The CAF Reference





The Main Page


1.1 Input Sensitivity

1.2 Voltage Gain

1.3 Maximum Output Voltage

1.4 Maximum Input Voltage

1.5 Noise Floor

1.6 Dynamic Range

1.7 Frequency Response Deviation

1.8 Latency

1.9 Common-Mode Rejection Ratio (CMRR)

1.10 Input Impedance

1.11 Damping Factor

1.12 Fan Noise Level


The I/O Matrix


 2.1-2.2 Burst Tests

 2.3-2.4 Tone Tests

 2.5-2.6 Noise Tests

 Matrix Overview


The Plots Section


 Plot A – Input vs. Output Level

 Plot B – Frequency Response (20 Hz – 20 kHz)

 Plot C – Frequency Response (2 Hz – 80 kHz)

 Plot D – THD Ratio vs. Frequency

 Plot E – IMD Difference Frequency Distortion

 Plot F – Excess Phase Response

 Plot G – Group Delay vs. Frequency

 Plot H – Crosstalk

 Plot I – Common-Mode Rejection Ratio (CMRR)

 Plot J – Damping Factor

 Plot K – Noise Floor Spectrum

 Plot L – Pin 1 Response

 Plot M – Fan Noise Spectrum

 Plot N – Short-Circuit Behavior


Appendix A – Input Impedance by Voltage Divider Method

Appendix B – Test Jig for CMRR Measurement






The Common Amplifier Format (CAF) is a report for presenting a set of specifications and measured performance metrics of audio power amplifiers. While much of this information can be found on manufacturer's specification sheets, each manufacturer presents the information differently. In addition, the methods used to test the amplifier may be undocumented or vague, making it impossible to do meaningful comparisons between various makes and models. The CAF provides a structured, consistent and well-documented report of an amplifier's performance.



The objective of the CAF is simple:


1. Provide the performance metrics needed to select and deploy an amplifier in a concise, well-defined and defensible format.


2. Provide performance metrics in a manner that facilitates apples vs. apples comparisons between makes and models.


The CAF is not a Standard, but many of the measurements therein are defined in Standards. Like its cousin, the Common Loudspeaker Format (CLF) the CAF is a tool created by those who need it, for those who need it. Whether or not it is, or becomes a Standard is irrelevant. Those who participate in the CAF program do so because they have a common interest in seeing audio systems properly designed and deployed.


The CAF is not perfect, nor is it the "end all" for amplifier specification sheets. It is a compromise reached by a long trial-and-error process, with input from multiple industry professionals. It is not as concise as we would have liked, but there are many more specifications that it could have included. All included metrics have stood the test of relevance, repeatability and clear interpretation. With regard to which metrics to include, and how to measure them, there are two driving principles:


1. We had to pick something.

2. We could not pick everything.


There are many ways to measure amplifiers and specify their performance. Many amplifier manufacturers have excellent specification sheets available to the system designer. The CAF is not meant to be a replacement for the manufacturer's data sheet. The report specification is free to amplifier manufacturers should they wish to use it. This saves them the time and expense of developing their own, and provides them with assurance that they are giving their customers the necessary body of information to properly use their products.


No set of specifications can replace the listening process for evaluating an amplifier, or comparing amplifiers. If you want to know how two amplifiers compare in reproducing an intense kick drum signal into a 2 ohm subwoofer, you should hook them up and conduct proper A/B listening tests. The CAF can serve as a front end for this exercise, or as a back end.  It bridges the gap between artistic evaluation and technical deployment. Both are vital to the sound system design process.



There are several choices for specifying the voltage and power levels produced by audio products. Voltage is commonly expressed in dBV or dBu. Power is expressed in dBW or dBm. All voltages in the CAF are root-mean-square unless otherwise indicated (Vrms). In the CAF, the abbreviation for voltage is E (electro-motive force) and the symbol is V.


It is certainly possible to include all of these level types, and early versions of the CAF did. This produced a report that some found confusing and bloated. In order to meet the objective of presenting the data in a concise, informative manner choices had to be made.


Since the CAF was designed primarily for use with commercial and professional amplifiers, the dBu was selected for voltage levels. This is the dominant level used for specifying professional mixers and digital signal processors (DSP). It has a long pedigree and has shown amazing resilience in an ever-changing technical landscape. So, in the CAF some voltage measurements are expressed as levels using the following formula.


dBu = 20log(E/0.775)


where E is the rms voltage.


It should be noted that dBu levels can easily be converted to dBV by subtracting 2.2.


dBV = dBu - 2.2



Ein = 1.23 Vrms

dBu =  20*log(1.23/0.775) = +4 dBu

dBV = 4 - 2.2 = 1.8


Thus the use of the dBu facilitates comparing the output level of a professional mixer or DSP with the input sensitivity of a power amplifier.


