PARAMETRIC MEASUREMENT OF CLASS-T AMPLIFIERS Revised: March 000 Copyright 997-000 Tripath Technology, Inc All Rights Reserved Introduction Audio amplifiers are commonly specified by and evaluated against generally accepted figures of merit, such as power output, bandwidth, distortion, and signal to noise ratio While most users agree that these criteria are important and useful, the methods employed to measure them have changed over time Many of the techniques used to measure tube-type amplifiers, for instance, had to be modified to test the new (at that time) solid-state designs Advances in the design and technology of both the amplifier and test equipment have necessitated changes in the standard test methodology This article describes additional changes required by the advent of Class-T amplifiers The possibility of confusion exists in measuring the THD+N performance of a Class-T amplifier and comparing that analytical result to the audible listening test Measured levels typically around 00% THD+N are best in class for a switching amplifier yet they don t fully represent the richness of the amplified sound on a Tripath device Some explanation is helpful An analog amplifier replicates an input signal and adds harmonics dependent on its non-linearities Figure shows a khz tone and its harmonics for a typical analog amplifier Generally, the slewing, clipping, and other problems of an analog amplifier manifest themselves as lower order harmonics that easily fall within the audio band Thus the EIA standard THD+N specification was written such that measurements could be made with an analog RMS meter after removing the fundamental with a notch filter This RMS voltage is then compared to the fundamental khz tone resulting in a THD+N measurement that approximates THD AN - Parametric Measurementdoc
Figure : A tone with typical analog amplifier harmonics Harmonics khz Frequency Uniqueness of Digital Amplifiers A digital amplifier, however, works on completely different principles than an analog amplifier The digital amplifier s goal is to faithfully replicate an input signal while driving its switching harmonics outside of the audible band to increase amplifier efficiency and to maintain fidelity Since these displaced switching harmonics and energy have been removed from the audio range as shown in Figure, Tripath s digital amplifiers are quite linear and have low noise in the audible band Harmonics Displaced harmonics and energy outside audible band khz Frequency Figure : FFT of Tripath Class-T Amplifier AN - Parametric Measurementdoc
Band Limited Output A low pass filter is usually placed on the output of a digital amplifier to efficiently remove the displaced harmonics and energy so they do not dissipate inaudible energy in the speaker This is where an understanding of digital amplifier principles can save considerable filter cost and design effort While intuition may suggest all of the displaced harmonics and energy need to be removed to maintain high fidelity, the truth is much of the inaudible energy can be passed to the speaker with only a degradation in efficiency Effects of Output Filters One contention in favor of completely removing the inaudible displaced harmonics and energy is that if two adjacent displaced harmonics are separated by khz, the result will be a khz tone modulated to the audible band by either non-linearities in the speaker, the air or the ear However, there is a very simple experiment to show that this is not a problem for even a nd order LC filter at 00kHz First, create a high order 0kHz filter that removes virtually all of the displaced harmonics and energy but does not droop in the pass-band Then create the relaxed filter desired to compare to the ideal filter Since the displaced energy exists even without an input tone, a listening comparison can check for audible energy in both cases There will not be a noticeable difference when switched between the filters because even while using a crude filter, the displaced harmonics are at least 0dB down And even if the speaker, the air and ear combined were nonlinear enough to mix two adjacent displaced harmonics down to the audible range at a % level, the resulting audible tones will only be at the -80dB level This is, of course, only as bad as a harmonic from an amplifier operating at 00% distortion One experiment that can be done to further alleviate the concerns of the skeptical is to hook the amplifier directly to a speaker with no filter at all In this case, the quality of sound going to the speaker is still indistinguishable from the example using an ideal filter This further indicates that a loosely designed filter that is flat in the pass-band will not affect sound quality When doing this experiment, great care must be taken to heat sink the amplifier, and a speaker with a high power rating must be used This is because the amount of power present at the speaker is roughly equivalent to connecting one terminal to and the other to, even if no audible sounds are sent to the speaker It is recommended only doing this experiment in short bursts to prevent overheating the amplifier AN - Parametric Measurementdoc
