CAPACITOR REPORT: A VISIT TO MUNDORF

Back
M AY 2 0 0 7
US $7.00/Canada $10.00

Tube, Solid State,

Loudspeaker Technology

FOCUS
ON TUBES Versatile Line Amp
Heathkit Rebuild
Tube Imp Mini Tester
Calculating Tube Parameters
Chassis Recipe
Amps from Athens

PLUS:
Expert Power Supply Tips
Continued

Testing Your Room’s

Acoustics

Test CD Review

www.audioXpress.com

Cover-507.indd 1 3/26/2007 9:11:20 AM

sound solutions By Walt Jung

Sources 101: Audio Current Regulator

Tests for High Performance
Part 2: Precise High Current/Voltage Operation
Measurement tests can help reveal which configuration is
best for your power supply application.

I
will conduct many additional mea-
surements here. Within this phase,
the focus is on current regulators
that operate at higher voltages, at
higher currents, and do so with a higher
degree of precision. This implies higher
initial accuracy, as well as good temper-
ature stability, for all circuits discussed
hereafter, with the exception of those
MOSFET based.
LM317 CURRENT SOURCE/SINK
One of the easiest ways to make a quite
good audio current source is to simply
connect an LM317 IC with a current
set resistor (Fig. 10A, left). This circuit,
which is simplicity personified, cannot
be reduced further in functionality. De-
tails of the LM317 operation are de-
scribed in References 7 and 8 (highly
recommended reading). The wide avail-
ability of this useful part in a variety of
packages at low cost makes it attractive.
The LM317 is a floating three-termi-
nal regulator, meaning it can be applied
quite flexibly, and no pin inherently
needs to be grounded. When operated
in a current mode, the internal 1.25V
reference voltage appears between the
OUT and ADJ pins, so a simple resistor
Rset programs the current into a load.
In this case a fixed 20Ω value sets up a
62mA load current. The 1.25V is held to
±50mV, and is stable over temperature.
Thus, an LM317-based current source
will be one of the more predictable and
stable types for DC. Of course, at such
higher currents power dissipation will
be an issue, so you should use a TO-
220 package part at these current levels,
along with the appropriate heatsink.
It may not be obvious at first, but the
LM317 can function as both a current
source (as in the left case) and as a cur-
rent sink, shown at the right. In either
case, the IC and its Rset resistor are
treated as a two-terminal circuit, which
is applied between the source and the
load. The LM317 current sink is imple-
mented with similar connections shown
at the right, with the load connected to
the IC’s IN pin, and using a negative
power supply. Note that in such cases a
small AC bypass capacitor may be nec-
essary at this pin, ~1µF.
The LM317 working in this current
output mode will require about 2.5V
across the IC, plus the 1.25V, for a total
of nearly 4V to make it operate. The IC
also needs a 10mA minimum of output
current for regulation. Practically speak-
ing, this means that Rset should never
be any higher than about 125Ω.

FIGURE 10B: Performance of the LM317 as a 62mA current

source shows 110dB or more rejection below 10kHz, but rapid
FIGURE 10A: Basic LM317 current source (left) and sink (right). deterioration at higher frequencies.

8 audioXpress 5/07 www.audioXpress .com

JungPart2-2779-2.indd 8 3/26/2007 9:59:43 AM

 LM337 formance (Fig. 11B). While the LM337
CURRENT SOURCE rejection is good below a few hundred
A companion device to the LM317 Hz, it degrades steadily above this, to
positive regulator IC is the LM337, de- the point where the rejection is less than
signed to operate from negative sources. 30dB above 100kHz. This is an example
It also has a 1.25V reference voltage and of the type of rejection not sought for
can be configured to regulate current higher performance audio circuits!
(Fig. 11A). The LM337 uses a similar set A detail worth noting at this point: If
FIGURE 11A:
resistor (Rset) to set up an output cur- complementary source and sink circuits
Basic LM337 rent Iout, but it also requires an output are needed for an application, it is actu-
current source. capacitor for fre-
quency compensa-
Once biased properly, the IC operates tion, C1. A typical
reasonably well, as shown in Fig. 10B, value for this ca-
an AC rejection performance plot of the pacitor is shown.
output. Here the low frequency (LF) While the
rejection is about 110dB, equivalent to LM317 and
an impedance of 316kΩ. There is, how- LM337 have
ever, noticeable deterioration at higher complementar y
frequencies. functionality, they
This is one aspect of the LM317’s achieve radically
performance that would be desirable different degrees
to improve, because the rejection at of rejection versus
200kHz is only about 60dB, meaning frequency as oper-
potentially increased sensitivity to high ated in a current
frequency (HF) intermodulation. A cou- mode. This is best
FIGURE 11B: Performance of the LM337 as a 62mA current source
ple of the following circuits address this appreciated by the shows 110dB or more rejection—but only at the lowest frequencies.
aspect of the LM317’s operation. LM337’s AC per-

audioXpress May 2007 9

JungPart2-2779-2.indd 9 3/22/2007 4:17:28 PM

 ally better performance-wise to use a
pair of LM317s as in Fig. 10A left and
right, than it would be to use an LM317
and an LM337.
Caveats: A further special point on three-
terminal regulator types is to simply be cau-
tious about replacement or “improved” 317-
type regulators, especially those designed for
low dropout. As a byproduct of their de-
sign for low DC dropout voltage, these
regulator types can have much worse
AC rejection characteristics vis-à-vis the
original. For example, two low dropout
versions of the 317 were tested for rejec-
tion in a current regulator mode similar
to Fig. 10A, and had responses more like
that of Fig. 11B than the more desirable
LM317 response of Fig. 10B. So, this is
definitely a case of caveat emptor!
DEPLETION MODE MOSFET
CURRENT SOURCE/SINKS
Power MOSFETs are both extremely
popular and widely available, and for
many years have seen widespread use in
audio amplifiers. Typically, these have
been the original format, which is that
of enhancement mode devices. This means
simply that they require an applied gate
voltage to conduct.
More recently, depletion mode MOS-
FETs have become available, which en-
ables easier use of such parts in audio
power supplies. Like the small signal
JFETs, a depletion mode MOSFET is
fully on with 0V bias, and is controlled to
lower degrees of conduction with the ap-
plied bias voltage. Thus far the depletion
mode MOSFETs that have appeared
are N-channel
parts. Two TO-220
packaged examples
are the DN2540
from Supertex and
the IXCP 10M45
from Ixys. See Ref-
erences 10 and 11
for further infor-
mation.
These TO-220
devices can oper-
ate at voltages up
to 450V, and at
currents from the
low mA range up
to about 100mA. FIGURE 12A: Basic depletion mode
They are already MOSFET current source.
being found in vac-
10 audioXpress 5/07
JungPart2-2779-2.indd 10
uum-tube-based audio projects where
high voltage capability is required. Ex-
amples can be found via Reference 12.
In application, a basic current source
using either part can be accomplished
( Fig. 12A ). This circuit is exactly the
same as with a JFET device, save the
addition of the gate-stopper resistor R1,
and the important fact that the applied
voltage can go up to 450V. And, like the
JFET counterpart current regulator, this
circuit is two-terminal, and so can be
used either as a source (shown here), or
as a sink, where the load is in the drain
lead and negative voltage is applied to
the bottom of Rset and R1. The tests de-
scribed here used an 18V power supply.
For a load current of 30mA, I found
that the two resistor values noted for
Rset were appropriate. This underscores
a basic point: These depletion mode
MOSFETs aren’t precision devices like
the LM317 and other ICs with their
fixed reference voltage(s). Rather, the
gate bias for these MOSFETs sample to
sample will vary, just as it does for other
JFET and MOSFET parts. Neverthe-
less, this circuit still has the utility of
extreme simplicity, and Rset is simply
chosen to get the required current.
Operated within the test circuit of Fig.
12A, the two sample parts produced the
data of Fig. 12B. Both devices show LF
rejections of around 110dB (~316kΩ),
with a gradual degradation beginning
in the 5–10kHz range. The DN2540 is
measurably better in terms of AC rejec-
tion at the higher frequencies. This is
apparently due to the lower parasitic
FIGURE 12B: Performance of two depletion mode MOSFET
30mA current sources shows ~110dB rejection below 5-10kHz,
then deterioration as frequency increases.
www.audioXpress .com
capacitance of the DN2540 versus the
IXCP 10M45, but I cannot precisely
confirm this (the latter isn’t specified for
capacitance).
Nevertheless, these general patterns
of AC rejection seemed to be typical for
the two devices, and were observed with
tests of other samples. The DN2540 is
preferred for operation in this circuit,
not only because of the better AC rejec-
tion at high frequencies, but because the
Idss of this part is 150mA, making it
more widely applicable.
CASCODE LM317
CURRENT SOURCES
These higher current regulators, like the
low-level circuits described in Part 1, can
also be enhanced for AC performance
by means of cascoding. As the DC cur-
rent carried by the regulator is increased,
the rejection performance inevitably de-
grades, making the value of an effective
cascode circuit more and more impor-
tant toward good results.
A circuit that can be used to cascode
the operation of an LM317 is shown
in Fig. 13A. This is similar to the basic
regulator of Fig. 10A , with an addi-
tional regulator added—stage U2. The
U1 LM317 operates just as previously,
producing an output current as noted,
which is proportional to 1.25V and in-
versely proportional to Rset. The input
drive for U1 is derived from cascode IC
U2, which floats atop U1’s output, 2.5V
higher by virtue of resistors R1 and R2.
C1 and R3 provide necessary stabiliza-
tion for the cascode.
3/22/2007 4:17:29 PM
 I tested the Fig. 13A circuit at a cur-
rent level of 62mA, to be consistent with
the basic LM317 operation of Fig. 10A.
The results are shown in Fig. 13B for
both the basic and cascode modes of
operation. Note that the addition of the
cascode reduces the noise down to a
level approaching the setup residual at
all but the very highest frequencies. Al-
though not shown here, for lower lev-
els of current operation (i.e., ~15mA),
this cascode scheme showed even lower
noise levels.
Some caveats for the Fig. 13A circuit:
Although the AC rejection properties
of this relatively simple circuit could be
considered exemplary in some regards,
I cannot recommend it unconditionally
for several important reasons. One, it has
a rather high dropout voltage, requiring
~6.5V across it—just to function! This
is due primarily to the basic characteris-
tics of the LM317, and can’t be easily re-
duced. Anticipating potential questions
here, using low dropout 317 regulators
isn’t any real help, either. I tried this, and
it does reduce the dropout—but at the
expense of rejection.
A second caveat is that the basic
LM317 dropout voltage is actually
specified as 3V for currents up to 1.5A.
Datasheet graphs show it to be typically
~1.7V at a current of 200mA at 25°C. So

FIGURE 13A:
FIGURE 13B: Performance of the LM317/LM317 as a cascode
Cascode LM317
62mA current source shows much greater rejection than in
current source.
basic mode, at all frequencies.

audioXpress May 2007 11

JungPart2-2779-2.indd 11 3/22/2007 4:17:31 PM

 the scheme here won’t really work well at
high currents and/or low temperatures.
But, there is still much latitude for
use at much lower currents and typical
temperatures from 25° C and up. Here
operation of U1 is at a fixed input/out-
put voltage of 2.5V, and because this is
still somewhat of a gray area, only load
currents of <100mA are suggested. Fi-
nally, and perhaps most important, this
setup can and will oscillate under cer-
tain conditions, so be wary. All cascode-
type schemes using additional high gain,
wide bandwidth parts have this potential
and should be rigorously checked. Input
bypassing should be used, with a film
capacitor such as C2 close to U2, and
the C1/R3 network always used.
Fortunately, for all of the cascode
schemes tested for this series, only a
couple of them showed oscillation ten-
dencies, this one included. The absence
of oscillation for an LM317 regulator
can be checked by the presence of a sta-
ble 1.25V ±50mV output (or, the exact
target DC current for this or other pre-
cision regulators). If a scope is used, the
output should be clean on a scale of a
few mV.
For some more carefully selected op-
erating conditions, a cascode LM317
arrangement can be implemented using
an LM317 as the control IC and a de-
pletion mode MOSFET as the cascode
part. This variation (Fig. 13C) can use
either the DN2540 or the 10M45 as the
cascode device M1. Note that this circuit
FIGURE 13C:
Cascode LM317 +
MOSFET current
source.
12 audioXpress 5/07
JungPart2-2779-2.indd 12
will simply not work with a conventional
MOSFET! For the two M1 device types,
it has the advantage of workability at
very high voltages, up to 450V, making
it quite attractive as a simple and precise
current source for tube circuits.
This circuit also has some caveats,
including the general ones for the 317.
For the LM317 to properly function as
a regulator, the input/output voltage, la-
beled here as V317, must meet the LM317
device dropout limits. In this circuit V317
is the Vgs of M1, and this should be
2.5V or more. Both the devices listed
for M1 typically meet this requirement
at lower currents of 10–20mA, and the
DN2540 holds up even higher. And,
don’t forget the RC stabilization net-
work, R2/C1.
AC rejection performance of this cir-
cuit operating at 16mA is shown in Fig.
13D, and for either of the cascode devic-
es it is nearly ideal. Only a tiny deviation
above the noise level at the very highest
frequencies can be noted. This excep-
tional performance makes this a very at-
tractive circuit for such lower currents.
At the higher current of 38mA (Fig.
13E), the 10M45 begins to approach the
sample device Idss. Therefore, V317 is
lower than the minimum required for ef-
fective LM317 operation, and as a result,
the data for the 10M45 shows noticeable
deterioration vis-à-vis lower currents. By
contrast, the DN2540, a higher current
device, still shows excellent rejection for
these conditions.
FIGURE 13D: Performance of the LM317 + MOSFET cas-
code 16mA current source shows excellent rejection com-
pared to basic mode at all frequencies.
www.audioXpress .com
A power caveat: While you should al-
ways be aware of power dissipation lim-
its for any of these circuits, this bound-
ary can quickly sneak up on you within
tube circuits—even at relatively low cur-
rent levels. For example, a 10mA current
in M1 of Fig. 13C with 150V across
it implies an M1 dissipation of 1.5W,
which will definitely require a heatsink.
Don’t operate under the assumption that
a datasheet rating of 1W at 25° C for a
TO-220 will guarantee a safe and long
life of the part, if it sees 1W of constant
power while the room is 25° C. Internal-
ly, the part will be much hotter, and it is
highly likely a hefty heatsink is in order
for a truly reliable design. See Reference
13 for further heatsink information.
TLV431 CURRENT SINK
The TLV431 is a three-terminal IC de-
signed to be used as a programmable
shunt regulator, from 1.24 to 6V14. It has
an uncommitted feedback path, mean-
ing that external active parts can be used
with it to extend the basic current and
voltage range. As you will see, this part
operates as a current regulator referred to
the negative rail, thus it is most suited to
make current sinks.
The TLV431 reference voltage of
1.24V has a tolerance of ±18mV (1.5%),
but A and B suffix parts tighten this to
12mV (1%) and 6mV (0.5%), respec-
tively. The TLV431 is related to the very
popular TL431, which offers similar
functionality at a reference voltage of
3/22/2007 4:17:32 PM
 2.5V. Because the TLV431’s lower volt- provide for a current of 100µA, mini- serving the total current in Rload1.
age of 1.24V is more desirable for a cur- mum. The simplest option is a 100kΩ The current in Rload1 has two com-
rent regulator (it means lower dropout), resistor (A); next most simple a current ponents, the output current flowing in
I chose it for this test. But note that source such as the J507 (B); and finally, Rset-Q1/Q2, and Iz1, the bias current
the same principles applied here for the for highest performance from the cir- of U1, which flows in Z1-U1. When the
TLV431 also work for the TL431, except cuit as a whole, functioning as a current current in Rload1 is monitored, both of
for the higher reference voltage of 2.5V. source, a J202 operating at ~280µA and these currents are, in fact, being mea-
Figure 14A is a basic TLV431 cur- cascoded with a 2N5486 (similar to Fig. sured. It would be desirable that only Iz1
rent sink that you can use over a range 8A, right option, Part one). be dominant, because this would mean
of voltages up to 40V, and currents up Because this cir-
to several tens of mA. The final voltage/ cuit is more aptly
current rating for this circuit is a func- used as a current
tion of the transistor type used for Q1 sink, the measure
and the heatsinking. Typically the load of how it per-
would be applied between the OUT1 forms would best
and OUT2 terminals. Note that the be told by a sense
OUT1 terminal need not be common resistor placed at
to the +18V supply as shown; it can (and OUT1-OUT2.
often will) be a higher voltage. But, as noted, the
The U1 IC, a TLV431, regulates with test setup here
a 1.24V developed between the R and A measures current
terminals as noted, so the Rset resistance in Rload1, which
determines the current flowing into Q1- is tied to ground.
Q2 and the external load. The feedback Interestingly, how-
path is via terminal K and the base- ever, you can still
emitter path of Q1–Q2. Z1 performs as infer some degree FIGURE 13E: Performance of the LM317 + MOSFET cascode
load impedance for IC U1, and can be of performance of 38mA current source still shows excellent rejection for the
one of three options, all of which should the circuit by ob- DN2540, but deterioration for the 10M45S.

audioXpress May 2007 13

JungPart2-2779-2.indd 13 3/22/2007 4:17:35 PM

 that the Rset-Q1/Q2 current path is
noise free.
To a great extent this is indeed true,
and is reflected by a related change in
Iout rejection, because Z1 is varied. This
is shown in Fig. 14B for various Z1 con-
ditions. Note that for a finite resistance
value for Z1, the net impedance is shown
by the Vout (100k) plot (as was true
for the calibration plots of Part one of
this article). And, as Z1 takes on higher
impedance characteristics, such as with
the Vout ( J507) plot, this condition is
reflected in a higher impedance display
(i.e., more rejection). The greatest rejec-
tion at the lower frequencies is provided
by the cascoded J202 setup, while the
J507 provides the most rejection with a
single component used for Z1.
So, while this test method doesn’t di-
rectly measure just the current flowing
in the collector of Q1/Q2, it still sug-
gests some aspects of relative quality—a
good thing, nevertheless. The bottom
line is that you can use the circuit as ei-
ther a current sink, in which case Z1 can
likely be the simple 100kΩ resistor, or,
alternately, as a current source, whereby
the higher impedance choices for Z1
are suggested, such as the J507 or the
cascoded J202.
You might ask what the need is for
this type of current sink, when previous
examples have provided good perfor-
mance at these currents. The answer
lies in the overall flexibility of Fig. 14A.
FIGURE 14A:
TLV431 current
sink (source).
14 audioXpress 5/07
JungPart2-2779-2.indd 14
Operated as a current sink, and with Q1
properly selected for power and voltage
handling, this circuit can handle cur-
rents of amperes and voltages as high
as the Q1 device rating. Although data
isn’t shown for this example, with a D44
series power transistor for Q1, output
currents of 350mA have been witnessed.
This is all available with relative simplic-
ity—Z1 a 100k resistor (Z1) and Rset
chosen for the current desired. Or, with
Q1/Q2 2SC2362XS, the OUT1 termi-
nal can operate up to 150V, at low cur-
rents, with proper heatsinking.
LM4041 CURRENT SOURCE
The LM4041-ADJ is a three-terminal
IC designed to be used as a programma-
ble shunt regulator, from 1.233 to 10V15.
Like the counterpart TLV431 series, it
also has an uncommitted feedback path.
And, as with the TLV431, this means
external active parts can be used with it
to extend the basic current and voltage
range.
A key difference in applicability is
that the LM4041-ADJ operates with
a positive rail common, as opposed to
the TLV431, which uses a negative rail
common scheme. The two devices can be
viewed as complements, performing sim-
ilar tasks. The basic LM4041-ADJ refer-
ence voltage is 1.233V, and the available
grades of C and D for this version have
initial tolerances of ±0.5% and ±1%, re-
spectively, for Vout = 5V.
FIGURE 14B: Performance of the TLV431 as a 38mA current
source depends upon Z1, but is excellent with a high-Z for Z1.
See text on current sink operation.
www.audioXpress .com
Inasmuch as the operation of the
LM4041-ADJ is with the positive rail
common, you can easily use it to make
current sources operating over a wide
range. An example is shown in Fig. 15A,
which is a mirror image of the TLV431
circuit of Fig. 13A.
In this current source circuit, the out-
put current is measured in Rload1, which
is in series with the Q1-Q2 collectors.
There is no error current from the inter-
nal amp of the LM4041, thus the rejec-
tion characteristics measured at Rload1
are indeed what you get. This is shown
in Fig. 15B, for conditions as shown and
a current of 38mA. The LF rejection is
approaching 130dB (3.16MΩ), which,
while good, is still well above the noise
level. However, the rejection deteriorates
above 1kHz.
Cascoding of Q1-Q2 in this circuit
did not improve the performance to any
great degree, only 2-3dB. At lower cur-
rent levels of a few mA, the rejection
improved to just above the residual noise
level. From this, you would conclude
that this particular circuit is better used
at the lower current levels.
TLV431 BOOSTED CURRENT
SOURCE/SINK
As noted, the TLV431 circuits are better
suited to use as current sinks, as opposed
to sources. But, with some key changes,
you can use a TLV431 current regula-
tor either as a source or sink, and/or at
3/22/2007 4:17:37 PM
 high voltages. One scheme to do this is
shown in Fig. 16A.
This circuit is like Fig. 14A, except Q1
uses a standard connection (non-Dar-
lington), and the current source portion
represented by Z1 of Fig. 14A is replaced
by a high current or high voltage equiva-
lent. This has the effect of regulating the
current in R1, making the error current
flowing from Rset and the TLV431-A
pin constant. Therefore, this circuit, op-
erating as a whole, can be used either as
a source or as a sink.
With an LM317 for U2, R1 establish-
es a current of ~800µA, providing drive
to Q1 for currents of 50mA or more.
Q1 is bootstrapped by the LM317 at
the collector and sees less than 2V C-E.
It thus does not dissipate high power at
38mA of output or even higher currents.
The LM317 will need the heatsink in
this circuit long before Q1!
For operating voltages higher than the
40V LM317 rating, you can also use a

FIGURE 15A: FIGURE 15B: Performance of the LM4041 as a 38mA cur-

LM4041 current rent source is good, but falls short of excellent, particularly
source. at the higher frequencies.

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. And MORE...
The results are outstanding !
• 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
• Switch your audio installation on or off without noise,
Outstanding kits for the DIYer thumps, or spikes • Turns on the various parts of your system
in a safe order • Avoids dangerous transients in your speakers
More information : www.dcx2496.fr / ww.selectronic.fr

Pub AUDIOXPRESS - Selectronic - File AXP 01-2007 - Mechanicals : 181 x 120 mm

audioXpress May 2007 15

JungPart2-2779-2.indd 15 3/22/2007 4:17:40 PM

 depletion mode MOSFET by substitut-
ing an M1 device at the points marked X
and Y. R1 can remain the same, and the
current limit for this mode will of neces-
sity be much less than 40mA. But, the
voltage limitation becomes that of the
M1 device used, or 450V as shown. Take
care to use a proper heatsink for M1!
Performance in terms of AC rejection
is shown in Fig. 16B for all three cascod-
ing options, operating at 38mA. Overall,
the best performance is achieved with
the LM317, where the errors are only
slightly more than residual noise, except
for the very highest frequencies. The two
MOSFET parts are nearly as good at LF,
but deteriorate more rapidly above 1kHz.
Of the two MOSFETs, the DN2540 is
favored due to lower noise at all frequen-
cies, plus its ability to handle more cur-
rent. To get higher output currents, Q1
can be operated as parallel devices driven
from R1-bottom end, with 10Ω current
sharing resistors in the emitters.
CONCLUSIONS, CAVEATS, AND
RECOMMENDATIONS
This concludes the testing portions of
this series. A future article will explore
some example applications of current
sources and sinks within audio circuits,
and discuss some general power supply
system noise reduction techniques.
Some general caveats are appropriate
here, beyond those specifically stated. I
believe the tests are valid for the condi-
tions cited, and in general can be used to
FIGURE 16A:
Boosted TLV431
current
source/sink.
16 audioXpress 5/07
JungPart2-2779-2.indd 16
differentiate among the various circuits.
Of course, there is an infinite set of dif-
ferent load, voltage, and current operat-
ing conditions that you may require. So,
you should not expect to duplicate any
measurements exactly for other condi-
tions. But, in general the observations
should hold up—cascodes work bet-
ter, JFETs need proper voltages to work
best, and so on.
To summarize, here are some prin-
ciples to keep in mind:
• Select single JFET parts from fami-
lies with lowest Vgs and thus highest
rejection. An example would be the
J201/2 series.
• Alternately, select from a specified
JFET current regulator device family,
such as the J507 series.
• Always operate current regulator cir-
cuits with sufficient voltage headroom
to maximize rejection.
• Above 4-5mA of current, consider cas-
code type circuits. At several tens of
mA, this should be considered manda-
tory for good performance.
• For any current regulator circuit, minimize
capacitance in whatever active devices are
used. This will enhance high frequency
noise rejection and minimize the possibil-
ity of high frequency intermod.
If I were asked to recommend which
of the many current regulators described
here to use, I’d try to keep it as simple as
possible. The maximum bang-for-the-
FIGURE 16B: Performance of the boosted TL431 as a 38mA
current source or sink ranges from good to excellent, depen-
dent upon the cascode device chosen.
www.audioXpress .com
buck is the cascode LM317 + MOS-
FET of Fig. 13C, assuming your current
requirement is 40mA or less. This one
worked great for me within a 12mA
current feed for a 24V shunt regulator.
For higher currents, the Fig. 16A circuit
is both flexible as a source or sink and
capable of much higher currents when
Q1 is appropriately selected. For low
currents of just a few mA, single and/or
cascoded JFETs are likely best (Fig. 8A).
Or, you could select the reference diode
circuit of Fig. 6A.
SOME HOMEWORK ASSIGNMENTS
One manuscript reviewer asked about
very high output currents, i.e., several
amps. My general answer is that yes, this
should be possible with minor revisions.
I said, “You could use Fig. 14A with
a conventional N-channel MOSFET
replacing Q1/Q2. The 1-2V Vgs of a
MOSFET will bias the K pin of U1
roughly 1-2V above the R pin (but don’t
forget a 100Ω snubber in the MOSFET
gate circuit). This should work OK for
ampere outputs. Pick the FET for the
required current, voltage, power, and,
preferably, lowest C. I’m sure you have
a favorite here. One possibility might
be the Fairchild FQP4N20L, a TO-220
part, available from Mouser. I think I’ll
put this idea in at the end of the Part 2,
as a reader ‘Homework’ assignment.”
So there you have one assignment
for some fun experiments. Let us know
what you find out with this MOSFET-
3/22/2007 4:17:41 PM
 boosted current source idea!
Another assignment is to explore a
hybrid vacuum tube/solid-state current
regulator. For example, you could also
use Fig. 13C with a power triode in place
of M1 (grid to U1-OUT, cathode to
U1-IN, and plate to the input voltage).
The LM317L might be a possible can-
didate for U1.
I’d be very interested to hear about
your results with these ideas. Write me
at audioXpress via conventional mail, or
contact me via my website, www.waltjung.
org/index, and happy current sourcing
and sinking! aX
Acknowledgments
A number of folks reviewed the
Sources 101 manuscript and made help-
ful comments toward improvement:
Ken Berg, Erno Borbely, John Curl, Bob
Fitzgerald, Clarke Greene, Chuck Han-
sen, Bruce Hofer, Mark Kovach, Rick
Miller, and Andy Weekes. My sincere
thanks go out to all of them.
REFERENCES
1. Walt Jung, “Regulators for High Per-
formance Audio, parts 1 and 2,” The Audio
Amateur, issues 1 and 2, 1995.
2. “LM134/234/334 3-Terminal Adjust-
able Current Sources,” National Semicon-
ductor, March 2005, www.national.com.
3. Arthur D. Evans, Designing With Field-
Effect Transistors, McGraw-Hill, ISBN 0-07-
057449-9, 1981.
4. “The FET Constant-Current Source/
JungPart2-2779-2.indd 17
Limiter,” Application Note AN103, Vishay/ 14. “TLV431, TLV431A, TLV431B Low-
Siliconix, March 10, 1997, www.vishay.com. Voltage Adjustable Precision Shunt Regula-
5. “J500 Series Current Regulator Diodes,” tor,” Texas Instruments, January 2006, www.
Vishay/Siliconix, July 2, 2001, www.vishay.com ti.com.
6. Selected and matched JFETs and 15. “LM4041 Precision Micropower
JFET current regulator devices, as well as Shunt Voltage Reference,” March 2005, Na-
other audio components, are available from tional Semiconductor, www.national.com.
Borbely Audio. See www.borbelyaudio.com/
audiophile_components.asp.
7. Bob Dobkin, “3-Terminal Regulator is
Adjustable,” National Semiconductor Appli-
cation Note 181, October 1975, www.national.
com.
8. Bob Dobkin, “Applications for an
Adjustable IC Power Regulator,” National
Semiconductor Application Note 178, Janu-
ary 1977, www.national.com.
9. “LM337 – 3-Terminal Adjustable Neg-
ative Regulator,” National Semiconductor,
www.national.com.
10. “DN2540 N-Channel Depletion-
Mode Vertical DMOS FETs,” Supertex,
www.supertex.com.
11. “IXCP 10M45 Switchable Current
Regulators,” Ixys, www.ixys.com.
12. John Broskie of www.tubecad.com has
written much on current regulators, both
solid-state and tube related. To reveal this
fascinating information, do a Google search
of his site using these terms: “current source
site:www.tubecad.com” (less the quotes).
13. Walt Jung, “Thermal Consider-
ations,” Section 7-5 within Walt Jung, Ed.,
Op Amp Applications, ISBN 0-916550-26-5,
2002, www.analog.com/library/analogDialog/
archives/39-05/op_amp_applications_handbook.
html.
audioXpress May 2007 17
3/22/2007 4:17:47 PM
 special report By Jan Didden
We Visit Mundorf
The Mundorf Company is perhaps best known for its high-performance
capacitors and coils for audio applications. The author visited the company
and found that it is active in many more audio fields and is hard at work
on new products for audiophiles.
M
undorf is located in Cologne,
Germany, just an hour’s drive
from my hometown. I was
cordially greeted by Raimund
Mundorf, the founder of the company,
and his brother Norbert ( Photo 1) . Over
a coffee with typical German Kuchen,
we talked about audio and the history
of the company, which started in 1985
when Raimund eventually built a coil
winding machine for the custom filter
coils he handcrafted one at a time. The
machine worked quite well and gradually
a business opportunity presented itself.
The old machine no longer exists but has
been followed by new models producing
new and better products.
PHOTO 2:
Large contact
area of the
MCap RXF.
18 audioXpress 5/07
Didden2788-1.indd 18
PHOTO 1: The Mundorf
brothers and the author
(right) discussing cross-
over issues over coffee and
Kuchen. (All photos by Lou
Jansen).
COMPANY BELIEFS
The goal of Mundorf is to design “ideal”
components. For example, it is known
that generally capacitors have not only
a capacitance, but also an ohmic series
resistance and a series inductance. Those
parasitic attributes have unwanted effects
on performance and can cause frequency
response deviations and even oscillations
in certain applications. The dielectric
that is used as isolation in capacitors is
also important: Less ideal dielectrics are
known to cause absorption and subse-
quent release of signal energy that can
distort the sound.
But there is no free lunch, of course.
For example, Mundorf foil caps are often
www.audioXpress .com
PHOTO 3: Raimund Mundorf giving away (al-
most) the secret of ultra-low inductance caps.
shorter, but larger in diameter, than other
caps of equivalent rating. Why? Well,
the usual wide, small diameter caps have
higher inductance because of the longer
pathways of the signal from one connec-
tion to the other. On short, large diam-
eter caps, not only is the path shorter, but
also there are more windings that appear
in parallel, further reducing unwanted in-
ductance. The ESR is also lower because
of the wide contact area. You can see this
easily in a type MCap RXF cap (Photo
2).
However, short, large diameter caps
have many more windings, which means
more expensive production. Mundorf be-
lieves that the extra cost is worth it for
better performance, and its customers
seem to agree! The latest trick is using
a series connection of two internal caps
with reversed internal current flow. That
means that to get, say 2µF, you need to
build two caps of 4µF in one package,
which is four times the material and ef-
fort otherwise required. The result, how-
ever, is a cap with almost zero inductance:
the Supreme series, also available in sil-
ver/oil and silver/gold. These are exotic,
expensive materials, but they do improve
the characteristics of the components.
Over the years, Mundorf crossover
components have earned their place
among discerning audiophiles and manu-
facturers alike. For example, if you own
some recent B&W speakers, chances are
that the crossover is by Mundorf.
3/26/2007 9:07:49 AM
 COIL CONSTRUCTION
People who know me are aware that I am often critical of
claims that seem to have no basis in physical reality. I was,
therefore, quite skeptical to hear that Mundorf produces coils
with special construction and impregnation treatment to re-
duce microphonics1. But to my amazement, when we went
over some third-party measurements, this actually became
quite clear! As Fig. 1 shows, some coils happily resonate me-
chanically in certain frequency bands and the choice of mate-
rial, construction, and impregnation can help to suppress this
effect.
You can argue that the microphonics as such would not
be audible with the component mounted inside an enclosure.
But remember: that mechanical resonance needs energy, and
that energy must come from somewhere, and it can only come
from the signal passing the components! The specific effect
on the sound depends very much on the particular place in
the circuit, but it is clear that this effect distorts the signal and
thus the sound.
The expertise Mundorf built up with home hi-fi compo-
nents also allowed it to branch out to car audio. Notable here
are very high value capacitors to hold up the car’s supply for
those watt-hungry systems–––electrolytics that can deliver
peak currents of several hundred amperes. And, oh yes, they
should be mounted on a heatsink. . . and if you need a 300A
13.8V DC supply to demo your car stereo at home, Mundorf
have those, too.

TUBECAPS
One of the disadvantages of electrolytics is that they age and
lose capacitance over their lifetime. The aging is accelerated
with elevated temperatures often encountered in (tube) power
amps. The solution is a foil cap, but traditionally foil caps have
been very, very bulky for the required capacitances and voltage
ratings for tube amps and supplies and consequently are quite
expensive. Using a very thin, rough foil can increase capaci-
tance for a given volume, but the reliability may suffer because
of production defects and foil puncturing.