Power levels are expressed in dBW. These are measured voltages, converted to watts (by squaring and division by the load resistance), and then converted to dBW. Here are the formulas:


Eout = 28 Vrms

Rload = 8 ohms (resistive)

W = E2/R = 282/8

W = 100 W continuous

dBW = 10log(100/1) = 20


The dBW is ideal for comparing amplifiers of different ratings. For example, the sound pressure level achievable from a 400 W amplifier vs. a 300 W amplifier is not apparent from the wattage ratings. But, expressing the wattages in dBW makes the difference obvious.


dBW1 = 10log(400) = 26.0 dBW

dBW2 = 10log(300) = 24.7 dBW

ΔdB = 26.0 - 24.7 = 1.3 dB


Amplifier Dummy Loads

An amplifier must be tested into a load. Most amplifiers will be used to drive a loudspeaker(s) with a complex impedance (both AC and DC resistance and AC reactance). It seems logical to test the amplifier into such a load. Unfortunately, with resistance, inductance and capacitance considered, an infinite number of possible combinations exist. We must simplify.


Amplifiers are typically tested into non-reactive resistive loads because:

- They are simple to define.

- They are simple to fabricate.

- There can be no universal agreement on the specifics of a suitable complex impedance.


A suitable dummy load for CAF testing could use sixteen (16) 8 ohm water heater elements and a switch matrix. The heater elements are cooled in a bath of water or mineral oil. This load bank would allow 4 amplifier channels to be tested to 2 ohms, or 8 channels to be tested to 4 ohms.


The CAF Report


The CAF is divided into three major sections.


1. The Main Page

2. The Voltage, Power I/O Matrix

3. The Plots Section


Here is an overview of each.


The Main Page

This "home" page of the CAF gives information on the general characteristics of the amplifier, including some "one number" performance metrics.



These are general characteristics of the amplifier as provided by the manufacturer.



These are physical characteristics of the amplifier as provided by the manufacturer.


Tested by

This is the party that performed the testing. There is no certification program for those producing the CAF report. The necessary tests are described in this reference. The requirement for plots for most of the metrics make it difficult to manipulate "one number" ratings of amplifiers.



This is a list of the instrumentation used to test the amplifier. It is highly recommended that an automated audio analyzer be used to perform these tests. Suggested models include the Audio Precision APx515 (or higher), and the Clio 10 measurement platform. Since the specifics for many of the tests depend on the analyzer used, the reader should consult the documentation for that analyzer for the fine details regarding the test setup. There is no attempt to reproduce those in detail in this document. Here are links for two analyzer platforms.


For the Audio Precision APx515 analyzer:


For the Clio 10 analyzer:


While it is possible to measure an amplifier and produce a CAF using discrete instruments rather than an analysis platform, it is discouraged due to the many variables involved in ad hoc measurements, and the (likely) lack of documentation for the exact methods used. Many of the distortion and burst tests are complicated and impractical without a purpose-built analysis platform.


Each metric as presented in this document is accompanied by one or more methods for how to measure it. These should help the reader understand the nature of each test.


1.0 Specifications

These are "one number" metrics used to summarize a characteristic of the device-under-test (DUT). Most of these metrics are taken from the indicated plots in the Plots section. While there is a separate Voltage, Power Matrix for each operating mode of the amplifier, these tend to be universal to all of them.


1.1 – Input Sensitivity

This is the input voltage required to drive the amplifier to its full output voltage, into the indicated load impedance (typically 8 ohms). It is provided by the manufacturer, and verified by measurement.  The drive signal is given in Vrms and dBu.


Method 1: The DUT is driven by a 1 kHz sine wave, and the output viewed on an oscilloscope.  The input sensitivity has been reached when visible flattening occurs on the sine wave.


Method 2: The input sensitivity can be take from the Input vs. Output Level plot (Plot A).


Note: Many amplifiers have user-selectable input sensitivity. The manufacturer specifies which setting is used for the CAF testing.


1.2 – Voltage Gain

This is the gain of the amplifier, found by


G = 20log(Eout/Ein)



G is the voltage gain of the amplifier in dB

Eout is the output voltage of the amplifier

Ein is the input voltage to the amplifier


Voltage gain is dependent on the setting of the input sensitivity control. Voltage gain tends to be independent of the drive level to the amplifier.


Method 1: The amplifier is loaded as per Ref. 1 – Input Sensitivity, and driven by a signal at or below the Input Sensitivity. The output voltage is measured with a high impedance voltmeter. If an input attenuator is present, it is set at maximum.


Method 2: The input sensitivity can be take from the Input vs. Output Level plot (Plot A).


Note: Many amplifiers have a user-selectable input sensitivity range. The manufacturer specifies which setting is used for the CAF testing.


1.3 – Maximum Output Voltage - Emax

This is the maximum linear sine wave output voltage of the amplifier at 1 kHz, given as Vrms and Vpeak. The peak voltage is the amplifier’s “rail voltage.” The DUT should be loaded as per Ref. 1 – Input Sensitivity.


Method 1: The DUT is driven by a 1 kHz sine wave, and the output viewed on an oscilloscope. The Emax has been reached when visible flattening occurs on the sine wave. The RMS voltage is measured.


Method 2: Read the Emax directly from Plot A - Input vs. Output Voltage.


