EASY WAY TO PLOT ANODE CHARACTERISTICS
S E P T E M B E R
2 0 0 7
US $7.00/Canada $10.00
Back
Tube, Solid State,
Loudspeaker Technology
SPECIAL ON
SPEAKERS
BIG BASS
FOR SMALL SPACES
TUBA:
THE HORN SUB
OF CHOICE
SOUND STROBE
(ONCE AGAIN)
TO THE RESCUE!
PLUS
DIY PHONO PREAMP WITH
ULTRA-LOW DISTORTION
ALL ABOUT
BENDIX TUBES
PC BOARD WORK MADE EASIER
www.audioXpress.com
Cover-907.indd 1
7/25/2007 3:55:37 PM
s o l i d s t at e
By Dennis Colin
The LP 797 Ultra-Low
Distortion Phono Preamp
A JFET input and ultra-low distortion AD797 op amps provide excellent
full-spectrum noise figure, headroom, and RIAA precision. Sound quality is
extremely natural, transparent, and dynamic.
PHOTO 1: LP797 phono preamp.
2.
3.
4.
5.
6.
7.
S
omeone writing to Stereophile
complained that vinyl can’t provide sonic fidelity, because this
“primitive system uses physical
contact.” Apparently this writer isn’t
aware that vibrating strings, and so on,
physically contact the air, which physically contacts your eardrums!
MM, 75dB with MC, and 89dB with
high-output (2.5mV ) MC, re 5cm/
sec.
Low noise over the full spectrum, not
just averaged.
0.002% THD, mostly 2nd harmonic,
full spectrum.
Instant MM/MC switching, no
“tweaky” circuitry, no adjustments.
7.2V RMS maximum output. Preamp
dynamic range is 118dB (MM), 98dB
(MC).
Typical headroom re 5cm/sec is 19dB
at 1kHz, 36dB at 20kHz.
Passes the sonic “intrinsic fidelity” test
(described later).
OVERVIEW
For those interested in obtaining the
naturalness and resolution that LPs are
capable of, this phono preamp provides
excellent reproduction and resolution
over a very wide dynamic range.
FEATURES
1. 81dB “A” weighted SNR with typical
The three-stage design (see gain partitioning in Fig. 1) is configured to avoid
noise degradation by circuitry following
the input stage, particularly at high frequencies where the RIAA rolloff results
in the lowest gain. Additional benefits
are independence between MM/MC
gain switching and the RIAA equalization, and lower distortion than a two-
FIGURE 1: Block diagram; overall 1kHz gain = 43.8dB (MM), 63.8dB (MC).
6
audioXpress 9/07
colin2774-2.indd 6
www.audioXpress .com
7/25/2007 3:55:08 PM
stage; this is because of the comfortably
lower required closed-loop gains. If desired, you can easily change the MM
and/or MC gains. As is, the 1kHz gains
are 43.8dB (MM) and 63.8dB (MC).
Note that the input stage provides
only the 2122Hz (75µs) rolloff portion
of the RIAA de-emphasis. This minimizes the gain range (over frequency)
required, allowing enough gain at 20kHz
to maintain a good overall noise figure
(NF).
Norm Thagard1 reports that Shure
Labs found velocity peaks of 25cm/sec
at 1kHz and 50cm/sec at higher frequencies. The input stage (and the entire
preamp) can accommodate this, while
the input gain (35.7dB at 1kHz) is high
enough to amplify MC outputs (0.5mV
at 5cm/sec typical) well above the noise
level of the following circuitry. Thus, the
MM/MC gain switching can be done
in the second stage, allowing the all-important input stage to be optimized as
a “fixed gain cell” of very high dynamic
range. Overall, the preamp’s own dynamic
range
(apart frompage
cartridge
noise
Classic
Tubes_half
5/10/07
limitations) is 7.2V RMS (maximum
output) divided by 9.13µV RMS (“A”
weighted output noise), which is 118dB,
in MM mode (98dB in MC mode).
DETAILED CIRCUIT
DESCRIPTION
The paralleled JFETs (Q1) are ACcoupled into the inverting input of U1,
whose output is non-inverting with
reference to the input signal. But the
connection through R6, R7, C9 to the
JFET source (and R3) is overall inverting, providing negative feedback (NFB).
R5 gives U1 DC stability, while R4,
C5, and C6 provide HF stability (the
AD797 is capable of 110MHz oscillation, so you must closely ground C7
and C8). R5, C6 form a 10kHz pole, but
the high loop gain through Q1 extends
this to well above 100kHz. R6, R7, C9
form the 2122Hz RIAA rolloff (where
the reactance of C9 equals R6 + R7 =
750Ω). However, R7 (and the non-inverting configuration with R3) cause
the -20dB/decade slope to flatten with a
transmission zero at 35.6kHz, where Xcg
= (R3//R6)
+ R7
2:10
PM Page
1 = 44.7Ω.
This flattening could be exactly com-
pensated for by the R9/C11 rolloff before the second stage U2. However, I set
this rolloff at 55.1kHz. The net result is
that at 20kHz the response is +0.65dB
with regard to the normal RIAA curve.
I say “normal,” because, as Norm Thagard1 points out, record cutter heads use
a 50kHz rolloff pole to protect the modern kilowatt liquid helium-cooled mechanics from RIAA pre-emphasized ultrasonic transients. (Who am I to argue
with a record-setting astronaut?) This
rolloff is -0.64dB at 20kHz, hence the
above (doubtfully audible) 0.65dB boost.
Close enough for vinyl record-setting!
2ND STAGE (U2)
C10 and R8 form a 3.0Hz high-pass
rolloff, as does the C15/R13 combination at the output of U2. Together,
they’re down 3dB at 4.7Hz, in addition
to eliminating DC offset accumulation.
U2 provides selectable gains (7.2dB for
MM, 27.2dB for MC, at 1kHz) and
has a flat response to audio; it’s -3dB at
2MHz in MC mode, from C12, which
ensures HF stability.
BEST OF THE CLASSICS
Tung-Sol 6L6GC STR
Mullard EL34
6L6G 6SL7 6SN7GTB 6V6GT
12AX7 12AX7/ECC803S
5881 6550 KT66
12AT7WA/CV4024
Genelex Gold Lion KT88
www.newsensor.com • info@newsensor.com
audioXpress September 2007
colin2774-2.indd 7
7
7/25/2007 3:55:26 PM
OUTPUT STAGE (U3)
R14 is needed to stop a 40MHz oscillation involving U2 and/or U3. The
output stage gain at 1kHz is 0.9dB. R16
and C16 provide the 50Hz (3180µs)
RIAA rolloff pole. The non-inverting
configuration produces the required
slope flattening, with the transmission
zero at 500Hz. This is (within 0.1dB)
where Xc16 equals R15//R16.
I added R19 and C20 as a “last minute tweak” to optimize the RIAA accuracy. C19 removes any DC offset. With
an (unlikely) low load at 4k7, 20Hz response will drop only 1dB. R18 sets the
output impedance to 200Ω. S2 mutes
the outputs if desired. You should activate it before applying power to the preamp if the audio system is on, to avoid
huge turn-on transients.
CAPACITORS
Some people claim they don’t like “capacitor sound.” But a phono preamp
needs them, unless you use inductors.
The film caps specified here are inexpensive, but of very high quality; I doubt
that they have any “sound.” With the
largest film cap (C19, 3.3µF), I measured an ESL of only 14 nano henries
(about that of a 1˝ wire) and an ESR at
750kHz of only 0.021Ω.
You should realize that as the music
waveform travels to the record grooves,
the signal “train” passes through many
“stations” named “Capacitor Junction”—
inside mike preamps, mixing consoles,
cutter head amps, and so on. In fact,
the best mikes are capacitors—no one
complains about “capacitor sound” there!
But, of course, as with other components, high quality is important.
19.4dB (20kHz), with a 5mV/5cm/sec
MM cartridge. Relative to the “standard”
5cm/sec velocity (RMS), the headroom
is 19.1dB (1kHz) and 36.4dB (20kHz).
Q1
For higher output cartridges, or for any
reason, you can increase the 1kHz maximum input to 92mV RMS (130mV
peak) by reducing the second stage (U2)
gain from 7.2dB to 0dB. To do this, simply delete R11 in Fig. 2. (This will decrease the MC gain by only 0.5dB.) The
maximum input at 20kHz (MM) will
then be 510mV RMS (721mV peak).
Doing this results in an overall preamp
MM gain at 1kHz of 36.7dB. Output
with 5mV RMS input is then 342mV
RMS.
With a 5cm/sec 5mV rated Shure
cartridge and the loudest part of a “hot”
LP (described in the “Showtime” section
at the end of this article), the highest
peaks were 2.9V at the preamp output.
This is about 10dB below clipping. With
very low output MC cartridges, you can
obtain state-of-the-art low noise with
Paul H. Rossiter’s head amp3 driving the
LP797 set for MM gain.
The LSK389 is advertised in audioXpress
as a “1nV (per √Hz) Low Noise Dual
JFET,” by Linear Integrated Systems,
(510) 490-9160 or (800) 359-4023. The
datasheet (www.linearsystems.com) specifies 0.9nV/√Hz typical, 1.9 maximum at
1kHz, 2mA, 10V.
I measured six samples; four had noise
less than 1nV/√Hz, but the other two
were 2.18 and 2.83 nV/√Hz. All had an
IDSS within the 6-12mA spec for the
“B” category (LSK389B TO-71 used
here).
However, considering the advertised
1nV/√Hz noise, you should request
units meeting this number. The price as
of last fall was $7.29 (1-99) and $5.87
(100-999).
HEADROOM
The input stage can handle 92mV RMS
1kHz input and 510mV RMS at 20kHz.
However, due to the following stages’
gain of 8.1dB at 1kHz (U2 and U3
combined), maximum preamp inputs are
45mV RMS (64mV peak) at 1kHz, and
330mV RMS (467mV peak) at 20kHz.
Relative to the 25cm/sec (1kHz) and
50cm/sec (HF) peak velocities reported1,
this is a headroom of 8.1dB (1kHz) and
IMPROVING HEADROOM
POWER SUPPLY
Not wanting a separate power supply
box, nor the induced hum of an internal
line transformer, I took advantage of a
readily available 24V DC “wall wart.”
An additional benefit is UL and/or other
safety approvals.
FIGURE 2: Phono preamp schematic (one channel shown).
8
audioXpress 9/07
colin2774-2.indd 8
www.audioXpress .com
7/25/2007 3:55:29 PM
The op amps, with their 130dB power supply rejection ratio
(PSRR), couldn’t care less what kind of DC you feed them,
provided that their HF bypass caps (C7, 8, 13, 14, 17, 18 in
Fig. 2) are closely grounded. So the simple RC ripple filtering of (Fig. 3) R1, C1, R3, R4, C2, C3 provides low enough
120Hz ripple (0.2mV RMS) such that the op amps attenuate
it to 63 pico volts. They’re fed the V+ and V- voltages.
Not so with the single-ended JFET input! This needs extreme hum rejection. Also, conventional regulators have far
too much noise. And while the FET bias doesn’t need to be
regulated to some precise value (the feedback structure stabilizes the gain), the bias must not be subject to AC line voltage
fluctuations.
The circuitry with Q1 and D3 provides excellent rejection
of line fluctuations and ripple. However, I observed some LF
noise from the D3 zener adding to the preamp’s noise floor,
from about 5-40Hz. This was fixed by the filtering of R7 and
the 10,000µF cap (C1 in Fig. 2), forming a 0.8Hz low-pass
rolloff.
Note the role of D2 (Fig. 3), which clamps the V- to -12V,
leaving +12V for V+ and the Q1 circuit. I couldn’t have two
zeners across the full DC input (about 26V, more on this
later)—as this input varied, the two zeners would shut off
and/or draw large currents. But two are not needed, because
the op amp supplies don’t need regulators, and the JFET supply has it.
Here’s how D2 functions: As shown in Fig. 3, the total positive supply current is 76mA, which is drawn from the TI wall
wart. Being a single (not bipolar) supply, TI must then draw
this same 76mA from its negative terminal. But the negative
supply output (V-) and the LED’s 2mA provide only 52mA.
The difference, 24mA (the Q1 circuit’s current), is pulled
through D2, which is thus happily kept on. Looks strange, but
works well.
A NOTE ON TI
The specified unit is rated at 24V, 400mA, but only 76mA is
used. This overrating has three benefits:
1. TI runs very cool.
2. The larger unit has a large internal filter cap, about 1000µF.
3. The output voltage is higher, about 26V.
R1 (Fig. 3) will act as a fuse if there’s a short or reverse voltage
applied.
MCap® RXF
Radial Xtra Flat Capacitor
NEW! MCap® RXF/Oil Impregnated
Featuring the ultimate winding geometry (edgewise) for
• extremely short, low-loss signal transmission,
• extremely reduced residual-resistance (ESR),
• remarkable low residual-inductivity (ESL).
Polypropylene capacitor-foil, alu metallized.
Grouted winding against microphonic effects.
• Fit-In-Adaptors now available.
f
e oAr t
t
a
St The
NEW! Silver/Gold/Oil
™ e
derprem
l
o u
MS S
e
Us
p
d
ze m
mib e A
i
t
Oopr Tu
f
e
e p Us
g
lta m
Vove A
h
Val
Hig
for
ALTERNATE POWER SUPPLY
If desired, you can use a regulated ±15V DC supply, which
you must isolate from the AC line ground, but connect its
common ±15V point to the preamp chassis ground. Referring
to Fig. 3:
1. Remove (or do not install) D2, C1, R1, and J5. You can install normally grounded jacks for the ±15V supply.
2. Connect the +15V to the left side of R3, and the -15V to
the left side of R4.
With this ±15V supply and the “stock” U2 gain of Fig. 2, the
preamp’s maximum input and output levels will increase by
about 3dB.
See more audio innovations on
www.mundorf.com
and subscribe for our newsletter
info @ mundorf.com
OEM and dealer inquiries invited
• MCap® Supreme - unique induction-free winding technology for
the utmost audiophile music
performance!
• MSolder™ Silver/Gold - ensures
the excellent sonic properties of
all concerned components in
your application – including, of
course, the involved Mundorf
parts.
• TubeCap® - high voltage MKP compact as an electrolytic
capacitor, as fast and voltage
proof as a film capacitor. Due to
its considerable low ESR and
lower residual inductivity the
TubeCap® contributes to great
fidelity. Advanced no drying out
and selfhealing properties!
• MLytic® HV - These electrolytic
capacitors have been developed
especially for the requirements
of tube amplifiers. In addition
to the utilisation in new developments, they are also outstandingly suitable for repairs and
tuning of existing amplifiers.
Recommended US Dealer
MADISOUND SPEAKER COMPONENTS
www.madisound.com
Exclusive Canadian Distributor
AUDIYO INC.
www.audiyo.com
Exclusive Australian Distributor
SOUNDLABS GROUP PTY LTD
www.soundlabsgroup.com.au
High End Components Made In Germany Since 1985
audioXpress September 2007
colin2774-2.indd 9
9
7/25/2007 3:55:33 PM
With the specified “wall wart,” MC
output hum is only 21µV RMS. “A”
weighted, that’s 108dB below the signal
with 0.5mV input. With MM gain the
figure is 128dB.
CONSTRUCTION
I built the preamp in a Hammond aluminum box (1590C, Digi-Key HM153,
4.72˝ × 3.70˝ × 2.07˝). Note that J5,
the DC input jack, must be insulated
from the box. I made a ¾˝ hole in the
back panel (Photo 2), using a Greenlee
chassis punch, P/N 730BB-3/4. Then I
glued a 1˝ square piece of 1/8˝ plywood,
with a 5/16˝ centered hole for J5, to the
inside of the box panel. Later, I found an
insulated jack, Mouser 163-4303EX.
As usual, I should have used a larger box! The circuitry (Photos 3 and 4)
is crammed onto two unetched copper ground-plane boards, and the power
supply components are glued to the
sidewalls.
The specified switches have a nice
feature: By lifting and rotating the handle, you can align the tabs to lock the
switch position selected in place. You
wouldn’t want S1 inadvertently thrown
to “MC” with an MM cartridge, producing a +20dB blast! The switches provide
excellent chassis grounding of the board.
You should connect the ground end of
R3 (Fig. 2) to a ground lug on the input
jack.
I recommend using a larger box, the
Hammond 1590D, Digi-Key HM154.
Dimensions are 7.38˝ × 4.70˝ × 2.05˝. If
you’d like a nice deep purple pilot light,
PHOTO 2: Rear panel.
PHOTO 3: Interior.
PHOTO 4: JFETs and op amps in sockets.
use the LED 5-UV-405-30 from Bivar
(949) 951-8808, and change R2 (Fig. 3)
to 3k3.
MM LOADING CAPACITANCE
I recommend reading Raymond A.
Futrell’s “The LP Terminator,”2 on how
to obtain the flattest frequency response
given the high inductance of MM cartridges. The LP797 preamp input ca-
pacitance (without CL in Fig. 2) is only
21pF, because the Q1 drains see the
“virtual ground” of U1 (no Miller effect),
and also because of the “bootstrapping”
of NFB to the Q1 sources. Mr. Futrell
shows how the right phono cables can
provide optimum capacitance.
RIAA ACCURACY
Figure 4 shows the deviations. As discussed, I incorporated compensation for
the 50kHz cutter head rolloff (which
causes a doubtfully audible 20kHz
preamp gain increase of 0.64dB). The
0.36dB rolloff at 20Hz is intentional,
because it’s due to the interstage infrasonic filtering/DC blocking. From
100Hz to 20kHz, the response variations
of both channels fit within a ±0.07dB
window. Channel balance tracks within
0.03dB across the band, and the two
channels’ gains at 1kHz are balanced
within 0.01dB.
This might have something to do
with the fact that I hand-matched all
resistors and caps in the signal path to
within ±0.1% between channels.
NOISE PERCEPTION
My article “A Low Noise Measurement
Preamp”4 covers the perceived effect on
noise level and spectrum, due to per-octave versus per-Hz analysis, “A” weighting, and the RIAA curve. At the (you
hope) low noise levels of audio components, “A” weighting is appropriate.
Also, our hearing (at any level) analyzes
spectral distribution on a logarithmic
(per octave or fraction thereof ) basis, not
FIGURE 3: Power supply.
10
audioXpress 9/07
colin2774-2.indd 10
www.audioXpress .com
7/25/2007 3:55:36 PM
FIGURE 4: RIAA accuracy.
FIGURE 5: Noise out, MM, input grounded. Total 20Hz
– 20kHz noise (unweighted) = 26.0µV RMS (-91.7dBV)
(MC = 20dB higher).
on the constant bandwidth (BW, per Hz) basis used in many
spectrum analyzers.
Figure 5 shows the preamp’s output noise spectrum (MM
mode, input grounded) two ways: noise per 1Hz BW (lower
curve 1) and noise per octave (upper curve 2). The upper
curve is higher because audio band octaves have bandwidths
much greater than 1Hz. But the total integrated noise (20Hz
– 20kHz) is the same; noise power is noise power no matter
how it’s (accurately) measured.
Because the JFET’s input spectrum is fairly flat on a perHz basis (white noise), curve 1 reflects the RIAA response,
plus the JFET’s gradual LF noise increase (the LSK389 data
shows 0.9nV/√Hz at 1kHz, 2.5 at 10Hz, typical).
“A” WEIGHTING
The lower curve in Fig. 6 (curve 1) shows this same preamp
output noise (MM mode; MC is 20dB higher) with “A”
weighting. Notice how the “A” curve’s strong LF rolloff, plus
the RIAA response’s strong HF downslope, result in a broad
peak around 1.5kHz. Thus, the perceived noise, if heard at all,
is “mellow” sounding, not “hissy” like white noise.
This is why I selected the preamp’s topology and ultralow noise op amps in all stages. Because the RIAA curve is
audioXpress September 2007
colin2774-2.indd 11
11
7/25/2007 3:55:44 PM
-19.62dB at 20kHz relative to 1kHz
(-18.98dB if the 50kHz cutter head
compensation is included), circuitry following the input stage could add HF
noise. For example, a preamp design
with only 20dB 1kHz gain in an MM
input stage would, after RIAA EQ, have
almost no gain at 20kHz. This means
that the following stages, even if they
have as low a noise level as the input
stage, would add 3dB to the input-referred preamp noise at 20kHz.
So it ’s apparent that SNR alone
doesn’t tell the whole story; it’s the “A”
weighted and per-octave analyzed spectrum that does. My article4 shows that
SNR values based on a preamp’s unweighted and nearly white input noise
are improved by about 6.8dB.
NOISE WITH AN
MM CARTRIDGE
I measured the impedance of a Shure
R27E cartridge. Although old and inex-
pensive, its impedance is typical: 635Ω
+ 631mH. Paralleled with this preamp’s
47k5 load, the impedance seen by the
preamp at 1kHz is 943Ω (resistive) +
j3k84 (inductive); impedance magnitude
is 3k95. But at 10kHz, z = 19k5 + j23k;
magnitude is 30k2.
The upper curve (2) in Fig. 6 shows
“A” weighted, per-octave preamp output
noise with this cartridge. The increase in
HF noise with regard to the lower curve
(1) is almost entirely due to the 47k5
loaded cartridge’s resistive thermal noise.
The preamp noise degradation (NF) is
only 0.3dB maximum across the band.
The curve broadly peaks around 4kHz,
so if heard at all, still doesn’t sound
“hissy.”
MM PREAMPS WITH
BIPOLAR TRANSISTORS
FIGURE 6: Noise out, MM, “A” weighted, per octave analysis. 1 = input grounded, as in
Fig. 5 ; total noise 20Hz – 20kHz = 9.13µV RMS (-100.8dBV). 2 = with Shure R27E cartridge; total noise 20Hz – 20kHz = 71.9µV RMS (-82.9dBV) (z = 635Ω + 631mH).
I had first tried an AD797 op amp alone
(no JFET) at the input. With a low
source impedance such as an MC cartridge provides, all was fine. But with
the MM cartridge, the noise was about
8dB higher than with the JFET input;
plus (rather, a negative) the noise peaked
around 10kHz. Too much and hissy!
The reason is the bipolar op amp’s input
current noise. Now, 2pA/√Hz might not
sound like much, but when multiplied
by the loaded cartridge’s 30k impedance
magnitude at 10kHz, a noise voltage
term of 60nV/√Hz undesirably appears!
This is 10.5dB higher (at 10kHz)
than the 18nV/√Hz noise of the loaded
cartridge’s 19k5 resistive component.
Combined, the resulting NF is 10.8dB
at 10kHz. The JFET reduced the NF to
0.3dB.
THE BOTTOM (NOISE) LINE
FIGURE 7: Noise out with cartridges. 1. MM, 635Ω + 631mH, 5mV. Total SNR = 80.8dB,
preamp mode MM. 2. MC, 50Ω, 0.5mV. Total SNR = 75.2dB, preamp mode MC. 3. hiout MC, 50Ω, 2.5mV. Total SNR = 89.2dB, preamp mode MM.
12
audioXpress 9/07
colin2774-2.indd 12
Figure 7 tells the whole perceived noise
story. The vertical dB scale is “A” weighted per-octave analyzed output noise (as
we perceive low-level sounds), relative to
the 5cm/sec outputs of three cartridges
driving the LP797 preamp. The total
20Hz – 20kHz SNR values are stated.
Some notes:
1. Above 7kHz, the typical 0.5mV MC
cartridge has slightly better SNR
(lower noise with reference to signal)
than the MM cartridge, even though
the total (full band integrated) SNR
is 5.6dB better with the MM. This
is from the HF noise increase, due to
www.audioXpress .com
7/25/2007 3:55:51 PM
the high MM inductance not effectively shunting the 47k5 load’s thermal noise, as I described.
2. Notice that the high-output MC cartridge has the best SNR of all: 89.2dB.
The example used is the Dynavector
10 × 5, available for $380 from Music
Direct. Art Dudley of Stereophile said,
“wildly, highly recommended.” Its use
of neodymium magnets generates
2.5mV at 5cm/sec; thus it uses the
MM mode of this preamp, where the
output is then 392mV RMS at 5cm/
sec, and 1.95V peak at the 25cm/sec
1kHz peaks reported by Shure Labs.
Peak dynamic range would then be
103.2dB, higher than that of any CD,
and, of course, pure analog!
Tables 1 and 2 are the parts list. Table 3
shows measured performance.
COMPARISON WITH
A $3000 PREAMP
Table 4 compares performance data
between the LP797 and the Sutherland “Ph.D,” reviewed and measured
by Stereophile (May '05). For $3000 you
get 8 to 20dB more noise, four times
the THD, 28dB less 20kHz headroom,
three times the RIAA error, and switching pop-producing DC output offset! (It
also needs batteries.)
Review comments include “slightly
soft bass. . . lack of dynamic punch. . .
dulling of piano. . . guitar lacked impact, muted and a little pale. . .” The last
words in the review were (no surprise)
“Highly recommended.”
As I mention in the last section
(“Showtime”), I compared guitars on LP
(with the LP797) to a live guitar I heard
the same day. The only “muting” I heard
was when I flipped the “mute” switch!
And my hearing is very good and well
experienced (see my speaker reviews in
aX).
REGARDING THE AD797
In response to a letter from David Elderton5, Gary Galo refers to the AD797
as a “stellar performer,” but ruled it out
for phono preamp inputs because of its
need for a fairly low source impedance.
He said it “would otherwise be a first
choice for phono preamps.”
As I mentioned, high source impedances are affected by the AD797’s current noise. Hence the JFET input in this
preamp. Thus, the noise figure is excellent over a wide range of source impedances, and the AD797s are happy.
Then their stellar performance shines.
It’s probably one of the most sonically
transparent audio amplifying devices
available, and its performance electrically agrees, including 0.0002% THD,
110MHz GBW, 20V/µS slew rate, and
130dB PSRR. The LSK389B/AD797
input stage cascade is completely stable,
both HF-wise and regarding RIAA precision.
INVERSE RIAA NETWORK
(IRN) TESTS
I built an accurate (±0.03dB) IRN to
facilitate future measurements, confirm
my direct RIAA test, and also to be able
to view waveforms. Figure 8 shows a
1kHz linear triangle wave (left) fed to
the IRN. The center response shows the
IRN’s output; this is what would drive
the cutter head. The recording pre-emphasis is evident.
audioXpress September 2007
colin2774-2.indd 13
13
7/25/2007 3:55:53 PM
This was fed to the preamp. The right
shows the preamp’s excellent reproduction of the triangle wave. Figures 9 and
10 show 1kHz and 10kHz square wave
responses.
INTRINSIC FIDELITY TEST
Having built a pair of IRNs, I thought
about trying what can be called an intrinsic fidelity test6. Driving a phono
preamp from an IRN feeds the preamp
the same pre-emphasized response used
TABLE 1
Ref
Description
C12 (2)
C6 (2)
C5 (2)
C7, 8, 11, 13,
14, 17, 18
C9, 16
C3 (2)
C2, 10, 15
C19 (2)
C4 (2)
C1 (2)
J1-4
Q1 (2)
R3 (2)
R7 (2)
R4 (2)
R10 (2)
R18 (2)
R14 (2)
R2 (2)
R6 (2)
R11 (2)
R12 (2)
R9 (2)
R15 (2)
R16 (2)
R5 (2)
R1 (2)
R8, 13
R17 (2)
S1, 2
U1-3
cap, 68pF
cap, 470pF
cap, 1nF
C20 (2)
R019 (2)
size
cap, 2n7, film
cap, 100nF, 250V, film, %
cap, 330nF, 100V, film
cap, 1µF, 100V, film
cap, 3.3µF, 100V, film
cap, 10µF, 35V
cap, 10,000µF, 16V
phono jack, gold
dual JFET
Res, 1% MF 10R0
Res, 34R8
Res, 47R5
Res, 48R7
Res, 200Ω
Res, 221Ω
Res, 499Ω
Res, 1% MF, 715Ω
Res, 768Ω
Res, 1k00
Res, 1k07
Res, 3k57
Res, 31k6
Res, 33k2
Res, 47k5
Res, 52k3
Res, 100k
switch, DPDT, locking
op amp
8 pin DIP socket
cap, 18nF, 100V, film
Res, 1% MF, 2k87
Mfr
Mfr./P/N
Dist. Dist. P/N
DK PS1272J
PE 027-200
DK EF1334
DK EF1105
DK EF1335
DK 493-1314
DK 493-1292
PE 091-1120
Linear Systems
LSK389B TO-71
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
¼W
TABLE 3 Measured Performance
AD
AD797AN
DK 450-1487
DK AD797AN
DK A9408
DK PS1183J
¼W
DK = Digi-Key, PE = Parts Express
TABLE 2
Ref
Description
C4
C1-3
D1
D2, D3
J5
LED1
cap, 1000µF, 35V
cap, 2200µF, 35V
diode
zener, 12V, 1W
power jack
LED, red
lens for LED
transistor, NPN
Res, 1% MF, 10R0
Res, 20R0
Res, 499Ω
Res, 10k0
wall transfomer, 24V DC, 400mA
box, 7.38˝ × 4.70˝ × 2.05˝
rubber bumper (4)
Q1
R1-4
R7
R5, 6
R2
T1
size
in record cutters. So I fed this combination high-quality SACD music. This
can be more stressful than LP outputs,
because the SACD format (DSD) has
a BW of 100kHz, and in this test any
ultrasonic components are strongly amplified by the recording (inverse) RIAA
response. The SACDs I used have peak
player outputs of about ±3V.
I used an A/B switch, feeding a highquality headphone amp and the excellent Sennheiser HD 650 phones, to
compare the direct SACD output with
that from the LP797 preamp. The two
signals were level-matched to within
0.02dB at 1kHz. I listened with much
A/B switching, ranging from hearing an
entire piece between switching, to just
listening to the first two seconds, switching, restarting the track, and repeating
the sequence 20 times. With a wide variety of excellent recordings, all direct to
DSD, I listened as above, plus frequently
switched rapidly in the middle of a sustained violin ensemble harmony, piano
chord, sung note, and so on.
My hearing is very good; Ed Dell and
Joe D’Appolito have praised my ability to hear fine details in my many aX
Mfr
Mfr./P/N
1N4001
1N4742A
Dist. Dist. P/N
DK
DK
DK
DK
M
DK
DK
493-1322
493-1323
DK
DK
N
T971-P6P
HM154
92N4782
1N4742ADICT
163-4303EX
67-1612
L30001
2N2222A
¼W
¼W
¼W
¼W
Hammond
SPC
1590D
2567
1kHz gain: 43.8dB (MM), 63.8dB (MC)
input z: 47k5//21pF, adjustable with R1, C1 (Fig. 2)
output z: 200Ω
maximum output: 7.2V RMS (1kHz), 6.3V RMS (20kHz)
maximum input: 45mV RMS (1kHz), 330mV RMS (20kHz), for
MM; 1/10 as much for MC.
maximum input (MM) with external ±15V supply (see text):
64mV RMS (1kHz), 460mV (20kHz).
maximum input (MM) with gain reduced by 7.2dB (see text):
92mV RMS (1kHz), 510mV RMS (20kHz).
RIAA error: ±0.07dB, 100Hz – 20kHz; -0.14dB at 50Hz; -0.36dB
at 20Hz
channel balance: ±0.03dB, 20Hz – 20kHz
channel separation: 70dB, 20Hz – 20kHz
slew rate: 20V/µS
output AC line hum:
2.1µV RMS, -130.6dBV “A” weighted (MM)
21µV RMS, -110.6dBV “A” weighted (MC)
“A” weighted output noise, preamp input grounded:
9.13µV RMS, -100.8dBV (MM)
91.3µV RMS, -80.8dBV (MC)
preamp dynamic range: 118dB (MM), 98dB (MC)
typical headroom above 5cm/sec: 19dB (1kHz), 36dB (20kHz)
SNR, “A” weighted: 81dB (5mV MM), 75dB (0.5mV MC),
89dB (2.5mV high-output MC W/MM setting)
THD: estimated 0.002%, 20Hz – 20kHz, 3V RMS out
power requirement: 24 to 30V DC, 76mA, ground isolated, or
±15V DC preamp ground referenced but isolated
from AC line ground (see text).
DK = Digi-Key, N = Newark, M = Mouser
14
audioXpress 9/07
colin2774-2.indd 14
www.audioXpress .com
7/25/2007 3:55:55 PM
speaker reviews. My sensitivity threshold is within 3dB of the standard “good
hearing for young people” audiogram.
My HF limit is 14kHz, but there’s little
spectral content in that top half-octave
in natural music.
Well, try as I did (and I wanted to
be the first to know about any deficiency), I heard absolutely no difference between the direct SACD output and that
through the inverse RIAA/LP797 phono
preamp. The HD650 phones are extremely resolving, such that I could hear
the very slight inferiority of the SACD
medium compared to vinyl (it’s much
better than 16/44 CDs, though). But the
LP797 output was indistinguishable from
the direct SACD. No audible distortion,
coloration, noise, hum, nor loss of resolution, tonality, spaciousness, instrument
focus, transient precision, or dynamics.
I tend to conclude from this that the
TABLE 4 comparison with a $3000
commercial preamp
cost
SNR, unwtd
SNR, “A”
RIAA error
THD
1kHz headroom
20kHz headroom
DC offset
LP797
Sutherland “Ph.D”
$130, parts
90dB, MM
70dB, MC
99dB, MM
79dB, MC
±0.1dB
0.002%
19.1dB
36.4dB
none
$3000, retail
63dB, MM
29dB, MC (!)
91dB, MM
59dB, MC
+0.3, -0.1dB
0.0086%
18.7dB
8.7dB
17.5mV
FIGURE 8: 1kHz triangle wave responses.
a: 1kHz linear triangle, input to inverse RIAA network, 1V/div
b: inverse RIAA output (signal FED
to record cutter), to phono preamp
input, 8mV/div
c: phono preamp output 1V/div
audioXpress September 2007
colin2774-2.indd 15
15
7/25/2007 3:56:00 PM
Tercel
Input to inverse RIAA
network
(IRN) 1V/div
Phono Kit
$499
preamp out
1V/div
FIGURE 9: 1kHz square wave response.
Input to IRN
1V/div
We’ve taken our very best phono stage
from the cult classic BlueBerry Xtreme
full function preamp and made a handsome kit as a stand alone phono stage.
Everything you need including this solid
chassis, tubes, and high-rel mistake-proof
PC board is included for $499. It takes
most people about 10 hours to
complete the Tercel.
Preamp out
1V/div
• 2 relay switched phono inputs.
• Vacuum tube rectified
• Low impedance output drives long
cables
• All triode design with passive EQ and
no loop feedback
• Green hammertone finish chassis
• Options: (1) Our best LOMC
transformer set can be added to one
input; (2) Full set of AuriCaps;
(3) Xtreme power supply
IRN output to
preamp input
0.1V/div
All JuicyMusic products are designed and
manufactured in the USA
and sold only by JuicyMusic direct.
Visit our website for the full story.
www.juicymusicaudio.com
or call 707-786-9736
JUICY
www.juicymusicaudio.com
16
audioXpress 9/07
colin2774-2.indd 16
FIGURE 10: 10kHz square wave response.
poor sound from some preamps (that
measure well), as well as the claimed
audibility of (good) resistors, caps, and
cables, are due to circuit design problems
(overly sensitive to component parasitics,
shifting bias points, power supply interactions, potential feedback loop instabilities, and so on). Another possibility is
the lack of rigorously controlled testing
with precise level-matching and rapid
switching ability.
LISTENING EVALUATION
I recommend the audiophile-quality
LPs from Music Direct and Acoustic
Sounds, because most popular music,
and even some classical recordings, have
been “produced” with EQ, compression,
www.audioXpress .com
7/25/2007 3:56:06 PM
synthetic reverb, and so on. This is because of the infamous “loudness wars,”
boosting frequency ranges to make everything “stand out,” and so on. Just ask
any mastering engineer.
Only with the best, honestly recorded
LPs (and, of course, excellent speakers,
power amps, and so on) can a phono
preamp be correctly evaluated!
SHOWTIME
I connected the preamp between a vintage Shure 600Ω MM cartridge (in an
equally vintage Elac/Miracord turntable)
and a 100W per channel solid-state amp,
driving the “Venue” speakers (aX Nov.
'06), or Sennheiser HD650 phones. My
best LP is “Blues, Ballads, and Jumpin’
Jazz” with Lonnie Johnson and Elmer
Snowden, Analogue Productions APR
3001. Their “Revival Series” used the
Wilson Audio Custom Tube mastering
facility (acquired by Acoustic Sounds
and RTI). As clear as it gets, guitars,
bass, and voices. As a local radio station
says, “no artificial ingredients, no preservatives, no pesticides!”
Meanwhile, earlier today I had the
pleasure of hearing Tristan Light, guitar
teacher/builder/player extraordinaire at
Greenlaw’s Music, Laconia, N.H., play
a beautiful Ibanez guitar. I heard it both
acoustically and electrically through the
HD650 phones. Full, rich, natural live
string tone right in your face!
Well, the (admittedly different) guitars
on the LP, heard on the same phones,
sounded real enough that I would have
had to instantly A/B the sound with the
live guitar to notice any record/playback
imperfections. The recorded acoustic
bass was equally full, clear, and natural.
And the voices were close enough to
live to not notice or care about the difference.
Even with the vintage cartridge, the
sound (on the best LPs) had that live
“fullness” or “roundness” of tone—coherent integrity of power down to the
smallest tonal details. Even SACDs
slightly degrade this, while standard
CDs make the sound somewhat hollow,
flat, and lightly “sandpapered.” But you
know this; otherwise, you wouldn’t be
reading about phono preamps—vinyl is
king!
Excuse me now; I’ve got to see a man
about a Dynavector. . .
Please respect the Legal Notice published
in audioXpress. This preamp design is the
copyright-protected intellectual property of
this article’s author. Commercial use including sale is prohibited. It is published here
for the personal use of those respectful of the
work of others. aX
REFERENCES
1. Norm Thagard, “A Phono Preamp for
the (SA)CD Age, Part 1,” aX Nov.
'05.
2. Raymond A. Futrell, “The LP Terminator,” aX Jan. '03.
3. Paul H. Rossiter, “A Head Amp for
Very Low Impedance MC Cartridges,” aX Sep. '06.
4. Dennis Colin, “A Low-Noise Measurement Preamp,” aX April '07.
5. Gary Galo, response to letter from
David Elderton, aX Aug. '05.
6. Dennis Colin, “Sonic Comparison
of Power Amp Output vs. Input,” aX
Dec. '04.
audioXpress September 2007
colin2774-2.indd 17
17
7/25/2007 3:56:11 PM
tubes
By Pierre Touzelet
Simple Approximations Of
Tube Anode Characteristics
No sophisticated software necessary. Try this simple
approach to determine anode tube characteristics.
M
any designers have proposed
analytical functions to approximate the anode characteristics of a tube. In general these
analytical functions depend on a certain
number of parameters which must be optimized, using a regression method, to
best fit the anode characteristics of a given
tube. Nevertheless, many audio amateurs
do not have such optimization software
in their toolbox. For these audio amateurs
this article introduces a simple alternative strategy, based on the solver available
in Excel software and gives a complete
description of the procedure to be used, as
well as an illustrative example of this procedure, with the triode type ECC81.
PROCEDURE
1. First you must collect the data of the
anode characteristics of the particular triode under consideration. You can derive
triode data collection from direct measurements, using a modern digital curve
tracer or tube data handbook. Triode data
is arranged in a table where horizontal
lines represent predefined anode to cathode voltage: Vak and vertical columns, predefined grid to cathode voltage: Vgk. At the
crossing of a given column and line—that
is to say, for a given Vgk and Vak—the resulting anode current Ia is indicated in
the table. In other words, the table defines a matrix a = I V V
i, j
-1,00E+00 -2,00E+00 -3,00E+00 -4,00E+00
-5,00E+00
-6,00E+00
-7,00E+00
0,00E+00
1,40E-03
2,97E-03
4,75E-03
6,41E-03
8,31E-03
1,02E-02
1,23E-02
1,44E-02
1,66E-02
0,00E+00
0,00E+00
4,75E-04
1,13E-03
2,02E-03
2,97E-03
4,39E-03
5,93E-03
7,57E-03
9,37E-03
1,13E-02
1,35E-02
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
1,90E-04
4,51E-04
8,31E-04
1,31E-03
1,90E-03
2,61E-03
3,49E-03
4,39E-03
5,46E-03
6,64E-03
7,83E-03
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
2,37E-04
4,75E-04
8,31E-04
1,14E-03
1,66E-03
2,25E-03
2,97E-03
3,68E-03
4,63E-03
5,62E-03
6,76E-03
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
1,19E-04
2,37E-04
3,80E-04
7,12E-04
1,07E-03
1,52E-03
2,06E-03
2,73E-03
3,44E-03
4,27E-03
5,10E-03
18
audioXpress 9/07
Touzelet2817-1.indd 18
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
9,49E-05
5,22E-04
9,49E-04
1,66E-03
2,49E-03
3,54E-03
4,70E-03
5,93E-03
7,36E-03
9,02E-03
1,07E-02
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
4,75E-05
2,85E-04
7,12E-04
1,23E-03
2,02E-03
2,85E-03
3,92E-03
4,98E-03
6,31E-03
7,59E-03
9,02E-03
gk j ; aki
)
2. The second step involves choosing the
appropriate analytical function. For triodes, I strongly suggest the one proposed
by Norman L.Koren1. It i s a simple formula with five parameters, flexible enough
to describe the anode characteristics of any
type of triode with rather good accuracy.

 
Vgk

Vak
 1
E1 =
Log 1 + Exp k p  +
kp
m

 
kvb + Vak2
 

(
If
E1 > 0 , then I a =
)


1 
2 

E1Ex
k g1
TABLE 2: Calculated tube data (Norman L. Koren analytical formula)
Triode type ECC81
0,00E+00
0,00E+00
0,00E+00
0,00E+00
0,00E+00
2,37E-04
6,88E-04
1,33E-03
2,25E-03
3,20E-03
4,51E-03
5,96E-03
7,47E-03
9,02E-03
1,08E-02
1,27E-02
(
∀i = 0,1,, N , ∀j = 0,1, M .
