David Smith worked as a
design engineer at JBL from 1980 to 1983. During that time, David was responsible for such systems as the L15 and L46 for the Consumer Division
and 4401, 4411, 4430, 4435, 4612 for the Professional Division. He would
later work for KEF Electronics, Meridian America, McIntosh Laboratories,
Snell Accoustics, and currently, Lenbrook Industries. The following are
anecdotes provided by David on his involvement at JBL.
I can remember some stories that
might interest your readers: I was thinking about the test setup. There was
a small anechoic chamber but it wasn't very good. Most measurements were
done outside on a large asphalt pad. You could bury a speaker flush in a pit
(which created big issues in the design of the L212, (see below) or you could
do ground plane measurements. With ground plane measurements the mic is
placed on the ground. Outdoor measurements, well away from buildings, are in
a semi anechoic environment the exception being a reflection from the
ground. If a speaker as placed in a typical position, say a couple of feet
above the ground with the mic at a typical listening height of 40 inches
there will be a second sound arrival from the ground bounce midway between
speaker and microphone. The second arrival is time delayed (it took the
longer path down to the ground and then back up to the microphone) and the
measurements will have severe comb filtering from this delay. Now as you
move the microphone nearer the reflective surface then the time difference
diminishes to the point where it is negligible if the microphone is in
contact with the surface. The system has to be tilted downward or stood on
its head to make sure the measurements are on the system axis that you are
interested in. The boundary surface adds 6dB to the measurement that you can
easily calibrate out.
Now there were two practical
problems to outdoor measurements: airplanes being the first. Unfortunately
he Northridge factory is just adjacent to the final approach for Van Nuys
airport, one of the busiest private airports in the country. Lots of C130
transports would take off and land there and usually in the middle of an
important measurement! Heat was the other issue. The hottest day I remember
hit 112 degrees. (Ah, but its a dry heat!) Woofers didn't seem to mind but
tweeters would have their materials and glues soften to the point where the
highs would roll off prematurely (giving nothing above, say 10kHz). If you
had an important project going in the middle of the summer you would start
measuring early in the day before the sun was at full height.
We did a lot of large horn
design in the early 80's. Don was working out most of his constant
directivity horns and needed a means of testing their directivity. He had
created a very clever measuring system. B+K had a turntable for speaker
rotation that would be used in synch with circular chart paper, but in the
end you had a picture of polar performance but no data (DI, beamwidth, Q
etc.) Don had built a more sophisticated system from scratch. He had taken a
huge multitoothed gear out of an old machine. It was probably 18 inches
across with several hundred teeth. A good sized motor would drive it while a
micro switch would count teeth and hence the rotation. An early computer ran
the device (DOS or earlier software?) and an old two-box B+K 1/3 octave
analyzer would take measurements and send the data to the computer program
for number crunching.
It ran something like this: Set
the device on a stand on top of the turntable with the microphone on the
zero axis. Tell the program how many measurements to take or how many
degrees per step and let her rip. It would turn to the first position and
blast away with pink noise. I think it did three bursts and looked at the
nose floor in the intervening silences to confirm the measurement validity.
If satisfied it moved to the next position. SHH...SHH...SHH, turn and
repeat. I remember we got complaints and a police visit from the apartment
building to the North once. They were convinced we were doing something
nefarious (Don just wanted a good signal to noise for the measurements.)
Any of the brochures that show
1/3 octave plots of Directivity Index or power response, such as the 4430/35
monitors or the theater horns, would have been measured on this apparatus.
The
picture to the left is of the rooftop measuring pad at
Northridge. I think it was reached at the top of the stairwell
situated by the old anechoic chamber. I think this preceded the
ground level pad and was used less after the ground pad came in,
as I can only remember going up there once or twice. The ground
plane pad had several test options: one was a 10 cubic foot box
with interchangeable baffles for any size woofer, midrange or
tweeter. Secondly was a larger pit (4 or 5 foot deep) that full
size systems would be placed in. 4x4's of lumber would be
stacked up to get the front flush and then a few pieces of
plywood would skirt around and give effective 2 pi (half space)
baffling.
Measuring a system flush
in the ground gives a very clean curve. AR (and others) used the
technique for years. Now the only problem with measuring in 2 pi is
that you've removed the box edges and baffle area from the equation,
factors that definitely exist when you get the speaker home. Text
books show a 4pi (full space) to 2pi transition curve that is
generally a low frequency rise of 4 to 6 dB that lasts up to a
frequency where the system's baffle area is adequate and the box
itself represents 2 pi. I say "generally" because there are usually
extra wiggles on top of the basic trend. Typically, although there
is a 4 to 6 dB gain below,say, 200 there is an octave with a few dB
of loss just above that frequency. So a system that is designed to
be perfectly flat in 2pi will have significant response aberrations
when free space mounted. It would have the mirror of the curve
described, that is it would be weak below 200Hz and overly prominent
above.
That brings us to the
L212.
The original concept
for the L212
was devised by Ed May, but he left JBL before the system was
developed. Lorr Kramer, Greg Timbers and Terry Sorenson were
responsible for engineering of the production model. Terry was a
fun guy, one of the crew when I was there. Terry was studying
for an MBA at night so he would frequently come in and preach
the Religion-of-the-Free-Market to us. Engineers like to debate
so that was as good a topic as any.
Anyhow the 212 had a
removable base that, once removed, allowed the system to be
buried flush in the pit. The crossover was designed to give a
highly flat system under those conditions. Unfortunately once
removed from the pit the 2pi to 4pi effect was enough to mess up
the response to an appreciable degree. Not that it was horrible
sounding, just that it wasn't the balance you would have given
it if you thought about it. I remember that the system was
reviewed by High Fidelity. They used CBS labs for measurements
and did a Floyd Toole type axial response/front hemisphere/total
sphere set of measurements. Sure enough the response showed a
pretty good mound around 200Hz (the secondary effect) that the
reviewers commented on. I don't remember if a running change was
made to fix it, but at least a lesson was learned that 2pi flat
was not the answer and no later systems fell into that trap.
This was one of the main incentives for developing the ground
plane measuring technique.
By the way I had one
of the 212 systems and liked it a lot. It used great drivers
including the very expensive 112A mid bass unit. This was the
only 8" JBL driver to use a 3" coil. Terry showed me a 3 element passive notch circuit
that was developed to fix the response and so it sounded quite
good. When I moved to KEF in the UK, this is the system I took
with me (I sold my KLH 9's to Randy Patton).
Instead of the 12"
subwoofer that came with it, I had a prototype of the domestic
18" sub that we were developing (the 2245 based 460). That was a
woofer!
Terry Sorenson was
the key guy on the SFG redesign. Terry wasn't a system designer
(crossover networks) to the extent that Greg, or even I, was.
But he was better with regard to theory and had a Physics
background. The SFG story was that the ferrite magnets
structures were designed, as you know, because of the rising
cobalt prices. Ferrite structures with the same parameters of
gap dimensions and flux density were made up but when listened
to they sounded different. Further measurements showed much
higher second harmonic distortion. The flux modulation ring
dealt with the 2nd Harmonic and the undercutting also helped the
low frequency distortion due to the symmetry. Mark Gander's AES
paper shows typical curves. The potential demagnetization of
Alnico was also noticed at the time. I distinctly recall Greg
Timbers having fun with modifying the parameters of Alnico
woofers at will during development: "I think the Q is a little
low on this one. Maybe a 2 dB drop in sensitivity would be about
right. Lets give it full output for a second from the Crown
DC600....Brrrapp. Curve it again and, yes, that's about right."