FIGURE 1: Vibration of coils of various constructions with signal frequency.

audioXpress May 2007 19

Didden2788-1.indd 19 3/22/2007 4:01:08 PM

 With typical creativity Mundorf at-
tacked the problem and solved it: Accept
the fact that after production the caps
may have some shorted spots, but then
use a special process that “burns off ” a
small conductor area around the short to
restore isolation.
This self-healing effect is not new, but
as far as I know has not been used in this
application area. The result is a capacitor
that is roughly equal in volume to an
electrolytic but with no aging, improved
reliability and less series resistance so
it functions much better and reliably in
PHOTO 4: The Mundorf Air Motion
Transformer diaphragm.
20 audioXpress 5/07
Didden2788-1.indd 20
(tube) power amps and supplies. It is
non-polar to boot, so is an excellent high
voltage, high capacitance coupling cap.
AMT DRIVERS
In a corner of the lab I spied an unknown
object, which turned out to be a prototype
Air Motion Transformer (AMT) speaker
driver (Photo 4)! Many readers remember
Oskar Heil’s famous Heil speakers of
the 70s with the patented AMT mid/
high driver, which consists of a pleated
diaphragm with a conductor etched on it,
placed in a strong magnetic field. As the
signal current flows across the diaphragm,
alternate sides of the pleats are attracted
to and repelled by each other: the pleats
open and close in a musical rhythm and
move the air outward and inward, gener-
ating a sound field.
Over the past three years, Mun-
dorf pushed this principle toward the
next level in terms of rethinking and
researching the designs leading to the
award of new patents. Today the com-
pany produces a wide range of varied
AMT drivers for OEMs, but I wouldn’t
www.audioXpress .com
be surprised if they also became available
to individual builders. The measurement
graphs I saw in Mundorf ’s own sound
dead room were impressive.
I am not surprised that Mundorf is
so successful in its endeavors. The com-
pany’s employees, who obviously enjoy
what they are doing, are tinkerers in
the right sense of the word, looking for
creative solutions to problems. Being a
family business, Mundorf has the flex-
ibility to do what it thinks is worthwhile
without the immediate worry of the
shareholders. Ultimately, that benefits all
of us audiophiles. aX
More info at www.mundorf.com
FOOTNOTE
1. Strictly speaking, the effect described
here is not microphonics. Micro-
phonics means that the component
picks up a signal from vibrations (air
or mechanical); in this case I describe
a component generating vibrations
from a signal. Maybe we should call
this speakerphonics?
3/22/2007 4:01:17 PM
O GLASS AUDIO • GLASS AUDIO
A Versatile Line Amp for Preamp,
Headphone, and Power Use
Here’s a do-it-yourselfer’s dream in tube amp design: four
operating modes, high quality, low cost, and low distortion.
T
his design offers two versions:
a ten-tube $160 line amplifier
and an eleven-tube preampli-
fier/headphone amplifier/power
amplifier. Either way, it is probably the
most versatile stereo amplifying device
ever offered to the DIY audiophile. It
offers four operating modes with out-
standing specifications: line amp, head-
phone amp, preamp, and power amp.
The 10W, 0.9% distortion single-ended
(SE) amp is the lowest distortion ampli-
fier I have come across. It also has lower
distortion than published tube manual
push-pull amps.
The damping of the 10W amp is in-
credible: 2.8V across an 8Ω resistor in-
creases to only 3.8V when the 8Ω load
is removed. This is accomplished in a
standard output circuit without the aid of
negative feedback. It is obvious new de-
sign ideals for vacuum tubes are still with
us. Behold! The magic of paralleling and
low mu tubes.
Note: If operating in the 10W SE
Stereo mode, you need to add two Ham-
mond output transformers (P-T1640SE,
$93 each) from Antique Electronics to
the cost.
THE CHOICE IS YOURS
The eleven-tube amplifier (Photo 1)
uses very linear low mu, high perveance
12B4s, and a 12AT7 preamp. Five paral-
Still-2756-1.indd 21
• GLASS AUDIO • GLASS AUDIO
leled 12B4s are used in each channel of
the stereo line amp (Fig. 1). In the line
amplifier mode only the 12B4s are used.
In the headphone/power amplifier/pre-
amp mode, the 12AT7 drives the 12B4s.
The line amp and multi-mode amp do
not employ negative feedback. The line
amplifier frequency response is flat from
10Hz to 50kHz, distortion is less than
0.15% at 2V RMS output, and output
impedance is 420Ω. The voltage gain of
the 12AT7 and five paralleled 12B4s is
125.
Note: The gain of a single 12B4 is 2.4
and five paralleled 12B4s have a gain of
3.8 (cathode un-bypassed) and a gain of
5.6 (cathode bypassed). The plate current
requirement of the five 12B4As is 50mA
when using a plate load resistor of 3.1kΩ.
The selection of the low mu 12B4 en-
ables this line amp to rival or surpass the
performance of the most expensive com-
mercial line amps because they use medi-
um mu triodes. For truly superlative sonic
performance a line amp requires a low
mu, high linearity, high perveance tube,
and the 12B4 meets this requirement. I
• GLASS AUDIO
By Joseph Norwood Still
PHOTO 1:
Finished unit.
know of no other tube more suitable for
this application. The amp is designed to
outperform $3k to $10k commercial line
amplifiers.
GENERAL INFORMATION
Great care is required when operating
tubes in a parallel configuration to pre-
vent oscillation or instability. To address
this problem most resistors in the amp
are carbon composition. The 3.1kΩ, 10W
plate load metal-oxide resistors have 2Ω
carbon composition resistors in series
with the plate load resistors. The cathode
resistors are 220Ω, 5W metal-oxide types.
The construction of the amp requires
short leads as used in R-F circuits, and
an aluminum chassis is mandatory to
ensure good grounding. A ground lug
between tubes 1-5 ensures short leads for
the cathode resistors. The “plate stopper”
2Ω resistors are connected to a termi-
nal strip located between tubes 2 and 3.
Single 2Ω resistors are connected to the
plates of tubes 2 and 3, and two series-
connected 2Ω resistors are connected to
plates of tubes 1 and 5, respectively. This
May 2007 21
3/22/2007 4:22:41 PM
 GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO

the headphone. I recommend the $20

65Ω headphones from Behringer, model
HPX2000, or the $50 Audio Technology
units, model ATH-M30, from Parts Ex-
press (800-338-0531).

Switch Settings:
1. Set S2 to H.P. position.
2. Set S3 to L.A. position.

PREAMP
Application (Optional): You may choose
the preamp mode if you prefer to build a
power amp that uses only a single ampli-
fying stage to drive a SE power output
tube or a phase inverter stage to drive
push-pull power output tubes. The pre-
amp mode is the same as the headset
mode except the output signal is ob-
tained from the plates of the 12B4s via
the 0.47µF capacitor. If you require direct
DC coupling of the phase inverter, add a
FIGURE 1: Multi mode amplifier. separate pair of RCA female jacks.
The +115V DC output f rom the
ensures a carbon block exists on the long audio output signal to drive the five par- 12B4s (3.1kΩ) plate load resistor is com-
lead required to connect tubes 1 and 5 to alleled 12B4s line amp. A 0.25V RMS patible with the DC grid voltage require-
the terminal strip and the metal-oxide input signal to the 12AT7 provides an ments of most direct-coupled phase in-
plate resistors. output of 9V RMS from the 12AT7- verters. The voltage gain of the 12AT7
Figure 1 shows that much thought output circuit to drive the headphone and 12B4s is 125 with cathodes of the
was given to ensure the stability of the 12B4s amp (maximum output). The out- 12B4s un-bypassed and 185 with cath-
paralleled 12B4s. The pin configuration put for the headphones is obtained from odes bypassed. The distortion at 10V
of the 12B4 grid with pins 2 and 7 con- the cathodes of the five paralleled 12B4s. RMS output is 0.4% at 32Hz, 0.34% at
nected to the grid ensured a short path You operate the headphones by simply 1kHz, and 0.6% at 20kHz with 12B4s
for the insertion of a “grid stopper” resis- plugging the headphone jack into the cathodes un-bypassed. The distortion at
tor between tube stages. Pins 1 and 9 of chassis mounted ¼˝ or 3.5mm jack. For 30V RMS output is 1.2% at 32Hz, 0.65%
the tube sockets face toward the middle best performance, a 55Ω-75Ω impedance at 1kHz, and 1.5% at 20kHz with cath-
of the chassis ensuring short runs for headphone is required. odes bypassed. The frequency response is
the cathode, plate, and grid resistors. It The amp has a distortion of 0.4% at flat from 20Hz-25kHz.
is important that you do not substitute 1kHz when using a 65Ω Behringer head-
other types of resistors for the carbon phone with a measured output signal Switch Settings:
composition resistors. These recommen- of 0.8V RMS. At a normal output level 1. Set S2 to H.P. position for a voltage
dations are required to ensure stability of 0.6V RMS, distortion is 0.15%. The gain of 125 or to P.A. position for a
when using parallel tube circuits. 0.6V and 0.8V RMS signals provide a voltage gain of 185.
very loud listening level. The frequen- 2. Set S3 to L.A. position.
HEADPHONE cy response is flat from 20Hz-50kHz.
Application: The headphone amplifier The distortion and frequency response POWER AMP
circuit uses a 12AT7 to obtain sufficient measurements are made at the coil of Application (Optional): To operate in
the power amp mode, set the DPDT
SPECIFICATIONS OF LINE AMPLIFIER switch (S3) from the line amp mode to
1. The frequency response is flat from 10Hz to 50kHz.
2. The 100Hz, 1kHz, and 10kHz square waves are perfect. the power amp mode and the DPDT
3. The distortion at 2V RMS (cathode un-bypassed) is 0.15% -20Hz, 0.12% -1kHz, and 0.14% -20kHz. switch (S2) to place the negative end of
4. The distortion at 5V RMS (cathode un-bypassed) is 0.25% -20Hz, 0.2% -1kHz, and 0.24% -20kHz. the 1000µF capacitor at ground. The S3
5. The distortion at 2V RMS (cathode bypassed) is 0.25% -20Hz, 0.15% -1kHz, and 0.36% -20kHz.
6. The distortion at 5V RMS (cathode bypassed) is 0.38% -20Hz, 0.25% -1kHz, and 0.5% -20kHz. switch setting connects the 3.1kΩ resis-
7. Clipping occurs at 30V RMS. tor from the plate circuit of the 12B4s
8. Plate impedance is 420Ω. to the 1200Ω 25W Hammond output
9. Voltage gain is 3.8 (cathode un-bypassed) and 5.6 (cathode bypassed).
transformer of the SE-12B4s. The power
Important: If you require only the line amplifier mode, omit switches (S2/S3), headphone jacks, and the 12AT7. output of the SE 12B4s is 10W, and the
distortion at this level is 0.9% with no

24 May 2007

Still-2756-1.indd 24 3/22/2007 4:23:28 PM

 GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO G

loop feedback. The frequency response is erating with a low plate voltage of 260V configuration by adding audio interstage
flat from 30Hz to 15kHz at 10W output DC. transformer A.E.- P-T124B (10K to 90K
level. A 0.7V RMS input voltage is re- C.T.) to the input of the amp and P-P,
quired for 10W output from the power Switch Settings: 120W, 1900 C.T. (A.E.-P-T1650T) to
amp. 1. Set S2 to P.A. position. the output of the right and left chan-
The plate current of the five 12B4s 2. Set S3 to P.A. position. nels. No changes are made to the 11-tube
in each stereo channel is 148mA. The amp. The power output of the monaural
damping criterion of the amplifier is ex- Note: A separate volume control is pro- amp is 24.5W, and the distortion at 1kHz
cellent. At 10W without feedback, the vided for R and L channels to prevent measures 1.2%. A large percentage of the
distortion at 32Hz is 2.0%, at 50Hz it crosstalk. Switch S1 (Fig. 2) turns the distortion is attributed to the input trans-
is 1.2%, and is 0.9% at 1kHz, 1.6% at heater supply on and off, and switch S2 former.
5kHz, and 2.2% at 15kHz, respectively. turns the high voltage supply on and off. The frequency response is relatively flat
At 5W the distortion is 0.8% at 32Hz, at The H.V. switch S2 only operates if the from 20Hz to 15kHz. Sensitivity is 0.56V
50Hz it is 0.7%, and 0.6% at 1kHz, 1.5% heater switch is set to the on position. RMS for 24.5W with 33kΩ resistors con-
at 5kHz, and 1.8% at 15kHz, respectively. nected across the secondary windings of
These specifications are outstanding for Of Interest: The right and left chan- the interstage transformer. I do not rec-
a single-ended amp, especially when op- nel amps are connected in a monaural ommend this amp for use as a monaural
amplifier, but I thought the information
AMPLIFIER PARTS LIST here might be interesting. The 35W/60W
Qty. Value Watts Part No. Type Mfg. amp presented in aX June 2004 is more
Note: Double Parts for Stereo cost effective and a more desirable choice.
14 2Ω ½W 30BJ500-2 Carbon Composition M
1 220Ω 5W 286-220 Metal Oxide M
2 330Ω ½W 30BJ500-330 Carbon Composition M POWER SUPPLY
4 680Ω ½W 30BJ500-680 Carbon Composition M The power supply (Fig. 2) contains a
2 1.0kΩ 1W 282-1K Metal Oxide M
2 6.2kΩ 5W 286-6.2K Metal Oxide M DC heater supply and a high voltage
6.2 K/5W resistors are paralleled to form 3.1 K/10W resistor. plate supply. The DC heater supply uses
2 100kΩ 2W 282-100K Metal Oxide M
1 470kΩ ½W 30BJ500-470K Carbon Composition M a 12V AC, 6A transformer connected
1 100K, single ½W 31VJ501 audio taper M to a 25A full-wave bridge rectifier. The
1 0.22µF, 400V orange drop 715P22454MD3 polypropylene M output of the rectifier is connected to a
1 0.47µF, 400V orange drop 715P47454MD3 polypropylene M
1 1.0µF, 250V, metalized 1430-2105 polyester M 4700µF capacitor, and the filtered output
1 220µF, 35V electrolytic ALEL voltage from the capacitor is connected
1 1000µF, 400V EC-1000 electrolytic ALEL
Note: Do not double the parts listed below: to the heaters of the 12B4s and 12AT7.
1 6-position, dual deck, 2 pole 275-1386 RS The ripple voltage of the supply is 0.9V
1 knobs, 4-per pack 275-415 RS
4 5-lug tie point, terminal strips 274-688 RS
RMS.
10 RCA, female, phono jacks 161-002 M The high voltage plate supply uses a
1 5-conductor, shielded cable, 10´ P.N. 5C-S22 ALEL 100 VA line transformer with the output
1 single conductor, shielded cable, 3´ P.N.42-2371 RS
11 T-61/2 tube socket, 9-pins P-ST9-137R micalex AE secondary winding wired for 230V RMS.
10 12B4A (NOS)-Order 12 AE This output voltage connects to a full-
Note: Parts shown below are not required for line amp mode operation.
1 DPDT rocker switch 112-R13-130B ¾˝ hole M wave bridge, and the 260V DC output
1 DPDT toggle switch 10TF115 miniature M from the bridge is connected to a 470µF
1 12AT7 (Note: I recommended you order 3) AE capacitor and a 1.0H 300mA choke. The
1 1,200Ω primary-4,8,16Ω secondary, 25W (P-T1640SE) AE
Note: Transformer is only required for SE 10W power amp. output of the choke is connected to a
470µF capacitor, and is additionally fil-
POWER SUPPLY PARTS LIST tered by a 2H choke and a 470µF ca-
pacitor connected to the 12AT7 plate
Qty. Value/Designator Part No. Type Mfg.
circuit. The ripple voltage of the 260V
1 100kΩ (R-1) 2W 282-100K metal oxide M
1 4700µF, 25V (C1) electrolytic ALEL DC supply is 0.3V RMS. The bleeder of
3 470µF, 400V (C2-4) 1.4˝ dia. × 1.63˝ EC-4740 electrolytic ALEL the power supply is 100kΩ, 2W.
2 SPST switch (S1-2) ¾˝ panel hole, 125V 112-R13-130A Rocker Sw. M
1 Bridge rectifier (BR-1) 25A, 200V FWB-252 1 1/8in2 ALEL
1 Bridge rectifier (BR-2) 4A, 600V FWB-46 ALEL CONSTRUCTION
1 1.0H, 300mA choke (L1) P-T158T AE The amplifier I originally built as a line
1 2.0H, 100mA choke (L2) P-T154M AE
1 115V-12.6V ct, 6A (T1) 546-166Q12 (Hammond) M amp on a 13.5 × 5 × 2˝ chassis is shown
1 115V-230V, 0.860A(T2) 100VA 553-N77U (Triad) M in Photo 1. The chassis was very crowded
1 3AG fuse block FHBL-3 ALEL (wiring) and required mounting the heat-
1 1 pack, 1A, 3AG, 5-fuses FS-1 ALEL
1 3 connector, AC power cord S-W206 AE er transformer on the rear of the chas-
1 Chassis box, aluminum 16˝ × 8˝ × 3˝ 548-1444-28 M sis. To avoid the crowding and to mount
1 1 pack Rubber feet-Self stick, rubber 64-2342 RS
1 115V AC Fan, 32 cfm, 80 × 80 × 38mm 433-3E-115B M the heater transformer on the underside
of the chassis and the two 25W output

26 May 2007

Still-2756-1.indd 26 3/22/2007 4:23:37 PM

O GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO

with the power trans- CONCLUSION

12AT7& 12B4 OPERATING VOLTAGES former, and position I believe this amplifying device is the
Headphone Preamp Line amp 10W Power Amp the five right-channel most versatile and outstanding in tech-
12AT7 12B4 12AT7 12B4 12B4s and five left- nical specifications for each of its four
Ebb 260V 260V 260V 260V
Eb 110V 115V 105V 220V channel 12B4s on ei- operating modes of any device currently
Ek 1.5V 12V 1.6V 33V ther side of the 25W offered to the audiophile. The headphone
Ib 1.5mA 50mA 1.6mA 148mA output transformer. and 10W SE amp modes are especially
Plate Dissipation (no signal) per 12B4 tube 6.5W Connect the four ideal for apartment dwellers. If you elect
Plate Dissipation (max signal) per 12B4 tube 4.5W
Efficiency of 10W Power amp 24% electrolytic capacitors to build this amp, you will enjoy many
in the power supply to hours of listening entertainment. aX
transformers on the top of the chassis, terminal strips on the underside of the
I recommend that you use a 16 × 8 × 3˝ chassis. Mount the fan facing the power WARNING: Dangerous voltages are
chassis for this project. Mount the power transformer and notch out the bottom of present, exercise extreme caution when
transformer at the end-middle of the the chassis to enable the fan to cool the working on the line amp and never leave
chassis. Locate the two output transform- electrical components on the underside of the line amp upside down when children
ers in the middle of the chassis in line the chassis. are present.

FIGURE 2: The power supply schematic. If you encounter any problems or have
any questions, contact me at 302E. Joppa
Rd., Apt. 1911, Towson, MD 21286.

Distributors:
RS - Radio Shack
ALEL – All Electronics 800-826-5432
AE - Antique Electronics 480-820-5411
M – Mouser Electronics 800-346-6873

May 2007 27

Still-2756-1.indd 27 3/26/2007 9:05:52 AM

 GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO G

Heathkit W-7M Rebuild

Follow this author’s successful method of
rebuilding vintage equipment from Heathkit.

By Bruce W. Brown

A
PHOTO 1: The finished rebuild. s the first power amplifier that
broke $1 per watt barrier of
listening power, the Heathkit
W-7 was a major breakthrough
for audiophiles. Introduced in 1958, it
carried a list price of $54.95. The W-7
is not as common as the W-5, and when
it comes up at online auctions from time
to time, does not seem to attract the in-
terest of many bidders. Their loss!

FEATURES
Each amp used two EL34s as output
tubes, and a single 6AN8 tube as a pre-
amp/phase inverter. The power supply
used four silicon diodes in a voltage dou-
bler. At 8½˝ deep by 6⅛˝ high by 15˝
wide, the unit was quite compact. Early
models featured a satin gold enamel
chassis with black wrinkle cages; later
models had a black chassis and gold
cages. The final models were designated
“AA-91,” and production ended in the
early 60s1.
Specifications were quite impressive.
Frequency response was ±1.0dB from
6Hz to 30kHz at 0.25W, and ±0.5dB
f rom 20Hz to 20kHz, at maximum
rated output of 55W RMS2. Harmonic
distortion was about 0.1% at 1W, rising
to 0.5% at 20W and 1.5% at 55W. (A
2% total harmonic distortion is accept-
able in music reproduction.) Hum and
noise was 80dB below 55W. The amp
utilized 25.5dB of negative feedback
and featured 4.8-16Ω outputs as well as
70.7V line output. It also featured a vari-
able damping factor switch on the front
panel, enabling the selection of either
maximum (20) or unity (one)2.
The output transformer in this unit is
huge, much larger than the Mark 3 Dy-
FIGURE 1: Heathkit W-7 circuit. naco amp that used 6550s, and is prob-
ably the reason this amp sounds so clean.