Method 3: The amplifier is tested as per CEA-2006 (see I/O Matrix, Ref. 1).


1.4 – Maximum Input Voltage - Ein-max

This is the input voltage to the DUT at which the input stage of the amplifier clips. The drive signal is given in Vrms and dBu. It is given because it may be possible to clip the amplifier's input stage, prior to clipping the amplifier's output stage. For example, if a mixer is producing +28 dBu, but the amplifier input stage clips at +20 dBu, the amplifier will have a clipped output signal even if there is plenty of headroom in the amplifier's output stage.


Method 1: The input sensitivity of the DUT is set at -20 dB ref. full output to prevent the output stage from clipping. A 1 kHz signal at -20 dB ref. Input Sensitivity is applied to the DUT and slowly increased. The output voltage is monitored on an oscilloscope for visible clipping, which terminates the test.


Method 2: The test for Plot A - Input vs. Output Voltage is repeated with the input sensitivity set at -20 dB to prevent the output stage from clipping. The Input Stage Clipping Level has been reached when the THD increases.


1.5 – Noise Floor

This is the residual or “self-noise” of the amplifier. All electronic devices produce output noise, even without the presence of an input signal.  Noise floor specifications are often A-weighted. This is justified because if it is audible, it would be at a very low sound pressure level (SPL). In the CAF, the "one number" noise floor specification is un-weighted. The Noise Floor Spectrum (Plot K) provides information on the spectral content of the noise.


Method 1: The DUT is optimally loaded, and it’s input terminated with a 150 resistor. Input Sensitivity is set as per Ref. 1 – Input Sensitivity. The voltmeter is band-limited to 20 Hz – 20 kHz. The output voltage is measured, and given in dBu.


Method 2: See Plot K - Noise Floor Spectrum. Measure or calculate the Vrms when making this plot.


1.6 – Dynamic Range

The Dynamic Range (DR) level difference between the Emax (Ref. 3) and the Noise Floor (Ref. 5). It represents the range of levels that can be output from the DUT.


Method 1: Convert the Emax (Ref. 3) to dBu. Subtract the Noise floor (Ref. 5).


Method 2: Perform a Dynamic Range test as per the analyzer indicated in the Instrumentation section.


Note: The results of Methods 1 and 2 can be slightly different due to the criteria used to determine the respective levels. Large discrepancies should be investigated.


1.7 – Frequency Response Deviation

The deviation of the frequency response magnitude over the indicated bandwidth (typically 20 Hz – 20 kHz) at the indicated load impedance.  This rating is determined from Plot B - Frequency Response.


Method 1: Measure the frequency response magnitude using the audio analyzer indicated under Instrumentation. The drive signal should be at -3 dB ref. Input Sensitivity to avoid clipping. The deviation is the minima and maxima of the response curve.


Note: Class D amplifiers may show a load-dependent response deviation. The load which produces the specified deviation should be indicated.


1.8 - Latency

Latency is the inherent, unavoidable delay of a signal passing through the amplifier.


Method 1: Measure the latency using a dual-channel oscilloscope. Ch1 of the scope is bridged across the amplifier's input. Ch2 is bridged across it's output. An impulse is fed to the amplifier, and the scope is set to trigger on Ch1. The latency is the measured delay between Ch1 and Ch2, usually given in milliseconds.


Method 2 (recommended): Perform a Group Delay (GD) vs. Frequency measurement as per Plot G - Group Delay. Use the cursor to find the group delay at 1 kHz.


Note: The GD plot should be flat at the frequency at which the "one number" latency is determined. Since the GD is often frequency-dependent due to the high pass response of the amplifier, a frequency other than 1 kHz may be used, and should be indicated.


1.9 - Common-Mode Rejection Ratio (CMRR)

For a balanced input, the output level of a common-mode input signal relative to a differential-mode input signal. This is a "one number" metric for 1 kHz.  See Plot I - CMRR in the Plots section for details.


Method 1: Take the 1 kHz CMRR from the CMRR Plot (Plot I). This should be the lowest CMRR figure between the "Balanced" and "10 ohm Imbalanced" methods.


Method 2: Measure the CMRR directly using the audio analyzer listed under Instrumentation. Note: See the details regarding CMRR measurement under Plot I - Common-Mode Rejection Ratio.


1.10 - Input Impedance (Zin)

The load impedance (electrical) presented to a source, such as a mixer or DSP by input stage of the amplifier. It is of interest when the system design calls for multiple amplifier inputs to be driven from a single output.


Method 1: Measure the input impedance at 1 kHz using an impedance meter.


Method 2: Use a voltage-divider method as described in Appendix A.


Note: Input impedance tends to be frequency-independent, so a plot is not included in the CAF.


1.11 - Damping Factor (DF)

The output impedance of the amplifier is the source impedance seen by the load (such as a loudspeaker), neglecting the resistance of the interconnecting cable. The Damping Factor (DF) is the ratio of the load impedance to source impedance.