Filter Design With digital amplifiers, therefore, filter design is done from an efficiency perspective rather than a fidelity perspective Designing for efficiency means inaudible energy at the -0dB level may be allowed to pass to the speaker for the sake of filter cost It also means using an instrument that notches out the fundamental and measures the remaining RMS voltage will not provide an accurate THD+N measurement for a digital amplifier; a spectrum analyzer is needed to faithfully capture the linearity of any switching amplifier Using a spectrum analyzer matches the EIA specification, that indicate the preferred way of measuring THD, allows for a spectrum analyzer with a bandwidth of 0kHz Nevertheless, a new specification would be better for use with digital amplifiers This specification could be called THD+N0 where the bandwidth used by the spectrum analyzer is exactly 0kHz thereby ignoring all erroneous displaced harmonics and energy Test Methodology Overview Testing the performance of the Tripath Technology Class-T amplifier is very straightforward In order to obtain valid comparative data, traditional Class-A, and Class-AB linear amplifier measurement techniques should be amended as described below: The output of a bridged Class-T amplifier is unlike that of most conventional linear amplifiers In general, where a single-ended linear amplifier s negative output terminal can be connected to it s chassis ground, the negative and positive outputs of bridged amplifiers must never be connected to ground The negative/positive output should be connected only (through the load) to its corresponding positive/negative output Consequently, the use of differential probes on all test equipment connected to the amplifier s outputs is mandatory Measurement of THD+N Unless your distortion analyzer is designed to reject out-of-band noise when measuring THD (and very few analyzers are), the THD measurement actually reflects harmonic distortion plus any noise present up to the bandwidth of the analyzer and so is actually a measurement of THD+N Since a Class-T amplifier has measurable energy far outside the audible frequency range (0Hz-0kHz), a test designed for a linear amplifier will not yield valid results when applied to a Class-T amplifier A sharp cutoff low-pass filter must be inserted between the Class-T amplifier s outputs and the test equipment s inputs prior to making any distortion measurements Tripath can supply an appropriate filter with our evaluation units Alternately, some measurement equipment (such as the Audio Precision ) have high-quality low pass filters which can be used to limit the test AN - Parametric Measurementdoc
bandwidth Figure shows the recommended test setup for measuring a Class-T amplifier with a typical distortion analyzer Test setup follows: Attach a load across the output jacks of one channel of the amplifier board Connect the IN+ and IN- terminals of the filter board across the load Connect the filter board s power terminals to a dual V power supply Connect a V supply to drive the onboard gain changing relay circuit Note: The V supply is not needed when testing a TA0B or TA00-00 Connect the THD+N test equipment to the 0kHz or 0kHz filter output Apply power to the Tripath component and then the filter board Apply test signals to the Tripath component Make the measurements at the output of the Tripath filter board V V + + V + Tripath Filter Board V Tripath Eval Board OUT+ R L IN+ VR 0K POT 0kHz THD+N Test Equipment OUT- IN- VR 0K POT 0kHz Rev H Figure : Recommended test setup for measuring a Class-T amplifier There are two outputs on the filter card: 0kHz and 0kHz The 0kHz output is the standard output The 0kHz filter output has been included for compliance with the EIAJ testing standard They will yield the similar results The filter board comes calibrated for use with the Tripath TA0B or TA00-00 amplifier There are two variable resistors, VR and VR, that the user may need to adjust when testing with TA0X based amplifiers: VR adjusts common mode rejection This board comes calibrated from the factory, but if the input attenuation circuit is changed (when using power supplies greater than +0V), VR may need to be trimmed To calibrate it, apply the same V rms sine wave in phase to both inputs Look at the output with a scope and trim VR until the output is minimized AN - Parametric Measurementdoc