28 May 2007

Brown-2786-3.indd 28 3/22/2007 3:39:37 PM

O GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO

OPERATION allows you to balance the current be- signed to 495V.

The circuit of the W-7 is simple and tween the tubes. This lets you use less- As I have mentioned in other articles,
straightforward (Fig. 1). The signal is than-perfectly-matched output tubes. this philosophy was much different than
fed through a gain control and 0.1µF The high voltage power is supplied that of Dynaco. In most Dynaco power
coupling capacitor to the control grid by a voltage doubler from the 180V amps, if the voltage used was 500V, then
of the 6AN8. The amplified signal is AC tap of the power transformer. The the capacitors were rated at 525V, and
coupled through an RC network to the 150µF and 200µF electrolytic capaci- multicap cans were used to further cut
grid of the triode section of the 6AN8, tors are each rated at 300V, and in series costs and save chassis space.
which is used as a split-load phase in- would be equivalent to 75µF at 600V, Heathkit comments: “The circuitry of
verter. and 100µF at 600V. This is a nice safety the W-7 was carefully engineered to en-
The signal at the cathode of this stage margin, because the power supply is de- sure an unusually high degree of stability
follows the grid, while the plate voltage
swings in the opposite direction. No PHOTO 2: The original unit.
amplification occurs at this stage. The
two out-of-phase signals are coupled
through 0.25µF coupling capacitors to
the control grids of the EL34s. The out-
put stage is operated in Class AB1; the
screen grids of the output tubes are con-
nected to taps on the primary of the
specially designed output transformer.
A unique and elaborate bias supply
and balancing circuit is incorporated in
the W-72. This ensures accurate bal-
ance of the output tube’s plate current.
Control R30 adjusts the bias voltage and
is set to yield about –38V on the control
grids of the EL34s, while control R38

May 2007 29

Brown-2786-3.indd 29 3/22/2007 3:39:42 PM

 GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO G

PHOTO 3: Capacitor replacements.

at both low and high frequencies which
may be verified by observing the 0.25W
frequency response curve—Graph A2.
The smooth rolloffs below 15 cycles and
above 50 kc are indications of the un-
usually wide stability margins. . . The ex-
cellent power supply regulation and low
impedance of the phase-splitter mini-
mize grid current effects as the rated
power is approached and exceeded. This
results in less than 1% IM distortion at
the rated 55W; the overload above 55W
is gradual, and clipping is symmetrical.”2
Another unique feature of this amp is
a sophisticated damping control. Heath’s
explanation of the utility of this control is
interesting: “Damping factor is defined
as the ratio of load resistance to the in-
ternal resistance of the amplifier. . . It has
been found that speaker systems which
have a high degree of acoustical damping
may be over damped when used with an
amplifier having a high damping factor,
with a resulting loss of bass efficiency.
On the other hand, too low a damping
factor will result in boomy or one note
bass which is undesirable”2.
Speaker systems of this era were often
efficient ported systems, or horns, but
newer “acoustical suspension” systems
were starting to appear on the market.
This may be the reason variable damp-
ing was provided.
Dynaco equipment provided an eco-
nomical entry into high fidelity. But
when comparing the Dynaco Mark
IIIs or IVs to the Heathkit W-7s, you
will find that the Heathkit betters the
Dynaco hands down. While the sound
is somewhat similar, the Heathkit has
more bass “kick,” and the treble sounds
cleaner.
Another difference involves the trans-
formers: Dynaco power transformers
run quite warm, and after several hours
of use are uncomfortable to touch. The
Heathkits run cool, and after several
hours are only warm to the touch. Side-
by-side comparison of the Dynaco and
the Heathkit shows where the “big iron”
is; while the power transformers are
similar in size, the output transformer of
the Heathkit is half again the size of the
Dynaco’s. While there is nothing wrong
with the Dynaco transformers (I have
many stock and custom amps that use
them), I would rather get my hands on a
W-7 output than Dynaco!
REPLACEMENTS
The unit as I received it is shown in
Photo 2. As with my other rebuilds, I
usually start with the power supply. In
this unit, all the diodes in the voltage
doubler were fine, but the electrolytic
capacitors were leaky. I was able to find
some 100µF 450V caps to replace the
100µF 300V versions in the first section.
The original caps were about 1½˝ in
diameter and about 4˝ tall. The new ones
are about one-tenth this size (Photo 3). I
removed the originals and mounted a
piece of fiberglass board in the cutout

�����������������
����������������������������������������
�����������������

������������������������������������������������������������������������
�����������������������������������������������������������������������������
��������������������������������������������������������������������������
���������������������������������������������������������������������������
������������������������������������������������������������������������
�����������������������������������������������������������������

���������������������������������������������������������������������
�����������������������������������������������������������������������
����������������������������������������
���������������������������������

30 May 2007

Brown-2786-3.indd 30 3/22/2007 3:39:47 PM

O GLASS AUDIO • GLASS AUDIO
and glued the new snap-mount ones to
this board.
I was also fortunate to find four
330µF 350V electrolytic caps to replace
the 200µF 300V originals. These were
short, but sized to fit through the origi-
nal chassis holes, and I mounted them
using some chassis mounts (Photo 3).
Also shown in this photo is the original
EL34 mount plate, which was extremely
rusty. I wire-brushed and painted it.
Note in Photo 1 the vented cages my
friend Larry built for me to cover the
new power supply capacitors. (Though
these cages were not necessary for op-
eration, he believes they give the amps
a finished look.) I replaced the selenium
diode in the bias supply with a silicon
diode, and also replaced the electrolytic
capacitors in the bias supply, using 100µF
160V units for all (Photo 4). I replaced
the remaining electrolytic caps feeding
the phase inverter and preamp sections.
I then powered up the first amp with
the Variac, and did some voltage mea-
surements. Everything was in spec ex-
cept voltage to the input grid of one
of the EL34s. This type of problem is
Brown-2786-3.indd 31
• GLASS AUDIO • GLASS AUDIO
usually due to a leaky coupling cap, as
was true in this case. I powered down
the amp, and after waiting for the power
supply to discharge, replaced all the cou-
pling caps with new polypropylene. (My
previous articles have covered capacitors
and my thoughts on them, so I will not
repeat here except to say I do not believe
in spending huge amounts of money on
“premium” caps. Use what you like.)
While you are under the chassis, use
your ohmmeter to measure the value of
all the resistor values. You will generally
find several that have drifted off value,
and this will affect the operation of your
PHOTO 4: Diode and electrolytic capacitor
replacements.
• GLASS AUDIO
finished product. The 100K (R15 and
R10) resistors that feed the EL34s should
be very closely matched. Also carefully
check the cathode resistors on the pre-
amp section of the 6AN8 (Photo 5).
These amps use a “surgister” as a surge
suppressor. Located in the corner of the
chassis near the fuse holder, it is about
PHOTO 5: checking the amp's innards
May 2007 31
3/22/2007 3:39:56 PM
 GLASS AUDIO • GLASS AUDIO
1½˝ long and looks like a small tube
with wire wrapped around it. On the
back side of the tube is a bimetallic strip,
which closes the contacts as the wire
heats up, acting essentially as a 100Ω
nichrome resistor that generates heat
and closes the contacts, bypassing itself
(Photo 6). I replaced this with a surge
suppressor (sold for 50 cents by Elec-
tronic Goldmine), which works well for
suppressing the turn-on surge of a tube
amp with silicon rectifiers and cold fila-
ments.
PHOTO 6:
Heathkit’s
surgister.
One of the interesting things I found
with my pair of amps was that one was
completely hand-wired as outlined in
the manual copy I had purchased. The
other amp had a factory-wrapped wiring
harness. I have not been able to find ad-
ditional information on this, but it was a
very professional-looking harness, much
like those seen in vintage military radios.
I contacted the manual copy source and
asked whether they had any information
PHOTO 7: Adjustment controls.
32 May 2007
Brown-2786-3.indd 32
• GLASS AUDIO • GLASS AUDIO
on a W-7 manual with a factory-made
wiring harness, but they had not seen
any other W-7 manuals. This may also
be a factory-wired unit.
UP AND RUNNING
The tubes used in this amp are very
common and readily available. NOS
6AN8s are quite reasonable. The choice
of available EL34s provides some out-
standing representatives. I often use
Svetlana EL34s, finding them very du-
rable, and producing a nice clean crisp
sound.
One of the nice things about the W-
7 is that it does not need a matched
set of output tubes, because the balance
adjustment allows you to balance the
current to the output tubes (so you can
even use vintage unmatched tubes). The
manual directs you to rotate the balance
control to mid position and connect a
voltmeter to the meter jacks on the back
of the amp (the bias control should be
fully counterclockwise and no inputs
or speakers should be connected at this
time), then turn on the amp and allow
it to warm up for several minutes. You
can see the two adjustment controls in
Photo 7.
Adjust the balance control to get a
zero reading on the voltmeter, continue
to adjust to the lowest range of your
meter, and readjust until it is zero. Once
this is done, remove one of the leads of
your voltmeter and connect it to the bot-
tom terminal below the 70V output on
• GLASS AUDIO
the speaker strip. Adjust the meter for a
reading of 0.36V, measuring across the
6Ω cathode resistor. This corresponds to
60mA current per tube (switch the lead
between the meter jacks and observe the
voltage on each tube).
I dug around in my used tube supply
and found some Mullards, Tung Sols,
and new Sovteks to try, and was fasci-
nated by comparing the sound differ-
ences between the new and old tubes. I
liked the Tung Sols the best, followed
by the Svetlana, then the Mullards, and
finally the Sovteks. (Not very scientific,
just personal preference.)
These amps will drive just about any
reasonable speaker system, they run very
cool, and are not the least bit fatiguing.
Very easy listening! aX
[An additional skill I developed in this
rebuild was replication of cosmetics, includ-
ing logos and cabinet feet. These items are
unique, and replacements are extremely hard
to find. With materials available today, it is
easy and fun to reproduce your own parts,
giving your restoration an authentic appear-
ance (please see my upcoming article, “Cast-
ing Replacement Parts”).]
REFERENCES
1. Vacuum Tube Valley, Issue 2, Volume 1,
Fall 1995.
2. Heathkit Model W-7M Assembly Op-
eration Manual, W7FG Vintage Manuals,
www.w7fg.com, 918-333-3754.
3. Schematic.
PARTS LIST FOR W-7 REBUILD, PER UNIT
2 100-200µF 450V electrolytic (C13-14)
2 300-400µF 400V (or better) electrolytic (C17-18)
3 100µF 150V electrolytic (C15, C18, C19)
1 1N4005 silicon diode (replaces selenium)
1 25µF 50V electrolytic (C2)
1 25µF 450V electrolytic (C6)
1 0.1µF 400V coupling cap (C1)
1 0.25-0.27µF 600V electrolytic (C7-8)
1 surge suppressor, 2A minimum
(replaces R27 surgister)
PARTS SOURCES
Electronic Goldmine
www.goldmine-elec.com
Antique Electronics Supply
www.tubesandmore.com
Mouser Electronics
www.mouser.com
3/22/2007 3:40:02 PM
O GLASS AUDIO • GLASS AUDIO
REVIEW
Tube Imp Mini Tester
By Charles Hansen
PHOTO 1: Tube Imp in its carrying case.
T
he Tube Imp mini tube tester,
designed and built in the UK,
measures the steady-state current,
gain, and transconductance char-
acteristics of any B9A double-triode tube
with the standard ECC83/ECC88 foot-
print.
Operating at 12V AC from the in-
cluded external power supply, the Tube
Imp offers matching and parameter test-
ing of a range of commonly used double-
triode tubes to individual tube equipment
enthusiasts as well as retail outlets and
small OEM concerns.
Specifications:
HT voltage maximum: 200V
HT voltage adjustment: 0-200V
HT setting accuracy: Typically better
than ±2%
HT current: 10mA maximum
HT current limit LED: >12mA
HT impedance on gain setting: >1MΩ
Grid voltage range: 0 to -10V
Grid voltage setting accuracy: Typically
better than ±2%
Heater voltage: 6.3/12.6V DC, switchable
Heater current 350mA, 500mA for less
Hansen2805-2.indd 33
• GLASS AUDIO • GLASS AUDIO
than 5 minutes
Measurement accuracy:
Cathode current: ±2%
Transconductance: ±4%
Gain: (Anode impedance in kilohms)
±5%
Valves that can be tested:
ECC81/12AT7
ECC82/12AU7
ECC83/12AX7
ECC88/6DJ8WA/6922
ECC189/ECC803
INSIDE THE TUBE IMP
The Tube Imp comes in a nice carry-
ing case that includes the tester, 12V AC
power adapter, and the manual (Photo
1). I plugged one of my vintage Mullard
ECC83 tubes, err valves, into the Tube Imp
(Photo 2).
The tester is quite rugged, constructed
of red powder-coated heavy gauge steel.
The 2.1mm × 5.5mm AC power jack is lo-
cated on the rear panel. Photo 3 shows the
Tube Imp with the bottom cover removed.
All chassis components are mounted on
two PC boards with some hard wiring also
involved.
• GLASS AUDIO
British Audio Products/Moth Group
10 Dane Lane, Wilstead, Bedford,
Bedfordshire, UK MK45 3HT
www.britishaudio.co.uk or
www.tube-imp.co.uk
e-mail: info@mothgroup.com
++44(0)123 474 1152
Fax: ++44(0)123 474 2028
Price: £299 UK
Test unit dimensions:
Net weight with carrying case: 4.4 lbs (2kg)
Available online at http://store.securehosting.com/
stores/sh204131/shophome.php?itemprcd=tubeimp
The transformer at the left is a 120/120
to 12/12V AC step-down unit that is
wired in reverse; the external AC adapter
feeds 12V AC to the paralleled transform-
er secondaries, and this is stepped up to
240V AC via the primaries connected in
series. The AC primary voltage is recti-
fied and used to supply the adjustable HT
(plate) DC voltage. The smaller heatsink
just below the transformer sits on the HT
regulator MOSFET. The adjacent 22µF
450V electrolytic filters the HT DC volt-
age.
The large heatsink in the middle holds
the filament DC regulator, which is pow-
ered directly from the AC adapter. A
switch on the front panel allows you to se-
lect either 6.3V or 12.6V DC. The manual
advises starting 12.6V tubes in the 6.3V
position, then switching to 12.6V after a
few seconds to minimize the stress on the
filaments.
The 9-pin gold-plated ceramic tube
socket pins are just to the right of the
transformer. Pin 9 of the tube socket is
not connected. Four small trimpots visible
through holes in the PC boards allow fac-
tory adjustment of the Tube Imp.
PHOTO 2: Tube Imp testing Mullard ECC83.
PHOTO 3: Tube Imp interior view.
May 2007 33
3/22/2007 4:14:13 PM
 GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO G

TOPOLOGY panel digital meter. With the Test Selector

The HT supply consists of a source fol-
lower MOSFET. The Test Selector allows
you to switch between a voltage source and
a current source in order to provide the
optimum loads for transconductance and
gain testing. The MOSFET is protected
by a 12mA current limit circuit, which also
lights a red LED when the HT supply is in
current limit. A switch on the front panel
allows testing of either the A or B section
of the dual triodes. The Section-1/2 switch
automatically switches the grid and cathode
connections. The tube section not selected
is idled with a -15V grid bias.
When the Test Selector is in the gain (µ)
position, the HT MOSFET is switched
into current source mode with an equiva-
lent impedance of about 2MΩ. An AC sig-
nal is fed to the grid, and the anode voltage
is rectified and displayed on the front-panel
digital meter.
When the Test Selector is in the mA/V
(transconductance, or gm) position, the HT
is supplied by the MOSFET as a regulated
voltage. A small AC signal is fed to the
grid, and the resultant AC current is used
to calculate transconductance for the front-
You're building the
world's best amp
.... why should it
look the worst?
ezChassis TM
in the mA position, the drop across the low
value cathode resistor is used to measure
the cathode current.
The Tube Imp does not directly measure
plate resistance, but you can easily calculate
it by Rp = µ/gm. By plotting a series of
cathode current versus plate voltage at a
fixed grid voltage, you can produce a plate
curve for your own tube.
MEASUREMENTS
I measured the actual heater, grid, and plate
voltages at the tube socket as selected by the
front-panel control settings. I used a power
resistor decade box set for 18Ω (6.3V) and
36Ω (12.6V) to check the heater voltages
at 350mA. I didn’t use any grid resistor to
measure the grid voltage because my Fluke
DMM (digital multimeter) has a 1MΩ
input impedance. I used a 220k resistor (A
to K) to simulate the plate resistance and
also checked the current limit point using
a 10k 5W 1% load resistor, raising the HT
until the LED was lit.
The filament voltages measured 6.44V
DC and 12.38V DC at 350mA, or within
about 2% of nominal, with pin 4 being
positive. I measured the grid and HT volt-
ages with respect to the scale markings silk-
screened on the front panel of the chassis. I
centered each line on the scale at the center
of the notch on the adjustment knob.
The grid voltages were consistently lower
than the scale markings, from -20% on the
low end to -10% at the high end. The grid
FIGURE 1: Audiomatica Sofia plate
curves for Mullard ECC83.
needed to use Va = 200V DC rather than
the 250V DC specified in the RCA man-
ual due to the Va limitation of the Tube
Imp. I also took these same data points
from the plate curve data file I ran on the
Audiomatica Sofia (Fig. 1).
The results of these tests on section 1 of
my Mullard ECC83 are shown in Table 1
and Table 2. The Sofia displays µ, gm, and
Rp directly along with plate current Ia. The
Tube Imp measures gain (µ), mA/V (gm),
and cathode current (Ik), so I calculated Rp
from the formula Rp = µ/gm.
Table 1
Parameter Data Book Sofia Tube Imp
µ 100 95.2 72.4
gm (mA/V) 1.3 1.17 1.3
Rp kΩ 80 81.2 55.7 (calc)
Measurements, Mullard ECC83 section 1
RC-30 Class A: Va = 100V, Vg = -1V Ia = 0.5mA
Note 1: Rp calculated from Rp = µ/gm for Tube Imp
Table 2

is the answer voltage at the maximum CW (clockwise)

