[net.audio] Crossovers - Throw the book away?

593aac@houxa.UUCP (S.JOHNSON) (10/17/85)

I read with interest a recent posting about crossover design circa 1952.
Bob, you're certainly on the right track by using film caps instead of
electrolytics, but unless you're very lucky, the design right out of the
book is not going to work very well.

The problem is that the book usually makes some simplifying assumptions:
the drivers present constant, purely resistive loads, and that the
acoustic frequency response of the drivers is perfect. In addition, most
experts in the field consider that 18db per octave is the gentlest
crossover slope to use with practical drivers (12 db networks put a nasty
notch at the crossover freq. unless you reverse the tweeter's phase, and
then you get some rather strange phase effects!) (6db networks are
theoretically perfect for ideal, physically coincident drivers but
don't work so well with most (all?) real drivers).

So in the past, most designers have either just used the book values,
or have empirically come up with other designs. Neither is acceptable
from my view. What is the answer???

1) Make each driver look like a constant, resistive load to the
crossover. By the way, assuming that the impedance is 8 ohms 
because the manufacturer says so just won't cut it. It's almost
certainly 6.2 or 5.1 or 9.7, etc. For the most part the Zobel network will
do this! It involves a capacitor and resistor across the driver in the
case of woofers  to compensate for the voice coil inductance
( no attempt to compensate for the fundamental resonance)
and a slightly more complex network for mids and tweeters to compensate
for the motional impedance at resonance as well as the voice coil inductance.
Note that ferrofluid often damps the fundamental resonance enough to simplify
matching problems for mids and tweets.

2) Don't assume that the frequency response of the driver is perfectly flat.
Tweeters do tend to rolloff at 1 Khz or 2 Khz, etc. Add the acoustic
rolloff to the electrical rolloff of the crossover. For example if the
tweeter is rolling off at 12 db per octave at 2000, a 6db electrical
network at 2000 will yield an acoustic crossover of 18 db per octave.
The other approach is to use a sharp rolloff rate (18 or 24db) per octave
far from where the driver rolls off and ignore the driver's rolloff.

Using these techniques, book designs can be made to work. And very well
indeed!

One last point, don't wind your crossover coils with small gauge wire.
The resistance, in many cases, will degrade performance appreciably
by throwing off your design. I use 18 guage in small coils, and 10
guage in my largest coils. Film caps are better than non-polar
electrolytics, never try to make your own non-polars by wiring
standard electrolytics back to back.

		Steve Johnson

(The above opinions are definitely my own)

rdp@teddy.UUCP (10/18/85)

In article <808@houxa.UUCP> 593aac@houxa.UUCP (S.JOHNSON) writes:
>I read with interest a recent posting about crossover design circa 1952.
>Bob, you're certainly on the right track by using film caps instead of
>electrolytics, but unless you're very lucky, the design right out of the
>book is not going to work very well.
>
In most cases, they don't work at all!

>
>2) Don't assume that the frequency response of the driver is perfectly flat.
>Tweeters do tend to rolloff at 1 Khz or 2 Khz, etc. Add the acoustic
>rolloff to the electrical rolloff of the crossover. For example if the
>tweeter is rolling off at 12 db per octave at 2000, a 6db electrical
>network at 2000 will yield an acoustic crossover of 18 db per octave.
>The other approach is to use a sharp rolloff rate (18 or 24db) per octave
>far from where the driver rolls off and ignore the driver's rolloff.
>
One of the most ignored characteristics of wide band drivers (woofer-midranges)
is the fact that the on-axis response oftyen shows a tendency to rise
with rising frequency, due to the fact that the radiation angle is reduced,
but show an integrated power response that drops with rising frequency.
A network can compensate for this, but the choice of response is a compromise.
Also (my favorite bitch), cones made from paper have the disconcerting
tendency to change these characteristics with changing humidity!

>One last point, don't wind your crossover coils with small gauge wire.
>The resistance, in many cases, will degrade performance appreciably
>by throwing off your design. I use 18 guage in small coils, and 10
>guage in my largest coils. Film caps are better than non-polar
>electrolytics, never try to make your own non-polars by wiring
>standard electrolytics back to back.

