4-Way, Symmetrical
Series Crossover Description
Paul Kittinger
September 5, 2001
When I started
thinking about what I wanted to achieve with this latest speaker
system design, the first goal was to have a higher than typical
sensitivity, around 93-95 dB relative to 2.83 v/1 m. I had on
hand four Morel MW266 8-inch woofers rated at 89 dB; with each
cabinet having a paralleled pair, their combined sensitivity
should be 95 dB. With crossover losses estimated at 1.0-1.5
dB, these Morel drivers would fit the bill. Another goal was
to minimize distortion by "sharing the load" amongst multiple
drivers, so I decided to go with a 4-way. Additionally, I wanted
the overall impedance versus frequency characteristic to be
as benign (flat) as possible. Lastly, since I don't have the
tools, knowledge or experience to optimize crossover design
with a computer, I decided to use a simple, series crossover
of the 1st-order flavour.
Looking at the way
a 1st-order, series crossover is usually wired, I realized that
a 4-way could be wired in either of two configurations, and my mathematical
paper analysis indicated that wiring it as shown in the accompanying
schematic would be beneficial. Referring to the schematic, a "normal"
wiring configuration would have the top leg of L3 connected to the
positive input terminal, just like L1 and L2, instead of being connected
as shown to the junction of the two midrange drivers. Also, the
bottom leg of C2 would be connected to the junction of the Morel
woofers and Vifa lower midrange (LMR) driver, instead of to the
negative input terminal as shown. Wired in this "normal" fashion,
the tweeter would "see" a 3rd-order roll-off, the LPG upper midrange
(UMR) driver would see a 2nd-order roll-off, and the LMR and woofers,
a 1st-order roll-off. [Note--when stating the order of a roll-off,
I'm referring to electrical roll-off, not acoustical roll-off.]
I deduced that wiring it as shown in the schematic would create
2nd-order roll-offs for the woofer and tweeter, and 1st-order for
the two midrange drivers. I built a breadboard using resistors as
drivers, and bench tested both wiring schemes. The breadboard measurements
confirmed my assumptions were correct, so I decided to use the configuration
as shown in the schematic and called it a "symmetrical" 1st-order
series crossover because of its equal (symmetrical) effects on all
four drivers. The main benefit of this is in providing additional
"protection" from low frequencies for the tweeter, and minimizing
the effects of break-up from the woofer above its normal pass-band.
With a pair of paralleled
8-ohm woofers, their combined impedance would obviously be nominally
4 ohms. Because I wanted to make the overall impedance be as flat
as possible, I initially designed and built the crossover with padding
resistors in parallel with the other three drivers to force their
individual impedances, as seen by the crossover components, to also
be nominally 4 ohms. To calculate the required values for L1-L3
and C1-C3, I used a Zeta of 1.0 and assumed driver impedances of
4 ohms. I ran impedance versus frequency for all drivers to determine
resonant frequencies and optimum Zobel values. With the Zobels as
shown, the paralleled woofers have an impedance of 3.8 ohms, each
midrange has an impedance of 6.0 ohms, and the tweeter has an impedance
of 5.2 ohms (at their relevant crossover corners).
Now, what
I didn't know and couldn't find anywhere was how to estimate how
much "gain" would be effectively imparted by a series crossover
to the midrange drivers. I also had a pair of Vifa P13WHs I wanted
to use, and had to choose the upper midrange driver and tweeter.
The tweeter choice was pretty easy; the Seas H881 has gotten quite
a lot of good press on several audio/speaker discussion forums,
and while its sensitivity spec is only 90 dB, it's an overachiever
as evidenced by its published response curve (an "S" of more like
92-93 dB). Going back to Bullock's original Speaker Builder articles
on multi-way crossover design equations, it would appear that in
a parallel 1st-order crossover, a band-pass (midrange) section
having corners two octaves apart would impart 8-9 dB of effective
gain. Even though I checked my calculations several times, this
seemed rather high, so I guessed I might get 5 or so dB. Thus, I
should be able to use the Vifa (88 + 5 = 93), and I then chose the
LPG 50FA, with 90 dB sensitivity, for the upper midrange because
of its overall response characteristic and a recommendation from
another hobbyist. Originally, then, I designed-in series attenuation
resistors for the two midrange drivers, with the paralleled padding
resistors previously mentioned placed in parallel with the series
combinations of attenuation resistors and midrange driver coils.
It's obvious that the schematic of the final crossover is missing
5 resistors; they were eliminated as testing progressed. The last
thing my design addressed was driver polarity; for lack of not knowing
any different, I simply wired all drivers in positive polarity.
The first
time I fired up these systems and listened, it was clear something
wasn't quite as expected. I was expecting a 3-6 dB (2-4 times) increase
in sensitivity over my previous speakers, but it was apparent from
the setting of the volume control, this wasn't happening. Also,
both low and high frequencies were very prominent compared to the
middle frequencies. Once I made some SPL measurements with my calibrated
MM2, it was obvious where the problem was; while pretty flat, the
range covered by the midrange drivers was very depressed. After
a little reflection this made sense; with 2nd-order roll-offs for
woofer and tweeter, the outputs of the midrange drivers were probably
out of phase. So I wired both midrange drivers in reverse polarity.