VR is a potentiometer that forms a voltage divider that allows the user to set the gain change trip point to maximize the performance of the filter board The onboard relay allows for two different filter board gains depending on the input signal level at IN+ The first range is a fixed db attenuation caused by the combination of R00 and R0 on IN+, and R0 and R0 on IN- The attenuation factor of the second range is determined by the additional parallel resistance seen by the inputs when the relay is closed (R0 on IN+, R0 on IN-) The value of VR determines when the filter board will switch from Range to Range Assuming a gate threshold of V for Q, VR should be adjusted to about 00 Ohms This attenuation will cause the relay to trip (therefore switching the filter board gain) before the input op amp starts to overload Table is used to choose an appropriate attenuation factor for the second gain range The required attenuation factor is supply voltage dependant The additional attenuation makes sure that the op amps on the filter board do not clip with large input signals Vpp is the magnitude of positive supply in a traditional split supply, single ended amplifier configuration (ie TA00XA designs) See Table for R0/R0 resistor value selection Amplifier Loaded Supply Voltage Resistor Value Vpp < 0V 99K* 0 < Vpp < 0V 9K 0 < Vpp < 0V 00K 0 < Vpp < 70V 0K 70 < Vpp < 80V K 80 < Vpp < 90V 00K Table : R0/R0 Resistor Value Selection *The filter board comes stuffed with 99KΩ for R0/R0 AN - Parametric Measurementdoc
When your analyzer s standard THD+N test is performed, the results will meet or exceed Tripath s published specs and will be valid for comparison with non-class-t amplifiers Typical THD+N performance for a Tripath TA0B amplifier is shown below in Figure THD+N versus Output Power 0 V S = v R L = 8Ω A V = 7 THD+N (%) 0 00 0 0 Output Power (W) Figure : Typical THD+N versus Output Power Performance for TA0B Measurement of IHF-IM Measurement of the Intermodulation Distortion (IHF method) proceeds just as it would with a linear amplifier, but with the output of the filter board replacing the output of the Tripath component The IHF method specifies that 9kHz and 0kHz sine waves be applied to the input of the Tripath component Measurements are then taken at the output of the filter board at 9kHz, 0kHz, khz, 8kHz, and 7kHz The IHF-IM distortion is the root sum square of these values taken as a percentage of the root sum square of the amplitudes of the 9 and 0kHz signals The EIA specification also recommends that a spectrum analyzer be employed to confirm (via an FFT display) that all distortion components are measured The formula for calculating IM-IHF distortion follows IM distortion % = 00 x ((f + f +f ) / / (f +f ) / ), where: f = RMS voltage @ 9kHz f = RMS voltage @ 0kHz f = RMS voltage @ khz f = RMS voltage @ 8kHz f = RMS voltage @ 7kHz AN - Parametric Measurementdoc 7
Measurement of Channel Separation Measurement of channel separation involves the detection of any energy at the output of one channel while a test signal is applied to the other channel Energy present in the quiescent channel is assumed to be cross-talk from the active channel and is measured to derive the channel separation figure As mentioned above, there is always energy present at the outputs of a Class-T amplifier, but it is outside the audible range This energy will throw off the cross-talk measurement, as the test equipment will assume that this energy is leaking through from the active channel This is not the case Again, however, applying the low-pass filter will remove this out-of-band energy and will allow valid results from a linear test unit This measurement requires that the input of the channel under test be AC grounded to prevent spurious noise from affecting the results Shorting the input of the undriven channel will minimize PCB coupling effects Before starting though, you must calibrate the output of the source channel with the filter board attached Connect the filter board to the source channel and set the amplitude on the signal generator so that the output of the filter board is V RMS Leaving