DF = Rload/Rsource


Method 1: Measure the open-circuit output voltage of the amplifier at -20 dB ref. Vmax (Ref. 3). Load the amplifier as per Ref. 3 and repeat the measurement. The Rout , and therefore the DF, can be found from


Rsource = Eload/(Eno-load/Eload)


The test frequency is 1 kHz unless otherwise indicated.


Method 2: Perform an automated output impedance/damping factor measurement as per the analyzer listed under Instrumentation. The "one number" DF shall be taken from this plot.


1.12 - Fan Noise (Min, Max)

Since amplifiers are often used in quiet environments, the acoustic level of the fan noise is of interest. These are "one number" frontal measurements taken from the Fan Noise Spectrum plot (Ref. M).


2 - The Input vs. Output Matrix



The most common question posed about any amplifier is "How many watts?" The answer is the often cited "It depends!" This section provides some meaningful answers.


Most amplifiers have multiple configurations. For example, a two-channel amplifier may be deployed as


- Single channel driven

- Two-channels driven

- Bridged-mono

- Parallel-mono

- 70.7/100 V


The voltage and power levels from the amplifier depend on the configuration used, so there is a separate Voltage, Power Matrix for each. This allows the report to grow to accommodate any amplifier.


The CAF does not test for combinations of configurations. For example an eight-channel amplifier may be configured as four (4) mono amplifiers and two (2) bridged amplifiers. In these cases it is safest to consider the worst-case (lowest) voltage output for each configuration (lowest load impedance).


The I/O Matrix gives the following information for each amplifier configuration.


1. The root mean square voltage output for various test stimuli.

2. The resultant calculated power into the indicated load impedance.

3. The input level (dBu) that produced the output voltage.

4. The power drawn from the utility company to produce the indicated audio power.

5. The thermal output of the amplifier for various signal types and drive levels.

6. Utility power draw and thermal output for ambient (resting) conditions.


The "Zoned-Distributed System" Scenario

An important aspect of amplifier performance is how it behaves with increasing load. In an ideal constant voltage interface, the voltage from the source remains unchanged as the load impedance is reduced. When the same voltage is developed across a lower load resistance, the current increases, as does the power delivered to the load. In the ideal case, each halving of the load resistance doubles the power from the amplifier.


At face value, it may appear that loading the amplifier to its highest power output is the way to go. This is seldom the case.


To understand why, consider a zoned-distributed system (Figure 1) where the amplifier drives a loudspeaker (Ldspk A) that is zoned to an audience area (Zone A).  Ldspk A produces an SPL to the listeners based on the voltage applied to it. Next, consider that a second, identical loudspeaker is paralleled with the first one (Ldspk B). Ldspk B is zoned to a different audience area (Zone B), and is not heard by the listeners in Zone A. The question is "What effect does adding Ldspk B have on the SPL in Zone A?"


In the ideal constant voltage condition, there would be no change to the SPL in Zone A, because the output voltage from the amplifier is unaffected by daisy-chaining the Ldspk B (and/or C) onto the amplifier's output. In the real world, it is highly likely that the amplifier cannot source twice the current required by Ldspk A only, so its output voltage drops, even though its output power may increase.


The I/O Matrix presents the change in SPL in Zone A, which results from additional loads being daisy-chained onto the amplifier (Column E). The SPL in Zone A is designated as "0 dB" since it is the reference for the one loudspeaker configuration. The voltage change that results from decreasing the load impedance (adding more loudspeakers) is given for each load resistance.




Figure 1 - A zoned-distributed system scenario for evaluating amplifier loading effects.


Test Procedure

Following is the test procedure used to produce the I/O Matrix. It is important to understand the procedure to properly interpret the results. Column letters are provided as a reference. The test sequence is from top to bottom, beginning with the CEA2006 burst test into the indicated impedance.


1. The amplifier is loaded as indicated in Column B.

2. The stimulus is applied as per Column C and the output voltage measured (Column D).

3. The load resistance is reduced as indicated in Column B.

4. The stimulus is re-applied and the output voltage measured.

5. If the lower load resistance causes a failure condition, the stimulus is reduced by 1 dB (Column C) and the test repeated. The failure condition for each stimulus is described later.

6. This sequence is repeated for each load resistance indicated in Column B.


Stimulus Types

Three different stimulus types are used in the CAF.  These include a burst test, a tone test, and a noise test. The burst and tone tests are based on sine wave signals that drive the amplifier to its full output. It is reasonable to expect that the amplifier can handle these signals at its rated input sensitivity. They will produce the maximum voltage and power ratings in the matrix, and are the basis for the amplifier's "rated power." The noise stimulus has peaks that drive the amplifier to full-scale output, but the rms level of the signal is reduced. This is much like speech and music program material. The noise stimulus is the basis for establishing the output voltage and power of the amplifier into a loudspeaker.


So, the I/O Matrix is ordered by the crest factor (CF) of the drive signal, with the lowest CF at the top and highest CF at the bottom.


Here is the basis for the use of each stimulus.


2.1 - 2.2 Burst Tests

The burst test in the CAF is as per CEA-2006. From the Audio Precision APx500 user's manual.


