0 6.48344

then I realized that a split-coil model is quite sensitive, it does matter were you put the coils and how:
attachment.php


so, here is a newer FEMM model for split coils, and the flux curve, the result is generally good:

attachment.php


attachment.php


attachment.php


...
 

Attachments

  • 2_sc2_blcurve2.png
    2_sc2_blcurve2.png
    50.2 KB · Views: 183
  • 2_sc2_blcurve.png
    2_sc2_blcurve.png
    44.2 KB · Views: 182
  • 2_sc2_flux.png
    2_sc2_flux.png
    10.8 KB · Views: 190
  • 2_sc1_blcurve2.png
    2_sc1_blcurve2.png
    43 KB · Views: 185
  • 2_sc1_blcurve.png
    2_sc1_blcurve.png
    43.5 KB · Views: 182
  • 2_sc1.png
    2_sc1.png
    66 KB · Views: 182

flowwolf

Member
2011-10-30 2:26 am
Good work with that keystone design Art,
Since i'm not exclusively interested in OBs, i may build the design one day


---------------------------------------

Next a JBL differential drive structure:

attachment.php


attachment.php


attachment.php


attachment.php


attachment.php


As expected, the even ordered harmonics are very low in level due to the symmetry,

Note that the flux curve is evaluated for 0A. If we set the current to 1A, we basically arrive to the same BL curve as would give by FEMM:

attachment.php


attachment.php


attachment.php


The BL curve is no longer symmetric as well as the flux curve.
(at post #1 i mentioned that the BL curve evaluated by FEMM is not symmetric, and also depends on the direction of the current. however, in real world we never use loudspeaker drivers with DC current, therefore - IMO - evaluating the BL curve without taking the DC currents into account is not problematic. i.e. we can evaluate the flux curves for 0A. so the application won't take the the coil's own field into account, this is why it's simplified)
 

Attachments

  • 3_jbl_blcurve3_1A.png
    3_jbl_blcurve3_1A.png
    42.8 KB · Views: 157
  • 3_jbl_blcurve2_1A.png
    3_jbl_blcurve2_1A.png
    47.1 KB · Views: 154
  • 3_jbl_blcurve3.png
    3_jbl_blcurve3.png
    43.2 KB · Views: 157
  • 3_jbl_blcurve2.png
    3_jbl_blcurve2.png
    46.4 KB · Views: 156
  • 3_jbl_blcurve.png
    3_jbl_blcurve.png
    48.4 KB · Views: 165
  • 3_jbl_flux.png
    3_jbl_flux.png
    7.9 KB · Views: 167
  • 3_jbl.png
    3_jbl.png
    77.1 KB · Views: 163
  • 3_jbl_flux_1A.png
    3_jbl_flux_1A.png
    7.9 KB · Views: 156
Last edited: