14 fasttest. Multitone Audio Analyzer. Multitone and Synchronous FFT Concepts

Size: px

Start display at page:

Download "14 fasttest. Multitone Audio Analyzer. Multitone and Synchronous FFT Concepts"

Howard Hampton
6 years ago
Views:

1 Multitone Audio Analyzer The Multitone Audio Analyzer (FASTTEST.AZ2) is an FFT-based analysis program furnished with System Two for use with both analog and digital audio signals. Multitone and Synchronous FFT Concepts Multitone testing with synchronous generator and analyzer waveform buffers provides many advantages: Testing speeds with multitone techniques are 10 to 100 times faster than conventional swept sinewave techniques. Multitone testing can provide total distortion and noise measurements, including both harmonic and intermodulation distortion products in one test Synchronous multitone testing is the only known technique for measuring noise in the presence of a test signal, providing for the first time meaningful noise measurements on dynamic processors such as compressors, modulation processors, and noise gates A multitone signal is much more realistic than a single sinewave, with both spectral and dynamic characteristics (crest factor) similar to music and voice A brief (fractional second) inserted burst of multitone signal in a broadcast program or recording can be recognized and captured by an analyzer for regular, unattended testing of broadcast transmission facilities Multitone signals provide realistic stimulus to low bit rate encoders and the output can be analyzed according to psychoacoustic models including frequency masking effects System Two User's Manual for APWIN version 2 Page 14-1

2 Chapter 14 Multitone Audio Analyzer Multitone and Synchronous FFT Concepts The major advantages of multitone testing depend on synchronous signals and window-less FFT analysis. As discussed in the FFT-Based DSP Program chapter (see page 12-8), windowing functions are required when analyzing non-synchronous signals by FFT. Application of a window function, however, causes energy to appear to spread over 5 to 10 FFT bins above and below the actual signal frequency. A synchronous signal is one with an exact integer number of signal cycles in the FFT analysis transform buffer. With a synchronous signal, each FFT bin represents a perfect rectangular selectivity curve one bin wide with no spillover to adjacent bins. Synchronous signals are created by: having a generator waveform buffer size equal to or an integer sub-multiple of the analyzer FFT transform length using the Utilities Multitone Creation menu command to create waveform files in which every multitone signal component is forced to have an exact integer number of cycles in the generator buffer length. See the Utilities Multitone Creation description on page Multitone Signals A typical multitone test signal consists of a number of sinewaves at frequencies distributed across the audio spectrum. It is usually convenient for all the sinewaves to have the same amplitude, but amplitudes may be set individually to more closely match typical spectral energy distribution of program material, for example. Also convenient is for the sinewave frequencies to be logarithmically spaced across the spectrum, such as at 1/3-octave intervals. However, any arbitrary spacing scheme may be used which does not violate the synchronous condition. The practical result of a synchronous signal is that all signal frequencies must be an integer multiple of a basic frequency which is the sample rate divided by the generator waveform length. For example, with a 48 khz sample rate and the maximum System Two DSP generator buffer length of 8192, the corresponding basic frequency is Hz. It is normally desirable not to locate signals at the exact 2nd or 3rd harmonic of lower frequency signals in order to leave those bins free for measurement of harmonic distortion products. Page 14-2 System Two User's Manual for APWIN version 2

3 Multitone and Synchronous FFT Concepts Chapter 14 Multitone Audio Analyzer Multitone Analysis The multitone signal, after passing through the device under test, is captured into the analyzer and an FFT performed. Rather than sending a complete FFT analysis to the computer for display, additional post-processing is done following the FFT to extract the most relevant audio performance information: Frequency response is plotted by the DSP unit sending to the computer only the amplitudes of the FFT bins containing the fundamental sinewave products of the test signal. Phase is plotted by the DSP sending to the computer the phase values at the FFT bins containing fundamental sinewave products. Total distortion and noise is plotted by the DSP unit integrating the amplitudes of all FFT bins except those which contain fundamental sinewave signals. Noise in the presence of signal depends upon the generator buffer length being exactly half the analyzer transform buffer length, creating empty bins in the analyzer which will contain noise from the device under test but no generator-related signals. For psychoacoustically-based analysis of coders and decoders, DSP post-processing after the FFT can generate a composite frequency masking curve for any multitone signal. Crosstalk function requires use of a stereo multitone waveform, with one or more unique frequencies on each channel in addition to any number of tones common to both channels. It determines the frequency of every generator waveform signal which appears in only one channel. Crosstalk then reports to the computer the amplitudes of all FFT right channel bins at unique left channel frequencies, and all left channel bins at unique right channel frequencies. System Two User's Manual for APWIN version 2 Page 14-3

