ax Fresh From the Bench

Attenuators for
TEST & MEASUREMENT FOCUS

Measurement

In this article, our author discusses external attenuators and

compares the impact different attenuators’ thermal noise will
have on measurements of distortion. At low test voltages, the
added thermal noise of the attenuator is of primary concern.
At high test voltages, the heat dissipation in the external
attenuator is of primary concern. For this reason, it can be
handy to have multiple attenuators of similar attenuation but
By with different impedances on hand.
Jonathan Novick
(United States)

External attenuators are an effective and choices for power amplifier measurements. These
cost effective way to measure signals above the attenuators are asymmetrical in nature. In other
maximum input voltage of an audio analyzer. words, they have a specific input and output
However, commercial off-the-shelf attenuators direction. Symmetric topologies, such as T-pads and
are usually not a good choice in this application Pi-pads and their variants are primarily used in RF
as they are typically designed for low-voltage use communications, where impedance matching and
only. There are several trade-offs to consider when bi-direction signal flow are of greater importance
designing attenuators for higher voltages, including than handling high voltages and minimizing added
topology, power handling and voltage ratings, noise.
resistor choices, and added measurement noise. An L-pad is a simple resistive divider network
Attenuation values are typically given as a ratio consisting of a series resistor attached between
expressed in decibels (dB). Table 1 represents the the signal source and the audio analyzer and a
ratio of the input voltage to the output voltage, shunt resistor across the audio analyzer input. The
which can be determined using Equation 1: output voltage of an attenuator can be calculated
using Ohm’s law:
⎛Voutput ⎞
Attenuation = −20 Log ⎜ Rshunt
⎟ L-pad output voltage = Vin ×
⎝ Vinput ⎠ Rshunt + Rseries

The U-pad is simply two identical L-pads placed

Attenuator Topology back-to-back in a mirrored configuration; each side
Attenuators can be constructed using many attenuating one side of the balanced signal (see
topologies, each offering different attributes to suit Figure 1). The two shunt resistors of a U-pad can
different needs. For audio noise measurements, an be combined into a single resistor but keeping them
attenuator with a low output impedance is desirable. separate provides a ground reference that can be
This makes the L-pad and U-pad topologies ideal useful to control common mode issues with floating
test systems. The ground point also enables the
Attenuation 6 dB 10 dB 12 dB 18 dB 20 dB 24 dB 30 dB 40 dB U-pad to be used in both balanced and unbalanced
configurations. The U-pad resistor values are shown
Voltage Ratio 2:1 3.2:1 4:1 8:1 10:1 16:1 32:1 100:1
here as half the L-pad values. This is so the same
Table 1: Attenuation and voltage ratios calculations used here can be applied to either case.

