T
he 4430 and 4435 Bi-Radial Studio Monitors were amongst the
most successful professional loudspeakers ever produced by JBL. They were in
production longer than any other JBL main studio monitor, being introduced
in 1981 and not discontinued until 1996 for the 4435 and 1999 for the 4430. David Smith
was the engineer responsible for the system design of these monitors. The
following is his account of the background and design of these
systems.
2, 3 or 4 way?
A lot of the systems
that preceded the 4430/35 were 3 or even 4 way designs. Adding an 8”
lower midrange would certainly improve power handling and also clean
up the sound at high levels where the woofer's excursion gets
significant. Putting a crossover between the main horn and a small
super tweeter type horn is problematic, though. The main horn
usually had a lot of depth whereas a super tweeter (such as a 2207)
would be considerably shorter. There would be several wavelengths of
separation if they are both front mounted on the cabinet, leading to
comb filtering in the crossover region. Pushing the supertweeter
back to the point where the voice coil planes are aligned can remove
the comb filtering although it isn't always practical and doesn't
work over much of a vertical range of angles. If your main horn has
the bandwidth then you are better off equalizing a bit than crossing
over to a super tweeter at a high frequency. (This is just my
opinion and I do realize that a lot of great JBL speakers were
designed contrary to this.)
|
I
came to JBL in September of 1980. It was my
second job in the industry. I had started out in an obscure OEM speaker
company called Essex Cletron in Cleveland. When Essex announced that they
were moving the engineering department to Indiana I thought it was a good
time to move on. Interestingly, soon after their move they were bought by
Harman and turned into Harman-Motive, a long lasting and highly profitable
sister company for JBL.
At the time JBL had a good line of studio
monitors including the 4350, the 4343 and the 4315 (a product I much
admired). The 4343, and especially the 4350, were very large and were
sometimes referred to ironically as "Japanese bookshelf speakers". This is
not a slur on their quality, but more reflected their great popularity in
the Asian market. As a practical issue their sheer size made them a little
over the top. Studios typically build their main monitors into soffets over
the control window and the 4350 was just too big. The biggest problem,
though, was that UREI was stealing market share with their 811 and 813.
These were based on an Altec 604, always a popular unit, with a "time
aligned" network and some enhancements to the horn. The sound was good
enough and the size was reasonable and it had a good story in the time
aligned network.
Don Keele had just gotten his constant
directivity horn design software going. You might know that he had really
pioneered the first constant directivity horns while at EV. He also patented
their design. Altec made thinly veiled copies with their "Manta Ray" horns.
Don, now at JBL, was to develop a line of horns and at the same time needed
to change the design enough to get around the other patents. The biradial
horns were the answer and they were extremely good. One of the first units
he developed was a 100 by 100 degree horn. The larger 90 by 40 and 90 by 60
horns were more for theater applications but this smaller, wider angle horn
was a natural for a studio monitor. If you look at the polar patterns that
are in the AES paper we wrote you can see that the polars are incredibly
uniform from 1000Hz to 16,000 Hz.
Why does the 4430's horn roll off?
It
doesn't, actually. All compression drivers have an inherent roll off
related to their diaphragm mass. This “mass breakpoint” usually
occurs around 3 kHz. In the lab we would measure a compression
driver on a terminated tube. In practice this is a long pipe
(perhaps 4 to 6 ft long) with a thin fiberglass wedge in it. A
small microphone is inserted into the side of the pipe near the
compression driver. This presents the compression driver with a
resistive acoustical load so that pressure in the tube represents
what the output of a perfect horn would have. Measured in this way
most of the JBL compression drivers would have flat response from
several hundred Hz to about 3kHz. At 3k they would start a gentle
rolloff at 6dB per octave until phase plug design and diaphragm
modes took over above 10kHz. You may know that Fanchur Murray did
all the compression driver design at JBL in the 80's. In my opinion
his greatest achievement was the diamond surround and getting the 2”
and 4” compression drivers to have smooth response out well beyond
15kHz.
Anyhow, this rolled off response is the raw response that is
presented to the horn. The horn then modifies this response via its
acoustic load at the low end its and via its directivity index at
the high end. (Directivity index represents the on axis gain. It
is essentially the “beaminess” of the horn expressed in dB.) Prior
to the constant directivity horns most people assumed that
horn/compression driver combinations should have inherently flat
response. Horns were evolved that in effect “equalized” the
compression driver. This ignored the fact that the combinations
could only be flat, and only on axis, if the horn exhibited a great
amount of high frequency beaming.