4 Chapter 14 Multitone Audio Analyzer Multitone Audio Analyzer Multitone Audio Analyzer The Multitone Audio Analyzer (FASTTEST.AZ2) is a specialized and augmented FFT (Fast Fourier Transform) program usable for analog or digital domain input signals. It combines the functions of the FASTTEST, FASTTRIG, and CODEC programs of System One. The Multitone Audio Analyzer provides time (oscilloscope view) or frequency domain (spectrum analyzer) views of the signal. This program cannot be used until an appropriate multitone arbitrary waveform file is selected on the generator panel (analog or digital) which is driving the device under test. With multitone test signals as the generator arbitrary waveform, the Multitone Audio Analyzer performs post-fft processing to measure frequency response, total distortion and noise, noise in the presence of test signal, and generate psychoacoustic masking curves in addition to providing conventional spectrum analysis and waveform display. Trigger modes include external, digital generator-synchronized and free running. It can also be made to trigger only upon receipt and recognition of the specific multitone signal presently stored in the generator as a reference. Variable trigger delay may be set to allow devices such as audio processors to settle before measurement. The Multitone Audio Analyzer is normally operated in a synchronous mode with the test signal so that no windowing function is required and maximum theoretical FFT selectivity is achieved. Synchronization is achieved even when the test signal frequencies have been shifted up to ±3% in passage through the device under test, by a frequency error correction technique. The Multitone Audio Analyzer tests low-bit-rate perceptual coders with multitone signals by summing quantization noise and distortion and comparing it to an imbedded psychoacoustic model of the frequency masking effect in humans. Input selection The FFT program operates from either Digital or Analog domain signals. To view and measure digital domain signals, select Digital in the Input field. To measure analog domain signals, select Low Bandwidth A/D (1x). The A/D converter sample rate will be equal to the Internal Sample rate, controllable (as Output Rate) from the DIO panel. Page 14-4 System Two User's Manual for APWIN version 2

Source Selection Chapter 14 Multitone Audio Analyzer Source Selection The Channel 1 (left) and Channel 2 (right) Source fields permit selection of specific signal pick-off points to be measured by

5 Source Selection Chapter 14 Multitone Audio Analyzer Source Selection The Channel 1 (left) and Channel 2 (right) Source fields permit selection of specific signal pick-off points to be measured by the Multitone Audio Analyzer, depending upon the basic Digital/Analog selection in the Input field above. When Digital is selected as Input, the Source field selections for both Channel 1 and Channel 2 of the analyzer are A, B, or None. A and B refer to the A and B subframes of the multiplexed two-channel digital signal. With normal stereo program material, A carries the left and B carries the right channel signal. The None selection should not be used. Figure 14-1 Multitone Audio Analyzer (FASTTEST), large version When the Low Bandwidth A/D selection is made at Input, the useful Source field selections for both channels of analyzer are Anlr-A, Anlr-B, and None. System Two User's Manual for APWIN version 2 Page 14-5

Chapter 14 Multitone Audio Analyzer Peak Level Monitors Figure 14-2 Multitone Audio Analyzer (FASTTEST), small version Anlr A and B are Analog Analyzer circuit points following all input ranging and

6 Chapter 14 Multitone Audio Analyzer Peak Level Monitors Figure 14-2 Multitone Audio Analyzer (FASTTEST), small version Anlr A and B are Analog Analyzer circuit points following all input ranging and balanced-to-unbalanced conversion, but prior to any filtering. These are essentially the same circuit points connected to System Two s front-panel BNC connectors labeled Analyzer Signal Monitors, Channel A and Channel B. The None selection should not be used. Peak Level Monitors The Peak Mon reading fields on the Digital Analyzer panel continually display the digital domain peak amplitude at the output of the Ch 1 and Ch 2 A/D converters. Only digital domain units (FFS, dbfs, %FS, or bits) available for these meters. Clicking the down arrow at the end of the display field, then clicking the desired unit will change the selection. The purpose of these Peak Monitors is to avoid overload of the A/D converters. When Automatic Ranging is in use in the System Two Analog Analyzer, converter overrange should never be a problem. If any of the Analog Analyzer Range controls is fixed, it is the user s responsibility to see that the maximum signal amplitude never exceeds digital full scale. Measurement The Measurement field controls the type of post-processing done to FFT results before they are sent to the computer for display and possible limits comparison. The six selections are Spectrum, Response, Distortion, Noise, Masking, and Crosstalk. To change the selection, Page 14-6 System Two User's Manual for APWIN version 2

7 Measurement Chapter 14 Multitone Audio Analyzer click on the down arrow at the right of the Measurement field, then click on the desired selection. Spectrum: this selection provides a normal FFT spectrum display with no processing except for peak picking. The Spectrum selection is typically used without a sweep table (.ADS file), and with a relative large number of Steps at Source 1 of the Sweep panel to provide good frequency resolution. Typical Steps values are from 250 to 500. If the transform length in use results in more FFT bins in the Start-Stop frequency span being plotted than the number of Steps, peak-picking takes place. With peak-picking, the DSP searches all FFT bins between the previous plotted point and the point presently being plotted and sends the highest bin amplitude in that range as the amplitude of the new point to be sure that no signals are missed. Response: this selection is always used with a sweep table (.ADS file) which lists the exact frequencies of the sinewaves in the multitone signal which are to be used for frequency response measurements. The DSP sends to the computer to be plotted only the amplitudes of the FFT bins containing those exact frequencies, resulting in a frequency response graph. There are typically from 3 to 30 sinewaves in most multitone signals. If the value in the Frequency Resolution field is greater than zero, the DSP performs an RSS (root-sum-square) integration of all the bin amplitudes within plus or minus the Frequency Resolution value around each sweep table frequency and sends the integrated sum value to the computer to be plotted. This mode is intended for frequency response measurements on devices such as analog tape recorders that introduce frequency modulation (flutter) to signals. Flutter spreads each tone s energy across a small region of the spectrum. This reduces the amplitude of the fundamental tone, since the total energy in the fundamental and all sidebands remains constant during frequency modulation. The RSS summation of FASTTEST combines this spread energy back into a single value, much as the human hearing system responds to signals with small amounts of FM. System Two User's Manual for APWIN version 2 Page 14-7

8 Chapter 14 Multitone Audio Analyzer Measurement Distortion: this selection excludes the amplitudes of the FFT bins known (from the generator waveform) to contain fundamental signals. All other bin amplitudes are summed (RSS) between each adjacent pair of frequencies requested from the DSP by the computer. It is thus not necessary to use a sweep table (.ADS file) listing the fundamental frequencies of the sinewaves in the multitone signal being used. Distortion and noise can thus be summed across spans determined by the Sweep panel Start, Stop, Log/Lin, and number of Steps, or the spans can be determined by a sweep table. If it is desired to sum the noise and distortion into critical bands, a sweep table can be used which defines the edges of the human hearing system critical bands. The resulting distortion and noise curve is normally compared to the composite masking curve generated in Masking function (see below). If the value in the Frequency Resolution field is greater than zero, the DSP also excludes all the bin amplitudes within plus or minus the Frequency Resolution value around each sweep table frequency before sending the integrated sum value to the computer to be plotted. This mode is intended for distortion measurements on devices such as analog tape recorders that introduce frequency modulation (flutter) to signals. Flutter spreads each tone s energy across a small region of the spectrum. If these close-in sidebands which fall outside the bin containing the fundamental are not to be measured as distortion, they must be excluded, much as the human hearing system masks low amplitude signals nearby in frequency to a stronger signal. Noise: this selection may be used with a sweep table (.ADS file) listing the fundamental frequencies of the multitone signal in use, but need not be. Noise mode depends on the FASTTEST FFT Transform length being set to the value twice the length of the waveform file that generates the multitone signal. The analyzer frequency resolution is thus twice the resolution of the generated signal. The result is that every alternate analyzer FFT bin falls between bins at which the generated signal could contain fundamentals or bins into which harmonic or intermodulation Page 14-8 System Two User's Manual for APWIN version 2