Now there was a much better balance to the overall sound and the
sensitivity was more or less where expected.
Over a period of about
2 months, mostly just on Sundays, I tried a number of changes and
made many sets of measurements. I found that the crossover imparted
maybe only 2 or 3 dB of "gain" to the midrange drivers and I eliminated
their series attenuation resistors in order to get as much as possible
from them. Measured at 2 meters on the tweeter axis, the overall
response was pretty good, except there was a 4 dB dip centred around
400 Hz that I just couldn't improve on, and I was beginning to worry
that the P13 couldn't keep up due to its 88-dB sensitivity. I tried
wiring just the woofers in normal polarity and the other three drivers
in reverse polarity. That was detrimental to the tweeter's output
and didn't get rid of the dip at 400 Hz, but it did cause me to
open up my thinking. Ultimately I wired the drivers in the polarities
as shown on the schematic, and the dip at 400 Hz disappeared. I
then eliminated the parallel padding resistors on midranges and
tweeter. At the end of my testing/listening, all of the component
values were as originally calculated.
Using warble
tones at 1/3-octave intervals and measured at 2 meters on the tweeter
axis, over the range of 125 Hz to 20 kHz the in-room response
fell within a +/- 1.5 dB window (mostly +/- 1 dB), with the exception
of a remaining 2-dB dip centred at 1600 Hz. [I think that dip
might be minimized by applying some felt or foam on the LPG and/or
Seas drivers' faceplates, which I'm going to try.] The impedance
curve over the whole frequency range is, indeed, very benign; its
maximum is 6.4 ohms at 42 Hz, its minimum is 3.7 ohms at 120 Hz,
and it remains between 4.4 and 5.0 ohms from 200 to 20 kHz. And,
based on measurements with pink noise input, the final sensitivity
came out to be 92.6 dB (re: 2.83 v/1m).
The driver arrangement
on the baffle is a bit unusual. Starting at the top, they're arranged
as LMR, tweeter, UMR and woofers. The reason for this is that the
cabinet was designed and built with each woofer having its own,
L-shaped, short hybrid transmission line, with both lines venting
out the back panel, above the mid-point. Because of this TL configuration,
I had to mount the P13 at the top. I don't believe this is the best
location for it relative to its output integrating well with the
other midrange. If, instead, the Vifa was located between the woofers
and the dome midrange driver, I believe both midrange drivers would
work better wired in reverse polarity (and woofers and tweeter wired
in positive polarity). Furthermore, I would expect the dip at 1600
Hz to disappear due to better integration of the outputs of the
midranges. Just for "kicks", at one point I wired up the crossover
in the "normal" configuration and made some measurements. While
somewhat different in the specifics, the results were not particularly
better (or worse), so I stuck with my "symmetrical" arrangement.
About the short, hybrid
transmission lines, these did not work at all as expected, even
when stuffed at the recommended high densities. Ultimately, I simply
installed Scan Speak resistive vents in the TL outlets on the rear
panel to convert the woofers' enclosures to aperiodically-vented,
"sealed" boxes with stuffing densities of 0.6 lbs/cu.ft. As configured,
I measured an f3 of 43 Hz, an f6 of 36 Hz, and an f10 of 25 Hz.
The cabinet is 48" tall, including base, 11-1/2 inches wide, and
15 inches deep. The woofers and P13 are mounted on 3/4" projections
from the baffle to approximately time-align them with the two domed
drivers, and each woofer sees a net cabinet volume of about 1.25
cubic feet. With these cabinet volumes and the stuffing, I calculated
that the final Qtc should be about 0.75.
I take
a very pragmatic approach to choosing components. Coupled with either
less-than-golden ears or, perhaps, having a more easily pleased
taste in musical sounds, I almost never get carried away with using
exotica. Therefore, I used Solen metallized, polypropylene capacitors
for C1 and C2. On the other hand, I used a bipolar electrolytic
for C3 because 140 uF of polypropylene is very expensive, not to
mention huge. I bypassed C3 with a small-valued Mylar capacitor
I had on hand. All of the Zobel capacitors are Mylar, and the three
inductors are wire-wound, air cores.
As to the
sound, I'm very satisfied. I'm unaware of the individual drivers
as the sound has a wholeness and seamless quality. The higher than
typical sensitivity allows my power amplifier to almost "loaf" most
of the time, and coupled with, perhaps, the attributes of a series
crossover, the dynamic capabilities have to heard to be appreciated.
This system should be a really easy load to drive for any amplifier,
and would be expected to be especially nice to a tube amplifier.
If I get an itch to build something in the next 6-12 months, I'll
probably just build a new cabinet for these same drivers and crossover
assembly, but optimize the driver location and enclosed volume for
the woofers.