TABLE 1: Collected tube data (Philips tube data handbook)
Triode type ECC81
Vak/
Vgk
0
20
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
a
As an example, Table 1 shows the data
for the triode type ECC81, using the
Philips tube data handbook.
Fitting parameters : t1 = 231,0, t2 = 81,5 t3 = 4280,0, t4 = 1,15, t5 = 155,0
Vak/
Vgk
0
20
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
0,00E+00 -1,00E+00 -2,00E+00 -3,00E+00 -4,00E+00 -5,00E+00 -6,00E+00 -7,00E+00
0,00E+00
1,30E-03
2,90E-03
4,63E-03
6,45E-03
8,34E-03
1,03E-02
1,23E-02
1,44E-02
1,64E-02
0,00E+00
1,56E-04
4,80E-04
1,08E-03
1,98E-03
3,15E-03
4,54E-03
6,09E-03
7,78E-03
9,56E-03
1,14E-02
1,34E-02
0,00E+00
4,13E-06
2,10E-05
8,40E-05
2,59E-04
6,24E-04
1,23E-03
2,08E-03
3,16E-03
4,42E-03
5,85E-03
7,42E-03
9,09E-03
1,09E-02
1,27E-02
0,00E+00
8,50E-08
6,67E-07
4,39E-06
2,16E-05
7,81E-05
2,17E-04
4,87E-04
9,30E-04
1,57E-03
2,40E-03
3,41E-03
4,60E-03
5,92E-03
7,38E-03
8,95E-03
1,06E-02
0,00E+00
1,73E-09
2,07E-08
2,19E-07
1,66E-06
8,63E-06
3,24E-05
9,41E-05
2,23E-04
4,53E-04
8,11E-04
1,32E-03
1,99E-03
2,82E-03
3,80E-03
4,93E-03
6,18E-03
7,56E-03
9,03E-03
0,00E+00
3,53E-11
6,42E-10
1,09E-08
1,26E-07
9,33E-07
4,66E-06
1,71E-05
4,95E-05
1,19E-04
2,46E-04
4,55E-04
7,67E-04
1,20E-03
1,76E-03
2,46E-03
3,29E-03
4,25E-03
5,34E-03
6,55E-03
7,86E-03
0,00E+00
7,20E-13
1,99E-11
5,44E-10
9,61E-09
1,01E-07
6,66E-07
3,07E-06
1,07E-05
3,00E-05
7,13E-05
1,48E-04
2,77E-04
4,75E-04
7,57E-04
1,14E-03
1,63E-03
2,23E-03
2,96E-03
3,80E-03
4,75E-03
5,81E-03
6,98E-03
0,00E+00
1,47E-14
6,18E-13
2,71E-11
7,31E-10
1,08E-08
9,50E-08
5,49E-07
2,30E-06
7,52E-06
2,03E-05
4,72E-05
9,71E-05
1,81E-04
3,12E-04
5,03E-04
7,67E-04
1,11E-03
1,55E-03
2,09E-03
2,73E-03
3,47E-03
4,32E-03
5,27E-03
www.audioXpress .com
7/25/2007 4:11:04 PM
S = ∑ ai , j − bi , j 
Else, I a = 0
Where: t1=kp, t2=µ, t3=kvb, t4=Ex, t5=kg1
are the five parameters.
This function is only valid for Vgk ≤ 0 .
For Vgk > 0 , you must model the grid
current, which is not considered here.
With this analytical function, and assuming you already have a rough idea of
the parameter values, you can set up the
same chart as in Table 1, but with the
calculated anode current values.
In other words, the table defines a matrix b = I V V
i, j
a
(
gk j ; aki
)
∀i = 0,1,, N , ∀j = 0,1, M .
As an example, Table 2 shows the triode
type ECC81 data, using the Norman L.
Koren analytical function and estimated
parameters.
3. This third step involves the setup of
an error function. Tables 1 and 2 for a
given grid to cathode voltage Vgk and
anode to cathode voltage Vak allow you
to build the following error function:
2
i, j
∀i = 0,1,, N , ∀j = 0,1, M
This error function is a Euclidian positive norm. Fitting parameters are generally determined by minimizing the error
function using the classical regression
method. The smaller the S, the better
the optimization of the parameters.
4. Next you need to determine the parameters. To do this, use the solver available in the Excel software and open a
worksheet devoted to the particular triode
of interest (see Fig. 1, the Excel worksheet
for the triode type ECC81).
The worksheet is arranged in such a
way that Tables 1 and 2 are placed close
to each other in order to have a natural
correspondence between equivalent cells.
Fitting parameters are declared in a dedicated area of the worksheet as well as the
error functionS. The optimization of the
fitting parameters is performed using
the solver available in the toolbox. The
cell to minimize is the one in which the
error function is available. The variables
which must be optimized are the cells in
which the fitting parameters are defined.
5. The final step involves the display of
the anode characteristics. This is done on
a graph attached to the worksheet. On
this graph are superimposed the anode
characteristics deduced from the tube
data collection (blue graph) and from the
calculated values according to the optimized fitting parameters (red graph). You
can add additional information, such as
the maximum anode power dissipation.
PENTODE APPLICATION
This procedure works very well with triodes, but you can also apply it with success to pentodes. The only difficulty is
the choice of the appropriate analytical
function. I suggest you use a combination of the analytical functions proposed
by Norman L. Koren1 and Menno van
der Veen2. Koren’s function models the
cathode current because it behaves like
the anode current of a triode. The van
der Veen function models the current dis-
audioXpress September 2007
Touzelet2817-1.indd 19
19
7/25/2007 4:12:57 PM
tribution function between the anode and
the screen grid. The result is a simple formula with five fitting parameters, flexible
enough to describe the anode and screen
grid characteristics of any pentodes, with
rather good accuracy.
E1 =
Vg 2 k
kp

  1 V 
Log 1 + Exp k p  + g1k   

  m Vg 2 k   
1
2
V n
a =  Arctg ak  , with 1 ≤ n ≤ 10
Vg 2 k 
 p
E1Ex
a
If E1>0, then I a =
k g1
and
Ig2
E Ex
= 1 [1 − a ]
k g1
Else Ia=0 and Ig2=0
Where: p1=k p , p 2 =µ, p 3 =E x , p 4 =k g1 ,
p5=n are the five fitting parameters.
This proposed analytical formula is only
valid for Vg1k ≤0. For Vg1k >0, you must
model the control grid current, which is
not considered here.
IMPORTANT
Proposed analytical functions are pure
designer’s imagination and, unfortunately,
FIGURE 1:
Excel worksheet to determine the anode
characteristics
of the triode
type ECC81.
20
audioXpress 9/07
Touzelet2817-1.indd 20
do not rely on the principles. If these
functions are convenient and accurate
to determine tube working points, they
are not as efficient when used to simulate tube dynamic behavior. This is particularly true when you try to quantify,
by simulation, distortions. Distortions
depend on partial derivatives of any order
that cannot be deduced from these analytical functions because no care is explicitly taken to achieve this goal. This is a
systematic mistake that designers should
avoid because of its dangerous design
consequences3.
I hope some of you will try this method, and sincerely wish you good luck.
REFERENCES
1. Norman L. Koren, “Improved Vacuum
Tube Models for SPICE Simulations”
5/96 Glass Audio.
2. Menno van der Veen, “Modelling power
tubes and their interaction with output
transformers,” AES preprint 4643; 104th
AES Convention 1998 Amsterdam.
3. Pierre Touzelet “Accurate non linear
models of valve amplifiers including output transformers,” AES preprint 6830;
120th Convention 2006 Paris. aX
www.audioXpress .com
7/25/2007 4:21:02 PM
R
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
The KISS Bass Project
•
SPEAKER BUILDER
By Tom Perazella
Sometimes smaller and simpler is better, as in the case of this sub project, which produces amazingly clean bass.
T
here is no doubt that the science
of speaker design has grown dramatically since the time I first
started in Hi-Fi. In the early
days, there were few tools to predict the
final results of a project. In many cases,
speaker design was a case of trial and
error. Although experience helped, using
what seemed to be similar drivers in the
same types of enclosures did not give the
same results.
Without going into the whole history of
speaker design, it is clear that the work of
Messrs. Thiele and Small provided a predictable basis for determining key driver
parameters and then using those parameters to tune the speaker enclosure. Since
then, many different books and software
programs have been written that can help
optimize the design of a speaker. However,
sometimes you just need to keep it simple.
If ultimate performance is not your goal,
but you want predictable performance, it
would be nice to have a KISS (Keep It
Simple, Stupid) method to achieve the
results you are looking for.
That was the case when I wanted to
produce a simple small subwoofer to take
with me on an extended trip to the UK.
My company was moving me there for a
five-month period to work on some special
projects. While there, I would be living in
a small flat. I was used to a sound system
that had prodigious amounts of clean bass
(see “True Bass,” Speaker Builder 5/96, p.
10). However, even if I could duplicate the
ability to produce 122dB of clean bass at
16Hz like my main system, when I moved
to the flat, relations would most likely soon
go pear shaped with the neighbors. A more
modest approach was called for.
BASIC DESIGN
While thinking about the project, I decided to see just how simple I could make
it. Basically I wanted to take a small driver
and put it into a sealed box. Forget vents,
passive radiators, transmission lines, servos,
and all the other wonderful methods that
offer some advantage but require more effort. Remember that the bass requirements
were quite modest. I would be using the
sub to augment a pair of Sequerra Met 7s.
Although I wanted reasonably deep bass,
the SPL levels would not be very high.
Having spent quite a bit of time in the
UK, I realized that housing space is not
only quite expensive, but it is also limited.
The flat I would be in was two stories,
with a kitchen, dining area, living room,
two bedrooms, and a bath, all totaling just
about 800ft2. All the rooms were small. In
addition, my wife would be accompanying
me and would not take kindly to a large
sub in such small quarters.
sponse and output level. If the driver is
not capable of producing enough linear
displacement at the lowest frequencies
and highest levels you wish to reproduce,
distortion will be high.
Volume displacement is a function of
the effective driving area of the cone times
its linear excursion. If the effective diameter of the cone, Sd, is not specified, a
reasonable approximation is the diameter
of the cone plus about one-third of the
surround. The linear excursion is specified
as Xmax.
Recent advances in driver design have
greatly increased the Xmax that can be
achieved, and, thankfully, this has extended
to smaller drivers. At first glance it would
seem that a small driver with a high Xmax
could produce the same linear output as
a larger driver with a smaller Xmax. This
is true to some extent, but as they say, a
funny thing happened on the way to the
Forum.
To achieve higher Xmax, you must have
PICKING THE DRIVERS
The first step was to pick the driver I
would use. Although I had previously used
larger drivers not only for their larger radiating area, but also for their usually longer
linear stroke, I decided to see whether I
could find a suitable 8˝ driver. It is important to never forget that you must have
sufficient linear volume displacement to
achieve your target levels of frequency re-
PHOTO 1: Typical large Xmax driver.
September 2007
perazella2364.indd 21
21
7/31/2007 11:57:13 AM
SPEAKER BUILDER
•
SPEAKER BUILDER
an increase in the ability of the surround
to travel in a linear fashion over greater
distances. That translates to a larger surround. Photo 1 shows the large roll surround on a typical large Xmax driver. So,
for a given outer diameter of the driver,
greater Xmax means a larger surround and
a correspondingly smaller cone diameter.
Because the surround does not contribute
as much to sound generation as the cone,
the driver will generate less SPL for the
same excursion.
For example, the Vifa M21WO-3908 8˝ driver, which is representative of
a conventional woofer, has an Xmax of
6mm and an Sd of 235cm 2. The Tang
Band W8-740C 8˝ driver (Photo 2) has an
Xmax of 12mm but an Sd of only 220cm2.
The TC Sounds 2574 8˝ driver (Photo 3)
has an Xmax of 16mm and an Sd of only
190cm 2. Therefore, although the Tang
Band has twice the excursion as the Vifa,
the displacement is not quite twice, but
only 1.87 times due to the smaller effective
cone area caused by the larger surround.
The TC Sounds has an Xmax that is 2.67
times the Vifa but only 2.15 times the
displacement.
There is no doubt that either the Tang
Band or TC Sounds driver has higher linear displacement than the Vifa. However—and this is especially true with smaller
PHOTO 2: Tang Band W-8-740C driver.
•
SPEAKER BUILDER
•
SPEAKER BUILDER
diameter drivers—increasing Xmax for
a given basket size does not result in a
proportionately higher displacement. It’s
something to keep in mind when comparing drivers.
At the time I was searching for drivers
for this project, these Tang Band and TC
Sounds drivers were at the top of the list.
Why did I choose two? First, there was a
huge difference in the pricing. The Tang
Band was $39 and the TC Sounds was
around $250. The price was an estimate
for the TC Sounds because it was not yet
available to the public. That’s a big price
•
SPEAKER BUILDER
difference for about 18% more displacement.
But, there was another reason to look
at the TC Sounds. In a sealed box, all else
being equal (which never really happens),
the stiffer the suspension, the smaller the
box that you can use. The equivalent volume of the driver, Vas, depends on the
effective area of the cone and the compliance. The relationship of the Vas to
the box volume determines the degree of
change of free air parameters of the driver such as Fs and Qts once the driver is
mounted in a box and results in the new
JANTZEN KITS WITH WORLD BRANDS
AUDIO TECHNOLOGY
AUDAX JANTZEN
RAIDHO
High end kit for
beginners.
Raidho tweeter FTT-75
together with
Audio Technology driver
18H52.
JA-8008 brand new
Jantzen Audio 8" driver,
8 ohm, 95 dB designed
by Troels Gravesen and
made by SEAS, Norway.
Ideally mated with high
efficiency, 34 mm dome
tweeter with
JA-waveguide.
JA-2806 Jantzen Audio
silk dome tweeter
together with our
drivers:
JA-5006
JA-6006
JA-6012
JBL L100 Century
Upgrate Kit with
Jantzen Audio
components.
A true renaissance
complete kit for the
many JBL lovers.
Please have a look at our website
www.jantzen-audio.com
PHOTO 3: TC Sounds 2574 driver.
to the various links to Jantzen Kits.
More to come.
September 2007
perazella2364.indd 23
23
7/25/2007 4:02:47 PM
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
parameters Fb and Qb.
For example, if you have two drivers
with the same free air resonance, but one
has a Vas that is twice the other, then for
the same change in resonant frequency
when mounted in a box, the driver with
the higher Vas will need a box volume of
twice that of the driver having half the
Vas. This also has a significant impact on
overall box size.
The Tang Band has a Vas of 23 liters
and the TC Sounds has a Vas of 14.2 liters.
Therefore, to achieve the same change in
parameters, the TC Sounds can go into a
smaller sealed box. Refer to the aX website
(www.audioXpress.com) for the specifications
of the TC Sounds and Tang Band drivers.
To understand the reasons for some
of the differences in the pricing of these
PHOTO 4: TC Sounds in
relationship to the Vifa.
24
perazella2364.indd 24
•
SPEAKER BUILDER
drivers, you must look at the construction.
Photo 4 shows the TC Sounds in relationship to the Vifa. For many reasons, the TC
Sounds is physically huge compared to
the Vifa, including those that have to do
with the need to keep efficiency up while
extending Xmax, having a relatively low
Fs and low compliance. Photo 5 shows
the Tang Band in relationship to the Vifa.
Here again, the difference with the Vifa is
dramatic if not as large as the TC Sounds
comparison. Finally, Photo 6 shows the TC
Sounds compared to the Tang Band.
Although beyond the scope of this
article, note that given the right motor
structure, electronic means can be used to
modify the passive effects that are caused
by box volume and Vas. Utilizing those
methods, however, would put this well be-
PHOTO 5: Tang Band in
relationship to the Vifa.
September 2007
7/25/2007 4:03:06 PM
S
R
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
yond a “keep it simple” project.
•
SPEAKER BUILDER
•
SPEAKER BUILDER
MCap® RXF
Radial Xtra Flat Capacitor
SIZING THE BOX
The next step was to determine the box volume. Again, to keep
it simple I just plugged in the T/S parameters from the manufacturers into a spreadsheet I developed. That spreadsheet, along
with instructions, is available from the audioXpress website (www.
audioxpress.com). The goal was to have an Fb of 40Hz. With the TC
Sounds, that resulted in a box volume of about 0.5ft3.
Because this was such a small volume and the driver motor was
quite large, I slightly upsized the box to 0.6ft3 to allow for the volume lost to the motor and basket structure. For the Tang Band, the
box volume was about 0.8ft3. At this point, it was such a close call
between the box volumes I realized that to really decide, I’d need to
build them both. It would be a good example of trade-offs.
POWERING THE SUB
Deciding the power source was a no brainer. We are blessed today
to have self-contained subwoofer amplifiers, sometimes called
plate amplifiers—not to be
confused with vacuum tube
terminology—that have
outstanding performance
and are excellent values.
Everything you need to
drive the sub is included in PHOTO 6: TC Sounds compared
one piece, including power to the Tang Band.
amp, summing network,
auto on sensing, crossover,
and equalization. All these
components come attached to a metal plate that
is mounted into a cutout
in the sub cabinet, making
life easy.
I chose the 300-794
from Parts Express (Photo
7). At the 4Ω impedance
of the drivers, it is rated at
250W, more than adequate
for the levels I was looking for. The circuitry for PHOTO 7: Parts Express 300-794.
this amplifier is shown in
Photo 8.
NEW! MCap® RXF/Oil Impregnated
Featuring the ultimate winding geometry (edgewise) for
• extremely short, low-loss signal transmission,
• extremely reduced residual-resistance (ESR),
• remarkable low residual-inductivity (ESL).
Polypropylene capacitor-foil, alu metallized.
Grouted winding against microphonic effects.
• Fit-In-Adaptors now available.
MCap® Supreme
NEW! Silver/Gold/Oil
Baked Varnish Litz Coils
NEW!
MCap® Supreme Silver/Oil
MCap® Supreme Silver/Gold
MResist™
Supreme
Audiophile Resitors
Varied Air Core Coils
OFC Pole Terminals
Gold Plated / Rhodanized
BUILDING THE BOX
Looking at the suggested
cutout for the amplifier, I
had an idea. Why have a
cutout at all? Because the
boxes were so small, if I
chose the dimensions correctly, the amplifier could
become the rear wall of
the box. So, that is what
I did. As a result, the box
consisted of only five pieces—four sides and a front
Silver/Gold Internal Wirings
MSolder™ Silver/Gold
MSolder™ Supreme
PHOTO 8: Amplifier circuitry.
MUNDORF EB GmbH
info@mundorf.com
mundorf.com
High End Components Made In Germany Since 1985
September 2007
perazella2364.indd 25
25
7/25/2007 4:03:24 PM
•
SPEAKER BUILDER
SPEAKER BUILDER
mounting piece.
Actually, after looking at the drivers, I
made one concession to aesthetics with
the TC Sounds box. I cut another piece
the same outline size as the front piece to
be used as a trim plate and make it look
as though the driver were flush-mounted.
That driver has a nice-looking rubber trim
ring on the front of the basket and the
idea worked well. The Tang Band is a little
more utilitarian-looking on the front with
a more traditional mounting piece, so I
decided to rear-mount that driver.
Because the rear opening was determined by the amplifier, the only differences in the boxes were their lengths. The
piece dimensions and final box sizes with
amplifier are shown in Table 1. I built the
box with ¾˝ MDF, using glue and biscuits.
I could have just as well used screws, but I
prefer biscuit joints.
Table 1 Box Dimensions
Note: All material is ¾˝ MDF.
Box
for TC
TC2574
2574
BoxPieces
Pieces for
Qty
Description
Length
2
Side panels
12˝
2
Top and bottom panels 12˝
2
Front panel and trim
10¾˝
Box Pieces for TB W8-740C
Qty
Description
Length
2
Side panels
16˝
2
Top and bottom panels 16˝
1
Front panel and trim
10¾˝
Finished Sizes Including Amplifier
Driver
Length
Width
TC 2574
14½˝
10¾˝
17¾˝
TB W8-740C
10¾˝
Width
9¼˝
10¾˝
10¾˝
Width
9¼˝
10¾˝
10¾˝
Height
10¾˝
10¾˝
•
SPEAKER BUILDER
•
SPEAKER BUILDER
If you have never built wood projects
using biscuits, you owe it to yourself to try.
Not only are the joints extremely strong,
assembly is very simple. In case you are
not aware of the technique, biscuits are
small flat pieces of wood in the shape of
a “squashed” biscuit (Photo 9). The alignment of the grain in the biscuits is such
that maximum strength results when they
join the plates. Slots are cut in the facing
plates that are to be joined to receive the
biscuits (Photo 10).
The use of a plate joiner to cut the slots
for the biscuits is not only fast, but it also
results in precise positioning of the pieces
when they are assembled. A typical plate
joiner ( Photo 11) has a rotating cutter
wheel mounted in a frame that allows
precise height, depth, and angular adjustments. Photo 12 shows the cutter wheel
in relation to the frame that positions it to
the piece being cut.
The biscuits—when inserted into these
slots—keep the pieces in alignment during
the assembly and add strength to the joint.