42 | March 2019 | audioxpress.com

The Test Load as an Attenuator
Power amplifiers are often tested with resistive

TEST & MEASUREMENT FOCUS

loads. These loads can also be configured as L-pads
or U-pads using multiple resistors. This is the
recommended way to attenuate an amplifier output
for measurement. index than other types and should be avoided for
Figure 1: The U-pad
An 8-Ω load can be made with eight 1 Ω resistors measurement purposes. is simply two identical
in series. Each node between the resistors provides The temperature coefficient indicates the L-pads placed back-
a differing attenuation value. The center node sensitivity of a resistor’s value to temperature to-back in a mirrored
(between the fourth and the fifth resistors) provides changes. Heating differences between the series configuration—each side
6 dB attenuation. The node between the sixth and and shunt resistors in an L-pad can produce changes attenuating one side of
the seventh resistors provides 12 dB attenuation in attenuation that can throw off measurements. the balanced signal.
and the node between the seventh and the eighth This can be mitigated by using resistors with
resistors is attenuated 18 dB. Amplifiers with push- lower temperature coefficients (≤50 ppm), using
pull outputs should be tested using a U-pad with higher wattage resistors in the series element,
each side having eight 0.5 Ω resistors. The total or constructing the series element from multiple
resistance is 8 Ω and there is still access to 6 dB, resistors. Small surface-mount resistors with low
12 dB, and 18 dB attenuation points (see Figure 2). thermal mass can exhibit nontrivial changes in
temperature within a single cycle of a low-frequency
Testing Without a Load signal and are not recommended unless used in
Testing amplifiers without a load requires large arrays to minimize heating.
more careful consideration of the attenuator’s Inductance in attenuators should generally
design. Most commercially available attenuators be minimized. Wire-wound resistors can handle
have an impedance of just a few hundred ohms tremendous heat loads, but they also exhibit higher
and are comprised of 1/8 W or 1/4 W resistors. At inductance than other resistor types. The actual
high voltages, the power rating of these resistors inductance is a function of the construction method
could be exceeded, causing them to overheat. The but vary from 0.1 to 400 µH depending on wattage
heat dissipation can be reduced by using higher and value. The maximum acceptable inductance
resistance values. However, higher resistance values should be considered for the specific device being
add excess noise that may impact measurement tested.
accuracy. The remainder of this article discusses Magnetic susceptibility occurs because wire
the design of higher impedance attenuators. wound resistors can both emit and pick up magnet
fields. If used, they should be placed away from
Choosing a Resistor Type sensitive circuits or strong magnetic sources such
There is no such thing as a perfect resistor. as power transformers.
There are many trade-offs to be considered for both Power ratings are an indication of how hot
electrical and physical properties as well as cost. resistors can operate without failing. However,
Thermal (Johnson) noise is the voltage noise operating the resistor well below its maximum
produced by a resistor that is proportional to the rated power insures the operating temperature
square root of the resistance value. The higher the stays closer to the ambient temperature.
resistance, the higher the noise voltage. The RMS Cost and availability can vary significantly
level of this noise is defined as: with resistor type. Metal-foil resistors approach

V = 4 × R × k × T × BW

where, R = Resistance value

k = Boltzman’s constant, 1.38 × 10-23
T = Operating temperature in degrees Kelvin
BW = Bandwidth in Hertz (Hz)

Current noise, sometimes referred to as 1/ƒ

noise, is caused by the electron flow through the
resistive material and varies with the voltage Figure 2: Amplifiers with push-pull outputs should be tested using a U-pad with each
across the resistor. Carbon-film, thick-film and side having eight 0.5 Ω resistors. The total resistance is still 8 Ω and there is still access
carbon-composition resistors have a higher noise to 6 dB, 12 dB, and 18 dB attenuation points.

audioxpress.com | March 2019 | 43

 ax Practical Test & Measurement

measurement is distortion. This is defined by the

TEST & MEASUREMENT FOCUS

Resistor Choices Actual Thevenin Thermal Noise

Attenuation Resistance equation for THD+N:
Series Shunt
Case 1 100 Ω 11 Ω -20.08 dB 9.1 Ω 0.057 µV/-145 dBv ⎛ residual noise and distortion ⎞
THD+N (in dB)=20 Log⎜ ⎟⎠
Case 2 1,000 Ω 110 Ω -20.08 dB 99 Ω 0.18 µV/-135 dBv ⎝ total signnal
Case 3 5,000 Ω 560 Ω -19.94 dB 504 Ω 0.41 µV/-128 dBv
Case 4 10,000 Ω 1,100 Ω -20.08 dB 991 Ω 0.57 µV/-125 dBv
Case 5 20,000 Ω 2,200 Ω -20.08 dB 1,982 Ω 0.81 µV/-122 dBv
Putting This to Work
The following two examples demonstrate the
Case 6 47,000 Ω 5,100 Ω -20.19 dB 4,601 Ω 1.23 µV/-118 dBv
types of trade-offs needed when designing an
the performance of an ideal resistor in many attenuator. I used AverLAB, the new audio analyzer
Table 2: L-pad resistor ways. However, they can cost up to 100 times from Avermetrics, to measure high-powered
combinations for 20 dB more than other resistor types. Availability can amplifiers. To keep AverLAB small, its maximum
also be a problem. Leaded thin-film resistors offer input was limited to 15.5 Vrms (balanced) and
an inexpensive, readily available compromise for 7.7 Vrms (unbalanced). Without an attenuator, it
constructing attenuators. Thin-film resistors can test up to 30 W into an 8 Ω load at the balanced
rated at 0.6 W, 1% tolerance and a 50 ppm/°C inputs or 7.5 W at the unbalanced inputs. For larger
temperature coefficient cost approximately 15 to amplifiers, 20 dB of external attenuation increases
30 cents each in individual quantities. These can those limits to 3000 W for balanced and 750 W for
be wired in series and parallel combinations to unbalanced connections. AverLAB has provisions
handle higher power dissipation at a reasonable to offset measurements for external attenuation
cost. which makes making these measurements easy.