Two signals are specified in the standard: for full-range amplifiers, the signal shall be a repetition of a burst of 20 cycles of a 1 kHz sine wave at 100%, followed by 480 cycles of 1 kHz at 10% (–20 dB). For limited-range amplifiers (subwoofer amplifiers), the signal shall be a repetition of a burst of 10 cycles of a 50 Hz sine wave at 100%, followed by 20 cycles of 50 Hz at 10% (–20 dB).

The CAF uses both the 1 kHz and 50 Hz burst tests for all amplifiers. The distortion threshold for the 1 kHz test is 1% THD. The distortion threshold for the 50 Hz test is 3% THD, unless otherwise indicated. Exceeding these distortion percentages constitutes a failure of the test.



Figure 2 = The CEA2006 test signal (1 kHz)


Burst testing is used to find the maximum root mean square output voltage without using a continuous sine wave.  These provide the largest legitimate wattage rating for the amplifier. This rating is useful as a starting point for amplifier deployment calculations, where it must be de-rated by the crest factor of the program material. Due to the instantaneous nature of the test, the electrical power consumed and thermal output are not given.


The crest factor of the overall signal is quite high (~17 dB) which means that the power drawn and produced is quite low. The crest factor of the burst itself is quite low (3 dB). The MIV and resultant power rating are based on the amplifier's reproduction of the burst. This is not a "peak" power rating.


2.3 - 2.4 Tone Tests

Continuous sine wave output is the most demanding task for any amplifier. Fortunately, it is rare for most amplifiers to ever have to output a continuous tone. This is an indicator of how the amplifier will behave under the most demanding signal conditions. It is also useful for observing the operation of protection circuits and protection algorithms.


A sine wave at rated sensitivity is fed to the amplifier at the indicated load resistance. The spectrum of the output voltage is monitored. If any harmonic is higher than -40 dB relative to the fundamental, a failure condition has occurred and the input voltage shall be reduced by 1 dB and the test repeated. Once the distortion percentage is acceptable, the output voltage is monitored versus time. If the voltage drops by more than 1 dB over the 15 second period a failure condition has occurred. The drive voltage is lowered by 1 dB and the test repeated. Technically, this means that the input sensitivity of the amplifier has changed, and is a function of load impedance. For simplicity, the input sensitivity is kept as specified on the main page, and the reduced drive level is indicated. The procedure is repeated until the amplifier can sustain the sine wave level for 15 seconds.


The tone tests represent the worst-case scenario for utility power draw and thermal output. Care should be taken in interpreting the results. An amplifier whose output voltage drops as a function of time is not necessarily a bad thing. It may be the intended behavior that is preventing the tripping of a circuit breaker. On the other hand, an amplifier that blows an internal fuse or cycles "off" during this test would be cause for concern.


2.5 - 2.6 Noise Tests

Noise tests are useful for determining the output of the amplifier for program material. The rms output voltage of the noise stimulus is useful for determining the continuous voltage the amplifier will present to the loudspeaker with speech or music program material. The electrical power drawn during these tests is useful for sizing circuit breakers in the electrical panel. The thermal output ratings are useful for designing a cooling system for the rack. Including measured data for these can free the system designer, electrician and/or architect from the task of performing calculations from sine wave ratings. This leads directly to power circuits and cooling systems that are appropriate to the task.


Normally the 1/8-power ratings are used for all of these applications. The 1/4-power ratings can be used for extreme conditions. The system designer must decide which is more appropriate for their application.


The "1/8th power" test feeds the device "typical" pink noise with a crest factor of 12 dB. The crest factor is the "peak to RMS" ratio. The CAF defines "full power" for an amplifier as that found by the CEA2006 burst test. Since this is a sine wave rating, and a sine wave has a CF of 3 dB, the 12 dB CF pink noise is -9 dB re. the sine wave output of the amplifier.


%W = 10(9/10) = 0.125 or 1/8


This is based on the amplifier's 8 ohm rating, unless otherwise noted. Typically the load impedance is halved and the test repeated, unless otherwise indicated.


The test duration is 60 seconds. If the output voltage of the amplifier drops by more than 1 dB over this interval, a failure condition has occurred. The drive level is reduced by 1 dB and the test is repeated.


The "1/4th power" test uses a pink noise stimulus that is +3 dB relative to the 12 CF pink noise used in the 1/8th power test. One method for creating this stimulus is to open the 12 dB CF noise in a wave editor, normalize it to 0 dBFS, and then increase the level by 3 dB. The noise is now slightly clipped, with 1.4x (2x the power) the RMS voltage of the 12 dB CF noise.


The Common Loudspeaker Format (CAF) includes the loudspeaker's maximum input voltage (MIV) based on the thermal limits of the loudspeaker. The noise rating of the amplifier should not exceed the loudspeaker's MIV.


The following graphic provides an overview of how the various metrics can be used in the I/O matrix.


2.7 Inrush Current

This is the surge current drawn by powering up the amplifier, as given by the manufacturer.