Glue is added to the slots for the biscuits
and the facing area of the plates. The biscuits are put into their slots and the plates
pressed together. If done with a little care,
the resulting joints are flush to the point
that you cannot feel the seam.
The first step was to lay out the pieces
and cut them out on a table saw. Photo
13 shows the four sidepieces for the TC
Sounds box. Photo 14 shows the front
plate after I cut slots for the biscuits, but
•
SPEAKER BUILDER
before cutting the mounting hole. Next, I
applied glue to the four sides of the box,
added biscuits, assembled the sides, and
clamped them in place. Photo 15 shows
the resulting box ready for mounting of
the front plate. I cut the driver mounting
hole using a router and then mounted the
plate to the box. For the TC Sounds driver
box, I cut the decorative trim plate, which
I routed and glued to the front plate. The
process was the same for the Tang Band
box except that no trim plate was used.
FINISHING THE BOX
Once each box was assembled, I rounded
all external edges with a router, except
those on the back face. After sanding, I
primed and painted the boxes. I had decided to use black paint, but while looking
at the various paints available, I ran across
a textured paint from Rust-oleum®. I decided to try it and purchased #7220, which
is textured black.
This paint was easy to apply, and after
drying had a nice, fine texture. The appearance is not like a flat finish, which
can sometimes appear to pick up different
tones depending on the light angle. It also
does not have the coarseness of the older
style wrinkle finishes. Overall, I was very
pleased with the results. The finished box
did not have a homemade look that you
sometimes get with normal paint finishes.
The next step was to drill the mounting
holes for the driver. The TC Sounds driver
has very substantial push-type mount-
PHOTO 9: Biscuit.
PHOTO 11: Typical plate joiner.
PHOTO 13: Four sidepieces for the
TC Sounds box.
PHOTO 10: Slots cut into faceplates.
PHOTO 12: Cutter wheel.
PHOTO 14: Front plate after slots cut.
26
perazella2364.indd 26
September 2007
7/25/2007 4:03:34 PM
S
R
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
ing posts for the wires. The Tang Band
had polarized spade lug connectors —a
larger one for the positive connection and
a smaller one for the negative connection. For convenience, I decided to use a
standalone push-on terminal strip fastened
to the inside of the box to make the connections to the unit using the Tang Band
driver, making it easier to assemble and
later remove the amplifier if needed.
Mounting the terminal strip directly to
the inside wall of the cabinet would have
made inserting the wires a little cumbersome because the entrance holes would
then be facing the inside of the box rather
than the rear. To solve this problem, I cut
a small piece of wood the size of the terminal base and fastened its edge on to the
wall. I was then able to mount the terminal
to this wood piece with the entrance holes
and push buttons facing the rear. Photo
16 shows the terminal strip inside the box
used for the Tang Band driver.
Next, I mounted the drivers. As previously mentioned, I mounted the TC
Sounds driver to its box from the outside,
and mounted the Tang Band from the
inside. Normally, I use square drive screws
PHOTO 15: Box ready to mount front plate.
PHOTO 16: Terminal strip inside the box.
•
SPEAKER BUILDER
from McFeely’s, but when I checked my
supply, I did not have the right size. After
using regular Phillips head screws, I decided not to let that happen again. In spite
of that, I did mount the drivers. Before
mounting the amplifiers, I connected the
Tang Band driver to the intermediate terminal strip.
FINAL ASSEMBLY
Mounting the amplifiers is a very simple
process. I first placed them in position on
the back of the subs and marked for the
pilot holes. After removing the amplifiers,
I drilled pilot holes in the back edges of
the boxes to receive the screws supplied
with the amps.
Sealing the amps to the back of the box
is ensured by the foam gasket mounted
around the edges of the amps. Power from
the amps to the drivers is provided by
one each red and black heavy gauge wires
soldered directly to the circuit board. The
wires were terminated with ¼˝ spade lug
connectors, which were removed for this
application. I connected the wires to the
appropriate terminals with excess wire
dressed neatly, replaced the amps in posi-
September 2007
perazella2364.indd 27
27
7/25/2007 4:03:39 PM
SPEAKER BUILDER
•
SPEAKER BUILDER
tion on the backs of the boxes, and screwed
them in place using a cross-tightening pattern to uniformly compress the rubber
gaskets.
The final construction step was to screw
four rubber bumpers to the bottoms of
each box. Photo 17 shows a three-quarter
view of the complete sub using the TC
Sounds driver. The rear view of the same
sub is shown in Photo 18.
The complete Tang Band sub is shown
in Photo 19, in which you can easily see
the extra length of this sub that was required to achieve the additional volume
that the larger Vas required. For a size
reference, Photo 20 shows the TC Sounds
sub with an LP resting against the side.
You can see that the box is lower than the
album cover and only slightly longer.
•
SPEAKER BUILDER
•
SPEAKER BUILDER
very close to the target goal of 40Hz. It
appears that the spreadsheet did a good
job of selecting box volume.
LONG-TERM RESULTS
The TC sub was the version I eventually
took with me to the UK, and it worked
extremely well, producing more bass than I
could reasonably use in the small flat that I
had. A few times I was playing it at a level
that I thought might provoke the ire of
my neighbors, but they exhibited the proverbial British patience and did not come
knocking on the door. The British are the
nicest people.
•
SPEAKER BUILDER
Several of my colleagues from the office came by and were astounded by the
volume of clean bass coming from such a
small box. They had never heard that level
of performance from such a small sub.
After that trip, our primary residence
moved, but I needed to keep an apartment
near company headquarters for the times
I was there on business. The sub has been
performing well there for almost three
years in conjunction with a pair of Infinity Primus 150 speakers. I’m using a pair
of old Orban parametric equalizers to roll
off the low frequencies going to the 150s
at 80Hz with a 12dB/octave slope. This
TESTING
For testing purposes, I brought both subs
to my friend, Tom Nousaine, who has
done a huge amount of speaker testing
both for himself and various publications.
The primary goal was to see how well the
actual box tuning matched the calculated
value determined by the spreadsheet.
Figure 1 is the response of the two subs
superimposed on the same graph. The sub
using the TC driver is shown with a dotted
line and the TB with a solid line. As you
can see, the curves are virtually identical.
The -3dB point is about 42Hz, certainly
PHOTO 17: Three-quarter view of the
completed sub using TC Sounds driver.
PHOTO 19: Completed Tang Band sub.
Table 2: Tang Band Q8-740C
specifications
Diaphragm MTL
surround MTL
nominal impedance
DCR impedance
Sensitivity 1W/1m
frequency response
free air resonance
voice coil diameter
air gap height
rated power input
maximum power input
force factor, BL
magnet weight (50 oz.)
moving mass
ferrofluid enhanced
suspension compl.
Effective piston area
Levc
Zo
Xmax
Vas
Qts
Qms
Qes
28
perazella2364.indd 28
paper
foam
4Ω
3.2Ω
84dB
28-1kHz
28Hz
50mm
8mm
120W
240W
13.35Tm
1427g
95g
No
333MN-1
0.022M2
3.18mH
113.99Ω
12mm
23 ltr
0.30
10.53
0.30
PHOTO 18: Rear view of the same sub.
PHOTO 20: TC Sounds sub with an LP
resting against the side.
FIGURE 1: Response of two subs superimposed on the same graph.
September 2007
7/25/2007 4:03:45 PM
S
R
SPEAKER BUILDER
•
SPEAKER BUILDER
makes their job much easier.
In addition, the Orbans did some minor
room tuneup. Again, the small size of the
sub makes it perfect for apartment use.
Photo 21 shows how I placed the sub
under an end table with the two Orbans in
front. It is effectively hidden by the table
and equalizers. I gave the other sub to a
friend, who is also using it with a pair of
small satellites.
There is no doubt that the design fulfilled the limited requirements that I had
for apartment use. If you have similar
needs, this is a good way to get reasonable
bass simply. Neither of these designs will
fill the need for high levels of very low bass
in large volumes. For that, you will need
much more linear volume displacement
than these drivers provide. But, if your
needs are modest and you want to keep it
simple, this is a good way to go.
RECENT DEVELOPMENTS
Since these subs were built, several changes
have occurred in the market. Tang Band
drivers are now available from Parts Express. The TB W8-740C, now listed as a
W8-740P, carries the Parts Express part
number 264-854 and sells for $65.37. The
T/S parameters are the same, so box size
•
SPEAKER BUILDER
•
SPEAKER BUILDER
should remain unchanged. The only apparent change in the new version is a higher
power handling capability.
The TC Sounds 2574 has been replaced
by the 8TC-1000. The good news is that
most of the parameters are the same, with
the exception that the Xmax is now an unbelievable 24mm, the Vas has been reduced
to 12.3 ltr, and the best news, the price has
been reduced to $149.
The specifications of the TC Sounds
8TC-1000 and the Tang Band W8-740P
are posted on the aX website (www.audioXpress.com).
This sheds a whole new light on the
decision process as to which driver to use.
The TC is now only slightly over twice the
price of the TB, but offers nearly double
the linear displacement. In addition, it can
be used in a smaller enclosure. It’s a tougher decision now, but both are viable choices
depending on your requirements.
The original amp from Parts Express
is no longer available, but a good substitute would be either the SA-240 or SA240B. The part numbers are 300-804 and
300-805, respectively, and each sells for
$128.88.
•
SPEAKER BUILDER
As this goes to press, Parts Express has
put the Tang Band W8-740P driver on
sale for $38.88 through 10/31/07. If you
are thinking about using this driver, now is
the time to buy. aX
TABLE 4: Tang Band W8-740P
Specifications
Diaphragm MTL
surround MTL
nominal impedance
DCR impedance
Sensitivity 1W/1m
frequency response
free air resonance
voice coil diameter
air gap height
rated power input
maximum power input
force factor, BL
magnet weight (50 oz.)
moving mass
ferrofluid enhanced
suspension compl.
Effective piston area
Levc
Zo
Xmax
Vas
Qts
Qms
Qes
paper
foam
4Ω
3.2Ω
84dB
28-1kHz
28Hz
50mm
8mm
150W
300W
13.35Tm
1427g
95g
No
333MN-1
0.022M2
3.18mH
113.99Ω
12mm
23 ltr
0.30
10.53
0.30
PHOTO 21: Sub under an end table
with two Orbans in front.
TABLE 3: TC Sounds Thiele/Small
Parameters
Qts
Qes
Qms
Fs
Res
Ls
Lp
Rp
Dia
Vas
mms
cms
bl
Spl
0.269
0.286
4.56
27Hz
3.70Ω
2.48mH
3.28mH
3.15Ω
145mm
12.31
105g
312µm/N
15.4Tm
81.6dB
tangband_ad.indd 1
perazella2364.indd 29
September 2007
7/13/2007 2:48:16 PM
29
7/25/2007 4:03:53 PM
•
SPEAKER BUILDER
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
The Tuba 24 II
You don't need to be a professional DJ to appreciate the performance of this noted speaker
designer's latest horn-loaded sub.
By Bill Fitzmaurice
M
uch has happened since I introduced the Tuba 24 prosound sub (aX, April 2004).
I’m now working full-time as
a loudspeaker designer, and my website, www.billfitzmaurice.com, is one of the
busiest DIY speaker sites on the web.
As for the Tuba 24, the first of what
turned into an entire line of DIY hornloaded pro-sound subwoofers, it and its
siblings have become the subs of choice
of professional DJs. Not just the DIY
subs of choice, but the preferred subs,
period. Attesting to their popularity is
a single thread at www.djforums.com devoted primarily to Tubas that stretches
back to January 2005, with over 3500
posts and 65,000 views. But as good as
the Tuba 24 is, there’s always room for
improvement, and thus the Tuba 24 II
(Photo 1).
PHOTO 1:
Completed Tuba 24II.
NEW FEATURES
The new version leaves behind the vertical baffle of the original, in favor of a
horizontal baffle that allows more leeway in your choice of cabinet widths
and driver complement. The side measurement, 24 × 24˝, is the same as the
original, as is the path length and phase
response, so you may use the old and
new versions together. But whereas the
original could only be built 24˝ wide,
you can build the T24II as narrow as
16˝ or as wide as 30˝. For best results,
you should use the T24II in multiples of
two. Because the 16˝ wide T24II weighs
only about 38 lbs, carrying two is a oneman job.
The recommended basic driver is the
Eminence Basslite S2010. Pertinent T/S
specs are fs 46Hz, Qts .31, Vas 63 ltr,
Xmax 4mm, and Pe 150W. An upgrade
30
fitzmaurice2845-1.indd 30
FIGURE 1: Comparison of S2010 and BP102 drivers.
FIGURE 2: T24II and Community XLT51e.
September 2007
7/31/2007 12:03:27 PM
S
R
SPEAKER BUILDER
•
SPEAKER BUILDER
is the Eminence BP 102, which has a
6mm Xmax and 200W power handling.
Figure 1 shows the 1m/2.83V halfspace response of a 16˝ wide cab loaded with S2010 and BP102 drivers. The
BP102 is less sensitive, but will handle
twice the power before exceeding Xmax,
so it’s the better choice for high output situations. You can get the Eminence drivers, along with all the other
hardware required for the project, at
www. partsexpress.com, www. bltsound.
com, and Langford Audio Systems (email LawLangford@AOL.com) in the US,
and www.loudspeakers.ca in Canada. The
150W rating of the S2010 may seem
low, but the T24II has far higher sensitivity than direct radiator subs.
Figure 2 compares a 16˝ wide T24II,
S2010 loaded, to a similar size 15˝ loaded direct radiating sub, the Community
Sound XLT51e. The T24II’s average
6dB sensitivity advantage gives it the
ability to produce as much output as two
direct radiator cabs driven with twice
the power.
The cabinet is primarily made from
½˝ plywood; the self-bracing design
makes thicker materials unnecessary.
You may use either softwood (spruce or
pine), plywood, or Baltic birch. With a
carpet finish I recommend softwood,
because it is lighter, cheaper, and easier
to find than quality Baltic birch.
Be sure your plywood has at least
five plies. With a 4 × 8´ sheet of plywood, you can build one 16˝ wide T24II,
fastening the parts with 1.25˝ drywall
screws, 1 5/8˝ ribbed shank paneling
nails, or a pneumatic brad nailer, using
1.25˝ brads. Screws and/or nails don’t
hold the cabinet together; they merely
serve to hold the parts in place while the
adhesive in the joints sets.
I recommend PL Premium polyurethane base construction adhesive, applied with a caulking gun. Polyurethane
adhesive expands as it cures, filling gaps
to ensure the airtight seal that a speaker
cabinet requires. However, urethane is
also messy to work with. After applying
a bead to a joint, release the pressure on
the caulking gun trigger and plug the
end of the tube before it oozes all over
your workshop, and you.
•
SPEAKER BUILDER
•
SPEAKER BUILDER
net sides. On one side draw the parts
layout pattern per Fig. 3. Figures 4 and
5 show the nominal parts lengths and
assembly order of the internals, and the
cut angles at the ends of the panels. If
no angle is shown, the end is square. The
parts lengths assume that the material is
exactly ½˝ thick, which is almost always
not the case, so cut each panel to final
size as it is installed, measuring the cabinet side to confirm the exact length.
The measurements shown at each
bend of the horn are to the edge of the
side. After drawing the joints on one
FIGURE 3: Parts layout.
•
SPEAKER BUILDER
FIGURE 4: Parts dimensions.
FIGURE 5: Joint angles.
CONSTRUCTION
Start by cutting out the 24 × 24˝ cabiSeptember 2007
fitzmaurice2845-1.indd 31
31
7/25/2007 3:59:13 PM
SPEAKER BUILDER
•
SPEAKER BUILDER
side, clamp the two sides together. Drill
1/16˝ pilot holes through the middle of
each joint. Do not drill closer than 2˝
or so from the end of the joint, with the
holes spaced 6-8˝ apart. Drill the holes
entirely through both sides.
If you’re using a brad shooter, just drill
two holes per joint, one at each end. Unclamp the sides. Draw connecting lines
through the holes on the outside of each
so you can easily see how the parts are
laid out on the opposite side.
On the first side, draw four lines 3/8˝
inside the joint lines of panels 1-4, parallel to those joints, as shown in Fig. 6.
These outline the driver access hole. To
cut it, clamp or screw a circular saw sled
or cutting guide to the side. Raise the
saw out of the shoe, place the shoe tight
against the sled guide, start the saw, and
make a plunge cut. You’ll want to practice this technique on some scrap wood
before trying it for real.
In turn, cut each of the four lines not
quite to the end, finishing the cuts with
a jigsaw. Be sure you have the side raised
on some scrap boards, so that you don’t
saw into the bench top. Put the cutout
aside, saving it to use as a cover.
Cut four 1¼˝ wide pieces of plywood,
sized to frame the access hole, forming
FIGURE 6: Driver access hole layout.
•
SPEAKER BUILDER
•
SPEAKER BUILDER
a mounting flange for the cover on the
inner face of the side (Fig. 7). Clamp the
strips in place when screwing or nailing
them. The resulting hole has plenty of
clearance with the S2010 driver, but the
BP102 might be a little tight. If you’re
using the BP102 or another driver, make
sure the driver fits through the hole,
trimming the flange as required.
Braces connect all panels except the
baffle. They may be made of ¼-½˝ plywood, their sizes determined by dead
reckoning, as shown in Photos 2 and 3.
Place a piece of plywood of sufficient
size against the inner part to which it attaches, lay a straightedge across it where
the outer part will joint to it, and cut the
line drawn. After sawing, lay the brace
in place again to check the accuracy of
its sizing.
Draw joint lines on the parts where
the installed braces will join to them.
The braces are not jointed end to end.
If you make the braces from ¼˝ plywood, dado a ¼˝ wide, 1/8˝ deep groove
down the center of each panel, to hold
the brace secure while the adhesive
dries. Make sure you put the groove on
the correct side of the panel; panel 6 is
grooved on both sides. Be sure to size
the braces to account for the dado depth.
•
SPEAKER BUILDER
With ½˝ plywood braces you can omit
the dadoes.
To install the braces apply adhesive
to their edges and slide them into place,
pushing in far enough to just be snug,
using a framing square to be sure that
the panels remain square to the assembly. Dado-mounted ¼˝ braces don’t need
fastening; ½˝ braces need a couple of
screws/nails/brads to hold them in place
while the adhesive sets.
The approximate standard cabinet
width is 16˝, so for minimal waste with
4 × 8´ plywood, make the remaining
pieces all 15¾˝ wide. If you are using a
table saw, cut all the pieces to width at
the same time without moving the rip
fence, so they’ll be identical. The roughcut lengths of the parts follow. Cut them
to these lengths initially, trimming to
their measured finished size as you install each. Be sure to label them.
1. 11.25˝
7. 17.25˝
2. 15˝
8. 23.75˝
3. 15.5˝
9. 23.75˝
4. 10.75˝
10. 23.75˝
5. 9.25˝
11. 5.5˝
6. 12.5˝
ASSEMBLY
Panel 1 is the baffle. Cut a hole through
FIGURE 8: T-nut/baffle detail.
PHOTO 2: Marking brace for cutting.
FIGURE 7: Driver access hole flanges.
32
fitzmaurice2845-1.indd 32
FIGURE 9: Panels 1 and 2 in place.
PHOTO 3: Checking brace for accuracy.
September 2007
7/25/2007 3:59:17 PM
S
R
SPEAKER BUILDER
•
SPEAKER BUILDER
it, sized according to the driver manufacturer’s specifications. With lightweight drivers such as the S2010, screws
are adequate for driver mounting, while
3/16˝ bolts and T-nuts are an option.
Place the driver on the baffle, centered
over the hole. Mark the screw/bolt locations through the frame holes. Put
the driver aside. Drill the baffle for the
screws or T-nuts.
When using T-nuts you sometimes
find there isn’t enough wood on the
baffle to seat them well near the baffle
cutout. A way around this is to mark
the hole the same diameter as the inner
diameter of the driver gasket, but cut the
hole with the jigsaw shoe set at a 30 to
45° angle, so that the hole diameter on
the opposite side of the baffle is smaller
(Fig. 8). This leaves much more wood for
the T-nut to grab.
A heavy-duty version T-nut is the
Hurricane Nut, available from Parts Express. Either will work best if you coat it
with some Gorilla Glue or the equivalent before driving it in, being careful
not to glue the threads. Drive the T-nuts
in place with a hammer, making sure
that they are located on the bottom of
the baffle. Trial-mount the driver to the
baffle to be sure that all the nuts are
properly installed, then remove it and
put it aside.
Attach panel 1 to the side. The best
way to fasten joints is to clamp a 2 × 2˝
guideboard along the joint edge, apply
adhesive to the joint, clamp the panel to
the guideboard, and, when all is right,
screw or nail it in place. Drill screws
through the pilot holes with a pilot/
countersink bit to prepare the hole before driving the screw.
With nails, drill through the predrilled pilot holes about ½˝ or so into
the panel to prevent splitting, using a
nail set to drive the head of the nail
below the face of the side. With panels
1 through 4 use the guideboard to hold
the panel tight against the flange. Drive
fasteners through the side, but don’t
drive fasteners into the flanges.
