Ironless Motor Assemblies and STEALLUS

[smowry]

Message from Purifi:

Hi Steve,

It’s a structure based on neodymium. Exact properties and method for optimization is a trade secret ;-)

Note that is one out of several IP’s working on concert to significantly reduce nonlinearities.

Read a bit more about it here: Ushindi | PURIFI and here Blog Posts | PURIFI



Thanks and regards,

Claus

 

[AOR]

STEALLUS required ~10x the volume of NdFeB of a typical motor assembly in FEA simulations.

Yeah, this was what I found from doing my simulations. It's got to be a really high end product to justify the extra BOM.

 

[smowry]

In 2005-2006, magnet prices were much lower. Magnet suppliers in China wanted to sell large magnets. Here's some Neo History (pun intended).

https://pearl-hifi.com/06_Lit_Archive/14_Books_Tech_Papers/Mowry_Steve/Neo_Supply_vs_Politics.pdf

 

[smowry]

AOR,

When there is no ferromagnet material in contact with the magnet(s), then the operating point of the magnet(s) will tend to be low. To counter this, the magnet(s) should be made THICK. In a sense, the B is related to the area of the magnet surface and the H is related to the thickness. If the operating point of the magnet is low, then the motor assembly will not be robust and the potential for demag may become an issue. In the FEA simulations we want to check the H within the magnet cross section. It's more than just B in the gap with ironless motors, one also should check H in the magnet and set a reasonable operating point for 90% of the magnet cross section by setting the limits on a contour plot of H in the direction of magnetization, typically the axial direction.

 

[Lars Risbo]

hi, I am here if you have questions about our transducers.

cheers

Lars

 

[smowry]

Hi Lars,

Do you use Isotropic or Anisotropic material properties for spider simulations? If you use Anisotropic material properties, how do you identify them?

 

[Lars Risbo]

Hi Steve, I assume isotropic. as you indicate, it’s very hard to identify the full stiffness matrix for the anisotropic model. Also, keeping track of the material orientation after pressing into shape.

cheers

Lars

 

[Heinrich]

Yeah, this was what I found from doing my simulations. It's got to be a really high end product to justify the extra BOM.

Some folks are known to throw 10 times the effort at a problem that is to all satisfaction 99% solved. Those then call it luxury or high end, if you will. I personally wouldn‘t argue with a rational caveat, but would say, it is just so ugly.

 

[smowry]

Thanks Lars,

I was curious because I was never able to identify the anisotropic material properties of a woven spider. So if you are not doing anisotropic models, then most likely nobody is!

Back in the day for me, it was surprising how few companies did suspension simulations at all before I worked with them. Frankly, other than Dr. Mark at BOSE, my mentor, I did not encounter any. I have been retired for more than 10 years now so I will make a listing for history sake.

1. Genelec
2. Seas
3. Eastech
4. TC Sounds and Sunfire
5. P.Audio

I believe that all of these companies have the capability to simulate suspension components today but of course using isotropic models.

 

[smowry]

In response to the above post, I worked with Patrick at PACSYS on the development of PafLS. PafLS (Loudspeaker analysis tool) ran inside PAFEC and was like "FINESuspension". The package consisted of a nonlinear FEA solver and a convenient GUI with loudspeaker suspension templates. I had a professional courtesy license but rarely used it. My solver of choice was ABAQUS for which I had written command files for spider, surround and diaphragm simulations.

From what I understand PafLS never really caught on. Have any of you guys out there ever used PafLS?

 

[Lars Risbo]

I use Comsol

 

[smowry]

I know that this is off topic but I just had to post this. @hartyewh1's comment gives support for Amir's references to proper ABX listening tests. However, folks like Everyday Audiophile and General* Danny Richie just don't understand subjective perception!

@hartyewh1​

1 day ago
I once spend two weeks trying expensive cables. Heard clear differences even when expecting almost assuming to hear nothing. Then did a blind ABX and all the effects dissappeared. Sighted tests imo are entirely worthless when it comes to these things.

My opinion on this topic is simply to utilize electronic crossovers and position the power amps as close to the transducers as possible and utilize short runs of 14 AWG OFC wire (active loudspeaker). Then let the longer runs between crossover/processor and power amps be balanced interconnects with the appropriate terminations. However, this does require power amps with balanced inputs and a processor with balanced outputs. Then inherently the balanced requirements eliminate the selection of obsolete electronics.