9 Measurement Chapter 14 Multitone Audio Analyzer distortion products could fall (assuming that the device under test does not shift fundamental frequencies or produce frequency modulation). The amplitude of these alternate empty bins consists of noise generated in the device under test, largely unaffected by fundamental signals or distortion. If the same sweep table is used in Noise mode that is used for response and distortion measurements, the resulting graph will be a spectrum analysis of noise in the presence of test signal. If a two-point sweep is made with Start at 20 Hz and Stop at 20 khz, for example, the plotted value at 20 khz represents the RSS integration of all empty bins across the audio band Masking: this selection generates a composite masking curve for the particular multitone signal in use. The shape of the curves is based on a model published by psychoacoustician Brian Moore in the Proceedings of the AES 12th International Conference, June 1993, pp The shape of the curves varies with frequency. The center frequency of each section of the composite masking curve is located at the fundamental frequencies present in the waveform file downloaded to the generator buffer. The reference amplitude at each frequency is determined by the measuring the amplitude at each fundamental frequency. The masking curve is normally used by saving it as a limit (.ADL) file, then comparing noise and distortion (Distortion selection) integrated across critical bands to the limit curve. For Crosstalk function to work properly, there must be at least one unique frequency in each of the generator channel waveforms in addition to sinewaves common to both channels. Crosstalk function automatically determines from the two generator waveform buffers which frequencies are unique to each channel. Then, Crosstalk measures the amplitude on the opposite (non-driven) channel at each of those unique frequencies. If Crosstalk is used without a Sweep Table, the resulting graph consists of a series of horizontal plateaus, each centered around the frequency of a crosstalk tone and scaled vertically to the measured crosstalk level of that tone. Results are easier to interpret when a System Two User's Manual for APWIN version 2 Page 14-9

10 Chapter 14 Multitone Audio Analyzer Frequency Resolution Sweep Table is used. The table consists of a list of the approximate frequency of each pair of unique tones, assuming that the left-only and right-only tones in the test signal are fairly closely-spaced pairs. Sample crosstalk waveforms have been included in the C:APWIN\WAVEFORM folder. XTLKLEFT.AGM, XTLKRIGH.AGM are monophonic waveforms for left and right channel buffers respectively. XLTK.AGS is a stereo waveform with 4 frequencies different on each of the channels. Frequency Resolution The Frequency Resolution field is a numeric entry field with % units. The user may enter values up to 13% which are used to control triggering and error correction and in Response and Distortion Measurement functions. For recognition and triggering on short bursts of externally-applied multitone signal, the Frequency Resolution field must be set to a value at least as large as the percentage that frequencies may have been shifted in the device under test. For example, to capture a multitone burst from an analog tape recorder whose speed may differ by as much as 2% from the tape machine which recorded the signal, a value of 2% or greater must be entered into the Frequency Resolution field. In Response function, the amplitudes of all FFT bins within plus and minus the Frequency Resolution value of each sweep table value are combined in RSS (root-sum-square) fashion and furnished to the computer as the integrated amplitude of the bins within that range. The purpose of this function is to provide accurate frequency response measurements of devices with wow and flutter. Wow and flutter spreads the energy from a single tone across a narrow spectral band. In Distortion function, the amplitudes of all FFT bins within plus and minus the Frequency Resolution value of each sweep table value are excluded from the RSS computation of energy falling between tones. Distortion function defines all signals other than the fundamental tones as distortion and noise. Entering a non-zero value of Frequency Resolution causes flutter sidebands to not be included in the distortion measurement. Page System Two User's Manual for APWIN version 2

SYSTEM ONE * DSP SYSTEM ONE DUAL DOMAIN (preliminary)

SYSTEM ONE * DSP SYSTEM ONE DUAL DOMAIN (preliminary) Audio Precision's new System One + DSP (Digital Signal Processor) and System One Deal Domain are revolutionary additions to the company's audio testing