Cut out panel 2; attach it to the assembly (Fig. 9). If you’re using ¼˝ braces,
dado panel two and all other panels for
bracing before installing. Always clamp
joints and check the parts alignment
with a framing square to be sure they are
true before driving screws or nails.
•
SPEAKER BUILDER
•
SPEAKER BUILDER
Use wood scraps temporarily tacked in
place to hold the free edges of the panels
in place for proper alignment (Photo 4).
•
SPEAKER BUILDER
Install panels 3 and 4, then panels 5 and
6, and the brace between them (Fig. 10).
Install panels 7 and 8 and their associ-
FIGURE 12: Lower back detail.
FIGURE 10: Assembly through panel 6.
FIGURE 11: Handle hole detail.
FIGURE 13: All panels in place.
September 2007
fitzmaurice2845-1.indd 33
33
7/25/2007 3:59:25 PM
SPEAKER BUILDER
•
SPEAKER BUILDER
ated braces.
FINISHING TOUCHES
The lower rear of the cab is a good place
to install a pair of cutout handles. Cut
them into panels 9 and 10 before installing the panels, as shown in Fig. 11.
The cutout is produced by drilling two
1½˝ diameter holes 4˝ apart on-center,
removing the material in between with
a saw.
Another option is 2 × 3˝ cutouts (Fig.
12) that allow mounting 2½˝ casters,
bolted to panel 11. Cut these also before
installing the panels. Install panels 9 and
10 and their braces, and then panel 11
(Fig. 13).
A good spot for jacks is on the lower
section of panel 9, just above panel 11.
Use Speakon jacks, which have the advantage of being airtight. A pair of jacks
will allow daisy-chaining of multiple
cabs. You can mount them with a commercial mounting plate, or you can make
your own from 3˝ diameter plywood
discs.
Drill a pair of 2˝ holes in panel 11,
attach the discs to inside of panel 11
PHOTO 4: Temporary braces keep panels squared.
FIGURE 14: One, two, four, and eight cabs.
34
fitzmaurice2845-1.indd 34
•
SPEAKER BUILDER
•
SPEAKER BUILDER
over the holes, then drill 15/16˝ holes
through the discs for the jacks. Drill a
hole through panel 2 for the wire to pass
through. Feed a piece of 14 or 16 gauge
speaker wire from the driver chamber to
the jack location, sealing the hole airtight with adhesive.
Lay the cabinet on its side. Apply
a generous bead of adhesive on all the
panel edges. Put the second side atop the
assembly, making sure that it is oriented
so that the pilot holes line up with the
panels. Nail or screw it into place, using
long pipe clamps to pull it into perfect
alignment with the rest of the cabinet.
Screws are the better option to pull the
side tight to the inner panels.
Look inside the box through the driver access and the mouth to make certain
that all the joints have adhesive squeezeout, applying more as required to seal
the joints airtight. When the adhesive
has set, true to flush the exterior edges
with a sander or router, rounding over
the edges as may be required for any
protective hardware.
If you’re painting the cabinet, fill the
holes over the nail or screw heads before
PHOTO 5: Speakon jacks.
•
SPEAKER BUILDER
sanding and finishing the box. I prefer
polyester auto-body filler for this job, because it sets fast, holds tight, and doesn’t
shrink. If you’re carpeting the box, filling
the heads is optional.
Put the access cover in place and drill
eight to ten pilot holes for mounting
screws or bolts/T-nuts through it and
the flanges. Apply your finish of choice.
DuraTex, available from www.AcryTech.
com, is what the pros use. This waterbase urethane acrylic goes on with a
roller or hopper gun, drying quickly with
a textured finish that’s very durable and
it’s easily recoated if you ding it. It costs
slightly more than paint, coverage is
quite good, and you can finish a cabinet
of this size in about 15 minutes exclusive
of drying time.
TESTING
Install and wire the driver and jacks.
Connect the +, - lugs of the driver to
the +, - lugs of the Speakon jacks. Seal
around the jacks with hot-melt glue
or caulk (Photo 5). Rim the porthole
flanges with weather-strip and screw the
porthole cover in place. The rear chamber is not lined or filled.
Test the cab, using a 30Hz test tone
f rom a generator or CD. Gradually
increase power, listening for air leaks,
which will seriously detract from performance, so be sure every joint is tight.
When using multiple cabs it’s critical
that they be wired in phase. To check
phase run a test tone, anywhere from 30
to 100Hz, through one cab, then plug
the second cab into the first. The level
should go up. If the level goes down,
you’ve either got the driver or jacks reverse-wired in one of the cabs, or the
cable connecting the two cabs is reversewired at one end.
The key to high output isn’t high
power handling, it’s high sensitivity,
which increases as more cabs are added
to the pile. Figure 14 shows the responses of one, two, four, and eight S2010
loaded 16˝ cabs with 2.83V input. For
even higher sensitivity, make the cabinet
wider, with panel widths up to 26˝. Figure 15 compares 16 and 24˝ wide S2010
loaded T24IIs. When building with 18˝
and wider panels, use two sets of braces,
evenly spaced.
With nominal cabinet widths of 24 to
30˝, you can use two drivers for increased
September 2007
7/25/2007 3:59:27 PM
S
R
SPEAKER BUILDER
•
SPEAKER BUILDER
power handling. Figure 16 shows a 24˝
cab loaded with dual S2010 drivers, parallel wired. Don’t bother going wider
than 30˝; it’s more practical to use two
smaller cabs with one driver each. Dual
drivers require a brace installed between
panels 1 and 5, midway across the box.
You can achieve even higher sensitivity with the use of a “V” coupling. Stack
the cabs in pairs at a 90° angle (Fig. 17).
Then actively couple their mouths (Fig.
18) with a plate (Fig. 19) that extends
the effective horn path. Attach the plate
to the cabs via thumbscrews and T-nuts.
The result, shown in Fig. 20, is an average sensitivity increase of 2.5dB, which
is almost the equivalent of a doubling of
power.
With more than two cabs, add ad-
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
FIGURE 15: 16 and 24˝ wide cabs.
FIGURE 16: One vs. two drivers.
PHOTO 6: Latching cabs together.
PHOTO 7: Two cabs latched for
transport, front view.
September 2007
fitzmaurice2845-1.indd 35
35
7/25/2007 3:59:31 PM
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
ditional pairs above the first, with just a
single V plate at the top of the stack. For
ease of transporting pairs of T24IIs, you
can clamp them together with PennElcom # LO925 clamps (Photo 6). Two
extra holes in a V plate allow you to
thumb-screw it to the existing T-nuts on
assembly for transport (Photo 7). With
wheels on both the top and bottom in
transport mode, you can wheel the assembly to a van or station wagon, then
tip it to place the upper wheels in the
vehicle for an easy roll-in load-up.
FIGURE 17: Two cabs arrayed.
SETUP TIPS
Figure 21 shows the impedance of an
FIGURE 18: Two cabs arrayed with V coupler.
PHOTO 8: Two cabs latched for
transport, rear view.
36
fitzmaurice2845-1.indd 36
S2010 driver in a 16˝ wide T24. The air
mass of the horn adds approximately
2Ω of acoustic impedance to the driver’s
nominal 8Ω impedance, resulting in a
nominal 10Ω load. Two cabs parallelwired will show a 5Ω load; four cabs
will produce a 2.5Ω load. A dual driver
cab parallel-wired will have a 5Ω load;
two dual driver cabs will give a 2.5Ω
load.
A dual driver cab series-wired will
have an 18Ω load. Two such cabs will be
9Ω, four cabs 4.5Ω. Do not use a total
load less than what your amp is rated to
drive.
While it’s customary with PA to have
speakers to either side of the stage, that’s
usually not the best way to place subs.
Subs work best when they’re placed either close together for mutual coupling,
or spread very wide to cover large areas.
The basic rule is to have them either
less than a quarter-wavelength apart or
more than two wavelengths apart for
their passband, which for 40 to 100Hz
means less than 2.8´ or more than 56´.
Use boundary loading whenever it’s
practical to do so. Having subs next to
a wall gets you 6dB of additional sensitivity, while putting them in a corner
an extra 12dB. In most cases you’ll have
best results aiming the subs toward the
wall or corner with the mouth about a
foot away from the boundary.
Horn-loaded subs have their own peculiarities, some of which led to this
posting by Eminence regarding the Lab
12/Labsub; these caveats apply to Tubas
as well:
You cannot hear the driver distort when
you push them too hard. So, most people do
not know when to turn them down. They
September 2007
7/25/2007 3:59:33 PM
S
R
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
BK-16 Kit
Madisound is pleased to offer the BK-16
folded horn kit.
F
O
S
T
E
X
FIGURE 19: V coupler detail.
FIGURE 20: Two cabs with and without V coupler.
push them till they break. It takes a while
to get used to the extra clean sound of this
cabinet and learn how hard you can push it.
These are designed to be used in groups
of 4 to 6 cabinets to get the desired SPL at
very low frequencies (below 45Hz). A lot of
folks are running them as singles and trying to EQ the bottom end to get more low
bass output. This pushes the drivers past
their safe operating range very quickly. If
you need a lot of very low bass, use more
cabinets.
Air leaks will kill the driver. Care must
be taken to get the chamber sealed and keep
it sealed. Before every show, check all the
screws that keep the cover on as they may
work loose and cause a leak.
You must use a high-pass f ilter set to
35Hz and that has a slope of at least 24dB
per octave to realize the real potential of the
design. Many people are using huge power
on these cabinets day in and day out, but
they are the ones who run steep high-pass
filters on them. aX
We have chosen the Fostex FF165K 6.5"
full range for use in the BK-16 cabinet. The
FF165K has a Kenaf fiber cone, inverted
foam surround and aluminum dust cap.
The FF165K is run full range with a
frequency response out to 15kHz. The
T90A super tweeter has been added to
cover the upper frequencies. The T90A is a
top-mount horn
tweeter with an
Alnico magnet.
The tweeter is
rolled off on the
low end with a
Fostex Tin &
Copper
foil
capacitor. The
system
frequency response is 55Hz to 35kHz at
95dB.
The BK-16 cabinet is made from Baltic
Birch plywood and is sold flat,
unassembled, unsanded and unfinished.
Cabinet dimensions are 9.75" W x 14.75" D
x 29" T.
The kit includes:
• Pair of Cabinets - Flat
• Pair of FF165K - full range
• Pair of T90A - horn tweeter
• Pair of DB-Cup - Input cup
• Pair of Crossovers
• Nordost 2-Flat wire for tweeter
• Instructions
Kit Price: $635.00 /pair
Parts without cabinets: $439.00 /pair
Cabinets only: $98.00 each
MADISOUND SPEAKER COMPONENTS, INC.
8608 UNIVERSITY GREEN
P.O. BOX 44283
MADISON, WI 53744-4283 U.S.A.
FIGURE 21: Impedance chart.
TEL: 608-831-3433 FAX: 608-831-3771
e-mail: info@madisound.com
Web Page: http:/www.madisound.com
September 2007
fitzmaurice2845-1.indd 37
37
8/1/2007 10:54:21 AM
•
SPEAKER BUILDER
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
More on the Sound Strobe
With the aid of a speaker analyzer and woofer mod,
this author gets his speaker system in sync.
By Ed Simon
PHOTO 1: Test setup.
O
ne of the rules of DIY audio is
that the best-sounding loudspeaker is the one you build
yourself. That rule holds true
until you start making measurements.
LATE ARRIVALS
A while back I tried building a twoway loudspeaker by ear using just the
Sound Strobe. One of the things I did
not mention is that when I voiced the
crossover by ear, there was a 2dB dip
at the crossover versus where I would
have set it using just the frequency response measurements. With a bit of
38
simon2847-1.indd 38
measurement and some
thought, it became clear
that the devious Sound
Strobe impulse f rom
the woofer came after
the impulse from the
tweeter.
When you look at
a f requency response
curve, you can often see
a discontinuity in the
graph at the crossover
point. Some curves hide
this by using a smoothing function. It is possible in theory to get a
perfectly flat frequency
response curve for a
single microphone position with just about any
combination of welltweaked drivers. That
would work, if you have
only one functioning
ear and want to hold
your head in one exact
position.
The f requency response in theory is the reciprocal of the
time response (1/f = T). When listening
to the pulses, my ear sensed that a slight
ducking of the crossover region gave a
cleaner sound. The final arbiter of this
would, of course, be extended music listening tests, the results of which would
be open to debate because listeners have
different preferences. The solution is, of
course, to make sure the impulse from
the tweeter and the woofer arrive at the
same time.
In order to achieve that solution, some
loudspeaker designers slope the front
and move the tweeter back to delay its
arrival time. Others put more elements
in the tweeters’ crossover path to add
delay.
I decided to try a different technique.
I thought I would see whether I could
get the sound out of the woofer sooner.
TEST SETUP
Although I have a reasonable assortment
of test equipment, it seemed the simplest
way to test this theory was with a microphone, an oscilloscope, and that infernal
Sound Strobe driving both the scope
and a small amplifier.
For the amplifier I used my singleended solid-state amplifier, which lives
on my test bench. The microphone was
an ACO Pacific 7012 capsule with a
4012 preamp powered by a Bruel and
Kjaer 2801 modified power supply. Of
course, just about any microphone would
do for this application. I just keep that
setup on my bench. The oscilloscope
advertises itself on every picture! Almost
any oscilloscope or computer emulation
would do for this test.
The setup was quite simple: I placed
the loudspeaker on a stand with the
woofer 36˝ off the ground and the tweeter 42˝ up. I placed the microphone 39˝
high and 19˝ away. The Sound Strobe
fed channel one of my oscilloscope and
the microphone channel two.
MEASUREMENTS
Figure 1 shows my results. The top trace
shows the pulse provided by the test
signal. This pulse has a sharp rise time
and a slower exponential decay. Some
internal oscilloscope noise shows up on
this trace.
Channel two shows there is a delay
of 1.525ms before the microphone gets
September 2007
7/31/2007 12:05:46 PM
S
R
SPEAKER BUILDER
•
SPEAKER BUILDER
the signal. At standard temperature
and pressure (STP) the delay should be
about 73.6µs per inch. So there is about
125µs of other delay. This comes from
the crossover, the transducers themselves, and smaller amounts from everything else in the chain. The first pulse is
from the tweeter, then there is a delay
of 338µs until the woofer reaches maximum output.
The tweeter pulse is a classic lazy S
curve. (For those of you who do not
read brands, lazy means the letter is on
its side!) The tweeter has some mass,
although very small. It cannot move out
instantly as the input pulse would require. To do that would require an infinite force (at just a few watts at most,
the amplifier falls a bit short). The top
of the pulse rounds a bit as the tweeter
piston slows down when the input stops.
The tweeter does not cover the low frequencies, so its work is done. Because
the piston moved out, it now must move
back. That is why you see a negativegoing spike.
If the amplifier provided better damping to the tweeter, the negative-going
spike would be smaller. That is because
as the tweeter moves back in it generates a voltage—a short circuit across the
voltage would increase the resistance to
this motion. To some extent this could
be helped by a Zobel network, a low
value shunt resistor, or an amplifier with
higher damping factor at high frequency.
The last sounds the best, but because
loudspeakers are voiced with normal
amplifiers, such a design would sound
dull.
In Fig. 1 the woofer begins to rise
70µs after the tweeter. It takes an extra
268µs for the cone to fully move. This
corresponds to an upper frequency response of about 1800Hz, which is not
too bad for a woofer. The woofer has
an interesting flattened top. I could
guess why, but there are instruments
that could prove my guess wrong, and
because I never make misteaks, I won’t
guess.
That the output curve does not match
the input signal may come as a surprise
to some, but these are real-world results. The woofer, as mentioned, cannot
rise instantly, so it lacks the sharp front
edge. The time delay because the tweeter
is mounted closer to the microphone
•
SPEAKER BUILDER
•
SPEAKER BUILDER
than the woofer means that the spike
from the tweeter is not riding on top
of the woofer so the combined results
only crudely approximate the input signal. That is why time response, which
should be the reciprocal of frequency
•
SPEAKER BUILDER
response, is an approximation that does
not hold up well for real (finite bandwidth) devices.
SUB MOD
Now it is time to play. The motion in the
FIGURE 1: Large cap.
September 2007
simon2847-1.indd 39
39
7/30/2007 8:47:47 AM
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
•
SPEAKER BUILDER
FIGURE 2: Small cap.
woofer is started by applying a current to
the voice coil. This moves the coil and
sets the attached cone in motion. The
cone then moves the air. The velocity of
propagation in the cone is faster than in
the air. I decided to move the area of the
cone forward. I chose to couple to the
existing cone about three-quarters of the
way to the front edge of the cone.
A few years back my friend Bill bet
the combination center of two loudspeakers was based at the voice coil, as
many believe. I picked the closest edge
of the cone based on computer simulations I had done before on my then new
PDP8. We measured the -6dB beam
width of a dual 15˝ loudspeaker. This
point would be 45° off the center of
combination. On 15˝ loudspeakers this
would be far enough apart to give a good
indication of the center.
It was not at the voice coil but actually only about three-quarters of the way
out on the cone. We suspected this may
be due to cone break-up.
I later concluded that it may not have
been due to cone breakup. That is because in my computer model the time
for the motion of the cone to reach the
edge was zero. In the real world it would
take some time.
My estimate of the cone density for
this loudspeaker based on the dust cap I
removed was .42 ounces per cubic inch.
That is very, very light, but denser than
air. My seat-of-the-pants estimate is
40
simon2847-1.indd 40
that this would give me a velocity of
propagation only four or five times that
of the air.
If I break the cone up into three parts
—the dust cap, the cone middle, and the
edge—I can simplify my model of how
the pressure is generated. The air by the
dust cap is first pushed forward, then the
air coupled to the middle of the cone,
and finally the edge air. However, the air
from the edge reaches the ear first because it is the closest, then the air moved
by the middle, and finally the air put
into motion by the dust cap. As a result,
I do not get a clean pulse but a stretched
pulse!
To get the loudspeaker to couple
three-quarters of the cone distance closer to the front edge was not too hard. I
did this by cutting out the original dust
cap and replacing it with a larger one.
I used a 5˝ dust cap that I cut down to
about 3¾˝.
The trick in cutting a dust cap is to
place it on a flat surface and rotate it
against a pencil to get a uniform cut line.
Place the pencil on a spacer to get the
diameter you want.
The increase in cone weight from removing the smaller dust cap and adding
a larger one was 2.58 grams. This would
change the response of the woofer and
require a new frequency response tuning for the loudspeaker, but I wanted
to see what it would do to the impulse
response.
September 2007
7/25/2007 4:04:50 PM
R
rs
g
Di
e
r
Fo cord
Re
l
a
it
Introducing
US
B A Fo
/D r
Sy
ste
PHANTOM POWER
ms
7052PH
Measurement
Mic
Measurement
SystemMic System
7052H Type 1.5TM
Titanium Diaphragm
3Hz to >20 kHz
<20 dBA > 140 dBSPL
A
MK224 (<14 dBA to >134 dBSPL) Optional
4048 Preamp
Superior
IEC 1094 Type 1
Long-term Stability
Temperature and Humidity
Performance
Phantom
to IEPE/ICP
Adaptor
Supplies 3-4 mA
Power
Accelerometers
Microphones
Now in Stock
ICP1248
MATT
MATTtm
Family
TM
Mic
Attenuator
Handle Higher Sound Pressure Levels
ACO Pacific, Inc., 2604 Read Ave., Belmont, CA 94002
Tel:(650) 595-8588
simon2847-1.indd 41
FAX: (650) 591-2891 E-Mail: sales@acopacific.com
Web Site: www.acopacific.com
A Be
C gi
O ns
uw
s it
t h
i A
c C
s O
TM
7/25/2007 4:04:58 PM
SPEAKER BUILDER
•
SPEAKER BUILDER
Figure 2 shows the new response. The
test signal remains the same!
The easy way to notice that the woofer begins to move forward in time a
bit sooner is there is less of a negative
spike from the tweeter. This decrease in
the woofer’s initial response lag is about
25µs.
This corresponds to moving the
woofer 1/3˝ closer to the microphone,
which is just about one-half of where
the new dust cap sits. (Leading me to
suspect the velocity of propagation in
the cone is twice that of air; however,
the motion in the dust cap is mostly in
the transverse mode!)
I tried using a larger closer dust
cap, but the increase in mass and cone
breakup prevented that from working
well.
Looking at Fig. 2 you should also notice that the woofer now has a sharper
spike and a slower rise time. In addition,
the pulse width now is longer. These results are probably due to the increase in
mass resulting in a lower frequency response. Of course, the modified woofer
now has a much nicer lazy S curve.
•
SPEAKER BUILDER
•
SPEAKER BUILDER
I did not get as much time delay as
I would like. I could try a longer voice
coil form for the tweeter. A deeper
tweeter or an all-pass filter would also
be solutions. Also, it is important to
note that even with the time set optimally, there still would not be as sharp a
rise time as the input signal. The woofer
does not talk to the tweeter to split up
the workload. Each does its own thing.
The combination will not be perfect.
Better is the goal.
HOW IT SOUNDS
Of course, to be fair for listening tests,
I now must re-voice the crossover to
account for the change in the woofer.