• Punched holes for inputs,valves,
power supply, speaker terminals
• With 30+ labels (inputs/outputs)
• Strong construction. 17" Front
• Black painted chassis & screws
• Hundreds of ways to assemble
• Also heatsinks, handles & plugs
• Internet or Fax order
www.designbuildlisten.com
rotation read -9.2V DC. Perhaps this could
be easily fixed with an internal trimpot ad-
justment or a tweak of the knob set screw.
The grid voltage of the idled tube section
had -11V DC rather than the specified
-15V DC.
The HT voltage settings were quite ac-
curate, varying only 1.6% above 20V on
the scale. The HT current limit LED just
began to glow at 12.6mA and was fully on
at 13mA.
Finally, with the correction factors in
hand for grid voltage, I checked the gain
and transconductance using the Mullard
ECC83 for the two Class A amplifier con-
ditions specified in the RCA Receiving
Tube Manual RC-30. For the Tube Imp,
I set the specified plate voltage and then
adjusted the grid voltage to produce the
specified cathode current. Note that for
the second Class A amplifier condition I
Parameter
µ
Data Book
100
Sofia
98.1
Tube Imp
72.1
gm (mA/V) 1.6 1.64 1.2
Rp kΩ 62.5 59.8 58.1 (calc)
Measurements, Mullard ECC83 section 1
RC-30 Class A: Va = 200V, Vg = -2V Ia = 1.2mA
Note 1: Rp calculated from Rp = µ/gm for Tube Imp
Note 2: Va held to 200V due to Va limit of Tube Imp
CONCLUSION
The Tube Imp consistently understated
the gain by about 26% in comparison to
the Sofia. There is some discussion of this
in the Tube Imp manual. The gm was 11%
high in the first test and 27% low in the
second test. Again, there is a discussion
of calculating the theoretical value of true
transconductance in the manual.
I repeated the Tube Imp tests by setting
the plate and grid voltages at the designated
values and accepting whatever cathode cur-
rent resulted from these settings. The µ and
gm results were essentially the same.
I believe the Tube Imp would be quite

34 May 2007

Hansen2805-2.indd 34 3/22/2007 4:14:15 PM

O GLASS AUDIO • GLASS AUDIO
valuable in matching tubes where the abso-
lute values of µ are less important than the
comparative values.
MANUFACTURER’S RESPONSE:
Thank you for the opportunity to comment
on the in-depth review of the mini TT.
The reviewer gives a fair assessment of the
mini TT and of its target market. There are,
however, a couple of points I need to com-
ment upon.
The grid circuit has a source impedance
of 1MΩ at DC, so reading the voltage at the
grid pin with a DVM (typically of 10MΩ
input impedance) would cause a drop of
around 10% from the actual set value, as
found by the reviewer. The safest place to
measure the grid voltage with a meter of
less than infinite input impedance is at the
wiper of the grid voltage pot, although not
accessible without taking the back off!
The gain readings of the ECC83 mea-
sured by the reviewer also need to be ad-
dressed. If we understand correctly the
Sophia calculates gain, from measure-
ments of the transconductance and anode
impedances. The mini TT cannot reach
these calculated values for gain. It measures
Hansen2805-2.indd 35
• GLASS AUDIO • GLASS AUDIO • GLASS AUDIO
gain directly, feeding a signal to the grid spect to their current source behavior. After
and measuring the amplified signal at the much measuring of different samples and
anode/plate. This requires a high imped- scouring datasheets, we found that unfor-
ance current source for the anode load. The tunately the ST IRF730—which we had
impedance of the current source will appear sourced—is probably one of the worst cur-
in parallel with the anode impedance, giving rent source MOSFETs there is, although
a slightly lower reading. the specs are all the same! Every other man-
For the mini TT we originally specified ufacturer’s IRF730 does better!
an IRF730 MOSFET, configured to act as a We have now found a good replacement
voltage or current source. The circuit is quite with a (dv/di) drain impedance of 240kΩ
simple and relies completely on the MOS- at 10mA (rather than 15kΩ of the ST
FET’s high impedance behavior in current IRF730), and have modified the mini TT
source mode to work effectively. MOSFETs test rig and test procedure to test for this
are known to work as an almost perfect cur- parameter.
rent source (theoretically), when the gate to With the new MOSFET fitted we get
source voltage is held constant. This simple the following results for gain, from a ran-
circuit works well; however, we were disap- dom selection of tubes:
pointed by the gain measurement results
Device Gain Cathode current Anode volts
found during the review of the mini TT. Aged Mullard 12AX7 90.5 2.08mA 180V
Checking the production unit, we found Sovtek 12AX7WXT 127.1 1.94mA 180V
that the current source impedance was lower Gold Dragon E83CC 94.8 2.88 mA 180V
Telefunken E88CC 34.8 5.00 mA 80V
than we were expecting, at about 200kΩ at Mullard 12AT7 64.4 10.00 mA 150V
typical ECC83 cathode currents. This in Brimar 12BH7 19.4 10.00 mA 100V
conjunction with the ≈60kΩ anode imped-
ance of the ECC83 results in the reduced Best regards,
gain reading seen by the reviewer.
On further investigation it transpires that Hamilton Cleare
not all “IRF730s” are created equal with re- Tube IMP aX
May 2007 35
3/22/2007 4:14:18 PM
 GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO G

An Improved Method of Finding

Vacuum Tube Model Parameters
Designers and others who work with tubes will find this program
handy to determine VT models.

By Bill Elliot

I
first became interested in vacu-
um tube models when I needed
good triode and pentode models
to use with my SPICE circuit
simulator to simulate an audio power
amplifier I was designing. Searching
the Internet, I discovered the vacuum
tube model work of Norman Koren1.
I then began to produce some model
netlist equations based on Koren’s
phenomenological triode equations
using Texas Instruments’ Derive 6
math software.
At first, I used trial-and-error
methods to find the parameter val-

Table 1.

36 May 2007

Elliot-2773-3.indd 36 3/22/2007 4:02:59 PM

O GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO

ues until they matched five chosen points on the plate

curves. Despite the tediousness, I managed to produce a
fair number of netlists for the tube types that interested
me.
I have been using Derive since its early DOS days
and have found it to be reliable, easy to use, and, above
all, it does everything I need and more. It is also very
inexpensive—about $115. I know other math programs
are very good, but they are way out of my price range.
I suggest anyone interested in Derive 6 read the article
on the Internet titled “Derive 6: Far too good for just
students.”2
Because of the tedious work and resulting inac-
curacies, I decided I needed to find a better way and
wondered whether I could use Derive 6 to calculate the
values of the parameters.

SOFTWARE SOLUTION
Examining Koren’s triode equations, it would seem that
the five unknown parameters can only be found using
trial-and-error methods or some sort of program using
iterative techniques.

Koren’s Triode Equations:

Ip = (E1X/kG1)*(1 + sgn(E1))

where:

E1 = (Ep/kp)*log(1 + exp(kp*(1/u
+ EG/√(kVB + Ep2))))

Upon closer examination of Koren’s equations, I saw

that there are really two separate sets of parameters.
The first set of two, kG1 and X, are independent of the
second set and are quite easily found using the zero bias
line of the triode plate curves. The second set of three,
kp, u, and kVB, are more difficult to find, but can be
directly calculated using algebraic and numerical pro-
gramming techniques.
After numerous attempts, I wrote three very simple
Derive 6 programs that directly calculate triode, pen-
tode, and diode model equation parameters based on
Koren’s triode equations.
The triode and pentode programs run in 15–60 sec-
onds on my computer, depending on tube type and data
point selection. Some tube types and data point selec-
tions may run on and on without giving a result. You
may even get negative or complex numbers for results.
In these cases, check your data entry numbers first. If
the numbers are OK, then try moving data points B1,
B2, and C to higher current levels. I cannot promise
that these programs will work on any tube type, but I
think they will work on most.

May 2007 37

Elliot-2773-3.indd 37 3/22/2007 4:03:01 PM

 GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO G

Table 2.

38 May 2007

Elliot-2773-3.indd 38 3/22/2007 4:03:06 PM

O GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO

The Pentode Program is shown in

Table 1 . All the programs plus the
step-by-step equation development are
available on my URL3 in Rich Text
Format (RTF) and Derive 6 format
(DFW ). You can download and read
RTF files with Windows Wordpad. If
you have Derive 6, you can download
the Derive files and run them on your
computer.
The pentode equations for a pen-
tode netlist are shown in Table 2. Nor-
mally, you would use only equations 3
and 4 in the netlist because a triode
can be obtained by connecting a plate
and screen together. The pentode-tri-
ode connection equations are used to
1/3 page ad for Glass Audio Insert
check the triode curves using the De-
rive 6 curve plotter. aX
All fonts are Helvetica
REFERENCES
Contact: Phil Marchand
1. www.normankoren.com/Audio/ Phone (585) 423 0462
Tubemodspice_article.html.
2. www.scientific-computing.com/
FAX (585) 423 9375
scwmarapr04derive6.html.
3. www.knology.net/~billelliot.
phil@marchandelec.com

crossovers. Tube Electronic Crossovers. MOSFET SE Amp. Tube Phono Preamp.Solid State

Phono Preamp. Moving Coil Transformers. SP/DIF Phono Options. Custom Solutions. We
Passive line Level

s
l
r
e e
c
v
Tube Electronic Crossovers . Solid State Electronic Crossovers.

t
o

r
o
s

n
s

i
o

c
r
C

Marchand Electronics Inc.

PO Box 18099, Rochester, NY 14618
(585) 423 0462
info@marchandelec.com www.marchandelec.com
can add notch filters, baffle step, etc..... Tube Electronic Crossoverrs.Kits and Assembled

May 2007 39

Elliot-2773-3.indd 39 3/22/2007 4:03:10 PM

 GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO G

AUDIO AID

Greek Gifts:
Amplifiers from Athens

A
lex Arion writes that his only
regret is discovering audioX-
press in “. . . the autumn of
[his] life.” We’re sorry too,
since his odyssey in audio construc-
tion appears to be exceptionally in-
teresting. He began life, including
his audio activities and his engineer-
ing education, in Bucharest, Roma-
nia. Today, Alex lives and works in
Athens, Greece, where he plies a
busy trade in exotic amplifiers, both
tubed and solid-state under the firm
name of Arco Sound. He makes ex-

PHOTO 1: Single-ended, stereo 3.5W/channel using 2A3s

for output, driven by 6SL7s.

PHOTO 2: Stereo 2 × 40W power amp using 211 outputs driv-

en by 6L6 in triode mode, plus 6SN7 inputs and an interstage
transformer.

40 May 2007

Darion2721.indd 40 3/22/2007 3:49:03 PM

O GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO • GLASS AUDIO

tensive use of wood and wood veneers You may reach Mr. Arion at Arco
in his productions. We include a few Sound, 13 Koritsas Str. 14561 Kifissia,
samples.---ETD Athens, Greece. 001 310 210 8083008.

PHOTO 3: Two 180W/

channel stereo power
amps using six EL34s,
driven by 6SN7s and
ECC81s and classically
supported, of course.

PHOTO 4: A hybrid
two-channel, 240W
amplifier with an
integral tube pre-
amp.

May 2007 41

Darion2721.indd 41 3/22/2007 3:49:14 PM


GLASS AUDIO • GLASS AUDIO
Cakepan Chassis
Have your cake and chassis, too!
By Ken Bird
B
uilding tube amplifiers has al-
ways required a good sturdy
chassis to handle the heav y
transformers and necessar y
drilling and hole punching for tube
sockets and other components. The so-
lution was always a Bud or Hammond
chassis from the local parts jobber or
mail-order house. Electronic parts job-
bers have gone the way of the steam
PHOTO 1: Pans ready to be put to a good use.
42 May 2007
Bird-2768-2.indd 42
• GLASS AUDIO • GLASS AUDIO
locomotive, while mail order is more
expensive.
I found an ideal source of aluminum
chassis at www.wilton.com, which is a sup-
plier of baking accessories for cake bak-
ing professionals and home based cake
decorators. Their line of square bak-
ing pans, which are perfect for even the
largest power amplifier chassis, is sold
through local dealers found in most cit-
PHOTO 2: Amp chassis.
F
• GLASS AUDIO
ies, and the variety of sizes will meet your
most demanding chassis needs. These
units are made of heavy ¹⁄₁₆ aluminum
construction but punch and drill easily,
and have flanges for attaching bottom
covers or feet if desired. After cleaning
and applying a coat of auto primer, you
can paint them or polish them to their
natural finish. Photo 1 shows two differ-
ent sizes of the square pans in their “raw”
format. Photo 2 shows the amp chassis.
The black chassis (left) is an 8˝ × 8˝ × 2˝
pan with a dual channel SE amp using a
6L6 in each output stage, and the “Red
Hot Amp” (right) is built on a 12˝ × 12˝
× 3˝ and is a dual channel PP amp using
a dual triode 6BX7 in the output stage.
The pans offer a lot of real estate for the
builder, and soon they will be baking up
some good sound in my workshop. aX
3/22/2007 3:19:57 PM
sound solutions By Ed Simon
Speech Intelligibility – Part 3
The author concludes his series with a look at problems—and solutions—
associated with reverb, resonances, and reflections.
PHOTO 3: Treatment too thin.
A
classic method of measuring re-
verb is to clap two boards to-
gether and use a stopwatch to
time how long the sound takes
to decay. The better method of the day
was to sound an organ pipe and time
its decay. More sophisticated measures,
of course, evolved as technologies ad-
vanced. Much of the early work was at
500 or 512Hz, which is still the design
center frequency for music; for speech
2000Hz is preferable. The modern ver-
sion of this is a small sound system,
such as the “Sound Strobe” (aX 3/06).
Now you can just twiddle a knob to get
the impulse you need for listening tests.
REVERB PROBLEMS
Most of the computer-based audio mea-
surement systems will also give a reverb
time number or graph. They typically
derive the reverb time from other kinds
of test signals. When you use computer-
based measurement systems to do this,
you encounter that important warning:
SimonPt3-2761-2.indd 45
“To err is human, to really foul things up
requires a computer!” (Anon.)
The reverberant field decay is usually
presented as a small matrix or graphs
of the actual frequency and the time it
takes to decay by 60dB. A typical reverb
time meter such as the Goldline GL60
model uses octave bands from 125Hz-
4000Hz. It is different in that it mea-
sures only the first 20dB of decay, then
presents the number as though it were
60dB of decay.
That is because when I designed it, I
was aiming for 20dB of signal-to-noise
ratio to ensure good speech intelligi-
bility. I knew then that it was the first
drop in the sound that affected this. The
60dB number is still the classic refer-
ence, so that was my way of presenting
useful data to the less experienced user.
The meter also has the “bad” habit of
not settling on a single reverb number
if the field does not have a uniform rate
of decay.
The model of speech I use is a burst
of sound followed by silence for twice as
long. I had a recording of Andrew Car-
negie speaking at the turn of the century
in a large reverberant room. He spoke
slowly by today’s standards—about three
syllables per second. A fast talker today
will speak six. For my model I use a rec-
tangular gate that is open five times per
second. Thus the on time is ¹⁄₁₅ of a sec-
ond and the off time is ²⁄₁₅ of a second.
Others who have studied actual speech
use different time intervals. As wrong as
they may be, my numbers work for me!
You can use a meter such as the Gold-
line to determine not only a matrix of
reverb numbers, but also how reverb
changes in a given room. The manual
(not mine) suggests you use a pink noise
source with the meter. The sound di-
rectly from the source drops off as you
move farther away. The reverberant field
in a room will continue to build until the
absorption and leakage out of the room
equal the rate at which energy is added.
This is the steady-state condition.
Turning off the noise while you are
far away from the sound source will trig-
ger the meter and give you a classic re-
verb time number. Just remember, this
is for only the first 20dB of the curve.
There are some cases in which this does
not happen. The meter bounces all over
and does not settle on a single number,
which is a strong indication that the
room is not linear and there may be spe-
cific echoes that interfere with speech
intelligibility. The room is most likely
not very music friendly either.
If you produce a sound impulse right
next to the meter, the number it com-
putes will be much lower than if you
repeat the experiment with the meter
across the room. If you use a Sound
Strobe or other impulse generator
audioXpress May 2007 45
3/22/2007 4:20:54 PM
 through an existing or portable sound
system and move away with the reverb
time meter, at some point the meter or
your calibrated ears will indicate a reverb
time of 1.5 seconds at 2000Hz. Note
this distance, which is most likely the
same distance you would get from the
speech articulation tests for an accept-
able loss distance.
Assuming that you speak at five syl-
lables per second and you require about
5dB of decay before the next syllable to
get acceptable understanding, then you
would like to see a decay rate of 25dB in
two-thirds of a second. The other third
of a second is your sounding time. This
gives you a classic good speech reverb
time of 1.44 seconds, basically what the
meter just showed you.
CRITICAL DISTANCE
As you move away from the constant
sound source, the level directly from the
source decreases. The level of the rever-
berant field for a steady source will reach
a constant level throughout a room. The
distance at which the direct field equals
the reverberant field is called the critical
distance, which you can measure with a
constant noise source and a sound level
meter.
Start at the farthest point and walk
toward the sound source. You are at the
critical distance when the level rises
by 3dB. With some practice you can
get close with just your ears and not-

FIGURE 1: Pre-absorption reverb. FIGURE 2: Post absorption reverb.