the signal generator at that level, connect the output of the undriven channel to the filter board input Sweep the signal generator from 00Hz to 0kHz and measure the worst-case output of the test channel The channel separation is 0 log (/(test channel output)) Measurement of SNR Measurement of the signal-to-noise ratio proceeds just as it would with a linear amplifier, but with the output of the filter board replacing the output of the Tripath component under both no signal and full scale conditions Measurement of PSSR To measure the Power Supply Rejection Ratio, you must use a power supply with a summing input or add the optional circuitry shown in the test setup in Figure (diagram is for TA0B and TA00-00 PSRR measurements) This test also requires that you measure the gain of the filter board To measure the gain, simply measure the RMS value of any sine wave at the output of the Tripath component, and then again at the output of the filter board The ratio of the two values (Vamplifier / Vfilter board) is the gain of the filter board and is denoted by k With no input to the amplifier itself, a V RMS sine wave is applied to the summing node The sine wave is then swept from 0Hz to 0kHz During the sweep, measure the highest amplitude at the output of the filter The PSRR is 0 log ((V x k)/v), expressed in db, where V = V RMS and V is the highest measured output Note that if this test is done on the TA0B reference board the VDD decoupling capacitors must be removed 8 AN - Parametric Measurementdoc
V V + + optional R L = Ω V 0,000 µf Tripath Filter Board Tripath Eval Board OUT+ + VDD of TA0B R L IN+ VR V 0K POT 0kHz THD+N Test Equipment OUT- IN- VR 0K POT 0kHz Rev H Figure : PSRR amplifier and filter board test setup AN - Parametric Measurementdoc 9
D C BANANA IN+ BANANA IN- POS_INP C00 0uF, V C0 0uF, V R0 / R0 Resistor Value Selection Table C0 0uF, V C0 0uF, V Amplifier Loaded Supply Voltage Resistor Value Vpp<=0V 99K 0<Vpp<=0V 9K 0<Vpp<=0V 00K 0<Vpp<=70V 0K 70<Vpp<=80V K 80<Vpp<=90V 00K R00 99k R0 99k RLY RELAY-DPST R0 99k R0 99k V R0 99k R0 0k + - R0 99k 8 POS_INP R08 99k 7 UA OPA0 R07 99k C0 0uF, V R09 0k C0 00pF C0 00pF R0 99k VR R 0k 0k POT Adjust VR to minimize output with common mode signal at IN+ and IN- C07 0uF, V R 0k Adjust VR to set gain change trip point VR 0k POT R NS CONOUT UA 8 OPA0 7 R 0 CONOUT D C B Q N7000 R 00k C09 uf, V C08 00uF, V D N8 B Nominal Supply voltage for OPA0 +/-V to BANANA BANANA A BANANA BANANA V V C0 0uF C 0uF BY-PASS CAPACITORS C 0uF C 0uF Title Input Section with Gain Change Circuit Size Number A Date: 7-Mar-000 Sheet of File: C:\LAYOUTS\0K-0~\FLTST-RASCH Drawn By: Revision A
R 0k OUT R0 k OUT R 8k D C8 00pF C 00pF C 00pF D C7 00pF C0 70pF C 00pF 0K-OUT 0KHZ BNC BNC(RP) CONOUT CONOUT R 0k R 9k 8 UA OPA0 R8 k R9 7k 7 R 8k R 7k 8 OPA0 UA R8 0 7 C 00uF R7 k R C9 k 00uF C 00uF R k R9 k C nd Order 0 khz LPF Q = 7 nd Order 0kHz LPF Q = 8 nd Order 0kHz LPF Q = 0 C 0uF C 0uF C 0uF C 0uF C BY-PASS CAPACITORS C7 0uF C8 0uF C9 0uF C0 0uF R8 98 OUT R 9 OUT R k C7 00pF C0 00pF C 00pF C 00pF C9 70pF C 00pF 0K-OUT 0KHZ BNC BNC(RP) B CONOUT CONOUT R 98 R7 0k 8 OPA0 UA R0 9 R 78k 7 R k R 7k 8 OPA0 UA R0 0 7 B C 00uF R9 k R C8 k 00uF C 00uF R7 k R k nd Order 0 khz LPF Q = 7 nd Order 0kHz LPF Q = 8 nd Order 0kHz LPF Q = 0 A A Title 0kHz / 0kHz Filter Stage Size B Number Revision Date: 7-Mar-000 Sheet of File: C:\LAYOUTS\0K-0~\STAGESSCH Drawn By:
C:\LAYOUTS\0K-0~\FILTMSTRBOM 7:0: 7-Mar-000 Bill of Material for C:\LAYOUTS\0K-0~\FILTMSTRPrj Used Part Type Designator Footprint Part Field ==== ========== ==================== ============ ================= 0 R R8 R0 0 %, METAL FILM 00uF C C9 C C C8 0 %, NPO C 0uF C0 C0 C C 0 %, NPO C C C C C C7 C8 C9 0k R R 0 %, METAL FILM 0k R7 0 %, METAL FILM k R8 R0 0 %, METAL FILM 78k R 0 %, METAL FILM 9k R 0 %, METAL FILM 00k R 0 %, NPO 00pF C0 C0 C C 0 %, NPO C C 00uF, V C08 RADPO0AX 0%, ELECTROLYTIC 0k R0 R09 R R 0 %, NPO 0k POT VR VR VR7 BOURNS 0uF, V C00 C0 C0 C0 SM D Size 0%, TANTALUM C0 C07 00pF C7 C8 C C C7 0 %, NPO C0 8 k R7 R R R9 R 0 %, METAL FILM R7 R9 R N8 D 0 %, NPO uf, V C09 RADPO0AX 0%, ELECTROLYTIC k R R 0 %, METAL FILM 7k R9 0 %, METAL FILM 70pF C0 C9 0 %, NPO N7000 Q TO9 FAIRCHILD 8k R R 0 %, METAL FILM Page
C:\LAYOUTS\0K-0~\FILTMSTRBOM 7:0: 7-Mar-000 7k R 0 %, METAL FILM 99k R07 R08 R0 0 %, NPO 99k R00 R0 R0 R0 0 %, NPO R0 R0 98 R R8 0 %, METAL FILM 7k R 0 %, METAL FILM 9 R0 R 0 %, METAL FILM BANANA V IN+ IN- GBNA00 * BNC(RP) 0KHZ 0KHZ GBNC00 * NS R 0 %, NPO OPA0 U U U U U U SO-8 BURR BROWN RELAY-DPST RLY RLYDPST HAMLIN