But my preliminary result was that a noticeable sibilance in the loudspeaker on
some female vocalists was now reduced.
I expect that sibilance is caused by
upper midrange energy hanging around
too long. In this case the tweeter is
clearly before the woofer, although the
cancellation of the rebound tail may be
what I am hearing.
Most folks would argue that I cannot
hear the 25µs time change, that it must
•
SPEAKER BUILDER
be my imagination or something else.
But then they probably have not tried
this.
The important question is, “Does it
sound better?” The stock answer is, of
course, “If you try this it will sound better to you!” If you wish to try this yourself, be safe, use removable glue.
The result of all this is that when I
build a loudspeaker system I will first
need to see the time delay introduced by
each proposed component and then pick
and choose my drivers. This is an interesting extra step because this data is not
aX
given on datasheets.
REFERENCES
1. “A Solid-State Single-Ended Power
Amp,” by Ed Simon, Apr. ’06 aX, p.
24.
2. “The Sound Strobe,” by Dennis
Colin,” March ’06 aX, p. 17.
3. “A Combination Horn You Can
Build,” by Ed Simon, Jan. ’07 aX, p.
10.
4. “The Sound Strobe,” by Ed Simon,
Apr. ’07 aX, p. 6.
NeoPro5i、NeoCD2.0、NeoCD3.5H、NeoCD3.0、NeoCD1.0 ......
Fountek Performance Ribbon Transducers
email:info@fountek.net tel:+86-573-3019220 fax:+86-573-3019221
42
simon2847-1.indd 42
September 2007
7/25/2007 4:05:32 PM
R
simon2847-1.indd 43
7/25/2007 4:05:41 PM
Ohm my.
VoiceCoil_ohmmy.indd
simon2847-1.indd
44 1
6/13/06 4:05:54
5:23:24 PM
PM
7/25/2007
tips & techniques
By Darcy E. Staggs
Solder Turrets
Discover solder turrets to make your PC board
work easier and more reliable.
M
PHOTO 1: PC boards modified for solder turret use.
y amplifier project has caused
me to make, install, modify, and repair a series of PC
boards, now involving a third
iteration.
Whenever I needed to change the value
of a component, I had to remove the entire
board and attached power transistors for
access to the foil traces, which deteriorated
rapidly to my great displeasure.
As I looked back on the frustrating
debugging efforts spent on the first two
boards, failing solder joints emerged as
the main trouble source. Consequently,
I decided to maximize the use of “solder
turrets” to improve the reliability of the
board-to-wire connections.
After much digging, I rediscovered “solder turrets” in the Mouser catalog at the
very end of the hardware section, where
they are called “standardized terminals”
and are strangely absent from the index.
These old friends from long ago offered a
slick answer to all my reliability problems.
SOLDER TURRET EXAMPLE
Photo 1 shows a pair of PC boards modi-
fied for solder turret use. I drilled the
boards with holes to fit the turrets, which
I swaged into place by forming a shop head
on the underside of the PC. After mounting the turrets, I tinned all foil traces and
soldered the turrets. The problem of lifted
foil traces thus disappeared entirely.
TOOLING
You can purchase a mandrel to put in a
bench vise to hold the upside-down turrets
while forming their shop heads. I found it
very expedient to simply drill, grind, and
file the end of a bolt to produce a perfectly
serviceable mandrel. The hole in the end
of the bolt is made to just fit the head of
the turret, but support it by its flange during forming.
The bolt was narrowed because my turret layout crowded them together rather
closely, mainly because the foils were originally spaced for simple wire connections.
FORMING THE SHOP HEADS
Place the turret upside-down in the hole
in the mandrel, lower the PC board foilside up over the protruding turret, and use
a hammer and center punch to form the
shop head on the turret. I finish by flattening that head with a few light taps of
the hammer.
My too-close turrets needed to be filed
to miss each other, and I used a fine saw to
separate some shop heads, simply because
the boards were adaptations. Any new
board layouts will include proper spacing
for the turrets.
SOLDERING
Because the turrets are hollow, it is a simple matter to solder jumper wires and component leads into them and later repeat the
process if any servicing is required. Wires
can, where necessary, be soldered to the
outside of the turret.
Several power transistors have their leads
attached to the underside of the boards, but
this is easy and reliable thanks to the hollow turrets. The boards are mounted with
¼˝ nylon spacers, so the underside transistor mounting maximizes lead length which
eases fastening them to a heatsink.
CONCLUSION
The small effort spent attaching turrets to
PC boards is repaid ten times over during their use. A certain smug satisfaction
pervades me whenever I work on the amp,
because it has once more become a pleasure to do so. aX
audioXpress September 2007
staggs2823.indd 45
45
7/25/2007 4:07:08 PM
tubes
By Charles Hansen
Excerpt from
A Brief History of
Bendix Red Bank Tubes
Author Charles Hansen offers a unique perspective of the
Bendix tube-manufacturing facility in his recent book.
. . . Bendix no longer exists. The plant at
Eatontown closed in 1999 and was torn
down in 2001. As of this writing, the
land hosts a Lowes home improvement
center.
The Bendix tube business ceased
operation only about 45 years ago, but
detailed information is extremely difficult to find. It was mainly through the
remaining retired employees, the local
Bendix retired employees associations,
and old Eatontown documents that I
could piece together this puzzle. Some
of it is anecdotal and may not be entirely
accurate. Keep in mind that by the time
I started at Bendix in 1966, there was
hardly even a mention of the tube business at Eatontown. Here we are in the
46
audioXpress 9/07
Hansen excerpt.indd 46
midst of the greatest information age
of all time, and we have difficulty keeping track of events that happened barely
half a century ago.
MILITARY VACUUM TUBE HISTORY
When the US armed services first
started using vacuum tubes just before
WWI, they devised a method for tube
identification that would meet the needs
of the individual services. In 1916, the
Army started assigning a series of “VT-”
numbers to commercial off-the-shelf
(COTS) equivalents of standard production tubes. The Navy system used
“C” designations for each manufacturer:
CG was GE, CRC was RCA, CW was
Westinghouse, and so on. The Associ-
ated Radio Manufacturers (ARM) was
formed in 1924, and was renamed Radio
Manufacturers Association (RMA) that
same year. They endeavored to standardize the peacetime radio industry.
As the US was dragged into WWII,
vital materials were required for the war
effort. A lot of firms in early 1941 began
seeking war contracts and even started to
manufacture some wartime items. The
US manufacturing industry was converted by Government order to full war
production in early 1942.
The different vacuum tube numbering systems used by the Navy and Army
(and the Army Air Forces) was a chaotic
mess, with a huge number of obsolete
tube types. The War Department instituted a Joint Army-Navy ( JAN) part
numbering system that, for the most
part, used standard industry tube numbers.
The extreme shock and vibration environments required that some of these
tubes be improved mechanically (ruggedization) to ensure the operability of
critical electronic equipment. The suffix
“W” was added to the tube designation
for identification of rugged tubes, i.e., a
ruggedized 5Y3GT became a 5Y3WGT.
A short time later, the numbering system was modified to use the four-digit
RMA/EIA (Radio Manufacturers Association/Electronic Industries Association) numbering system. The 5Y3
became the 6106.
As a result of the rapid industrial conversion toward the war effort, the government found that quality varied widely
across what were previously industrial
products. After the war, the military de-
www.audioXpress .com
7/25/2007 4:01:40 PM
cided it needed better control over the
quality of all materials and components
that went into critical military equipment. It instituted a series of Military
Specifications (Mil Specs) that defined
the parameters, performance, test and
quality requirements for military items.
All military contracts required that
the contractor have in place a Quality
Assurance program. It also prohibited
the re-marking of tubes or any other attempts at third-party manufacturing. All
military tubes had to be manufactured,
tested, and qualified at the same facility.
The very first Mil Spec was for electron tubes, called MIL-E-1. It consisted
of 77 pages of specifications, tests, and
quality requirements for electron tubes.
Each unique tube type was given a
MIL-E-1 “slash number,” and this same
specification is still in use today.
While the identity of the MIL-E-1/1
tube type is lost in obscurity, MIL-E1/2 was issued for the 6SK7WA remote
cutoff pentode. Other electronic components covered by that first group of Mil
Specs were:
MIL-C-5
capacitor, fixed, mica dielectric;
type CM
MIL-R-11 resistor, fixed, composition
(insulated); type RC
MIL-R-19 resistor, variable, wirewound
(low temperature); type RA
MIL-C-20 capacitor, fixed, ceramic, temperature compensated; type CC
MIL-R-22 resistor, variable, wirewound
(power type); type RP
MIL-C-25 capacitor, fixed, paper dielectric,
DC; type CH
MIL-R-26 resistor, fixed, wirewound (power
type); type RW
MIL-T-27 transformers and inductors
(audio, power and pulse)
MIL-C-62 capacitor, fixed, electrolytic (dry
electrolytic); type CE
MIL-W-76 wire, hookup, electrical, insulated
MIL-C-81 capacitor, variable, ceramic
dielectric; type CV
MIL-C-92 capacitor, variable, air dielectric
(trimmer); type CA
MIL-R-93 resistor, fixed, wirewound (accurate); type RB
MIL-R-94 resistor, variable, composition;
type RV
Interestingly, the early germanium
and silicon diodes were also listed in
����������
��������������
������������������ �������������
�����������������������������������������
����������������������������������������
���������
�� ���������������������������������������������
� ������������
�� ����������������������������������������������
�� ������������������
�������������������������
����
�������������������
������
�
����������������������������
��������������������
�����������������
��������������
������
�
������������������������� � �������
����������������������� � �������
����������������������� � ������
����������������
�����������
�������������������������������������������
�������������
�
���������������������������
�
����������������������������������
�����������������������������������
audioXpress September 2007
Hansen excerpt.indd 47
47
7/25/2007 4:01:47 PM
MIL-E-1, starting with MIL-E-1C
(“electron tubes and crystal rectifiers”) in
1955. The first Mil-Spec diode I could
find was the JAN-1N253, which was officially designated MIL-E-1/1024 with
the oxymoron title of “Electron Tube,
Silicon Power Rectifier” in 1956. Once
solid-state diodes and transistors began
to supplant vacuum tubes in military
designs, the Department of Defense
(DOD) recognized the need to produce a Mil Spec that was dedicated to
semiconductors. The first device in this
new Mil Spec was the 2N220 germanium PNP transistor, designated MILS-19500/1 June 14, 1957.
On Sept. 12, 1961, the 1N253 was
made inactive on MIL-E-1 and moved
to MIL-S-19500/194, soon followed by
all the other semiconductor diodes in
MIL-E-1. The title for MIL-E-1 then
reverted to just “electron tubes.” Over
1500 unique tube types were eventually
listed in MIL-E-1. As far as I can determine, the last number used was MIL-E1/1772, a ku-band magnetron.
ECLIPSE-PIONEER AND
BENDIX RADIO HISTORY
The Fleming “valve” and DeForest “audion” were just beginning to have an impact in radio in 1909. About this time,
the Bijur Motor Appliance Company
of New York City formulated the idea
of cranking a car engine with a batterypowered electric motor. In 1910, Vincent Bendix patented what later became
48
audioXpress 9/07
Hansen excerpt.indd 48
known as the “Bendix drive” for electric
starters. Instead of having the starter
motor continuously engaged like the
starter Delco designed for the 1911 Cadillac, it used a solenoid and sliding gear
to engage teeth around the outside of
the flywheel only during engine start. It
did not require the Delco reduction gear.
The Bendix starter soon became standard on all cars produced in the US.
Bijur went on to design the electric
starter used on the famous Liberty aircraft engine used in WWI.
In 1923, Vincent Bendix started the
Bendix Brake Company in South Bend,
Ind., to market the Perrot four-wheel
braking system. The Eclipse Machine
Company of Elmira, N.Y., acquired the
Bijur Company, also in 1923. Bendix
went public in 1924 and issued shares to
finance the brake business. In 1928 Bendix, with $6.8 million in notes granted
by General Motors, acquired control of
Eclipse, which had been producing the
Bendix automotive starter since 1914.
Three former employees of Sperry
Gyroscope had formed the Pioneer
Instrument Corporation in Brooklyn,
N.Y., in 1919. Bendix bought the Pioneer Instrument Corp. in 1929, and introduced the earth induction compass
that Charles Lindbergh used for his epic
flight across the Atlantic.
In 1929, a subsidiary of Eclipse, in
East Orange, N.J., began building starters and generators for aircraft. Vincent
Bendix changed the name of his com-
pany to Bendix Aviation, purchased
Scintilla Magnetos, originally a Swiss
firm; and set up manufacturing facilities in Sidney, N.Y. Then he bought
the manufacturing rights to Stromberg
carburetors.
At the start of the Depression in 1929,
Bendix developed power brakes and
power steering for cars and trucks. By
1937 he had successfully developed the
pressure carburetor for aircraft engines.
Bendix built a large plant in Teterboro,
N.J., across from Teterboro Airport, in
1938, moving Eclipse Aviation and Pioneer Instruments into the new building.
Bendix built a casting foundry and extensive engineering and test facilities.
Vincent Bendix acquired an interest in the Radio Research Company in
Washington, D.C., in 1936. They licensed their home radio designs from
RCA and Hazeltine Industries (a label
on the radios listed the licensing notices). In 1937 all radio operations were
consolidated in Baltimore, Md.
When the US entered WWII in
1941, another plant was opened in nearby Towson, Md. Bendix Radio made aircraft communications sets as well as the
first automatic direction finder and radio
detection systems. By 1938, Bendix had
a complete line of navigation and communications systems.
In 1942, Vincent Bendix resigned as
president and chairman of the board of
his company (he died in 1945). Meanwhile, back in Teterboro, on May 4,
www.audioXpress .com
7/25/2007 4:01:48 PM
1943, Eclipse and Pioneer were combined into the Eclipse-Pioneer Division
of Bendix Aviation Corp.
TALOS MISSILE PROGRAM
Like many companies of the era, Bendix’s production philosophy determined
that it be vertically integrated; that is,
they kept as much of their production
capability in-house as possible. What we
today call “outsourcing” was a dirty word,
because you were taking work out of the
factory by buying outside material and
services. Each Bendix division was also
fully autonomous. The sales and service
staffs were located in the same plant
where the engineering and manufacturing took place. This philosophy propelled the Eclipse-Pioneer division of
Bendix into the vacuum tube business.
The trigger event was the Talos missile program. The Talos (along with Terrier and Tartar, the “3-T” missile systems) Naval missile program was an outgrowth of the Naval Ordnance Bureau’s
Bumblebee program, which sought to
provide medium- and long-range surface-to-air (SAM) missile protection for
the Navy’s surface ships in the face of
increasingly faster jet aircraft. Jet aircraft
were essentially immune to conventionally aimed antiaircraft artillery (AAA).
Bumblebee was initiated at Johns Hopkins University’s Applied Physics Lab
( JHU-APL) in 1945.
Bell Labs had begun work on radar
systems in 1937 at the request of the
Navy. At this time radar waves were
generated with conventional power triodes and were limited to relatively low
frequencies. That same year the Varian
Brothers in the US designed the klystron, and the British developed the reflex
klystron in 1939. One year later British
physicists invented the magnetron.
In August of 1940 the Battle of Britain began. The Germans destroyed the
town of Coventry in the fall of 1940,
where virtually all British aircraft radios were made. Winston Churchill
dispatched a team to the US to find a
company that could build radios for the
British. Bendix was the only company
that had any experience with VHF (very
high frequency) radios, and, as a result,
the company’s production suddenly in-
creased to 20 times its pre-war level,
using RCA versions of the British tubes.
The British also needed to find a
safer location to continue developing
the “Home Chain” ground radar systems
being used to detect German bombers, and smaller high-frequency radars
for use aboard fighter interceptors. Bell
Labs and the MIT Radiation Lab in the
US took on the development of these
high-power devices. Thus, by the end
of WWII, Western Electric had gained
extensive experience in high power radio
and radar transmitting tubes.
Near the end of WWII, Western
Electric was awarded the Army Nike
long-range antiaircraft missile defense program, and developed both the
ground-based tracking and targeting
radar and the Nike missile guidance system. Convair was awarded the Terrier
and Tartar missile programs. APL/Philco and Motorola were assigned the guidance system design tasks, and Bendix
Aviation won the Talos project with the
Naval Ordnance Test Center (NOTC).
The Talos, along with the Nike and
Terrier, were to be the first guided rock-
audioXpress September 2007
Hansen excerpt.indd 49
49
7/25/2007 4:01:49 PM
ets. The earlier rockets were all ballistic types with no guidance system. The
“WAC” Corporal (designed after the
supply of captured WWII German V-2s
ran out) stood for “Without Any Control.”
The Talos program presented a bit
of a logistics problem because Bendix
was neither in the missile nor the rocket
engine business. After the war, Bendix
Radio in Towson turned to home radios
and TV sets, and then automobile radios. The military navigation and communications product lines were moved
to Eclipse-Pioneer in Teterboro. About
this time, Eclipse-Pioneer developed the
first electric starter for jet engines.
JHU-APL had developed a ram jet
engine that eventually became the final
propulsion stage of the Talos. Bendix
acted as the prime contractor. The missile airframe and ram jet engine were
built by McDonnell, the solid rocket
first stage was developed by Allegheny
Ballistics Laboratory, and the automatic missile loading and launcher system
program went to GE. The Guidance
Systems group at Bendix in Teterboro
teamed with Sperry Gyroscope to develop the missile guidance system (the
shipboard tracking and guidance radar
was probably a version of the Western
Electric system). R&D work on the
Talos program began in earnest in 1948.
The Talos team built a supersonic test
vehicle (STV) to evaluate the guidance
system at high speeds, but the high acceleration forces developed by the solid
rocket booster were destroying even the
ruggedized “W” tubes that were available at the time from GE, Tung-Sol,
and Sylvania. All the previous guided
“missiles” were basically unmanned subsonic winged aircraft—high g forces
were never imposed on the guidance
system vacuum tubes.
During WWII, British and American
scientists had co-developed the top-secret variable-timing (VT) radio proximity fuse1 used in 75mm and larger
antiaircraft shells. The five subminiature
metal tubes developed for the fusing circuits could tolerate 20,000 axial g’s and
a 30,000 rpm spin rate, and then trigger
an explosion within 30′ of the target.
But these tiny devices did not have the
���������
������
��������������������������������������������������
��������������������������������������������������
������������������������������������������������
�
��������������
��������������
���
��
��
��
�
��
���
���
��
���
�������
�������
���������������
��������������
�
��
�
��
��
��
���
��
��
�����
����
��������������������������������������������������
�����������������
�����������������
����������
��
�
��������������
��
���
��
��
��
��
���
��
��
���������
����������
��
��
���
��
��������������
�����
�������������������
����������������
�
��
��
��
��
��
�����������
����
��������������������������������������������������
���������������
����������������
�
��
��
��
��
���
�������������
����
����������������
�����������������
��
��
��
��
��
��
����
�������������
���������������
����������������
�
��
�
���
����
�����������������
������������������������������������������������
�����������
���
��
���
��
�����������
�������
����
���
��
��
��
��
���
�
��
����������
��������
�����
��
�������
��
��
��
�
��
��
��������������� �������������
����������
�
��������������������������������������������������
��
��
��
��
��
������������������
�����������������
��������
��
��
��
��
��
��
��
������
������������������������������������������������
����������������������������������������������������������
�������������
��������������������������������
�
��������������������������������������������������
50
audioXpress 9/07
Hansen excerpt.indd 50
power needed to direct the course of a
3.5 ton, mach 2.5 missile fired against a
fast, maneuverable target.
Bendix had no choice (because of its
vertical integration philosophy) but to
develop expertise in the vacuum tube
business, and do it fast! The EclipsePioneer group at Teterboro was tasked
to develop the vacuum tubes required
for the Talos program. Bendix received
vacuum tube production code ID number 125 from the Electronics Industries
(EIA) in 1948.
The task was certainly easier than the
proximity fuse program, whose tubes
needed to withstand much higher gforces along with the high spin rate imparted by the rifled barrels of the artillery.
The guidance system selected for
these early missiles is known as a guidance radar beam rider. The Talos was to
be launched into a capture beam that
steered it to the guidance beam. Using
conical radar scan techniques, the Talos
needed to keep itself inside the guidance
beam all the way to the target.
This type of external control guidance
required only one radar that tracked
both the target and the missile, and sent
steering commands to the Talos that directed it into a collision course with the
target. The system minimized the electronics that the missile carried, but accuracy tended to decrease with range. The
tubes in the early test missiles, which
must operate for only a few minutes in
flight, would tend to fail at launch or
soon after.
TUBE DESIGN ENHANCEMENTS
Bendix Aviation and Eclipse-Pioneer
threw all its expertise into the tube durability problem. It used its best metallurgists, chemists, electrical and materials engineers, along with whatever tube
experts it could hire (especially from the
nearby RCA and Tung-Sol factories).
Tube designer John Wyman headed up
the division.
Rather than the Bakelite or phenolic used in commercial octal tube bases,
they first tried a material called Micanol,
a mica-filled phenolic. After further experiments with some new materials, they
settled on a tough brown mineral-filled
melamine formaldehyde resin. For further insurance, they filled the base with
www.audioXpress .com
7/25/2007 4:01:55 PM
a polyester potting compound while the
tube envelope was connected to the base.