With
the 100 by 100 horn the polars are so consistent and the directivity
index so flat that the on and off axis response looks just like the
terminated tube measurement of the compression driver. It is up to
the crossover network to take this rolled off response and makes it
flat.
|
To explain the biradial name, there were a
lot of radial horns at the time. Probably the best known examples were the
Altec 511 and 811. They took a vertical cross section designed for
exponential area growth and rotated this cross section around an apex back
in the throat (in effect a radial swing). Slice such a horn from front to
back at any angle (through the vertex) and the cross section would be the
same. Designing a horn in this way gave good extension to its theoretical
cutoff, but poor polar response and a lumpy frequency response that
contributed to “horn colorations”. Two independent variables were at play
here. The first, the rate of area growth, defines the acoustic load that
governs the radiated power curve for the first octaves above cutoff.
Independently, the wall contours determine the polar pattern versus
frequency. It works out that the side wall angles near the compression
driver primarily determine the high frequency directivity. Contours farther
out the horn, as the dimensions grow, progressively set the polar curves for
lower and lower frequencies. Don developed a sidewall contour that gave a
great polar pattern for a wide range of frequencies. These contours (or one
contour used twice in the case of a 100 by 100 degree horn) were then used
for the horizontal and the vertical shape. With his new designs the special
contours were rotated twice around two radial points, hence the name
biradial.
One part of the mechanical development of the
4430/35 was of the 100 by 100 horn castings. I remember working with Mark
Gander on their construction. First samples were from a low tech molding
process (reaction injection molding?) and always a little warped. The back
surface that would have to seal to the cabinet tended to bow forward.
General quality was much improved by the time production rolled around.
As an aside, a lot of the evolution of horns
has been tied to their construction processes. Early horns were usually of a
multicell design because a tinsmith could solder them up. More complex horn
shapes wouldn't become practical without molding techniques. Horns that were
expected to sell in volume could be cast in aluminum, such as the Altec 511
and 811. Don's big theater horns would have been cost prohibitive (and
heavy) if aluminum diecast due to their large size and relatively low
projected sales volume. They ended up being made in fiberglass with
reinforcement panels molded in. The 4430 horn was a little tricky because it
needed side extraction of the tool for the lateral pockets. It also had a
separate sand cast throat section that was bolted on from behind and linked
the front to the compression driver. Getting the cross section and the
juncture between the two just right impacted the response so we played with
that variable a fair bit.
Don had just started on a crossover network
when I took over the project. I remember that there was a lot of work to get
the midrange and tweeter controls to work sensibly. Also a lot of work to
get the octave to octave balance right and also to insure the balance of the
4430 and 4435 were identical even though the 4435 was three dB more
sensitive. There were a number of listening sessions with Gary Margolis and
John Eargle which I, as a young engineer, found very instructional. They had
good ears and could identify what octave needed to go up or down a dB to get
the balance just right and I wanted to be able to do that!
How to improve
the 4430/35
The only real negative
of the biradial horn designs was that the concentration on
great polar response sometimes was at odds with the ideal of exponential
area growth. This resulted in some ripple to the first couple of octaves of
the horn's response.
The bottom end of a horn/compression driver
combo is largely determined by the rate at which the horn's area grows. Back
in the acoustical gramophone days mathematicians had figured out that
Exponential area growth gave the most extended acoustical load to the
transducer. Exponential growth simply means that the cross sectional area
grows a constant percentage for a fixed unit of length along the horn. The
slower the area grows the lower the cutoff frequency. For example a 500
cycle cutoff would dictate that horn area should double every 38
millimeters. A 250 cycle horn would grow half as fast with area doubling
every 76 mm. For a horn to work well to a certain frequency it will need
appropriate exponential growth and also a certain minimum mouth area. You
must have both. Having a low flare rate but with inadequate mouth area will
lead to choppy response in the horns first octaves.
In the original biradial horns the only truly
exponential part was the short section from the exit of the compression
driver to the gap that fed the horizontal flare. Making the gap narrow gave
great horizontal polars up to a high frequency but pushed the exponential
cutoff frequency down to the point that the area didn't support it. The
consequence was a periodic ripple in frequency response. It wasn't real
bad...about 2 dB total ripple.