2.8 Idle Current

This is the current drawn when the amplifier is "on" and under load, but no input signal is present.


2.9 Standby Current

This is the current drawn by the amplifier if it is in standby or "sleep" mode.


Column Description

The columns in the I/O Matrix have reference letters. There are three sections in the matrix. The Audio section characterizes the input and output signal voltage from the amplifier, and the calculated audio power. The Supply section characterizes the electricity drawn from the utility outlet when reproducing the audio signal. the Thermal section characterizes the heat produced by the amplifier when reproducing the audio signal. Here is a description of each column.





A - Crest Factor - The peak-to-rms ratio of the test stimulus. See "Burst Tests" for details regarding the CF of the burst stimulus.


B - Load R - The load resistance used for the test.


C - dBu In - The input level to the DUT that produced the output voltage.


D - Vrms Out - The root-mean-square output voltage of the DUT. See "Burst Tests" for details regarding the rms voltage of the burst stimulus. This is always for "1-of-n" channels tested, since the remaining channels will be connected to different loads.


E - Loading Effect - The reduction in Vrms Out caused by increasing the load on the DUT.


F - Rated Watts - The power rating of the DUT as given by the manufacturer.


G - Meas. Watts - The output power of the DUT calculated from the measured output voltage and load resistance. This is always for "1-of-n" channels tested, since the remaining channels will be connected to different loads.


H - Meas. dBW - The output power of the DUT stated as level in dB ref. 1W.





I - Amps - The AC line current drawn during the test. It is the sum of all channels tested, since the amplifier is connected to a single power source.


J - Watts - The AC line power drawn during the test. It is the sum of all channels tested, since the amplifier is connected to a single power source.



K - Power Factor - The Power Factor of the AC line power drawn during the test.



Thermal Output


L - BTU/hr - The heat produced by the DUT. It is the sum of all channels tested.



M - kCal/hr - The heat produced by the DUT. It is the sum of all channels tested.



Figure 3 = The Voltage, Power I/O Matrix


To summarize, the input vs. output matrix yields the power rating needed to select an amplifier for a loudspeaker (burst tests), the worst case power continuous output power (tone tests) and a "real world" test for sizing circuit breakers and cooling systems (noise tests).

4 - The Plots Section


Some metrics don't lend themselves to "one number" representations. Since a picture is worth a thousand words, the CAF puts strong emphasis on presenting the performance metrics as plots.


An example of each plot is presented here, along with a short explanation of what it is and how it is measured.


Plot A - Input vs. Output Level

The DUT is driven by a 1 kHz sine wave at -10 dB re. input sensitivity. The output voltage and THD are monitored. Both are plotted vs. time. The sine wave is increased in level until the slope of the THD breaks from linear, forming a “hockey stick” response curve (see Plot A, below). The input sensitivity is the input voltage at which the THD curve changes slope.


Some amplifiers have circuitry that prevents clipping and the resultant rise in THD. The output voltage ceases to rise at some applied input voltage. This is the input sensitivity of the DUT.




Plot A - Input voltage vs. output voltage, THD


Plot B - Frequency Response (20 Hz - 20 kHz)

This is the magnitude of the frequency response for the rated passband as per the manufacturer. It is measured at -3 dB ref. the input sensitivity into the indicated load impedance to prevent clipping or current limiting. The initial test only loads the amplifier with the input impedance of the analyzer. This "open circuit" response can be useful if the amplifier is used as a line driver into an electronic input, rather than to drive a loudspeaker. The test is repeated for the mfg-recommended amplifier loads (See Plot B, below).


Method: The response is measured by applying a log-sine sweep to the DUT. The test should be run as per the audio analyzer indicated on the main page of the CAF report.



Plot B - The frequency response magnitude of the amplifier into various load impedances.


Plot C - Frequency Response (2 Hz - 80 kHz)

This is the same test used to produce Plot A, but with extended bandwidth.


The out-of-band response can be of interest, since it can affect the in-band response. Amplifiers are band-limited, and there are many choices for high and low pass filters. Class D amplifiers employ brickwall low pass filters to avoid aliasing.


Note: The 2 Hz to 80 kHz bandwidth of this test is based on the Audio Precision APx515 analyzer. If a different bandwidth is used, it should be indicated. If the bandwidth of the DUT exceeds the bandwidth of the analyzer, the maximum bandwidth of the analyzer should be used.


Method: Repeat the test for Plot B, but with the bandwidth of the analyzer extended to 2 Hz to 80 kHz.



Plot C - Frequency response magnitude - extended bandwidth


Plot D - THD Ratio vs. Frequency

This is the total harmonic distortion at - 3 dB ref. rated sensitivity. It is found by driving the DUT with a log sine sweep. The sweep tone is filtered out of the response by the analyzer, leaving the distortion products. In general, < 1% THD (-40 dB relative to the fundamental) is considered to be inaudible. This test is a good indicator of the lowest suggested load impedance, as THD can rise dramatically when the amplifier reaches its power limits (see Plot D, below).


Method: This test should be performed as per the audio analyzer indicated on the main page of the CAF report.




