AudioBanter.com - View Single Post

(Dick Pierce) wrote in message ...
(Svante) wrote in message news:hO3db.435049$Oz4.241458@rwcrnsc54...
I have been reading older posts in this group regarding the inductance
of the voice coil, and I realize that the inductance is far from
ideal, that its inductance drops with frequency (in one way of looking
at it). I also read an article by Marshall Leach (JAES no 6 2002, p
442-) which describes this phenomenon quite nicely and presents a
model that I have succeded in fitting very well to a small number of
real speakers I have had at hand. I actually carved the cone and coil
away from one element and measured the impedance curve again, and
indeed the impedance was quite different. Apart from the obvoius (the
peak corresponding to the mechanical resonance disappeared) the
inductance now looked like an inductance normally would, ie with a 90
degree phase shift and a +6dB/oct slope above some frequency (for the
disected loudspeaker, nb).

First of all, your data is inconsistant with Leach's and also with
Lip****z and Vanderkooy's, as well as my own, where it is the
mechanism of lossy eddy current generation in the pole piece and other
moederatey conductive parts in proximity to the voice coil that is
responsible for the non-ideal behavior of the inductance.

The phenomenon we observe is that as we go higher in frequency, we
see an increase in eddy current coupling in the metal parts, which
have a fairly high ohmic loss. The result is a decrease in inductance
and an increase in resistance that pretty reliably follows a 1/sqrt(f)
for inductance and sqrt(f) for resistance.

Yes, that is what I see too, approximately. So, I don't think our data
are that different.

However, I describe what happened when I cut a loudspeaker element
apart, removing the coil and cone from the magnet. When I did that,
the coil behaved very much like an ordinary coil, so my point was that
there IS a huge difference between the inductance of the normally
mounted coil and the "free-air" coil.

Now, there are secondary and tertiary effects due to cone resonances
and the like that can be visible, but they tend to be fairly narrow-
band phenomenon that do little to modify the overall trend of the curve.

If you want to experiment to verify what's being clained, try the
following: take your "modified" driver and remove the voice coil entirely
from the magnet and measure the inductance. Here, you'll find almost
perfect agreement with a simple series RL circuit model.

That's what I did.

Now, place
the voice coil back in its proper position in the gap, and fill the
gap with epoxy, solidly locking it in place, so it cab't move at all,
then measure the impedance again. You'll see the very strong non-ideal
behavior, with no cone resonances or other mechanical phenomenon.

Haven't done that (have you? ;-) ), but I agree that that should
happen.

One thing that is often done is to put a thin copper cap over the pole
piece. This acts as a shorted turn and prevents the generation of the
eddy currents in the iron pole piece and you see a more ideal inductance
behavior. I did a similar albeit pointlessly heoic experiment where I
had a pole piece sliced with a thin slitting saw about 4-5 mm apart
across its face, and saw a similar improvement in the bahavior
(improvement in the sense that I no longer saw the 1/sqrt(f) dependency).

(No, really, the guy running the machine shop owed me the favor,
I had access to raw speaker parts and assembly facilities, and
had several assemblies where I could swap out the basket/cone
assembly in one piece from magnet to magnet with little effort!)

This brings up the question what the manufacturers say in their data
sheets. They mostly specify an inductance. They rarely specify the
"lossiness" (the n value in the article above) of this inductance or
the frequency at which the inductance was measured.

That's because, as mentioned above, the "lossiness" is due to a
common mechanism for all drivers. They all suffer from these eddy
current losses, they all use low-carbon mild steel pole pieces
and fron plates which have the same bulk electrical/magnetic
properties.

So, what does the "voice coil inductance" figure that the
manufacturers supply us with mean, really?

In many case, since they are unable or unwilling to specify the
inductance any further, they're pretty useless. In general, most
manufacturers measure the inductance at 1 kHz. If you assume the
sqrt(f) behavior due to eddy current losses, you'll find that
predicting that the inductance at 10 kHz is about 1/3 that at 1
kHz (for woofers or for any other driver where we are well above
mechanical resonance) gets you real close.

I certainly agree it is pretty useless. Unless they specify at which
frequency the inductance is measured, the impedance cannot be matched
to the sqrt(f) behaviuor, and the figures cannot be used. But of
course, such a figure may not help the sales, since the speaker looks
more "non-ideal"... ;-)

I think you answered my question in this paragraph; "1 kHz" was what I
was looking for. Have you seen this frequency in datasheets or how do
you know this?

Thank you for your elaborate answer, I have appreciated your earlier
postings (quite a few!) too.