I just spent the last ten minutes searching my office for the notebook
holding about 100 AES and JAS reprints, and, to my anger, discovered
that someone has walked off with it. DAMN!

But, I will try to relate this as best I can from memory. In the late '70s
an article appeared in the Audio Engineering Society about inductors for
use in crossovers. In it the author, A. N. Thiele (yes, that Thiele) 
described a novel and simple method for designing and producing coils
for crossover use. The name of the article is "AIr Core Inductors for
Audio Use), or something very much like it. Without the article in hand,
it is difficult to summarize things, but the method described allows
accurate production of coils where all parameters (inductance, resistance,
etc.) are controllable as needed by the user. When I find the b*****d
who took my articles, I will post more information.

Dick Pierce

rdp@teddy.UUCP (10/18/85)

In article <1452@teddy.UUCP> rdp@teddy.UUCP (Richard D. Pierce) writes:
>
>I just spent the last ten minutes searching my office for the notebook
>holding about 100 AES and JAS reprints, and, to my anger, discovered
>that someone has walked off with it. DAMN!


Well, sure enough, I'm getting to the bottom of this heinous crime! I
just discovered that the culprit did, indeed, steal my book, but he had
the chutzpah to sneak an EXACT DUPLICATE into my briefcase! The nerve of
some people today!! He even went so far as to cover it with MY fingerprints!


>But, I will try to relate this as best I can from memory. In the late '70s
>an article appeared in the Audio Engineering Society about inductors for
>use in crossovers. In it the author, A. N. Thiele (yes, that Thiele) 
>described a novel and simple method for designing and producing coils
>for crossover use. The name of the article is "AIr Core Inductors for
>Audio Use), or something very much like it. Without the article in hand,
>it is difficult to summarize things, but the method described allows
>accurate production of coils where all parameters (inductance, resistance,
>etc.) are controllable as needed by the user. When I find the b*****d
>who took my articles, I will post more information.
>

The exact title of the article is: "Air-Cored Inductors for Audio",
initially printed in Proceedings of the I.R.E.E (Australia), vol. 36,
pp. 329-333 (Oct. 1975), later reprinted in the Journal of the Audio
Engineering Society, vol. 24, no. 5, pp. 374-378 (Jul. 1976),
with a followup in the Journal of the Audio Engineering Society, 
vol. 24, no. 10, pp. 830-832 (Dec. 1976).

Dick Pierce

mohler@drune.UUCP (MohlerDS) (10/20/85)

I agree whole heartedly with your posting, however I thought I might add
one quick comment. If the driver whoose impedance you are trying to linearize,
(with a shunt) is mounted in the enclosure it will be used in while testing the
effectiveness of that shunt. You will get a more accurate compensation effect.

		David S. Mohler
		AT&T - ISL @ Denver
		drune!mohler

rdp@teddy.UUCP (10/21/85)

In article <63@drune.UUCP> mohler@drune.UUCP (MohlerDS) writes:
>
>I agree whole heartedly with your posting, however I thought I might add
>one quick comment. If the driver whoose impedance you are trying to linearize,
>(with a shunt) is mounted in the enclosure it will be used in while testing the
>effectiveness of that shunt. You will get a more accurate compensation effect.
>
>		David S. Mohler
>		AT&T - ISL @ Denver
>		drune!mohler

rdp@teddy.UUCP (10/21/85)

In article <63@drune.UUCP> mohler@drune.UUCP (MohlerDS) writes:
>
>I agree whole heartedly with your posting, however I thought I might add
>one quick comment. If the driver whoose impedance you are trying to linearize,
>(with a shunt) is mounted in the enclosure it will be used in while testing the
>effectiveness of that shunt. You will get a more accurate compensation effect.
>

No, this is not so. The shunt capacitor/resistor method of linearizing
impedance takes care of the driver's voice coil inductance. This is
pretty much unaffected by mouniting/enclosure, as it is well above the
region affected by the enclosure. For example, one notices a (nearly)
montonic rise in impedance in a typical 8" woofer above 500-1000Hz due 
to the voice coil inductance, where the enclosure affects the impedance curve
due to the resonance, which is usually restricted to below 100 Hz.

Dick Pierce