Impact of Noise on a Measurement Measuring a 200 W Amplifier

The thermal noise from an attenuator can This hypothetical amplifier is rated at 200 W
impact the readings of certain measurements. This into 8 Ω with a THD+N of -100 dB (0.001%). The
will be most evident in noise-related measurements amplifier output is 40 V at the rated power and
such as signal-to-noise ratio (SNR) or total harmonic solving Equation 4 shows that there is 400 µV of
distortion + noise (THD+N), particularly at lower residual noise and distortion. A 20 dB L-pad will
signal levels. reduce the output signal to 4 V, which is well within
The thermal noise of an L-pad is determined AverLAB’s input range, and will also reduce the
as a function of its Thevenin equivalent resistance. residual to just 40 µV. The analyzer will still show
Assuming the amplifier under test has a negligible a THD+N reading of -100 dB unless the attenuator
output impedance, the Thevenin equivalent circuit has added too much noise of its own. The question
is simply the parallel combination of the series is how much of this added attenuator noise can
and shunt resistors of the attenuator. Table 2 be tolerated.
shows some common resistor combinations that If the maximum acceptable measurement
can produce ~20 dB attenuation as well as the error due to added noise is 0.1 dB, the distortion
thermal noise produced by that combination over reading must be ≤-99.9 dB. Equation 2 is solved
a 20 kHz bandwidth at 27°C. for -99.9 dB to show the total residual seen by the
The significance of this thermal noise depends analyzer cannot exceed 40.46 µV. Given that the
on the needed measurement accuracy, the signal amplifier is producing 40 µV of this residual, the
level of the test and the residual noise and distortion RMS allowable noise from the attenuator can be
of the product being tested. Suppose the desired determined. Since noise sources are uncorrelated

Resistor Choices Dissipation at 40 V Dissipation at 141 V

Series Shunt Series Shunt Series Shunt
Case 1 100 Ω 10 Ω 13.22 W 1.322 W 165 W 16.5 W
Case 2 1,000 Ω 110 Ω 1.30 W 0.143 W 16.2 W 1.8 W
Case 3 5,000 Ω 560 Ω 0.259 W 0.029 W 3.23 W 0.36 W
Case 4 10,000 Ω 1,100 Ω 0.130 W 0.014 W 1.62 W 0.18 W
Table 3: L-pad power Case 5 20,000 Ω 2,200 Ω 0.065 W 0.007 W 0.81 W 0.09 W
dissipation at 40 V and
Case 6 47,000 Ω 5,100 Ω 0.028 W 0.003 W 0.35 W 0.04 W
141 V

44 | March 2019 | audioxpress.com

 Figure 3: Any resistor
of value R may also be
constructed from multiple
elements to further
reduce heating.

RMS summation, the equation:

TEST & MEASUREMENT FOCUS

Total noise 2 = amplifier residual 2 +
at tenuator noise 2
0.1 dB accuracy is now just 0.43 µV. Table 2 shows
is used to determine the allowable attenuator noise. that Case 3 offers the highest impedance resistor
This shows that the maximum allowable attenuator combination that would satisfy that accuracy target.
noise for 0.1 dB accuracy is 6.1 µV. Any of the
attenuators shown in Table 2 are thus sufficient to Measuring a 1,500 Watt Amplifier
measure the amplifier. However, this is only true The previous example was only for a 200 W
for measurements at 40 V. amplifier, such as those found in home sound
Amplifiers are often tested by sweeping from systems. Professional live-sound amplifiers frequently
a low amplitude to maximum power. An audio have outputs in the kilowatt range. A 1500 W
analyzer would normally auto-range its input amplifier produces approximately 141 V into an 8 Ω
as the signal level changed to ensure maximum load. A 20-dB attenuator would reduce this voltage
measurement accuracy. However, a fixed external to ~14 Vs (within AverLAB’s balanced input range)
attenuator cannot be similarly switched in and but heat dissipation within the attenuator becomes
out as needed during an automated level sweep. a significant factor to consider. Table 3 shows the
Therefore, the noise of the attenuator must be same resistor combinations as Table 1 along with
considered at the lowest level being tested. the power dissipated by each resistor at both
At 1 W (2.83 V into 8 Ω) from the amplifier, the 200 W and 1,500 W.
analyzer sees only 0.283 V with the 20 dB attenuator It is easy to see that significant power can be
in place. The same equations as above are used dissipated in an attenuator at 141 V. Case 6 is the
again to find that allowable attenuator noise for only attenuator that could be made using 0.6 W

MEASUREMENT MICROPHONES

T OMER N
CUS FACTIO D
IS EE
SAT RANT
GUA

www.iSEMcon.com sales@isemcon.com

audioxpress.com | March 2019 | 45

TEST & MEASUREMENT FOCUS
ax Practical Test & Measurement

More Helpful Information

The following calculations are helpful in the design
and use of external attenuators:
Variable Description Units
• Equations for design: Determine the ideal resistor Level The output level of the device under test volts (V)
values based on test level and desired accuracy (DUT)
• Equations for use: Determine the measurement Distortion The rated distortion level of the DUT decibels (dB)
accuracy based on actual resistor values, test level Temperature Temperature in degrees Kelvin K
and distortion reading
Bandwidth Bandwidth of measurement Hertz (Hz)
The residual noise and distortion of the
These equations are simply variations of equations VnDUT
DUT at Level and Distortion
volts (V)
shown earlier solved for different variables. These can
Attenuation The value of the external attenuation decibels (dB)
be conveniently entered into a spreadsheet as a handy
k Voltage divider ratio of a given attenuator k:1
reference. Such a spreadsheet is available for download
at www.avermetrics.com. Value of NDUT following the external
Vnpost volts (V)
attenuator
Project Files Accuracy Tolerable measurement of accuracy decibels (dB)
To download a spreadsheet with all the equations to
Residual seen by the analyzer at limit of
calculate the attenuators, visit www.audioxpress.com/page/ Vnmeas volts (V)
accuracy specification
audioXpress-Supplementary-Material.html
Noise contribution of the attenuator at
VnAtten volts (V)
Variable Equations for Design limit of accuracy
Highest Thevenin equivalent resistance
VnDUT VnDUT = Level × 10Dist/20
for an L-pad attenuator that will meet
VnAtt VnAtt = VnDUT × 10Atten/20 RThev the accuracy specification at the rated ohms (Ω)
Vnmeas Vnmeas = VnAtt × 10Acc/20 distortion and level (at 27°C and 20 kHz
BW)
VnAtten VnAtten = Vnmeas 2 − VnAtt 2 Highest value of a series resistor for an
Rseries ohms (Ω)
L-pad that meets accuracy specification
RThev (
RThev = VnAtten / 4 × Temp × BW × 1 .38 E
2 −23
) Highest value of a shunt resistor for an
Rshunt ohms (Ω)
L-pad that meets accuracy specification
Rseries Rseries = RThev / 10 Atten / 20

Power dissipated in the series resistor

Rshunt (
Rshunt = RThev / 1 − 10 Atten / 20 ) Pseries
at Level
ohms (Ω)
2
⎛ Level ⎞ Power dissipated in the shunt resistor at
Pseries Pseries = ⎜ × Rseries Pshunt ohms (Ω)
⎝ Rseries + Rshunt ⎟
⎠ Level

2
RshActual Chosen shunt resistor ohms (Ω)
⎛ Level ⎞
Pshunt Pshunt =⎜ × Rshunt RseActual Chosen series resistor ohms (Ω)
⎝ Rseries + Rshunt ⎟
⎠
Precise attenuation given the chosen
AttenActual decibels (dB)
Equations for Use resistors