� � � � � � �������������������������
����������������

��������������������������������
by John Linsley Hood
NEW REPRINT OF THE 2ND EDITION. Starting with lucid descriptions of the parts involved
(resistors, capacitors, inductors, tubes, and transistors), the author then begins joining them together
to form circuits offering the reader an accessible approach to linear electronics design. Focusing mainly
on audio amplifiers, the book features two of Linsley Hood’s classic designs which have become popular
with the DIY crowd. Includes excellent coverage of amplifiers,
oscillators, power supplies, filters, and feedback using clear examples
and easily solved equations. Reprinted 2006, 1998, 348pp., 7 3/8” × 9
5/8”, softbound, ISBN 1-882580-49-4. Sh. wt: 2 lbs.
������������������������������ ������
“For many years ‘do-it-yourself’ audio enthusiasts have asked me,
‘What’s the best book for learning amplifier design?’…Thanks to
Edward T. Dell [and Audio Amateur Inc.], this valuable work is once
again in print, and now I have an easy answer to the ‘What’s the best
book?’ question.” — Nelson Pass
� ���������������������������������������������������������������������
� ������������������������������������������������������������������������
� �������������������������������������������������������������������������
� ������������������������������������������������������������������

46 audioXpress 5/07 www.audioXpress .com

SimonPt3-2761-2.indd 46 3/22/2007 4:21:02 PM

 ing where the volume be-
gins to increase. The rise
means the sound level has
doubled. One-half is the
direct field, the other half
is the reverberant field.
In a really bad room this
distance should be about
the same as your speech
articulation loss distance.
That is why when
you move away from the
pulsed noise the reverb
time meter shows agree-
ment with the articulation
tests. This is the maxi-
mum distance for unaid-
ed acceptable speech in
the room. This will be at
least the critical distance.
At the critical distance
the signal-to-noise ratio
is by definition 0dB for
continuous sounds.
Because we speak in
bursts, the reverberant
field will not build to the
same level as for continu-
ous sound. With the model I use one-
third on, two-thirds off—the ratio at the
critical distance is at least 4.8dB signal
to 0dB noise. This would be the worst
case with an almost infinite reverb time.
If you know the reverb time is less
than infinite, you can calculate how
much decay will occur between syllables
and add that to the signal-to-noise ratio.
You also know that the talkers’ speech
drops off with distance squared. You can
now calculate where the signal-to-noise
level drops to 5dB in your room. You
may wish to raise this goal depending on
the expected use of the room.
The critical distance can change if you
approach from different directions when
you measure the critical distance with
a pulsed noise source. It is possible that
this is due to room geometry or con-
struction producing a nonuniform re-
verberant field. The most common cause
is that the sound source has directional
characteristics.
You can use this ability of loudspeak-
ers to be more directional than a plain
human talking to your advantage. The
idea is to put sound on the audience only
where it will be absorbed. That way you
SimonPt3-2761-2.indd 47
Altitude 3500
Integrated Valve Amplifier
The Altitude 3500 is the
culmination of many years of
development and refinement by
Fountek Electronics. Component
quality and circuit design have been
carefully chosen to deliver the highest
quality of musical reproduction, with
accurate and emotional delivery of the
PHOTO 4: sound stage.
Round trap The Altitude 3500 Integrated valve
in use. amplifier features only the best
available components with the shortest
and cleanest signal paths possible with
direct coupling used on the input stages
to improve signal transit response.
do not feed energy into the reverberant
field. The directional loudspeakers will
increase the effective speech intelligibil-
ity distance. It is possible to deliberately
design very directional loudspeakers.
One theory states that you can make
a sound system overcome bad acous-
tics. As some ancients believed, the gods Specifications:
made us vain so they can enjoy the fall! Channels 2-channel
You can make some improvements— Inputs CD/Tuner/Aux1/Aux2
horrible to bad or bad to okay, for ex- Output Power 32WPC@1KHz /8ohm
ample. But horrible to good will just not Output Class AB1 Push Pull
happen with loudspeakers alone. You Output 4 or 8 ohm speaker load
must improve the room. T.H.D. less than 1%
One of humans’ annoying traits is our S/N Ration 89dB/A
preference for simple answers. Is the Valves 2 x 12AT7 Twin triode
room good or bad? Exactly what does it 2 x 12BH7 Twin triode
take to be good? In this case, you could 4 x EL34B Pentode
hire one of my acoustician friends, pay Chassis heavy gauge brushed
them lots of money, follow their advice, aluminum on all sides
pay attention to the details, and you will Dimensions 350mm*190mm*320mm
have a good room. But for those of you Net Weight 18Kgs (39.7lbs)
who want to know more. . . Conformity CE Rating
Introductory Price $1150.00
A CHURCH STORY
A local large church had a problem with
speech not being well understood after a
recent building renovation. The church,
as far as I could tell, was originally built
audioXpress May 2007 47
3/26/2007 9:06:18 AM
 with an interior finish of plaster and
stone. When this proved to be far too
reverberant, the interior plaster ceilings
were sprayed with a sound-absorbing
plaster and the main ceiling covered with
absorbing panels. This left the room ad-
equate to good for speech but very un-
pleasant for music.
When 30 years had passed and no
one left could remember the original
problems, the room was renovated to
improve the musicality of the room. No
acoustician was consulted. The designer
plastered over the absorbent tile ceil-
ing and painted the absorbent plaster.
Now the room had lots of reverb and a
few folks incorrectly thought it was im-
proved, but it was now good for neither
speech nor music. In addition, the room
was much noisier.
The noise problem involved an organ
blower chamber isolated from the re-
modeled room by a masonry wall except
where it had been cut to accept a heat-
ing radiator. Only a piece of sheet metal
blocked the visible opening. Closing the
opening with a double layer of drywall
and sealing reduced the noise.
Many folks proposed installing large
and very expensive loudspeakers to proj-
ect speech the length of the room. Three
different systems were tried and none
proved adequate. I had a few ideas re-
garding the problem, so I contacted Sam
Berkow (Sia Acoustics-Jazz at Lincoln
Center, Hollywood Bowl, and Smaart
Software) to visit the church. He took
measurements using Smaart software
and determined the major problem was
that all the sound energy was not de-
caying uniformly. In fact, the energy at
the 250Hz octave band was decaying so
slowly that it was masking speech in the
room. In addition, he thought that the
curved roofs on the transepts (side seat-
ing areas) were the primary cause.
He suggested adding enough ab-
sorption to drop the reverb time of the

THE DECIBEL

www.Alltubetesters.com
Tube Tester Repair
Repair, Calibration, Upgrades
 35 + years experience
 All Makes and Models
 Circuit Upgrades/Improvements
 Testers sales
 Technical Support
 Bogey tubes available
 New Meters available
 Tester Refurbishing
6 month limited warranty on all work
performed including calibration.
Roger Kennedy
21143 Hawthorne Blvd., #354
Torrance, CA 90503 USA
www.alltubetesters.com
Email: alltubetesters@msn.com
PH:(310) 323-3281
T he decibel is one of those won-
derful bits of jargon that we use
to confuse the uninitiated. Please do
not read further unless you already
know the secret sign and password.
The base unit is the bel, which is the
logarithm base 10 of the ratio of the
sound being measured to the threshold
of hearing. Thus, if the sound energy is
ten times louder than our threshold of
hearing it is log(10/1)=1. If the sound
is 100 times louder, it is 2 bels, 1000
times louder is 3 bels, and so on. Just
count the number of zeroes after the 1.
You can hear over a range of greater
than 12 bels (120 decibels), which,
when written out, is 1,000,000,000,000
to 1. Writing 120dB is much easier.
The decibel becomes confusing
when you have, for instance, only twice
the energy. The math works out to 10
(to convert from bels to decibels) ×
logarithm (2/1). Use a calculator to
do the logarithm, which works out to
a two-fold increase, or 3.010299957. .
. decibels, which you can round off to
3. With ten times as much energy (1
bel), you get 10 decibels. If you double
the sounds’ energy again, it cannot be
20 decibels because that is 100 times
as loud.
You now know the secret of loga-
rithms: add the 3 from the doubling
to the 10 from the ten-fold increase to
determine that the energy increases by
13dB. Thus, 40 times as much energy
would be 3 + 3 + 10, or 16dB. Eighty
times, of course, is 3 + 3 + 3 + 10, or
19dB.
If you know that 20dB is 100 times
and 19dB is 80 times, then what is
1dB? Well, 100 divided by 80 is 1.25.
The log of 1.25 is .0969100. Ten
times that is close enough to 1 that
you should see how this works. The
precise answer is 10 to the exponent
(1dB/10 decibels to bel), which equals
1.258925412. . ., or 1.
Another consideration is the ac-
curacy to which you can measure. It is
fine measuring a voltage on a digital
meter that is accurate to four digits.
A change in barometric pressure of 1˝
will change some sound levels by about
.2dB. Wind and air movement also af-
fects the readings. That is why sound
pressure levels are rarely used with a
resolution of greater than 1.0dB.—ES

48 audioXpress 5/07 www.audioXpress .com

SimonPt3-2761-2.indd 48 3/22/2007 4:21:09 PM

 room to 1.5 seconds with a reasonable
crowd; in particular, putting absorbers
in the transepts. As soon as we added
two absorbers to a transept, it became
immediately clear that he had identi-
fied the source of the resonance. Treat-
ing both transepts and adding material
along the side walkway ceilings provided
the acoustic treatment needed to change
the room from very bad to good. In ad-
dition, because of the location of the
absorbers, the aesthetics of the church
remained unchanged. The moral of the
story—find out exactly what the prob-
lem is before trying to fix it.

RESONANCES
Any two parallel walls will bounce sound
back and forth. Most rooms have at least
three pairs of such surfaces. If the walls
are good reflectors, this will show up at
the frequencies which have a half wave-
length that is a multiple of the spacing
distance. There are other kinds of reso-
nators besides parallel surfaces.
It is easy to understand that a wind
instrument such as a flute uses the air
motion inside it to produce a tone.
Sound travels at a speed of about 1132´
per second. If you have the air bouncing
between two acoustic nonlinearities, it
will form a standing wave. Even random
air motions will excite the wave and add
energy to it as it bounces back and forth.
The other common form of resonator
is like a whistle in which the air rolls
around in a circle and reinforces the re-
sulting sound wave.
All rooms have many resonances. So
the concern involves resonances that
occur at particular frequencies to which
we are especially sensitive or are very
strong.
Finding out you have a problem reso-
nance is simple. It will show up as a
frequency or band of frequencies that do
not decay as rapidly as the surrounding
bands of frequencies. You can measure
this with test equipment or use a musical
instrument such as a piano or organ. You
just listen for notes that hang around too
long. This is sometimes called “color.”
Figuring out which part of the room
causes the problem resonance is a bit
harder. Once you are able to produce
or measure the resonance on demand,
you can move a large absorber around
until you find out where it has the most

audioXpress May 2007 49

SimonPt3-2761-2.indd 49 3/22/2007 4:21:12 PM

 effect. Sometimes you will need to use
many absorbers and cover most of a wall
to produce a noticeable effect. The dif-
ficult times occur when you have a bell
tower that may be causing problems and
you want to try material on the ceiling
120´ above you!
This may sound difficult, and it is.
Finding out the cause of a problem is
more art than science. Although there
are some instruments that help, they are
not common or readily available. The
good news is that from my experience,
resonances that cause a real problem
with speech intelligibility are present
in about 10% of the rooms I examine.
Many small resonances at the right fre-
quencies can enhance music.
The common mistake in trying to fix
a resonance is to place foam on the walls
until the problem goes away. If the foam
is not thick enough to absorb the fre-
quencies of interest, it is not doing what
you want and is just killing the mids and
highs. This also colors the sound. If you
want to use movable panels, be sure they
are really big and thick.
Once you have identified the area that
causes the resonance, you can add just
enough absorption to damp it. This is
different than trying to add absorption
to get a desired reverb time. Because of
the nature of resonance, the absorption
has twice as much or more effect than it
would if it were just reverb.
A second approach would be to
change the surface. Building out, adding
angled surfaces, or features of interest
such as reliefs, or even using shelving to
break up the characteristics that allow a
single frequency to build are often used.
If you can’t absorb it, bounce it around
enough so that it spreads out and dis-
sipates.
REFLECTIONS
Of course, by now you are thinking, this
is nuts. “How can this be right if you
don’t need to use a computer or any lu-
dicrously expensive test gear?”
That’s right! With simple measure-
ments and calculations, you can achieve
good reverb times in a room with non-
uniform absorption and nonlinear reverb
time curve. The only equipment you
need is a pair of ears, paper, and a pencil.
A calculator helps.
Of course, sometimes this is not
50 audioXpress 5/07
SimonPt3-2761-2.indd 50
enough. You have until now been look-
ing at statistical averages of the sound
field. Another problem area to address
involves echoes or sharp distinct reflec-
tions that are not muddled in the statis-
tical average.
If you add bookcases to the side and
front walls, a small room would sound
much better even with the same reverb
time. The decay curve would now be
much smoother due to the diffusion you
have added.
Helmut Haas (1958) determined that
a single echo can reinforce speech intel-
ligibility or destroy it. He concluded that
if the repetition of a sound comes within
30ms, you perceive it as part of the orig-
inal sound, perhaps adding a bit of spa-
ciousness or volume to the original. The
second sound must be 10dB louder than
the first to be perceived at all.
When the second sound comes after
this period, it disturbs the speech. The
effect is greater for faster speech. Some
forms of late energy can even destroy
speech intelligibility. Increasing the vol-
ume of the speech does not change this
effect because the echo also increases.
The direction of the echo does not have
a great effect on sounds that originate
from the front.
As a demonstration, I ask two singers
to sing a capella the same piece while
walking away from each other. When
they get to around 30–35´ apart, they
find they cannot sing together. There is
a slight variance due to the tempo of the
piece.
If you have ever tried Sound Strobe
on a loudspeaker that has a path length
difference at the crossover frequency,
under some conditions you can hear a
distinct double click even though the
path length difference may be less than a
foot. Haas used speech for his model; it
does not encompass dissimilar sounds or
frequency effects. It does work great for
speech reinforcement design if you do
not act silly trying to stretch the results
of his work.
RAY TRACING
When you try to predict what the echo
pattern of the sound will be like, you use
the ray tracing model of sound disper-
sion. Simply put, you draw arrows point-
ing out from the sound source. When
the rays strike a wall, ceiling, or other
www.audioXpress .com
surface, you draw them bouncing on-
ward as though they hit a mirror. Con-
tinue this path until the rays (arrows)
reach where the listener(s) should be.
You can then assume that the length
of the ray is directly proportional to the
time it takes the sound impulse to reach
the listener. It is useful to plot the first
direct ray, the second rays that bounce
once off a wall or ceiling, and the third
rays that bounce off two surfaces. This
is usually enough to get an idea of what
the ray response will be like. You can do
more if you like, but the task becomes
almost impossible for five to seven
bounces.
You can also scale the amplitude of
the sound rays by multiplying the in-
verse square law value by the absorp-
tion coefficient of the surface materials.
Sometimes the reflecting surface is small
enough or angled so that it reflects only
part of the ray and spreads out part of it.
The wavelength or frequency of interest
determines what size of non-uniformity
becomes effective.
Sometimes the room is too small to
use the ray simplification at the lower
frequencies. I stop using the ray model
if the room is too small to allow seven
wavelengths of the frequency of interest.
This allows you to produce a graph of
amplitude versus time.
If you want to cheat a bit, you can also
plot the reverb time curve on the same
scale. Because the reverb time is usually
based on steady-state energy excitation
and the ray tracing assumes a unit im-
pulse, you may wish to move the reverb
time curve down a bit on the X axis. If
you plot enough ray bounces, you will
find it merges into a giant jumble that
about matches the classic reverb time
and that is where you can merge the
curves. Computer programs are useful
for this, but you can still do this by hand!
All the unit spikes up to 30ms are
combined as though they are one great
sound. Some folks combine the rest
of the mess for a second value. When
the first value is 5–10dB greater than
the second, speech intelligibility will
be good. The exact value is influenced
by your exact methodology, the noise
level, a few other semi-mythical issues,
and the intended audience.
When the combination curve shows a
nice drop of about 5–7dB for the initial
3/22/2007 4:21:12 PM
 break and a smooth linear tail without
any major jaggies, you probably have a
good room for music. When the reflec-
tions coming in past 30ms are stronger Easy Ordering In Nanoseconds
than desired, you can add absorption to
the reflecting surfaces. One mistake is to
treat one sidewall and not the other. The
reflections off the sidewalls contribute to
the spaciousness of a room. If sidewall
treatment is needed, use it sparingly and
split it up onto both walls and stagger it.
Sometimes this will decrease the overall
reverb time too much.
The trick, then, is to use a diffusing
surface—anything from columns along
a wall, moldings, or even strips of absor-
bent material. This will spread out the
reflection, decreasing its level and adding • The ONLY New Catalog Every 90 Days
to the reverberant tail. • NEWEST Products & Technologies
Sometimes you wish to reinforce the • Over 755,000 Products Online
early energy of the first 30ms, which • More Than 325 Manufacturers
tends to improve the intimacy of a • No Minimum Order
space. This is where a low ceiling angle • Same-day Shipping
The Newest Products
to maximize projection to the audience
mouser.com (800) 346-6873 for Your Newest Designs
works wonders. I once was involved in a
project with acoustician David Klepper,
who designed a glass reflector system The NEWEST Semiconductors | Passives | Interconnects | Power | Electromechanical | Test, Tools & Supplies
that was on a remote control. At the first 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.
use Mimi Lerner was the soloist.
I did not like the sound and asked her
what she thought. She was diplomatic
Mouser AudioXpress 4-1-07 indd 1 3/12/07 4:09:13 PM
but agreed to sing a very brief bit (con-
serving her voice for the second perfor-
mance). I lowered the reflectors by about
3´ and heard an amazing difference as
the path length locked into the magic
30´. The second performance was much
better!
In the 1940s studies of movie theaters
surprisingly concluded that some of the
“good” examples had worse reproduction
system frequency response than some
of the “bad” examples. An excellent ex-
ample that the nature of the sound re-
flections, reverb, and noise can be more
important than just “fidelity.”
One question that comes up from
time to time is, “If we know so much
these days, how come all the old concert
halls are good halls and many newer
halls are not as good?” The answer is
simple: When you build a bad concert
hall you either fix it or tear it down
and build another one! I hope you have
picked up enough useful information
to improve the acoustic spaces you care
about and will not need to resort to the
wrecking ball! aX

