The piezo plate adds the top octave. The JBL 2440 on its 2350 horn is flat to about 9 kHz, then drops fast, down 6 dB at 10 kHz and 16 dB by 11.6 kHz. The plate is a slim strip of eight piezo tweeters that sits level on top of the 2350 and covers the same 90 degrees side to side.
It has no amplifier or crossover of its own. It rides on the horn channel, wired in parallel with the 2440, and the piezos' own rising impedance does the crossing over.
Fig. 8-2. Front view, 31-1/2 in wide by 6 in high. Dashed lines are the facet seams.
BUILD
Measure a piezo flange before cutting anything. The original datasheet gives a 3.34 in square flange; the GRS clone is listed at 3-1/4 in round. The 4-3/16 in facets leave about 7/16 in each side of a 3-1/4 in flange.
Cut eight facets 4-3/16 in wide. Bevel every seam 5-5/8 deg on each edge, so neighbors turn 11-1/4 deg. Cut a 3-1/16 in hole centered in each facet.
Top and bottom panels follow the arc and stand 1 in proud of the face. Side walls stop 1 in past the outer facets. Join the outer facets to the side walls with glue blocks.
Round every exposed plywood edge with a 3/8 in corner-round bit. Leave the beveled facet seams square: they are glue joints.
Mount the four resistors on a metal strip inside the box. They get warm.
Fig. 8-3. Piezo plate wiring. Each capacitor symbol is one PZ1005; (1-2) etc. are the facets in each bank. The plate is in parallel with the 2440 on GX5 #4 B.
Horn jumper. A Speakon from the 2440's LINK jack to the plate IN. Plate + goes to every resistor; plate - goes to every piezo's other leg.
Banks. Four banks of two: adjacent facet pairs, left to right (facets 1-2, 3-4, 5-6, 7-8). Each bank is one 16 ohm 25 W aluminum-housed wirewound (gold, Dale RH style) in series, bolted to metal, then two piezos in parallel.
Polarity. Red is + on every piezo. Mixed polarity within a bank cancels where the two patterns overlap.
Load. At 20 kHz the whole plate is about 10 ohm. With the 2440 in parallel the GX5 sees about 6 ohm, which is fine.
Stand-in until the 25 W parts arrive: four 16 ohm 3 W per bank, two in series and the pairs in parallel (still 16 ohm, 12 W).
CHECKS BEFORE FIRST USE
All eight red leads to the resistor side; all eight black leads to plate -.
The -20 dB pad is in at GX5 #4 input B before the horn amp sees signal.
Play quiet music and put an ear to each facet in turn: all eight should sound.
We kept the single row subtle on purpose. It sits about 10 to 15 dB below the 2440's flat range above 11 kHz, enough to add some air at the top. We checked that against the horn's corrected 111 dB sensitivity and it is enough. For more top end, add a high shelf on DCX Out 6 above about 11 kHz rather than rewiring the plate.
Side-to-side variation+/-45 deg, simulated
at 5 kHz8-1/2 dB
at 8 kHz12 dB
at 12.5 kHz10 dB
Verticalone piezo's own pattern
at 8-12 kHzabout 30-47 deg wide
Level vs 2-row plateabout 6 dB lower
Variation is the spread between the loudest and quietest direction within +/-45 deg, smoothed over 5 deg; smaller is more even. Model: each piezo as a 3 in flat piston, third-octave averages, real facet positions. We have not measured it.
One row is the widest vertically of all the layouts considered, because no piezos are stacked above one another to narrow the beam, so it can sit level with no shim.
We drew and simulated four arrangements of 16 piezos before choosing the single row of 8 (2026-10-16). Box sizes are W x H x D in inches; variation is across +/-45 deg at 5 / 8 / 12.5 kHz.
Layout
Box
Variation, dB
Vertical
Concave fan, 4 x 4
15 x 18 x 8
8 / 8 / 3
narrow, shim 6 deg
Convex arc, 4 facets x 4
31-1/2 x 18 x 8-1/4
13 / 9-1/2 / 4-1/2
narrow, shim 6 deg
Convex arc, 8 facets x 2
31-1/2 x 10 x 9
8-1/2 / 12 / 10
wide, works level
Convex arc, 8 x 1 (chosen)
31-1/2 x 6 x 9
8-1/2 / 12 / 10
widest, level
Flat 4 x 4 (reference)
-
19 / 37 / 33
narrow
Concave fan. Four strips aimed inward (15 and 45 deg); the beams cross a few inches out and fan across the room. Most even and most compact, but four rows make a narrow vertical beam that needs a 6 deg downward shim.
Convex arc, 4 facets. A 90 deg arc, radius 21-1/4 in, matching the 2350's width. Drops off sharply past 50 deg, and also needs the shim.
Convex arc, 8 facets x 2 rows. Same side-to-side coverage as the chosen plate, about 6 dB louder, wide enough vertically to work level (better with a 6 deg shim).
We went with the single row because it keeps the 8-facet arc's coverage, is the widest vertically, needs no shim, and stays subtle.
No polar data exists for the KSN1005A or its clones (GRS PZ1005, Goldwood GT-1005): not from CTS or Motorola, and not from independent measurements. The "90 deg dispersion" on retailer pages is copied from the rectangular 2 x 6 in horns (KSN1016 and KSN1025), not this one.
Estimate. Treating the roughly 3 in (75 mm) mouth as a piston, plus or minus 30 percent, one piezo's -6 dB width is about:
5 kHz80 deg
8 kHz47 deg
10 kHz37 deg
12.5 kHz30 deg
16 kHz23 deg
It is the same up and down as side to side, since the horn is round. A forum report on a similar Motorola bullet tweeter ("falls away past 15 deg" at 20 kHz) fits this.
Can they sit straight? One piezo alone, yes: it is about 30 to 47 deg wide at 8 to 12 kHz. Stacking them is what narrows the beam. Four in a column act together, so the column's vertical beam is only a few degrees, hence the shim on the 4-row layouts. Side to side, 20 to 25 deg between neighbors is what each piezo can cover above 10 kHz; the 8-facet arc's 11-1/4 deg spacing overlaps comfortably.
To settle it for real: one piezo on a lazy Susan, REW and a measurement mic, a reading every 10 deg.