Plot D - THD Ratio vs. Frequency for the indicated load impedances.


Plot E - IMD Difference Frequency Distortion

Two tones fed to an amplifier can produce sum and difference frequencies of the two tones. These are output frequencies that were not present in the input signal, and are referred to as intermodulation distortion (IMD). Here are descriptions of  IMD from two authoritative sources.


An audio measurement designed to quantify the distortion products produced by nonlinearities in the unit under test that cause complex waves to produce beat frequencies, i.e., sum and difference products not harmonically related to the fundamentals. For example, two frequencies, f1 and f2 produce new frequencies f3 = f1 - f2; f4 = f1 + f2; f5 = f1 - 2f1; f6 = f1 + 2f2, and so on.  Rane Pro Audio Reference


The DFD stimulus is two equal-level high-frequency tones f1 and f2, centered around a frequency called the mean frequency, (f1+f2)/2. The tones are separated by a frequency offset called the difference frequency. The two tones inter-modulate in a distorting DUT to produce sum and difference frequencies. For analysis DFD selectively measures the 2nd and 3rd order intermodulation products, combines their values arithmetically and provides a result that is the ratio of the sum of the products to a reference voltage defined as 2x the voltage of f2 (effectively, the sum of f1 and f2). In the APx500 implementation, the 4th and 5th order products are also measured and reported in the distortion product result.  Audio Precision AP500 Users Manual




Plot E - IMD distortion at the indicated load impedances.


Plot F - Excess Phase Response

The Excess Phase shows the phase shift over the audible spectrum, after removing the linear component (device delay). In theory, a perfectly flat frequency response magnitude plot (Plot A) and excess phase plot (Plot F) would mean that the DUT could pass a square wave without changing its shape. The importance of that is subject to ongoing debate, but it remains the "holy grail" response for electronic devices such as amplifiers.


A positive phase angle at low frequencies is due to the (necessary) high pass response of the amplifier. The amplifier's response would have to extend to DC to not exhibit this (a bad thing). A negative phase angle at high frequencies is due to the low pass response of the amplifier. The amplifier's frequency response must be extended well beyond the audible spectrum to not exhibit phase shift within the audible spectrum. Alternately, linear phase low pass filters (just outside the audible spectrum) can allow the phase to remain flat though the audible spectrum.


Note: The results of this test do not tend to be load-dependent. The load impedance used should be indicated in the plot.


Method: This test should be performed as per the audio analyzer indicated on the main page of the CAF report.




Plot F - Excess Phase vs. Frequency plot


Plot G - Group Delay vs. Frequency

Group delay is an alternative way of presenting the phase response. In fact, it is calculated from it. It may be preferred over the phase response when large delays are present in the DUT. It is more intuitive than the phase response, since delay manifests a slope in a phase plot. An alternate name for group delay is "frequency delay."


Technically the time interval required for the crest of a group of waves to travel through a 2-port network [IEEE]. It is the rate of change of phase shift with respect to frequency. Hence, constant group delay, or linear group delay, describes circuits or systems exhibiting constant delay for all frequencies, i.e., all frequencies experience the same delay.  Rane Pro Audio Reference


The Group Delay plot can be used to determine the latency of the DUT. The cursor in the plot is placed on a flat portion of the GD response. The latency is read at the Y: field in the Cursors box (lower right). This latency is reported on the Main CAF page (Ref. 8).



Plot G - Group Delay vs. Frequency plot


Plot H - Crosstalk

The crosstalk plot answers the question "How much signal comes out of Channel 2 if I only drive Channel 1?" The DUT is driven by stepped reverse log sine sweep at -3 dB ref. rated sensitivity. There are 10 data points. The plots shows the output of Channel 2 when Channel 1 is driven. The test can be repeated for any channel pair, and for various load impedances. Crosstalk tends to increase with frequency, so the 500 Hz to 20 kHz range is tested.


Crosstalk is unwanted leakage or bleed of a signal from one or more channels to other channels within a device. The Crosstalk, One Channel Driven measurement result provides a measurement of the crosstalk into the un-stimulated DUT channel(s), when one channel is stimulated. Audio Precision AP500 User's Manual


Crosstalk is expressed as a ratio. The measured crosstalk level in a channel is divided by the measured level of the stimulus tone in the source channel. Crosstalk is typically stated in decibels (dB).  Audio Precision AP500 User's Manual


The plot shown is smoothed to improve readability.



Plot H - Crosstalk vs. Frequency plot


Plot I - Common-Mode Rejection Ratio (CMRR)

CMRR describes the ability of the DUT to reject a signal that is present to both terminals of a balanced input (i.e. "pin 2" and "pin 3" or "+" and "-"). In a balanced interface, the desired signal is differential between the two input terminals. Interference is common to both, or "common-mode." Here is an overview of the CMRR test sequence. See Plot I for the results.


1. The DUT is driven at -3 dB ref. input sensitivity with a log-sweep. This is its response to a differential signal (Diff trace).