audioXpress May 2007 51

SimonPt3-2761-2.indd 51 3/22/2007 4:21:19 PM

 Classified
VENDORS
Audiophile components: JFETs, MOSFETs,
Tantalum, Caddock, Dale resistors,
Black Gate, ELNA Cerafine, SILMIC II,
TONEREX, Nichicon Fine Gold Muse,
KZ, Jensen Four-pole capacitors Stepped
attenuators, Teflon cables, Connectors
Custom designs for OEM customers
www.borbelyaudio.com
Selected BORBELY AUDIO kits in Japan:
http://homepage3.nifty.com/sk-audio/
52 audioXpress 5/07
Classified-Ad Index507.indd 52
Low cost custom power and output
transformers made by Trafomatic.
www.engineeringvista.com
Lots of audio information on our website.
Please visit WHAT’S NEW on our homepage,
www.tdl-tech.com (505-382-3173)
Yard Sale
Yard Sale
WANTED
Pioneer Elite cassette tape deck model CT-A9XBK.
Luther Wright
16901 Parksdale Road
Petersburg, VA 23805
“Yard Sale” is published in each issue of
aX. For guidelines on how subscribers
can publish their free ad, see our website.
www.audioXpress .com
WANTED
stereo gear old/new/quality, speakers,
amps, preamps, turntables, reel to reels,
vacuum tubes, tube testers, guitars, parts,
etc. 850-314-0321
Email: sonnysound@aol.com
What’s New
on the aX website?
Web-exclusive content:
Word Test files from Ed Simon’s “Speech
Intelligibility, Part 1” (aX 3/07).
“Modifying Mighty Mouse,” By Bob McIntyre
(aX 3/07).
Jesse W. Knight’s review of Sony CD/DVD
Player Model DVP-NS55P (aX 1/07).
Articles from past issues:
“A Phase Meter Calibrator,” By Charles
Hansen (aX 11/06).
Dan Ferguson’s review of Installer car stereo
installation software (aX 12/06).
“Low-Level Analog Switching,” By Dennis
Hoffman (aX 1/07).
“Grounding and System Interfacing,” By
Gary Galo (aX 1/07).
Dennis Colin’s review of The Art of Linear
Electronics (aX 1/07).
“How Lound Is Real?” By Larry Klein (aX
2/07).
“Transferring LPs to DVDs in High
Resolution,” By Victor Staggs (aX 2/07).
“The Rocky Mountain Audio Fest,” By Bob
Cordell (aX 2/07).
Dennis Colin’s review of High Performance
Loudspeakers Sixth Edition [plus table of
contents] (aX 2/07).
“A Prototyping System for Passive
Crossovers,” By Ramkumar Ramaswamy (aX
3/07).
David Milford’s review of PNF Audio Cables
(aX 3/07).
“A Low-Noise Measurement Preamp,” By
Dennis Colin (aX 4/07).
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!
3/22/2007 3:58:33 PM

review By Tom Perazella
Revue Du Son
Test CD Number 17
T
est CDs fill a basic need to eval-
uate system and room perfor-
mance. They come in various
f lavors—some with lots of test
signals, others with mostly music. Each
has its place.
A CD with properly designed and re-
corded/produced signals can save a lot
of money otherwise spent on specialized
test equipment. However, at the end of
the day it’s all about music, so picking
recordings that can reveal your system’s
performance in various categories is in-
valuable.
There is no shortage of musical test
CDs. I have quite a few and use them
extensively. Some are commercially avail-
able, others were custom made.
A test CD should have a wide variety
of music to test frequency response, tran-
sient response, dynamic range, freedom
from distortion, and spatial characteris-
tics. By the way, it should be listenable. If
the music sounds like test tones, I’d rather
use test tones.
Test CD No. 17 was produced by the
French magazine Revue Du Son. Usually I
first read the track descriptions to get an
idea of what to expect. In this case, I not
only had liner notes, but also an article in
the January/February 2005 Revue Du Son
that gave more complete descriptions.
54 audioXpress 5/07
Perazella2803.indd 54
PHOTO 1: Revue du Son.
Great! Well, maybe not. Everything is
in French, which I can’t read. I’ve done
blind equipment tests, but this is my first
blind CD test. I would like to have pro-
vided insight from the magazine descrip-
tions, but not being able to read them
certainly minimized any listener bias.
My response to a quick run-through
was quite positive. The CD has 22 tracks
and runs a total of 71:26. Deciphering
what I could from the notes written by
Jean Hiraga, the disc was mastered at Le
Studio Acoustique de Passavant and is
www.audioXpress .com
the 17th in a series that started in October
1993. The music is highlighted by female
and male vocals (soloists and groups), old
woodwind instruments, strings, piano,
orchestra, organ, bells, large drums, and
gongs. Tracks 16, 21, and 22 are not mu-
sical, but include applause, street sounds,
and the killer helicopter.
Track 1: The Magic of
Kasarova—“Deh! Tu, Bell’anima”
This track opens with very smooth wood-
winds and strings. The vocalist is mezzo-
soprano Vesselina Kasarova, whose dy-
namic range is very large. Your speakers
must be able to handle very sotto-voce
inflections all the way to huge peaks.
My Tenma sound level meter mea-
sured from 61-to-88dBsplA. If you have
any midrange or tweeter peaks or dis-
tortions, this recording will sound very
harsh. On a good system it will sound
hugely powerful and engrossing.
Track 2: Kari Bremnes: Svarta Bjorn—
“Sangen om fyret ved Tornehamn”
Close-miking can make recordings sound
in-your-face, especially with mikes that
have a presence peak. However, done
properly it can impart a sense of imme-
diacy and detail. This track walks the fine
line in between.
The voice is very clean and detailed
without being overly bright. It is a great
test piece because there are some high-
level low-frequency drums in the back-
ground that center around 40Hz and hit
peaks of 102dBsplC, while the average is
78dBsplA. Anyone for an IM distortion
test? A small two-way speaker will be
hard-pressed to handle the high low-fre-
quency levels while simultaneously keep-
ing the voices clean.
Track 3: Kari Bremnes:
Svarta Bjorn—“Byssan lull”
Strong, but not very deep-bass percus-
sion opens this track. Peaks of around
95dBsplC centered around 50Hz provide
a backdrop for more smooth vocals. The
vocals are clear and separate from the
background, resulting in a very intimate
feeling. There are also some delicately
struck bells that float above the vocals.
At the end of the cut, Kari’s voice
evolves to what sounds like quiet synthe-
sized surf. This is a good test of separa-
tion of a clean vocal from a strong per-
3/22/2007 4:19:24 PM
 cussive background.
Track 4: Trésors Moyen-Age: Musique
sacrée et Profane ensemble Sequentia—
“Bonne amourette me tient gai”
This is a short vocal track that was re-
corded in a hall with good acoustics. A
solo voice provides the chance to hear
its definition along with hall reflections.
Other voices join, adding to the complex-
ity and testing the ability of your system
to separate the voices and their harmon-
ics not only from each other, but from
the hall reverberance. Each voice should
sound rich and spacious, yet distinct.
Track 5: Trésors Moyen-Age: Musique
sacrée et Profane ensemble Peceval—
“Aisso es viadera”
Ancient instruments add distinctive fla-
vors to this track. The opening wood-
winds have a strong reedy sound. A fe-
male vocal enters and is accompanied by
background percussion. The separation of
the voice from the percussion is excellent.
Track 6: Trésors Moyen-Age: Musique
sacrée et Profane ensemble Peceval—
“Ja nulls om pres”
The opening strings are quite distinct
with well-defined plucking and delicate
harmonics. A male voice with realistic
tonal balance and clear vibrato sepa-
rates cleanly from the accompaniment.
Sibilants are well controlled. A flute en-
ters solo and is later added to the mix.
During the vocals it should be audible,
though in the background.
Track 7: J. Brahms: Sonate Nr 1 op 1
pour piano, M.J. Jude—“Scherzo”
This is an OK piano recording that shows
good spatial and tonal balance. Harmon-
ics are well presented. However, I have
heard recordings with a much greater
sense of dynamics. It’s a good recording,
but would not be my primary reference.
Track 8: Stefano Bollani:
Smat-smat—“La Vita Intensa”
This is a good track to show how de-
manding a piano can be to reproduce. It
is quite dynamic with excellent hammer
sounds and natural resonances from the
lower strings. Being able to separate the
chords’ harmonics while sounding tonally
balanced is a definite challenge. All this
happens in an atmosphere of transients
Perazella2803.indd 55
and rhythms that can easily lead to a
sense of musical confusion if your system
cannot resolve all that is going on. It may
not be everyone’s choice in music, but it
is a great test piece.
Track 9: Chopin: Ballades et Scherzos,
A. Rubinstein—“Scherzo nr 2 op 31,
Siminear”
This piano piece is all about dynamics,
transients, high- and low-level detail, and
room acoustics. There is a good sense
of the instrument while still having hall
ambience that is not overly reverberant.
Attacks are good. Low-level detail is ex-
cellent.
Track 10: Orchestre de Contrebasses:
CH. Genet—“Bass, Bass, Bass, Bass,
Bass & Bass”
As you might gather from the title, this is
all about bass. And bass there is, plucked,
bowed, and struck—string bass heaven!
The detail is extraordinary. Harmonics
extend to all frequency ranges. This is
clean bass with enough higher frequen-
cies to add spice. It is a great test of the
ability to separate transients from the
background bass lines.
I ran it through my Behringer
DEQ2496 RTA and found some of the
bass line centered around 28Hz. There is
quite a bit of energy all across the audio
band with an interesting change from
bass to midrange at one point. At a little
over three minutes into the track several
voices appear that move around the stage.
It’s an entertaining change of pace. The
piece ends with a sharp impact.
Track 11: Vivaldi: Concerto nr 2, La
Stravaganza, C. Todorovski—“Largo”
As with a piano, it is often very difficult
to make an organ recording that really
sounds right. This one does justice to the
instrument—a complex, powerful, subtle
combination of tones that separates the
organ from all other instruments. Bal-
ance is very good with a sense of position
that is neither too close nor too far. The
hall sounds believable.
Upper registers are clean with good
separation of upper-midrange and treble,
especially in the presence of bass. The
notes seem to float above the ambience.
This is not a fireworks piece, but a believ-
able simulation of being with the instru-
ment.
Track 12: Beethoven: Concerto pour
violon, Heifetz/Munch—“Larghetto”
Strings, beautiful strings. How they can
make your spirit soar. Though if repro-
duced badly, they can send you running
for the door.
Here they are done well. The strings
are very smooth without sounding soggy.
Ambience adds to the realism. The vio-
lin is clean and well defined. Massed
strings sound like massed strings, not
some fur-ball of strident sound. Indi-
vidual harmonics are easily heard. The
violin’s intonation, bow sounds, and vi-
brato are excellent and well separated
from background instruments. Plucked
strings are believable.
There is a lot going on in this record-
ing. It is an excellent test of definition in
which the result is a composite rich in
texture rather than a lot of impressive,
individually dramatic sounds. Detail is
important to producing a realistic image
even if not accompanied by striking dy-
namics. This is an excellent piece to see
how well your system performs. It’s ei-
ther right or the magic will be gone.
Quad speaker owners will love this track.
Track 13: Rachmaninoff: Concertos
Nr 3 pour piano, Horowitz/Ormandy
—“Allegro ma non tanto”
I’m torn over this track. As I review my
notes, I see the phrase “can’t put my fin-
ger on the problem.” There is something
about this recording that I don’t like.
The sound of the solo piano is good
and clean, except for weakness in the
lower register. However, when listen-
ing over speakers or headphones, there
seems to be some glare when it is going
full tilt with other instruments. Dynamic
range also does not seem to be as wide
as other piano recordings. This is the
weakest track as a reference recording. If
not in such strong company as the other
tracks, it might be considered very good,
but it pales in comparison.
Track 14: CD Test Nr 7: “Presentation
des Grandes Orgues de Saint Eustache
à Paris par Jean Guillou”
This is another great organ recording,
not only because of the instrument, but
also because of the hall and the ability
to hear a male voice so clearly and with
great intonation. Ambience is excellent.
The different voices of the organ are
audioXpress May 2007 55
3/22/2007 4:19:26 PM
 explained with short excerpts following
each spoken section.
There is first-rate definition in a highly
reverberant space, displaying good high-
frequency clarity and tremolo. The low
pedals provide a nice rumble. The article
mentions a stop capable of producing a
16Hz fundamental, but if it is there, it is
at a low level.
Never fear, this track is not about earth
shaking bass, but rather about definition
and ambience. The house shaker comes
in track 22. Enjoy this one for its realism.
Track 15: CD Test Nr 10: “clochettes”
The bells are an extreme test of the high-
frequency system performance. It is ripe
with harmonics that can easily smear to-
gether into high-frequency oatmeal. The
strikes produce strong transients and the
decay sounds natural with gobs of har-
monics and beat frequencies. Good luck
if you have tweeter or crossover problems.
Track 16: CD Test Nr 10:
“applaudissements”
The first of the non-musical tracks, this
one features almost three minutes of ap-
plause in a large hall, with an ability to
identify individuals yelling at different
distances from the mikes. It is a very
complex sound that will lose its realism
if the playback system is lacking in any
critical aspect.
At a little over a minute into the track
there is so much going on that the sound
almost produces pink noise. Howev-
er, through it all, the handclaps have
a sharp attack, the voices are real and
distinct, the hall size palpable, and the
enthusiasm of the audience apparent.
This should not sound overly sharp or
muted. Pay special attention to how the
voices separate from the applause.
Track 17: Percussions XX—CD Test
Nr 14: “Appendice alla perfezione”
Delicate bells introduce this track. While
the level is low, the attack and decay
should be quite clear. The intensity in-
creases and the sound, especially the at-
tacks, should become louder without
becoming harsh.
With the harder strikes, note the in-
crease in the ratio of the initial impact to
the decay. There is extended harmonic
content. You should also hear movement
across the stage. As the track ends, there
56 audioXpress 5/0
5/077
Perazella2803.indd 56
are complex sounds at high volume. Re-
gardless of the volume, they should al-
ways sound like bells.
Track 18: Percussions XX—CD Test
Nr 14: “Towards”
Drums, cymbals, gongs! A friend once
said I would stop while walking down
the street to listen to a garbage can fall
down a flight of stairs. I guess that is be-
cause highly dynamic, wide-bandwidth
sounds can tell a lot about what is right
and wrong with your system.
This track is a great broadband tran-
sient test. I played it back at average lev-
els around 100dBspl and peaks as high
as 111dBspl, where everything must be
right. Your speakers need a lot of linear
displacement at all frequencies, you need
enough clean power to avoid clipping
the peaks, and a good radiation pattern
will help it sound realistic. Use caution
when playing this track because it can
damage wimpy speakers, and your hear-
ing if you play it for extended periods.
Track 19: CD Test Nr 10: “Gongs”
This is the gong show. They can really
sound nasty if you have problems any-
where in the frequency band. The tran-
sients are brutal. There is energy over
much of the audio spectrum. Harmonics
are abundant.
Check out the levels of the strikes at
23 and 32 seconds. Also check out the
shimmering effects later in the track. If
your system has good low-level resolu-
tion, you will hear voices close to the end.
Track 20: CD Test Nr 10:
“Grosse caisse”
Forget the dog, beware of the drum! This
track is brutal. There is a huge drum-
strike impact. At my listening position,
I recorded a peak level of 115dBsplC. If
your speakers survive, listen for the detail
in the decay before the next hit.
Track 21: CD Test Nr 12:
“Bruits de rue et de moto”
This is one of only a few recordings I
have of an outdoor venue that sounds
real. It was made on a street and the
sounds are amazingly real in terms of the
image width and depth. The superim-
position of the voices with the footsteps
sounds uncannily real.
At around 23 seconds there is some
www.audioXpress .com
kind of mechanical sound I cannot iden-
tify that is clearly separate from the rest
of the foot and voice traffic. The place-
ment of different voices is excellent. At
around 1:06 a motorcycle starts up in
the background. It comes into the fore-
ground moving to the left and across to
the right, with low and high frequencies
as the rider blips the throttle.
As with the earlier applause track, the
sound of footsteps is the delineating fac-
tor in the realism. The heel attacks are
crisp without being overly sharp, and you
can hear the slap of the soles as they hit
pavement. The natural voices float above
the foot sounds providing the frosting on
the cake. This is a very good test track.
Track 22: CD Test Nr 8:
“hélicoptère, decollage”
OK, you have been warned about other
tracks, but if you have not paid attention
before, do so now. This track is brutal. If
you value your speakers, please use re-
straint when setting the volume. I suggest
extra caution if you have a small two-way
speaker because it just won’t cut it. In
fact, it will be a total letdown; you will
probably hear nothing partway through
the cut except gobs of distortion and the
sound of your speakers self-destructing.
This track sneaks up on you because
it starts out pretty normal. You hear the
sounds of a turbine-powered helicop-
ter preparing to lift off. The compressor
whine is realistic, plus the hiss character-
istic of gobs of air being drawn into a jet
engine. The hiss becomes dominated by
the sound of the rotors.
If you cannot reproduce frequencies
below 20Hz, you will miss the action.
There are very high levels at low frequen-
cies. To determine just how low, I ran the
signal from my preamp to my Tektronix
oscilloscope and recorded the waveform.
Figure 1 shows a large low-frequency
component around 19Hz.
This is serious low-frequency energy.
Even with my two very long-excursion
15˝ woofers, I started to get ugly noises
at 107dBspl. Death to small speakers!
I needed to reduce the peak levels to
around 105dBspl to stop the really seri-
ous uglies.
After running the tests, I consulted
the calibration chart for my sound level
meter and saw that the response at 20Hz
was 5dB down. Therefore, the sustained
3/26/2007 9:03:54 AM
 level was really 110dBspl. The sub I had
built into my former house would have
been a better match for this track. It had
eight 12˝ drivers (see “True Bass,” 5/96
Speaker Builder). That is the kind of vol-
ume displacement necessary to reproduce
serious bass at high levels.
At around 2:55 there are some high
frequencies that are either the rotor tips
exceeding Mach 1 or the digital record-
ing instruments exceeding 0dBFS. In any
case, this track is an interesting high-fre-
quencies test if you are familiar with com-
pressor whine. It is certainly a source of
very high levels of low frequencies from
the rotors. Small woofers need not apply.
SUMMARY
The real question is did it pass as a useful
musical test disk? The answer is definite-
ly yes. There are enough different musi-
cal and non-musical but natural sounds
to provide a wide range of stress tests for
any system. More important, there are
many pieces that have sufficient content
quality to test not only whether it sounds
real, but real enough to elicit emotional
responses. Highly recommended! aX

FIGURE 1:
A large low-frequency
component around
19Hz.