*In audio the designation of General identifies one who Generalizes.

 

[dfuller]

This is one of the more interesting threads on ASR as of late.

 

[AOR]

In response to the above post, I worked with Patrick at PACSYS on the development of PafLS. PafLS (Loudspeaker analysis tool) ran inside PAFEC and was like "FINESuspension". The package consisted of a nonlinear FEA solver and a convenient GUI with loudspeaker suspension templates. I had a professional courtesy license but rarely used it. My solver of choice was ABAQUS for which I had written command files for spider, surround and diaphragm simulations.

View attachment 440087

From what I understand PafLS never really caught on. Have any of you guys out there ever used PafLS?

We use PAFEC, with custom tools for defining geometry and results post processing.

 

[Anton S]

@smowry

Would you be kind enough to provide an example of what you consider a good Bl (X) curve for a 10" woofer with an Xmax of 6mm? Thanks in advance.

 

[smowry]

Anton S,

I have not practiced for more than 10 years, so I don't have Bl(x) example curves on my desktop. I am a hobbyist now. However, I consider your question/request to be a paradox. A 10" woofer with an Xmax of 6mm is simply not good. If we use the definition of Xmax as Bl(Xmax) = 0.7Bl(0.0) @ 10% THD, this would be at least moderately nonlinear without any significant flat plateau about x = 0. What I consider equally important is the symmetry of the even Bl(x) function about x = 0. Any offset about x = 0 is what Dr. Klippel calls Xdc, (DC Offset). So a good Bl(x) curve should have Xdc ~ 0.0. This is another paradox. In most cases, the motor geometry is highly asymmetric about the magnetic gap but we expect the Bl(x) curve to be symmetrical and this includes Purifi's low frequency transducers. With respect to typical motor cross-sectional geometry illustrations, this is obvious. If we didn't know better, one would thick that the motor was designed to operate in only one direction, DC, but that's obviously not the case either. The input signal is AC. The typical asymmetric geometry dates back to Rice-Kellogg's 1924 invention of the moving coil loudspeaker. Yes 100 years ago. Frankly, I consider this to be a technological embarrassment to the industry. How many transducer invention Patents have been granted since 1924?

Dr. Klippel's transducer non-linearities poster summarizes the above. https://www.klippel.de/fileadmin/kl...rature/Papers/Klippel_Nonlinearity_Poster.pdf

So having said all this, STEALLUS may have been ugly but the geometry was symmetrical about the gap and so is the geometry of the Borresen ironless motor assembly. Doug Button introduced a motor assembly that is geometrically symmetrical to JBL and is similar to the old Velodyne subwoofers that were referred to as Push-Pull drive but Button used Neo magnets. B&W made an attempt too. You can find documentation of these alternative topologies within this thread.

 

[smowry]

This is one of the more interesting threads on ASR as of late.

Thanks! Yes this is a transducer discussion. So if the ASR community finds transducers interesting, then perhaps we will see more.

 

[Anton S]

@smowry

Specifically, this is the curve I'm looking at, which shows a manufacturer's new version of a driver (red trace) vs. the previous version of the driver (black trace):

It looks fine to me. Perhaps you can give me your thoughts. Thanks in advance.

 

[smowry]

Firstly, the curve is extraordinarily linear; it looks like a line. However, something seems amiss here. The acquisition is hosed because the displacement limits were improperly set by the individual operating the Klippel DA. You indicated Xmax = 6.0mm when in fact this curve does not identify Xmax. Frankly, I have never seen a Bl(x) curve like this one, either from experience or from anything published. Having said that I must question the validity of the acquisition and believe that the curve above is controversial!

Suppose you post some information on the transducer that was measured. The curve is atypical and this implies that the transducer is atypical.

 

[Lars Risbo]

ideal Bl(x) is constant. period. symmetry does not mean linear ( seems to be a common misconception) but absence of even order distortion. The motor does not need to be symmetric to have a constant Bl(x)

Ironless: main advantage is that Lgen(x) is constant which eliminates force factor modulation, ie Bl is independent of the coil current. Secondly, the magnetic hysteresis distortion is eliminated once iron is removed. Two huge advantages.