If you look at the impedance curve you see
impedance peaks that correspond to the response peaks. This gave a challenge
in the network design. You needed to keep the networks driving impedance low
otherwise the voltage at the compression driver would start to follow the
ripple of the impedance curve and make the response worse.
With the 4430 network the driving voltage had
a ripple of about 1 dB and so the horn's total ripple was about 3 dB. The
consequence of the ripple was a slight coarsening of the sound of the
midrange, most noticeable on pink noise. I always thought that biamping
would help a little here (directly coupling the amplifiers low output
impedance to the compression driver) but what would be really neat would be
to drive it with and amplifier with a slightly negative output impedance.
This is can be achieved with various amplifier feedback schemes. Then for
every frequency where the impedance bumps up the amplifier output voltage
would actually drop. A negative output impedance amplifier would act to
equalize the first couple of octaves of the horn's response to something
really smooth. Any experimenters out there up to the task??
|
Equalization of the horn/compression driver was the only out-of-the-ordinary
issue for the 4430/35 network. As mentioned above the high frequency
section will have an inherent 6dB per Octave rolloff above 3kHz. A
passive network can equalize that as long as the inherent sensitivity stays
high enough to the highest frequency needed. The sensitivity of the
2425 compression driver on the 100 by 100 horn was about 108 dB from 1000 Hz
up until its rolloff above 3K. By 16K the response had dropped
to about 94dB, just enough for the 93dB target of the 4430 and near enough
for the 96dB 4435.
The upper section of the network started out as a second order filter
followed by an L-Pad for overall treble level and with a first order bypass
(a small capacitor around the L-Pad) to give the highest frequencies a
path around the L-Pad. Basically the L-Pad would pull the lower treble
down about 16dB, the bypass would push the highs back up and equalize the
horn.
A
couple of issues needed to be dealt with: First, the first order
bypass approach still lost a couple of dB at 16kHz, leading to a softer top
end than was desired. Secondly, it would be nice to have sensible
response controls for the system. The L-Pad, because it was bypassed,
would only have effect from 1 to 5kHz. It would become a “lower
treble” control. I wanted separate controls for both the lower treble
and upper treble, more in keeping with our other 3 and 4 way systems.
A
solution for both was to use a series resonant bypass network. 1
microfarad in series with about .08 milihenries would resonate around 15K
and give about 2 dB more output for the highest frequencies. It also
gave a little less around 5k where the 2425/biradial combo was a little hot.
A variable resistor in series the resonant leg gave a nice “upper treble”
adjustment so we now had a pair of controls with really useful control
centers and range.
After getting the horns working the woofer section was tackled. The
woofer section was a straight second order network with a conjugate to
flatten out the woofer's inductance. Using a conjugate lets you
achieve a more “classic” looking second order rolloff since the network
inductor and the woofer's inductance wouldn't be interacting. Most of
the work in this network was in getting the best blend between woofer and
horn. When the woofer and network shape looks about right you have to
see how the sections add. Do they add well in phase? Or in
reverse phase? At what axis do they sum best, and is that where you
want them to sum best? It turns out that with the crossover slopes
used and the relative depths of the woofer and horn that the units gave best
summing connected in phase (++ we would say) on a measuring axis that was
straight out or slight rising. The 0 degree and 15 degree up curves
were both about equally good. This would work well if the system was
floor standing or even if inverted and mounted over a studio window.
At this point we now have a system with all the sections working and giving
a reasonable response curve. We still aren't done though. At
this point you call over the marketing guys with the golden ears and
everyone has a listen. The curve will be massaged a dB here and a dB
there until the group (primarily Gary Margolis and John Eargle for this
product) is happy.
I was keen to give a paper on the monitors.
Since they were the first to use a constant directivity horn there was some
merit to a paper and so I received permission to write it. I did most of the
writing, although John Eargle corrected a lot of my grammar and finessed the
introduction and my wife captioned the illustrations! If you read between
the lines of the paper you can see that we were taking some potshots at the
UREI product.
I was pleased that the products were well
respected by the Pro community and stayed in the line for a long time. The
4400 series grew after these models. I designed a 4401 and 4411 and after I
left JBL Greg designed a baby 4430, the 4425. One thing that was amusing at
the time was how people reacted to the shape of the new biradial horns. I
remember more than one studio engineer referring to them (affectionately!)
as the “Dolly Parton” horns or “Baboon butt monitors”.
© David Smith, February 2005
|