(Eatontown developed a proprietary
high-temperature silica-filled brown
epoxy called RB-159 in 1962 to encapsulate generator and transformer windings, but it came along too late to ever
be used for octal tube bases.)
They used Corning Nonex2 envelope
glass, with higher amounts of silica and
boron oxide than the soda lime glasses
used in commercial tubes. This borosilicate glass withstood much higher temperatures, and is known as a “hard” glass.
This necessitated the use of platinum or
tungsten lead-in wires, and tungsten or
kovar alloy pins on the miniature tubes,
because Nonex had half the expansion
rate of “soft” commercial glass. Commercial glass tubes used either dumet alloy or
nickel for the stem wires and pins.
A rugged support system for the tube
elements was absolutely critical in order
to withstand the high g-forces imposed
on the missile guidance system tubes.
Standard mica was inert and could tolerate the high internal envelope temperatures, but it was too weak for guided
missile applications. The Eclipse-Pioneer engineers turned to ceramic materials. One readily available material was
alumina (aluminum oxide).
Champion Spark Plug had developed
the aluminum oxide insulator in the
1930s for aircraft spark plugs. Drawings were made to procure press-molded
alumina equivalents of the mica wafers
then in use. Metal eyelets were added
wherever the nickel support rods passed
through the alumina, to lock the entire
support structure together and maintain the critical inter-electrode spacings
(these metal eyelets were also fitted to
Bendix tubes with mica spacers). There
were some early problems with the different expansion rates between alumina
and the metal tube elements.
The coiled tungsten heater was
threaded through an extruded alumina
heater insulator block, and this heaterinsulator assembly was then inserted
into the cathode. Covering the legs of
the heater wire below the alumina insulator block with nickel sleeves provided
a strong swaged weld, and heavy heater
bus wires provided firm anchoring to
the base.
In addition to the inherent ruggedness
of this structure, it provided two other
advantages. The heater-cathode breakdown voltage was much higher than
that allowed by the standard aluminacoated heater. And the heatsinking capability of the alumina insulator helped
prevent local hot spots in the heater that
might shorten tube life during the turnon surge, but this also lengthened the
warm-up time.
The cathodes were larger than commercial tubes of the same types to increase the cathode area and accommodate the alumina heater blocks. Two
large cathode tabs were used on each
alumina spacer to hold the cathode in
place despite the high vibration and
shock.
The getters were way overdesigned.
The thinking seemed to be that if some
getter surface area is good, more is better,
and too much is just right! Some tubes
included as many as four getter loops,
one on each of the support posts. An
extra ceramic spacer was used at the top
of the tube as a getter shield to ensure
AROUND the DCX-2496...
A DCX-2496 is THE affordable high performance audio DAC coupled with a powerful loudspeaker
management processor. It is the ultimate tool for those seeking the audio perfection.
To improve on the audio performance of the DCX, Selectronic offers a range of high end
kits for the DIYer. Very easy to assemble (no SMD) and minimally invasive on your DCX.
The results are outstanding !
And MORE...
• 6-channel R/C volume control
I/O board ��
• Upgrade your DCX-2496 with this
I/O board. Completely transparent,
just Plug & Play. Uses top of the line
circuitry.
�� Precision analog power supply
• Decrease noise and improve stability
with this supply designed for DSP circuits.
• Highly transparent Technology uses pure "Class A" I/O audio
J-FET buffers and special ALPS motorized pot • ±3dB level
adjustment on each channel • Infrared R/C • Zin : 47kΩ
• Zout : 600Ω • Easy assembly • Compatible with any DSP or
5.1 system
• Sequential mains switch
Ultra-low jitter clock ��
• This clock circuit uses a high performance TCXO with
less than 10ps jitter resulting in extremely detailed sound.
by
Outstanding kits for the DIYer
More information : www.dcx2496.fr / ww.selectronic.fr
Pub AUDIOXPRESS -
Hansen excerpt.indd 51
• Switch your audio installation on or off without noise,
thumps, or spikes • Turns on the various parts of your system
in a safe order • Avoids dangerous transients in your speakers
Selectronic - File AXP 01-2007 - Mechanicals : 181 x 120 mm
audioXpress September 2007
51
7/25/2007 4:02:01 PM
that the flashed metallic vapor did not
splash against the stem or the insulated
lead-in wires. It was also anchored in
place with metal eyelets.
Various exothermic and nitrogendoped alkaline and rare-earth metals
and compounds were tried in order to
remove all traces of nitrogen, hydrogen, and especially oxygen following
initial pump-out, as well as throughout
the life of the tube. The chemists experimented with barium, barium-aluminum, barium-magnesium, cesium
chromate, calcium, hafnium, magnesium, rubidium chromate, strontium,
zirconium, zirconium-iron, and mischmetall (an alloy of cerium, lanthanum, and other rare-earth elements).
While some of the more exotic getter
materials were probably used in microwave, high power, and gas tubes, the
electron tubes (TE) used conventional
barium flash getters.
The grid lathes were modified to
use smaller diameter wire, with a finer
winding pitch than standard tubes. The
copper grid side rods were also heavier
than usual. Bendix engineers developed
a winder that automatically swaged the
rods so the grid wire was embedded
into them. Heavy radiators connected the grids to their through-supports
with multiple welds. Extra spacing was
added between the cathode and control
grid to prevent voltage breakdown.
Using custom tooling developed at
Eclipse-Pioneer, workers assembled
the tube elements, supports, and other
parts inside a set of alumina (or sometimes mica) spacers. The entire assembly was then crimped and spot-welded
to the tube base. The glass envelope for
each particular tube (with an integral
exhaust tube) was installed over the
assembly and sealed to the base disk
using a natural gas flame.
Next the tube/envelope assembly was
sent from the clean assembly area to
Sealex rotary workstations. A vacuum
line was connected to a glass exhaust
tube at the top (miniature tubes) or
bottom (octals) of the tube envelope.
The air was pumped out with a multistage vacuum pump, and the metal
parts of the assembly were heated with
an induction coil. The pump removed
most of the gases
inside the envelope and the heat
activated the cathode coating.
Next, the exhaust tube was
heated in another
gas flame and automatically sealed
(tipped-off ). If
you look at the
Bendix miniature
tube in Photo 1
PHOTO 1: 6754
(a 6754 TE-36
TE-36 miniature
with the HY-Gtube tip-off.
300 marking and
gold-plated pins), you can see that the
tip-off is larger than commercial tubes,
because of the heavier hard-glass envelopes.
Next, the Sealex turntable rotated
the tube to the getter-flash RF induction coils, where the barium getters
were heated white hot to absorb the
last vestiges of oxygen, nitrogen, and
hydrogen. After the Sealex machine,
the tubes went to inspection and test.
�
����������������
�
�
������� ���������� ����
��������������������
�������������������
������������
Produces six selectable spectrum shape pulse
signals to help you identify problems with your
speakers and their interactions with your listening room.
Quickly hear your speaker’s degree of image
focus, transient precision, and tonal neutrality.
52
To find out more about The Sound Strobe™contact customer service by
calling 888-924-9465 or e-mail custserv@audioXpress.com.
The Sound Strobe™ was designed by Dennis Colin, currently working as
an Analog Circuit Design Consultant for microwave radios and a frequent
contributor to audioXpress magazine.
��������������������������������������
A great product for DIY speaker builders who
want to tweak their designs and a must-have for
anyone installing loudspeakers professionally.
������ ����������������������� ������������ ����������������������
�����
�������������������������������������� ����������������������
������� �������������������������������� ����������������������
�������������������������������
�������������������
Old Colony Sound Laboratory, PO Box 876, Peterborough NH 03458-0876 USA
Toll-free: 888-924-9465 Phone: 603-924-9464 Fax: 603-924-9467
E-mail: custserv@audioXpress.com www.audioXpress.com
audioXpress 9/07
Hansen excerpt.indd 52
www.audioXpress .com
7/25/2007 4:02:04 PM
Missiles were always on hot standby
(powered up), so, to ensure there were
no infant mortalities (early failures),
Bendix gave each tube a 45-hour runin under various overload, vibration,
and shock conditions to eliminate any
tubes with defects that might lead to
failure under operational use. Vacuum
tubes and gas regulator tubes were also
used in some of the Eclipse-Pioneer
generator voltage regulators and supervisory panels (the black boxes with
the generating system control and protection circuitry), but I was unable to
find out whether those tubes were ever
made in Teterboro. aX
ABOUT THE AUTHOR
Charles Hansen is an Engineering Consultant and holds five patents in his field
of electrical engineering. He began building vacuum tube audio equipment in high
school. He plays jazz guitar and enjoys
modifying guitar amplif iers and effects
to reduce noise and distortion, as well as
building and restoring audio test equipment. He has another book to his credit,
The Joy of Audio Electronics, and over
200 magazine articles on electronics and
audio. He joined the Bendix Red Bank
Division in 1966, and retired in 1998
as Supervisor of the Systems and Controls
engineering group.
REFERENCES
1. “Deadly Accuracy,” Inventions and
Technology, Spring 2001, D. Colley.
2. Nonex was a slightly different borosilicate formulation than the Corning
Pyrex or Kimble Kimax glasses that
were used for laboratory test tubes and
flasks.
A Brief History of Bendix Red Bank
Tubes is available for $24.95 from Old
Colony Sound Lab (PO Box 876, Peterborough, NH 03458, 888-924-9465,
custserv@audioXpress.com). The book,
which contain over 80 pages, clearly
details the role this N.J.-based company played in the history of tube manufacturing. Includes photos of the Red
Bank tubes manufactured, a list of key
Bendix personnel, and dozens of pages
of valuable datasheet information about
these tubes.
audioXpress September 2007
Hansen excerpt.indd 53
53
7/25/2007 4:02:08 PM
Classified
What’s New
on the aX website?
Articles from past issues:
“Sources 101: Audio Current Regulator Tests
for High Performance, Part 1: Basics of
Operation,” By Walt Jung (aX 4/07).
“Sources 101: Audio Current Regulator Tests
for High Performance, Part 2: Precise High
Current/Voltage Operation,” By Walt Jung
(aX 5/07).
“We Visit Mundorf,” By Jan Didden (aX
5/07).
“Testing Amp Peak Power at the RMAF,” By
Peter J. Smith (aX 6/07).
“Radio Shack Sound Level Meter
Characteristics,” By Daisuke Koya (aX 6/07).
“An A/B/I Switch,” By Dennis Colin (aX
6/07).
Charles Hansen’s review of the Monarch
Audio M24 Dac + Tube Line Amp (aX 6/07).
Ed Simon’s review of The Sound Strobe (aX
4/07).
Tom Perazella’s review of Revue Du Son Test
CD Number 17 (aX 5/07).
VENDORS
WANTED
High Performance kits: Phono, DAC, Buffers,
Line amps, Headphone amps, Power amps,
Home Theater amps, Crossovers, Power
supplies, Regulators
Custom designs for OEM customers
www.borbelyaudio.com
Selected BORBELY AUDIO kits in Japan:
http://homepage3.nifty.com/sk-audio/
stereo gear old/new/quality, speakers, amps,
preamps, turntables, reel to reels, vacuum
tubes, tube testers, guitars, parts, etc. 850314-0321 Email: sonnysound@aol.com
Our phono preamps are popular for copying
45s, 78s, LPs. One customer recently said: “I
can’t believe it’s so quiet!” www.tdl-tech.com
(505-382-3173)
AudioClassics.com Buys - Sells - Trades Repairs - Appraises McIntosh & other High End
and Vintage Audio Equipment 800-321-2834
ezChassis® easily the best DIY amplifier
chassis. Pre-punched holes and slots.
www.designbuildlisten.com
Charles Hansen’s review of the Tube Imp
Mini Tester (aX 5/07).
Yard Sale
Gary Galo’s review of Three New
Prototyping Boards (aX 6/07).
For Sale
“MATAA: A Free Computer-Based Audio
Analysis System,” By Matthias S. Brennwald
(aX 7/07).
Six Longines Symphonette albums (Music
of the 20s, 30s, Blues, International
Dance, Movie Musicals). Eleven Time-Life
albums—accurate big band re-creations
of 1930s into 1950s. Three stereo LPs in
each album, 51 total. Descriptive literature
included. $200 for the lot, plus shipping.
Approximately 50 lbs.
John Markwalter
1661 Britannia Blvd.
Port Charlotte, FL 33980
(941) 764-8713
“High End 120W MOSFET IC Driven Amp,”
By Jack Walton (aX 7/07).
“The Gamp: Adaptable Power for Tube
Amps,” By Paul J. Stamler (aX 7/07).
For information on these and others, visit
www.audioxpress.com. And don’t forget to
check out the links to our other magazines,
Voice Coil and Multi Media Manufacturer!
Complete set of Speaker Builder 19802000. rumreich@gmail.com
Eton 11-581 11˝ woofer
Co-Drive 12˝ subwoofer
Eton ¾˝ tweeter
Seas W22EX 8˝ woofer
Raven R2 tweeter with passive crossover
Marchand XM26 two-way tube crossover
(4) 10˝ × 72˝ electrostatic panels with
two interfaces
two Plitron 50:1 transformers
Snell EC200 electronic crossover 60dB hi/lo
pr. Rethm 2, one broken.
(609)397-8315. jrmq@verizon.net
“Yard Sale” is published in each issue of
aX. For guidelines on how subscribers
can publish their free ad, see our website.
54
audioXpress 9/07
classy-adindex-yrds.indd 54
www.audioXpress .com
7/25/2007 3:57:50 PM
Ad Index
ADVERTISER
Full-range drivers made in the USA by people
passionate about sound reproduction
PAGE
ACO Pacific Inc ...............................................41
Antique Radio Classified.............................. 53
Audience ..........................................................58
Audio Amateur Corp
audioXpress Subscription .......................54
Old Colony Sound Lab Sound Strobe .......................................... 52
Audio Transformers....................................... 61
Audiomatica SRL .......................................... 59
Avel Lindberg..................................................58
Bandor Miniature Loudspeakers ............... 31
Common Sense Audio .................................33
DC Gold Audio ............................................... 55
Eminence Speaker, LLC ...............................44
Flat Earth Audio .............................................35
Fountek Electronics Co.,Ltd. ..................... 42
Front Panel Express, LLC .............................40
Hammond Manufacturing............................. 3
Jantzen Audio Denmark ......................... 11,23
Juicy Music Audio......................................... 16
KAB Electro-Acustics ....................................60
Liberty Instruments ....................................... 27
Linear Integrated Systems .......................... 57
Madisound Speakers .................................... 37
Midgard Audio AS .......................................... 5
MISCO/Minneapolis Speaker Co.,Inc. .... 43
Monarchy Audio ............................................49
Mouser Electronics ....................................... 55
Mundorf EB GmbH .................................. 9, 25
New Sensor Corporation .......................... 7,25
Orca Design & Manufacturing ................... 17
Parts Connexion............................................. 15
Parts Express Int’l., Inc. ......................29,CV4
Penn Elcom, Inc. ........................................... 24
RAAL Advanced Loudspeakers .................39
Rocky Mountain AudioFest – RMAF ........ 47
SB Acoustics ................................................... 20
Selectronic .......................................................51
Sencore ............................................................ 13
Smith & Larsen Audio ..................................40
Solen, Inc. ........................................................ 24
Tang Band ....................................................... 19
Tektron-Italia .................................................... 62
Test Equipment Depot .................................. 53
Usher Audio .................................................CV2
Vacuum Tubes, Inc. .....................................48
Vidsonix Design Works ................................60
WBT-USA/ Kimber Kable ................. 22, CV3
CLASSIFIEDS
All Electronics .................................................54
Audio Classics Ltd. ......................................54
Billington Export, Ltd. .................................54
Borbely Audio .................................................54
Design, Build, Listen, Ltd. ..........................54
Sonny Goldson ...............................................54
TDL Technology .............................................54
We are offering a 10% discount (off MSRP) to AudioXpress readers. Please
call (661) 618-0221 between 9am and 6pm Monday through Friday
Available drivers:
Model /
Imp
Size
Power
N5R
4 or 8 Ohm
5 1/4”
80 W cont.
N69R
4 or 8 Ohm
N9.5R
4 or 8 Ohm
Efficiency
1W / 1m
Freq
+/- 3dB
MSRP
Pair
92
50Hz - 20kHz
$306.01
6 X 9”
80 W cont.
94
35Hz - 20kHz
$337.15
9.5”
80 W cont.
96
30Hz - 19kHz
$359.33
N7R
7”
Coming August 2007
$337.15
N7S
7”
Coming September 2007
$540.65
Easy Ordering In Nanoseconds
• The ONLY New Catalog Every 90 Days
• NEWEST Products & Technologies
• Over 845,000 Products Online
• More Than 330 Manufacturers
• No Minimum Order
• Fast Delivery, Same-day Shipping
mouser.com
(800) 346-6873
The Newest Products
for Your Newest Designs
The NEWEST Semiconductors | Passives | Interconnects | Power | Electromechanical | Test, Tools & Supplies
Mouser, Mouser Electronics, and Mouser Electronics Pte. Ltd. are registered trademarks of Mouser Electronics, Inc. Other products, logos, and company
names mentioned herein, may be trademarks of their respective owners.
audioXpress September 2007
Mouser AudioXpress 9-1-07 indd 1
classy-adindex-yrds.indd 55
55
7/12/07 10:39:12 AM
7/25/2007 3:57:58 PM
XPRESSMail
SOURCES 101 TESTS
With regard to Walt Jung’s recent article
(“Sources 101: Audio Current Regulator
Tests for High Performance,” April ’07
aX, p. 10), I am curious how he managed to get a one Vbe current source to
perform worse than a two Vbe current
source. My hand calculations and SPICE
simulation show the one Vbe source to be
5 to 6dB better than the two Vbe source.
In addition to having better power-supply
rejection, the one Vbe source has higher
output impedance for signals originating
at the collector of the output transistor.
The one Vbe source also lends itself well
to simple modifications that can put both
its power-supply rejection and output
impedance below the noise threshold of
Mr. Jung’s test.
It is important to note that while
power-supply rejection may indicate high
output impedance, there is no guarantee
that it is so. To fully characterize something as simple as a current source would
require much more detailed testing. Some
of the later examples in part two of the
article (May ’07), specifically Fig. 13C,
look good for power-supply rejection because the source of M1 is bypassed to
ground by stability capacitor C1. However, I believe that the output impedance
of U1 will be quite a bit lower for signals
originating in the output leg.
Thomas Bohley
Colorado Springs, Colo.
Walt Jung responds:
To first respond to paragraph one of this
letter, it should be noted that others have
already pointed out the poor-performance
discrepancy of the Fig. 3A circuit. Chris
Paul had first noted this to me in an e-mail;
namely, that the “One Vbe” circuit has a
theoretical advantage, vis-à-vis the “Two
Vbe” type of Fig. 4A. We have since had
several back-and-forth exchanges on this
topic, aimed ultimately toward some sort
of clarifying piece. The request from audioXpress for a published response to this letter
warrants a reply, even if more information
comes later. There is simply no way all of the
pertinent technical points can be adequately
addressed in a short form.
The essence of Chris’ point was that the
56
audioXpress 9/07
xpressmail907-2.indd 56
basic reference impedance of Fig. 3A should
be lower than that of Fig. 4A, by a factor of
two. Thus it should theoretically and practically be better than Fig. 4A, as is also noted
by Mr. Bohley. And, a SPICE simulation does
support this point using the values of the respective figures—the Fig. 3A circuit is about
a factor of two better than that of Fig. 4A.
For SPICE, that is.
Unfortunately, lab measurements don’t
show this advantage for the Fig. 3A circuit,
vis-à-vis the Fig. 4A circuit, at least not using
the example values. When recently repeated,
the lab data for the “One Vbe” current source
were found as published. But, it should be
noted that higher values used for R1 might
yield better results (by operating Q2 at less
current). When R1 is set to 100kΩ, about
10dB better results were noted, as illustrated
in the plots following below.
So, it is thus suggested that users of the
Fig. 3A circuit might wish to operate Q2 at
lower current relative to Q1, for two reasons.
One is potential improvements in rejection,
as noted. A second is for stability reasons, as
originally noted.
So, my brief answer to Mr. Bohley about
how I managed to get worse results for the
Fig. 3A circuit is that I just plugged in the
specified values, and I observed the cited results. Both originally, and also more recently.
I agree that these results aren’t spectacular,
so I may have been bitten by a poor example. Mea culpa. Raising R1 does help the
performance of the circuit as originally published. But, there is much more to be said on
this circuit.
I received a further e-mail exchange on
this part of the article from John Popelish,
with a suggestion of a performance enhancement. I note from the above that while Mr.
Bohley alludes to improvement modifications
to this circuit, he offered no specific information. But John Popelish did, as follows:
“I am wondering if you have tested the
simple enhancement to the two-transistor
source shown in Fig. 3A. It is based on adding a second driving resistor, with a current
approximately proportional to the total supply, but connected to cancel most of the
effect of supply voltage variation, over some
small range.