AttenActual = 20 × Log ( RshActual / RshActual + RseActual )

Thevenin equivalent resistance of the
AttenActual RThevActual ohms (Ω)
actual attenuator
⎛ 1 ⎞ VnActual Noise produced by the actual attenuator volts (V)
RThevActual RThev Actual = ⎜ ⎟
⎝1 / RseActual + 1 / RshActual ⎠ Measured distortion reported by the
DistMeas decibels (dB)
audio analyzer
VnActual VnActual = 4 × Temp × BW × 1 .38 E −23 × RThevActual Residual noise and distortion measured
VnMeas volts (V)
by the analyzer
Dist Meas / 20
VnMeas VnMeas = LevelActual ×10 LevelActual Signal level seen by the audio analyzer volts (V)
Residual noise from DUT (without
VnDUTActuual volts (V)
VnDUTActual VnDUTActual = VnMeas 2 − VnActual 2 attenuator noise)
DistActual Actual distortion of the DUT decibels (dB
Dist Actual = 20 × Log (
VnDUTActual / Level Actual )
DistActual
Accuracy of the measured distortion
AccActual based on the test level and attenuator decibels (dB)
AccActual Acc Actual = DistMeas − Dist Actual
used

46 | March 2019 | audioxpress.com

 About the Author
Jonathan Novick is an electrical engineer with more than 25 years of test and
measurement experience, the last 15 exclusively in audio. He most recently served
as VP of Sales and Marketing for Avermetrics, LLC. Prior to that he was a Director of
resistors. However, Table 2 shows that the noise Sales for Audio Precision and a Senior Product Manager at Agilent Technologies (now
Keysight). He is also a former vice president and governor of the Audio Engineering
will be 1.23 µV.

TEST & MEASUREMENT FOCUS

Society. Jonathan co-chairs the R3WG2 audio standards committee of the Consumer
At full power, the analyzer will be seeing a 14.1 V Technology Associations and contributes on other working groups as well.
signal and the amplifier’s residual would be 0.14 µV
for a distortion of -100 dB. The 1.23 µV of additional
noise from the attenuator would not cause the
reading to change significantly. In fact, it would The shunt resistor dominates the noise produced
take 21.6 µV of noise to change the measurement by the L-pad. In Case 3, the temperature of a 0.6 W
by 0.1 dB. However, at 1 W, it takes only 0.43 µV thin-film shunt resistor will rise almost 60°C whereas
of noise to change the measurement by 0.1 dB. in Case 4 it will rise less than 20°C. Any resistor
With 1.23 µV of attenuator noise the new THD+N of value R may also be constructed from multiple
reading would be -99.25 dB, a difference of 0.75 dB. elements to further reduce heating (see Figure 3).
The series resistor in Case 3 could be built
using multiple 0.6 W resistors to handle the heat Conclusions
load. For instance, eight 10 kΩ resistors could be This article discusses the impact an attenuator’s
wired as four series-pairs in parallel to produce the thermal noise will have on distortion measurements.
5000 Ω series resistance. This brings the dissipation At low test voltages, the added thermal noise of
in each resistor to ~0.4 W. However, at 0.4 W, these the attenuator is of primary concern. At high
resistors will see almost a 70°C temperature rise, test voltages, the heat dissipation in the external
which may not be tolerable. attenuator is of primary concern. For this reason,
The resistor combination in Case 4 changes it would be handy to have multiple attenuators of
the measurement reading by less than a 0.2 dB similar attenuation but with different impedances
and Case 5 would cause less than a 0.4 dB. These on hand. The use of 0.6 W 1% thin-film resistors
other solutions make it easier to handle the heat makes this a very practical and cost-effective
load in the attenuator. solution. ax

Measure sound like a pro,

modern, easy and low-cost!

AudioTools
offering an extensive range of ready-to-go
toroidal transformers
to please the ear, but won’t take you for a ride.

Avel Lindberg Inc.

47 South End Plaza, New Milford, CT 06776 5-star App Store rating!
p: 860.355.4711 / f: 860.354.8597
www.studiosixdigital.com

audioxpress.com | March 2019 | 47