2. The analyzer then feeds the same signal to the + and - inputs. The green trace is the DUTs response to this signal. Since it is "common-mode" it is rejected. The dB difference between the Diff and CMRR traces is the common-mode rejection ratio vs. frequency.

3. The blue trace shows the effect that a 10 ohm imbalance in the source impedance has on the CMRR, as per Bill Whitlock and IEC standard 60268-3.

4. The  red trace shows the CMRR when the DUT is driven from an unbalanced source. The imbalance is 1 kΩ between the output legs of the source. This is of interest if the amplifier is driven from an unbalanced consumer source, such as a surround sound processor.

The "one number" CMRR (Main Page Ref. 9) is the 1 kHz value taken from this plot, for the 10 ohm imbalance condition. Both the + and - terminals should be tested with the imbalance, and the worst of the two included on the CMRR plot and used for the "one number" rating.



Plot I - Common-Mode Rejection Ratio (CMRR) vs. Frequency plot


Plot J - Damping Factor (DF)

While damping factor is nearly always presented as a "one number" rating, a plot of DF vs frequency is supported by many audio analyzers. Here are some factoids regarding DF:


1. It is usually specified for an 8 ohm load. This yields 2x the DF than for a more typical 4 ohm load.


2. It is generally accepted that it is of more significance for LF components (i.e. woofers, subwoofers) than for HF components.


3. It is usually specified for 1 kHz, unless otherwise indicated.


4. It's measurement is VERY sensitive to cable resistance. Your audio analyzer should include a method for compensating for cable resistance.



Plot J - Damping Factor vs. Frequency plot


Plot K - Noise Floor Spectrum

This plot (Plot K) shows the spectrum of the DUT noise floor. The analog input is terminated by 150 ohms, the input sensitivity set at maximum, and the output is monitored on a spectrum analyzer over the amplifier's rated bandwidth.


The noise floor of a resistor or piece of wire is white - a flat line as a function of frequency. If spectral components (spikes) are present in the noise floor, they should be low in level. This plot can be integrated (summed) to give a broadband noise level (Item 5 on the Main page).


The FFT size, number of averages, and window type should be indicated. At least 128K points and 10 averages are recommended.



Plot K - The noise floor spectrum of an amplifier.


Plot L - Pin 1 Response

This is the same plot as K (Noise Floor Spectrum) with with a square wave (50-60 Hz at 80 ma) injected into the shield connection of the analog input. The input (+/-) are terminated by a low valued resistor (e.g. 150 ohms).


If this test produces harmonically-related spikes in the noise floor, it shows that the injected signal is finding its way into the audio path. This indicates susceptibility of the DUT to current flowing in on the cable shield, which can be caused by ground loops and other sources.


The "Pin 1 Problem," as coined by Neil Muncy, is well documented in the publication AES48 - Grounding and EMC Practices - Shields of connectors in audio equipment containing active circuitry.




Plot L - Response of DUT to "Pin 1" test, indicating susceptibility to EMI on the cable shield.


Plot M - Fan Noise Spectrum

If the amplifier is to be used in a quiet environment, the fan noise level and spectrum is of interest.


If the amplifier is not fan-cooled, then this test should be omitted.


1. The amplifier is placed in a quiet test environment, on a pedestal at least 1 m from the nearest room boundary. The test environment should have a noise floor at least 10 dB below the fan noise level.



2. The following acoustic measurements are performed using a calibrated spectrum analyzer. The measurement is made normal to the front of the amplifier at a distance of 1 m.


Lamb - Ambient noise level of test environment


Lfan-min -  Noise level at idle.


Lfan-max - Noise level with fan at maximum speed


The noise spectra are overlaid on this plot. The A-Weighted Slow maximum level is stated on the Main page (Ref. 12).



Plot M - Fan Noise Spectrum



Plot N - Short-Circuit Behavior

This shows the response of the DUT to a simulated short-circuit across its output. The "short" is actually 0.16 ohms, which emulates a short at the end of 50 ft of AWG 12 wire. Not all amplifiers have short-circuit protection, and this test is only performed with permission from the manufacturer, as it can cause permanent damage to the DUT.


The DUT is driven at 1/8 power with pink noise (a simulated program condition) and the short applied. The plot shows the output signal as a function of time. Additional annotation describes how the device behaves, i.e., does it remain in Stand-By until the short is removed, or does it periodically try to recover and continue playing? There is no consensus on the appropriate behavior, but it is of interest to the sound system designer.



Plot N - Short-circuit behavior


Appendix A - Input Impedance by the Voltage Divider Method


1. The DUT  is driven through a 500k ohm potentiometer from a low impedance sine wave generator (1 kHz), as shown in this diagram.



2. With the DUT disconnected, adjust the generator to produce 1 Vrms on the voltmeter.


3. Connect the DUT, and adjust the potentiometer until the voltmeter reads 0.5 Vrms.


4. The potentiometer resistance is equal to the input impedance of the DUT. Measure the combined resistance of both pots, or, measure one of them and multiply by 2. This is the input impedance of the amplifier.


Appendix B - Test Jig for CMRR Measurement