Appendix
Home System:
CD player Sony 707ESD
Preamp Custom-built, based on AD797 op amps and
BUF03 output buffers
EQ Behringer DEQ2496
Crossover Behringer DCX2496 (Linkwitz-Riley 48dB/
octave slopes at 71 and 303Hz)
Amps—high One two-channel Crown Macro Reference amp
Amps—mid Three—AudioSource Amp3
Amps—low Two—KG-5230 “plate” amp
Speakers—mid/high Two—each consists of a Bohlender-Graebener
RD75 in a custom baffle. See “On Angel’s
Wings” in the January 2001 audioXpress.
Speakers—mid bass Two—each consists of six Peerless 831727 10˝
woofers in a custom baffle. See “A Dipole
Midbass” in the June 2004 audioXpress.
Speakers—sub bass Two—each consists of a Dayton 15˝ DVC
woofer in a 5ft3 sealed box
Portable System:
CD player Panasonic SL-CT520
Headphone Amp Headroom Total Airhead amp
Headphones Etymotic Research ER4S
Test Equipment:
Sound level meter Tenma 72-860, calibrated against an
ACO 7012 microphone
RTA Behringer DEQ2496
Oscilloscope Tektronix TDS210

audioXpress May 2007 57

Perazella2803.indd 57 3/22/2007 4:19:33 PM

 XPRESSMail
LESS WAR, MORE PEACE
(OR VALVET AMPLIFIER)
Due to a slight mis-communication, I
didn’t see the final proofs of my article
“Amplifier War and Peace” about the
Valvet™ hybrid amplifier before print-
ing (March ’07, p. 14). Thus, potential
builders should note that C7 is shown
in Fig. 1 as 47µF, but should be 0.47µF
(plastic film).
Also, if built as shown, there is a nasty
(but not fatal) “turnon thump” through
the speakers. This is due to the HT
turn-on upsetting the MOSFET bias-
ing momentarily. The simple fix is to
add another step to the power switch, so
heaters, then HT, then MOSFETs are
progressively turned on.
Simon Brown
simon@designbuildlisten.com
OSCILLATOR PARTS
Reading Dennis Colin’s “A Wide-Range
Audio Sweep Oscillator” (aX 2/07, p.
26), I found that he is, as usual, making
it difficult to obtain parts! Harris Semi-
conductor no longer exists as a separate
entity; they were eaten by Intersil, long
ago.
Allied Electronics used to carry this
part (under Harris), but no longer carry
Intersil. A google search turns up loads
of suppliers—in Hong Kong, and other
exotic places—bad news if you live here.
I did find one stocking distributor in the
US: www.1sourcecomponents.com. I have
requested a quote for small quantities.
We’ll see what happens.
John Nickerson
johnnickerson@verizon.net
Dennis Colin responds:
My sincerest apology; I hadn’t known that
the CA3280AE wasn’t available. Digi-Key
has some Intersil parts, but not this one.
What a shame; this part is extremely versa-
tile. I have six Harris units and one original
RCA; I’d better preserve them in a time
capsule!
Please let aX know if your search is suc-
cessful. If not, and there’s no substitute,
I could try to design a discrete transistor
equivalent, or change the oscillating circuit
58 audioXpress 5/07
xpressmail-1.indd 58
to use analog multipliers. That could reduce
the frequency range because of voltage
offsets.
Thanks for your letter; in future projects
I’ll make sure all parts are available.
DVD/SACD PLAYERS
I decided to document my recent experi-
ences with SACD/DVD player modifi-
cations. I first pursued this back in 2003
when I procured a Philips SACD1000
unit. I then had a lot of trouble obtain-
ing technical information (tech service
manual) until I wrote your magazine,
detailing my experiences.
With help from your readers, I even-
tually did obtain a CD-ROM contain-
ing the information I wanted. However,
that’s “ancient history” right now—I was
diverted from that path by involvement
in other projects and put it aside. My
experience now is with a pair of cheap
players—Sony DVP-NS500V and the
oppo-digital DV-970 universal disc
player.
Back in 2003, I had also (easily and
quickly thanks to Sony’s professional
services) procured the tech manual for
the Sony unit. At that time, I did an
ultra simple (and simple-minded) re-
vision, installing “pull-down” resistors
(4.32k, RN55D) at the outputs of the
main stereo output op amp and also by-
passing the output coupling caps (47µF/
16V) with small polypropylene caps (the
well-known 50V Panasonic units avail-
able from Digi-Key). The audio output
channels in the Sony unit are all handled
by the cheap and ubiquitous 4558 dual
op amp.
The results were modest but appar-
ent, and listening quality did improve
a bit. I put that aside for a while until I
remembered how useful the AD827 had
been in my previous projects. This fast-
settling video op amp works very well as
a substitute for other unity-gain stable
devices in common audio circuits—it
has always outperformed the garden va-
riety chips I have encountered in other
pieces of consumer gear (line-level ap-
plications).
Its one drawback is that it is not avail-
able in an 8-pin surface-mount package.
www.audioXpress .com
Luckily a friend came to the rescue and
mentioned the AD828, which is stable
down to a gain of 2 (no problem in these
players I am referring to) and is other-
wise reasonably similar in performance
to the AD827. It is also cheaper and is
available in an 8-pin SOIC package
(surface-mount).
The temptation was too great. I tried
it out in the Sony DVP-S500V (IC202
position) and the results were what I’d
been dreaming about for SACD play-
back (and the 24/96 Classic Records
DAD discs—which are DVD-V in two-
channel stereo).
All of the problems with the SACD
and the DAD were eliminated, including
the new “Living Stereo” hybrid SACD
series, the older Sony Classical SACD-
only discs dating back to 1999/2000 and
later, and the original circa-1998 Classic
Records DAD discs. By comparison,
the playback results are thrilling. My
resources are very limited, so I don’t have
access to more than about 20 discs with
which to demonstrate these results; oth-
ers will need to go further afield than I
can right now.
How do these hi-res discs compare to
CD? (My current CD reference is any
transport used with the heavily modified
Philips DAC960. . . work done on it in
1993 and in 1998—all documented in
the essential POOGE book by Jung and
Galo.) They are decisively superior—in
every way. And this is with an el-cheapo
(but now modified with the AD828)
SONY DVP-NS500V DVD/SACD/
CD player. It’s possibly the most dra-
matic transformation I’ve experienced in
several years.
My suspicions that SACD and DVD-
V (at 24/96) formats were capable of
outstanding results are now confirmed
(and almost nobody I know is aware of
this!). I am surprised and pleased that
my suspicions were correct. However,
beware of producers who issued SACD
discs made from CD masters ( John
Atkinson—and others?—have exposed
these charlatans in the past. . . and I can
tell you now, the differences are amaz-
ingly obvious).
I also performed this “trick” on my
3/22/2007 3:50:07 PM
 oppo-digital DV-970 player (the main ing condition and configured them so inently in this story because she was
stereo outputs only—and yes, it also uses the children could use them with ease. always very supportive of his involve-
the 4558 chip too—perhaps everything He gave away dozens and dozens of ment in his many and varied projects.
does). . . and with some trepidation, be- entire computer systems, complete with She supplied the grille material for his
cause I have no documentation for it. monitors, mouse devices, printers, and D’Appolito speakers and actually helped
Luckily a scope and the old CBS CD-1 speakers, to children of all backgrounds. him stretch the material over and around
test disc come in handy. The sonic im- I recall one very poor Mexican family the speakers. She also provided the fore-
provements are similar, and are also ob- who had previously been living in a van. going biographical information on Matt.
vious when playing-back a movie DVD They’d finally gotten a home to live in, I miss my friend more than words can
(PCM - stereo option). This is a good and the little girl who was the recipient say. He will be forever in my mind and
all-round improvement, with good re- of Matt’s generosity of heart was recom- in my heart.
sults for every application. mended as an outstanding student to Angel Luis Rivera
Matt to be a computer recipient by her angelluisrivera@excite.com
R. K. LeBeck. Jr. school teacher. Another fortunate ben-
Kirkland, Wash. eficiary of Matt’s kindness was a little CD/VINYL TESTING
Mexican-American girl whose mother After scanning the Feb. '07 issue of au-
MATT HAMILTON—IN my wife had seen through four pregnan- dioXpress, especially the CD/vinyl sys-
MEMORIAM cies (my wife is an Advanced Registered tem in Indonesia (p. 31), it occurs to me
In January 1994, I placed an ad in Speak- Nurse Practitioner—Certified Nurse- that an overlooked issue in the entire
er Builder and Audio Amateur. It brought Midwife), and she has mentored the controversy may be related to whatever
Matt Hamilton and me together. I wrote entire family. Matt was happy to give her occurred in mastering.
about my relationship with Matt as my a computer system, too. He never asked I roundly criticize the current practice
mentor in the May 2004 issue of au- all the dozens and dozens of children of severe peak limiting on many CD
dioXpress (“Letters,” p. 68). and their families for anything more releases in a misguided attempt to make
It is with great sadness and a profound than thanks. something with a brickwall peak ceiling
sense of personal loss that I report Matt’s His wife, Wanita, says of Matthew: play louder. I thought this practice came
passing on October 29, 2006, in Braden- “He was very capable—no job was im- about with the adoption of CDs as our
ton, Fla., after a prolonged illness. possible for him.” Wanita figures prom- primary music transport media.
William Matthew Hamilton was born
on May 21, 1925, in Audra, W.Va. While
growing up, he and a friend built radios,
and never lost his love for electronics.
He graduated from Buckhannon Upshur
High School, W. Va., in 1943; attended
Fenn College in Cleveland, Ohio, and
married Wanita Waugh in 1947. He
worked as a mechanic on both automo-
biles and buses until 1955.
Matt then worked for the Thew
Shovel Company in Lorain, Ohio, in
the engineering department until 1959,
when he started working for the Bur-
roughs Corporation. During the period
he worked for Burroughs, he assisted in
selling and installing many large com-
puter systems in Ohio, Michigan, Wis-
consin, Pennsylvania, California, and
also in Amsterdam, The Netherlands.
He returned to the US in 1971, retired
from Burroughs in 1986, and moved
from Pennsylvania to Bradenton, Fla.,
in 1988.
Sometime in the 1990s, he started
giving computers to academically moti-
vated, needy children. He accepted com-
puters and associated equipment from
individuals and companies, and made
sure the computers were in good work-

audioXpress May 2007 59

xpressmail-1.indd 59 3/22/2007 3:50:13 PM

 In the process of transferring some
vintage vinyl to CD, I was surprised
when I looked at the audio envelope of
the transferred disc on my audio work-
station and found it looked like tooth-
paste, as most contemporary CD releases
look.
My conjecture, at this point, is that
at least part of the differences heard in
CD/vinyl comparisons are differences
in mastering practice rather than dif-
ferences between the two methods of
delivery. It is folly to assume that the
master tapes are transferred verbatim
to either medium, because they aren’t.
The record companies have finally rec-
ognized the value of their back catalog
and are bringing more and more of their
old recordings back into circulation. But
when they do this, the recordings are
inevitably subjected to “mastering,” and
in the process they may be caressed or
abused.
I plan to explore this further by com-
paring the long-term envelopes of the
recordings between vinyl and CD releas-
es of the same material. I know that I
have an original Capitol Records release
of Sgt. Peppers, a British pressing of the
same, and the CD release of the same. It
will be interesting to compare the three.
Rick Chinn
Sammamish, Wash.
SAFETY FIRST
I would like to comment about certain
technical viewpoints expressed in the
article “Grounding and System Interfac-
ing” ( Jan. ’07, p. 26).
While I certainly agree with most of
Mr. Galo’s statements and those he at-
tributes to Mr. Whitlock, I must take
exception to one major issue: Attribu-
tions to Mr. Whitlock that “. . . earth
grounds are for lightning protection,”
and that “. . . . earth ground plays no role
in protecting people from electrocution.”
These seem to be contradictory state-
ments. If earth grounds are for lightning
protection, then they certainly play a role
in protecting people from electrocution
(from lightning and from a little-known
effect I describe later).
I can only assume that Mr. Whitlock
was only considering his example shown
in Fig. 1 on page 26, in which defective
equipment develops an internal short
60 audioXpress 5/07 www.audioXpress .com
xpressmail-1.indd 60
and, if the equipment’s chassis were to
be not grounded, the chassis would be
“hot” and thus dangerous. While light-
ning strikes occur very rarely in most of
the country, Mr. Whitlock’s figure shows
the components of a more menacing,
24-hours-a-day threat to equipment and
humans. This threat is the utility distri-
bution system itself; and this is where
the building ground, in my opinion, does
its “thing.”
The utility transformer does not have
perfect isolation, as depicted in the sche-
matic, between its primary (typically
4kV, 12kV, or more volts) and the build-
ing’s load (typically 120, 208, 240, or
480V). This lack of isolation “in the real
world” produces leakage currents that
are primarily due to capacitance effects
(proximity of windings and wires) and
resistance (dirty insulators, poor insu-
lation, and so on). While the leakage
current may be relatively small as far as
building loads are concerned, it can eas-
ily provide more than enough amperage
to electrocute someone. This current is
always there to varying degrees, depend-
ing upon transformer and wiring condi-
tion and age. Admittedly, the ground rod
(or any other grounding system) is not a
perfect connection to “ground” (it has
a low, but not zero resistance), and this
is one of the reasons that it is possible
to measure voltage differences between
grounds.
Stated another way, if the ground rod
(or ground system) were not present,
people would be electrocuted every day
in their homes, office buildings, or facto-
ries just by touching perfectly functional
equipment that has no internal faults.
Another point I’d like to make is that
utility companies ground their power
transformer secondaries in addition to
the customers’ grounds. This is due part-
ly to the fact that utility transformers
rarely feed just one customer. If they
had multiple customers, they would have
multiple grounds anyway (and could not
depend upon one customer to provide
an adequate ground for all of the other
customers). Additionally, there are times
when they might need to disconnect the
lines to customer(s), so they certainly
need to be concerned about the safety of
their own personnel and wiring.
Imagine an ungrounded transformer
with a 12kV primary and its secondary
3/22/2007 3:50:14 PM
 is not grounded. The utility’s wiring and
the customer’s wiring are rated at 600V.
Without a ground, the wiring is subject
to voltages (thought to be zero) but that
could be many kilovolts, maybe 12kV.
Over a prolonged time, even new wiring
would fail, and—more important than
this—personnel could be subjected to
the same voltage (very, very dangerous).
Ironically, new wiring would lead to a
more dangerous condition, because more
leakage current would be available.
With 50 years as an electrician, I
am personally aware of several danger-
ous—and lethal—instances of this type
of problem. In one case, a utility worker
was killed near a pumping plant I was
maintaining when he was working inside
a relatively small ground-mounted trans-
former feeding several houses nearby.
He was working on the secondary when
he received a “high-voltage” burn even
though the 12kV was not accessible.
Another instance occurred when an
electrician working at a pumping plant
received a call that the pumps were not
working in the facility and decided to
measure the voltage-to-ground of a
xpressmail-1.indd 61
480V service to the pump building. This
was a three-phase system with a 12kV
primary and 480V secondary (he was
not aware at the time that the secondary
was ungrounded). Normally, the voltage-
to-ground is 277V on each phase; but
this is only true in a grounded system.
The utility source was in trouble (appar-
ently one or two of the three transform-
ers' primary fuses had blown due to in-
sulation breakdown in the old secondary
wiring as described previously).
Even though the blown fuses would
not permit any pumps to run, that one
good fuse was allowing leakage current
to flow from the high to the low volt-
age connections. When the electrician
touched the leads of his voltmeter, which
was set to the 750V scale, the meter ex-
ploded in his hands and he was slightly
burned and very surprised and fright-
ened. If he had touched any electrical
connections in the building with any
part of his body—thinking that power
was off—he could have been killed by a
current that might not have been strong
enough to even light a light bulb.
As far as the article’s discussion of
good and bad ways to try to eliminate
noise in audio gear, I couldn’t agree
more. Three-prong to two-prong adapt-
ers (used as “ground lifters”) is a misun-
derstood, and thus, dangerous item to
deploy by persons testing or trying to
eliminate ground loops. They are used
because they are convenient; and alter-
natives may not be known or available at
the time. Perhaps their discontinued use
for these purposes, thanks to Mr. Galo’s
article with information from Mr. Whit-
lock, may prevent someone’s tragedy.
Gene Davis
gdavis219@hawaii.rr.com
CAR AUDIO
I thank you for publishing audioXpress,
without which a large part of my life
would be missing. I do, however, have a
few complaints.
Please, more modern topics and less of
the “good old tube amps.” I understand
why you run so many tube articles, and
empathize. But, may I suggest you take
even a brief look at the automotive amp
industry (repair, maintenance, building)?
audioXpress May 2007 61
3/22/2007 3:50:19 PM
 I am an automotive installer, but prefer
the permanence of “home” equipment/
installs. I’ve seen Class BD, Class D,
Class T, and so on each hit big in the
automotive field (and last), years before
home stereo even considered them. Also,
nearly all of the innovations for speakers,
especially subwoofers, have come from
auto needs/desires.
So please, do music a favor by getting
more young people involved in how all
this stuff works. (I’m 47 and have saved/
repaired massive quantities of amps,
EQs, subwoofers, and so on through
simple repairs and total rebuilds.) My
daily driven “grocery getter” van has
done 160dB at the dash.
Topics of most interest, I think, might
be:
1. The amp power converter/supplies.
2. Amp topologies and special needs in
autos.
3. Matching subs with amps. (None of
our customers seem to understand
that you don’t need 3000W RMS
and subs to match, when 300W with
efficient drivers will rattle the mirror
off the windshield.)
I’m not sure whether you’re aware—or
care—but it is interesting that in com-
petition concrete filled cars are nearing
190dB. Personally, I prefer sound quality.
Again, thank you for what you do.
Glenn Ray Dõrzök
Dimock, SD
NIGHT SCHOOL
I love your magazine. I also think there
are a few subscribers like myself who
love the hi-fi hobby, love to solder and
rebuild equipment, but may have needed
extra work in areas of mathematics. In
that respect many of your articles go
over my head. But one did not.
The February issue featured an article
by Jan Didden about a visit with some
Indonesian audiophiles. It was a really
great article! Keep those special reports
coming, they make your magazine more
adaptable to all audio nuts. In turn, I will
go back to night school to brush up on
decimals, fractions, and graphs. Thanks
for a great mag.
Mark Korda
mark.korda@verizon.net aX

������������������������������������
��������������������������������������
���������������������������������������������������������������������������
���������������������������������������������������������������������������������
��������������������������������������������������������������������������������
�������������������������������������������������������������������������
����������������������������������������������������������������������������
������������������������������������������������������������������������������
�����������������������������������������������������������������������������
����������������������������������
��������������������������������������������������������������������������������������

��������������������������������������������������������������������������

������������������������������������������������������������������������������������������������������������������
��������������������������������������������������������������������������������������������������������������
�������������������������������������������������������������������������������������

���������������������������
���������������������������������������������
���������������������������������������������������������������
����������������������������������������������������������

62 audioXpress 5/07 www.audioXpress .com

xpressmail-1.indd 62 3/22/2007 3:50:22 PM