“For instance, if you reduce the Rset resistor from 332Ω to 316Ω (to compensate
for the small current reduction this change
causes) and connect 174kΩ between ground
and the bottom of Rset, you get much higher output impedance over the whole audio
spectrum, but especially at the low end.
“Adding this compensating resistor, R3,
makes the circuit a little less general, because its value must be optimized for each
application, depending on the expected
range of the supply. But if well chosen, the
improvement in supply rejection can be impressive—about 60dB improvement at low
frequencies. I am pretty confident that this
version can compete with any non-cascoded
design on which you reported.”
I tested John’s suggestion for performance enhancement both with SPICE and
in the lab. The SPICE analysis was done
with Linear Technology’s “LTSpice” package
(www.linear.com/designtools/software/switchercad.
jsp/). The schematic of the circuit is similar
to the original Fig. 3A, but with the addition
of R3 to ground, at 174kΩ, and the slight adjustment of Rset. R1 is set to 100kΩ, for the
reasons cited previously. This modified circuit
is shown in Fig. 2, as it was lab tested.
Figure 1 is a plot of a DC simulation in
LTSpice, with the supply swept from 16–20V,
while R3 is stepped through a range of values, to illustrate the nulling properties. Note
that there is a null in output change with
supply voltage, which is here centered within
a range around 18V for the 185kΩ trace.
This null range constitutes a region of very
high supply rejection, as can be seen from
the data. A value for R3 which minimizes
sensitivity for a given supply voltage can be
used as a starting point for lab tests.
The Fig. 2 lab results also illustrate relevant points about the enhancements to the
“One Vbe” circuit discussed. With R1 set to
100kΩ and R3 open, a modest improvement
is made over the originally published data of
Fig. 3B. This is about 10dB, as can be noted
on the intermediate curve.
When R3 is added and optimized with an
in-situ trim, the errors fall down to the residual noise level at low frequencies. If this trim
is done, it should consist of multiple steps.
The first would be to use a high-resolution
multi-turn film trimmer as a portion of R3, so
as to find the exact null point. This would be
using a low frequency measurement point,
while measuring the null. In this case, the
www.audioXpress .com
7/25/2007 4:22:08 PM
null observed was -142dB at 100Hz. Then,
the closest value film resistor can be used in
circuits built, i.e., 150 or 154kΩ. Note that
although this example was tested with the
Audio Precision system, an ordinary shop
audio source and a high gain AC preamplifier
could also be used to find this null point.
Note also that even if an exact equivalent value resistor isn’t available, substantial
improvement can still be obtained, vis-à-vis
R3 open. This preliminary run with SPICE
may or may not be helpful toward narrowing
down the truly optimum R3, depending upon
the specific type and vendor of transistors
and models used. But, in any event, it should
give some insight into the mechanism causing the very high supply rejection properties.
Leaving the Popelish enhancements, this
brings us to the points raised in paragraph
two of Mr. Bohley’s letter. He says: “. . . while
power-supply rejection may indicate high output impedance, there is no guarantee that it
is so.” I simply disagree with this, for most of
the intended measurement context. Are we
really on the same page here? Virtually all of
the Sources 101 tests have been specifically
aimed toward uses in power systems, as, for
example, the shunt regulator cited. This was
discussed under “Whys and Wherefores,”
and “What Tests.” I don’t believe that a valid
critique of this content should be extended
to include all of the many more general usages possible. In designing these tests, I
was aware of the limitations in testing for
fixed loads, but decided that, even with this
constraint, the information would still be very
worthwhile. After all, who could argue with
the merits of audio power supply systems
with low RFI sensitivity?
Such applications use a series-connected
current source of some impedance, Z, and
drive a shunt-connected load. This situation
is emulated in these tests with the 1Ω load
and the various circuit impedances tested,
with calibration data shown. So, I believe the
tests are valid for the conditions cited. But
there can be exceptions to this—see “intrinsically high impedance output nodes” discussion below.
All that said, I think I do understand what
Mr. Bohley is getting at as a potential weakness of some of the circuits. So, in principle I’d grant the general point that, yes,
the behavior of some of these current source
circuits can be application dependent. Many
current source circuits behave differently if
fed to medium or high impedance loads with
voltage swings present, as opposed to the
virtual short of the Sources 101 test cases.
As far as other possible tests, I did allude to this, under “Measured Noise,” so no
one should interpret these results to be a
final word on audio current source circuits.
Of course, I do agree with Mr. Bohley that
many other useful tests are possible. Perhaps
he could explore some of these points in a
future article.
Finally, on the performance of Fig. 13C.
Yes, this circuit will act differently if the Rset
output node is allowed to move in voltage
terms. This current source circuit (and many
others, I should add) can show different
behavior depending on which node is used
as the load, and the relative impedances
seen there. Few current source circuits have
completely symmetric two-terminal behavior,
but as was noted in the article, if that’s what
you need, the JFET (or MOSFET, for higher
current) types should be tried. See “Current
Source or Current Sink?”.
Other current sources useful where load
dynamic swings are required would be ones
with intrinsically high impedance output
nodes (transistor collectors, FET drains, and
so on). Examples here include the much
discussed Fig. 3A (with variants), as well as
Figs. 4A, 5A, 6A, 9C, and so on. For these
types of applications at the lower current
levels, I will admit that my Sources 101 test
methodology doesn’t necessarily show an
entire picture, as Mr. Bohley says.
Such as they are, the tests nevertheless
still give indications which circuits are useful! The tests show that they differ in basic
performance; they demonstrate cascode effectiveness, the importance of low-C, and
the deterioration with current, and so on. The
better performing ones (for example, Fig.
6A, using 2SA1016K transistors) should also
do well in amplifier signal paths, either as
an input diff pair tail current source, or as a
driver stage dynamic load.
But, the above caveats weren’t a consideration for the Fig. 13C circuit as originally
used with a shunt regulator, with the output
fixed at 12V or 21V. So, for these conditions,
the test data can be considered valid. This
circuit can also be used as either a source or
a sink, and will be featured as part of a future
shunt regulator article.
Other errata: There are references in the
text and figures to the MOSFET circuits using
the IXYS IXCP10M45S and the Supertex
DN2540, with operation “up to 450V.” To
audioXpress September 2007
xpressmail907-2.indd 57
57
7/25/2007 4:22:13 PM
clarify this point, readers should note that
only the IXCP10M45S has the 450V rating;
while the DN2540 is rated at 400V. My apologies for any confusion this may have raised.
Finally, my thanks to readers Bohley, Paul,
and Popelish for sharing their thoughts on
these articles, allowing an opportunity for
further discussions. I have particularly enjoyed interacting with John Popelish on his
enhancements to Fig. 3A. Also, John Larkin
posted comments about similar enhancements to the “Two Vbe” type of current
source on the USENET forum sci.electronics.
design (message ID 1q3013huejba8d51v9kgn9
n2spjgl96dbh@4ax.com) and also in an e-mail
to me.
It is hoped that a future “Sources” update
can address some further circuit developments along these lines.
Walt Jung’s recent articles (April and
May ’07 aX) were very informative. I only
wish there had been space to explore a
couple of additional areas.
First, a few years back Doug Self briefly touched on small signal current sources
and concluded that the differences in rail
rejection would largely be mooted by use
of a decoupling cap across the current
source. I would have been interested to
see whether the conclusion was repeatable and what benefit, if any, there might
be when applied to the other current
sources schemes described by Walt Jung.
FIGURE 1: A DC SPICE simulation of the enhanced circuit shows low errors, with an optimized null that occurs at one R3 value, here 185kΩ. Note that this null centers on a narrow
range of supply voltage, in this case 18V. Note also that R3 values too low (155kΩ) result in
a downward error slope, while values too high (221kΩ) result in an upward slope.
FIGURE 2: A lab test of the enhanced circuit (see inset) shows performance for Rset = 332Ω,
R1 = 100kΩ, and R3 open, then with R3 added at 155kΩ. There is a distinct setting for R3
that results in the lowest errors, but unfortunately one not exactly predicted from SPICE.
58
audioXpress 9/07
xpressmail907-2.indd 58
www.audioXpress .com
7/25/2007 4:22:17 PM
Second, his measurement setup used
a fixed load. Perhaps in a future article,
he could explore the performance of the
various current sources with variable or
even reactive loads. This, too, may be a
concern when selecting one approach
over another requiring a decision to trade
off some of one benefit to get more of
the other.
M. Whitney
mwhitney6@cox.net
Walt Jung responds:
First, my thanks to Mr. Whitney for his interest in the articles.
To respond to the first item on Doug
Self’s form of the “One Vbe” current source,
let me say that this is a worthy point. As it
turns out, his variation, similar to the circuit
shown in Fig. 1, really has excellent performance for line rejection. Readers of the
original “Sources 101” articles will recognize
this circuit as another variant of the “One
Vbe” circuit, which was originally published
as Fig. 3A in Part 1 of the article.
For specific details of Self’s circuit and
his overall context, I refer readers to his
Audio Power Amplifier Design Handbook,
Fourth Edition, Newnes, 2006, ISBN: 978- Rset value shown, producing an output cur0-7506-8072-1. A circuit which contains rent just under 2mA.
the current source in question can be
Note that when applying this circuit to
found as Figure 7.5 (Note: this is avail- power amplifiers operating at voltages higher
able online from http://books.elsevier.com/ than 18V, the R1 value(s)/operating point of
Q2 may need attention. For reasons cited
companions/9780750680721).
In the circuit of Fig. 1 shown here, Q1 previously, the higher values for R1 might
and Q2 are 2SA1016K transistor types, yield better results, by virtue of operating
which have a 150V rating. These transistors Q2 at less current. Here, the target is about
are not only suitable for power amplifiers 160µA. Self’s Figure 7.5 circuit operates the
in terms of this voltage rating, but, impor- transistor comparable to Q2 (his TR14) at
tantly, they also feature
better performance in
this circuit, vis-à-vis the
PN2907A general-purpose counterparts. Self’s
circuit uses MPSA56
types for Q1-Q2, which
have noticeably higher
capacitance than do the
2SA1016Ks (about a factor of 3 or more at Vcb =
10V). Lab measurements
were done on this circuit
operating at a supply of
18V, under conditions
otherwise similar to the FIGURE 1: A “One Vbe” type current source using a C1
previous tests, with the bootstrap capacitor, similar to the form used by Doug Self.
AUDIOMATICA
� IEEE-1394 audio interface
� 24-bit up to 192kHz
� Balanced input and output
� Maximum performances
� Maximum portability
� The CLIOfw FW-01 audio interface sets new hardware precision standards for the
CLIO system. The CLIO 8 system software is its perfect companion being the synthesis
of more than 15 years of excellence in electrical and acoustical measurements.
audioXpress September 2007
xpressmail907-2.indd 59
59
7/25/2007 4:22:21 PM
~2mA. For Fig. 1, the total R1 resistance is
100k. . . providing the 160µA. The equal values for R1a/R1b allow the AC bypass capacitor C1 to perform almost identically to the
values of Self’s circuit, where the cap value is
47µF and the total resistance is 20kΩ.
As can be noted from the data of Fig. 2,
the presence/absence of C1 makes a remarkable difference toward operation. Without
C1 (as in the intermediate curve), the line
rejection is about 105dB, consistent with previous performance of this same circuit using
PN2907As. But, with C1 active, the line rejection is on the order of 140dB at low frequencies, and actually challenges the test setup.
The C1 capacitance works to maintain
high effective AC impedance for R1a, similar
to the use of an active current source in place
of R1a-R1b, but using passive parts only.
It is worth noting that this technique also
works with other current sources of this type;
among these are the “Two-diode” and “LED”
variants discussed in Figs. 4A and 5A of the
original article. The key step is to split bias
resistor R1 into two equal parts, and apply
the coupling cap to the midpoint. I hope to
discuss these circuit types in a follow-up article (see final point below).
As for Mr. Whitney’s second query about
other load conditions for current source tests,
I can only hope that this has been at least in
part addressed with my reply on this same
point, within the reply to the Thomas Bohley
letter.
My thanks again to readers Whitney,
Bohley, Paul, and Popelish for sharing their
thoughts on these articles, allowing opportunity for further discussions.
I hope that a future “Sources” update can
address some further circuit developments
along these lines, and bring these many-faceted points of audio current source performance into a more complete discussion.
FIGURE 2: The addition of bootstrap capacitor C1 to the One Vbe current source
circuit of Fig. 1 source provides a substantial line rejection improvement.
60
audioXpress 9/07
xpressmail907-2.indd 60
CABINET BUILDING
I was so taken by William Eckle’s article (“Building a Center Channel for
an Altec A7-800,” April ’07, p. 24) that
I started looking for and have located a
pair of Altec A7-800s and a pair of Altec
755s, I think, type “C” 8˝ Altec drivers.
I haven’t found the 3000H tweeter or
N3000 crossover yet.
Does Mr. Eckle have any dimensional
information on how to build/manufacture the Altec “Malibu” style cabinets?
They look great!
Edward J. Rusnak, Jr.
roidua@aol.com
William Eckle responds:
I thank Mr. Rusnak for his interest in my article.
The outside dimensions of our “Malibu” clone
are: top section 24˝ wide, 12¼˝ high, and
20¾˝ deep. The bottom section is 24˝ wide,
30˝ high, and 20¾˝ deep. They are bolted
together with a 1˝ spacer covered with black
Formica (we had scraps on hand). We made
the cabinets in two pieces for easier handling
and construction.
The bottom cabinet sides, front, and back
extend beyond the cabinet bottom 1¼˝ to
hide the casters. There are holes between the
top and bottom cabinets to provide a common
interior volume. We used 1˝ MDF with an extra
½˝ MDF on the front, and covered the entire
cabinet with ¼˝ oak plywood and installed
quarter-round solid oak molding on all edges.
Two ports 4˝ in diameter and 3˝ long are
on the bottom of the back panel. The ports are
black ABS plastic sewerpipe. Good luck with
your vintage home theater, and have fun.
VERSATILE AMP
Could Mr. Still clarify his schematic for
the Versatile Line Amp (aX, May ’07, p.
21) only with no headphone output nor
12AT7 tube nor output transformers? Is
T2 the correct triad number? Mouser’s
description appears to be a single 115V
secondary transformer in my catalog. It
appears to me that in the power supply
both DC outputs will be higher than
12V and 260V. Am I missing something
here?
Thanks for the article. I look forward
to building it.
G. Max Carter
Woodland Park, Colo.
Joseph Norwood Still responds:
The 12V DC and 260V DC are the output
www.audioXpress .com
7/25/2007 4:22:22 PM
voltages obtained from the power supply.
These output voltages are correct.
The triad power transformer T2 (553N77U) is reversed; the primary is operated
as the secondary with the two 120V AC
windings connected in series. The secondary
of T2 is, of course, operated as the primary.
To limit the multimode amplifier to line
amplifier use, delete all parts associated with
power amp and headphone amp operation.
This will simplify construction of the line
amplifier.
Tube socket P-ST9-137R is no longer available. I suggest replacing with ceramic 7/8˝
diameter P-ST9-511 ($1.95) from Antique Electronic Supply. Good luck on your project.
I do not understand the function switch
S2a in Joseph Norwood Still’s versatile
line amp article. In one position it looks
as though all audio is shunted to ground
before it ever gets to the 12B4 tubes. Also,
in the other position switch S2a shorts
out all of the headphone signal, but this
may allow more cathode current to flow
for the power amp function. Maybe the
switch is somehow misdrawn.
I noticed that the 12B4 tubes are not
allowed to draw equal current. I say that
because it looks as though they are not
all sharing the same current in plate
and cathode circuits because the diagram shows some plates and cathodes go
through two and some tubes go through
one resistor in both plate and cathode circuits. I am wondering whether I’m missing something.
James Gilmore
jmgilmore3@sbcglobal.net
FIGURE 1: Corrected versatile line amp circuit.
Joseph N. Still responds:
Thank you for reading my article and the
critical review. You are correct, the drawing
I submitted to audioXpress had an error
pertaining to switch S2A. Switch S2 has
a wire corrected from the H.P. position to
the 220N capacitor (C3). This wire must be
eliminated from the circuit and all is well.
The corrected circuit is shown in Fig. 1.
The second part of your question concerned unequal value of resistance between
12B4 stages. These resistors are used as
carbon blocks to prepare oscillation. The
small values of these cathode and plate
resistors (2Ω) have no effect on currents of
the 12B4s. Thus no unequal electrical balance problems are produced.
Thank you very much for finding the wir-
audioXpress September 2007
xpressmail907-2.indd 61
61
7/25/2007 4:22:37 PM
ing error of switch S2. A critical review of
articles is what audioXpress is all about. Between the author and a discerning reader,
such as yourself, a finished DIY product is
finally obtained!
PROTOTYPING BOARD
I would like to offer some additional information on the D-4 prototype circuit board
that Gary Galo wrote about in the June
issue (“Three New Prototyping Boards,”
p. 38). This board has several features that
make it useful for building audio circuits.
Some of the more important ones are:
1. The two power supply traces run the
full length of the board, making it very
easy to supply power to multiple ICs.
2. The ground trace runs down the middle
of the board, providing a very short path
to ground for the pins of an IC. This
allows bypass capacitors to have very
short leads, which can reduce the likelihood of oscillation in some circuits.
3. The above-mentioned power and
ground traces are double width, to keep
resistance as low as possible.
4. The “5 hole pad” design makes it easy
to connect multiple passive components
to each IC pin. This is very handy for
RIAA stages, which often have multiple
resistors and capacitors connected in
parallel.
Photo 1 shows a stereo phono preamp
that I built using this board. The circuit
was published in the May 2003 issue of
audioXpress (page 62). As you can see,
both channels (four ICs) easily fit on
one D-4 board. The other circuit board
in the photo is an Old Colony PCBK13B power supply board. aX
Darren Hovsepian
DH Labs
PHOTO 1: Phono preamp application
for D-4 prototype circuit board.
Tektron-Italia
�������������
�����������������������
����������������������
����������������������
����������������������
�������������
�����������
������������������������������
��������������������
��������������������
����������������������������
��������������������������������
��������
���������������������������
���������������
��������������������
������������������������
����������������������
����������������������
�������������������������������
�����������������������������������
����������
�������������
�������������������������������������
����������������
������������������
������������������
������������������������������
���������������������������������
62
audioXpress 9/07
xpressmail907-2.indd 62
CONTRIBUTORS
Dennis Colin (“The LP797 Ultra-Low Distortion
Phono Preamp,” p. 6) has demonstrated the audibility
of phase distortion at Boston Audio Society, and
has designed the “Omni–Focus” speaker (bipolar
coincidental with phase–linear first–order crossover),
ARP 2600 analog music synthesizer, 1kW biamp and
PWM supply at A/D/S, and Class D amps.
Pierre Touzelet (“Simple Approximations
Of Tube Anode Characteristics,” p. 18) is an audio
enthusiast who resides in Vezily, France. His Excel
Circuit Simulation article appears at www.audioXpress.
com.
Tom Perazella (“The KISS Bass Project,” p. 21)
is the Director of Information Systems for a national
retailer of professional photographic equipment
headquartered in the midwest. His prior experience
includes work as a Criminalist in the San Diego and
Long Beach California Crime Labs and Director of
Marketing for a photographic wholesale distributor. In
addition to speaker design, Mr. Perazella has designed
commercial high–powered electronic flash equipment
as well as numerous pieces of audio electronics for
his own use. Other leisure activities include cooking,
golf, scuba diving, and motorcycles.
Bill Fitzmaurice (“The Tuba 24 II,” p. 30)
has been a professional musician since 1966 and
has been constructing instruments, amplifiers,
and speakers for just as long. Vice president of
DeltaSounds Loudspeakers Inc., Fitzmaurice is the
author of over 30 magazine articles dealing with
speakers and electric instruments. Bill and his wife
reside in Laconia, NH.
Ed Simon (“More On The Sound Strobe,” p. 38)
received his B.S.E.E. at Carnegie-Mellon University.
He has installed over 500 sound systems at venues
including Jacob’s Field, Cleveland, Ohio; MCI Center,
Washington D.C.; Museum of Modern Art Restaurants,
New York; The Coliseum, Nashville, Tenn.; The Forum,
Los Angeles; Fisher Cats Stadium, Manchester, N.H.
Darcy Staggs (“Solder Turrets,” p. 45) is a
retired engineer. His work experience began with
the Apollo project, then defense analysis, writing
management software at IBM in Stockholm Sweden,
and later more defense analysis back in the US. His
later years were occupied as a consulting engineer for
a firm that dealt with product design, machine design,
business development, and any other fascinating
challenge that came in the front door.
Charles Hansen (“A Brief History of Bendix
Red Bank Tubes” excerpt, p. 46) is an Engineering
Consultant and holds five patents in his field of
electrical engineering. He plays jazz guitar and enjoys
modifying guitar amplifiers and effects to reduce noise
and distortion, as well as building and restoring audio
test equipment. He has another book to his credit, The
Joy of Audio Electronics, and over 200 magazine articles
on electronics and audio. He joined the Bendix Red Bank
Division in 1966, and retired in 1998 as Supervisor of
the Systems and Controls engineering group.
b
S
www.audioXpress .com
7/25/2007 4:22:49 PM