Knowledge BaseHOME/SERVICE & SUPPORT/KNOWLEDGE BASE

Audio.TST July 2008 Multitone theory & Using your AP to do a Power Supply Rail Health Check

Created 02 Jul 2008

Notes from the Test Bench

By Bruce Hofer, Chairman & Co-Founder, Audio Precision

Hello addressee_placeholder and welcome to the July issue of Audio.TST.

First, I'd like to welcome our newest international distributor, HITECH Eletrônica of Brazil. With 34 years of experience in engineering and T&M, I'm confident they'll be able to serve AP's Brazilian customers extremely well.

Following last month's releases of APx v2.2, APx520/525 and the HDMI option for the APx585, we're now working hard on APx v2.3 and new hardware modules to make APx even more versatile.

In late August I will be giving a half-day seminar at the Technical University of Denmark in Lyngby (a suburb of Copenhagen) as part of their week-long graduate level course in Switch-Mode Amplifier Design. This course attracts engineers from many different countries, and I expect to see some of you there!

Finally, Audio Precision will also be at IBC in Amsterdam in September. We'll be in Hall 8, stand 8.C78.

Bruce

Output: Tech tips and new applications from AP

Using multitones in audio test

AP's manager of Technical Support, Joe Begin, looks at the theory behind multitone audio test.

This month, we’ll review some of the theory behind multitone analysis and its application to audio testing. Multitone testing is the foundation on which HST, Audio Precision’s High Speed Test Application for the 2700 series and ATS-2 audio analyzers is built. HST was featured earlier this year in the January issue of Audio.tst.

As the name implies, a multitone stimulus signal consists of multiple sine waves (or tones) at different frequencies combined together. Any number of tones can be used, but 3 to 30 tones spaced logarithmically across the audio band is typical. This feature alone offers several advantages over traditional single tone testing:

It allows the performance of a DUT to be evaluated over a range of frequencies with a single measurement.
The characteristics of a multitone waveform (e.g., spectral content, histogram, crest factor, etc.) give it a much closer resemblance to typical audio program material like music or speech, than a single sine wave.
Multitone distortion is generally better at detecting real-world problems involving clock jitter and sample rate conversion than traditional THD+N measurements.

But the real power of multitone testing comes from the speed advantage that it provides. Multitone testing can result in test speeds that are 10 to 100 times faster than conventional swept sine techniques.

Multitone testing uses FFT analysis and some special digital signal processing techniques to derive a number of measurements from a single acquisition. First, the multitone signal is made “synchronous” with the FFT transform buffer. A synchronous signal results when the frequency of each sine wave present is chosen such that an integer number of cycles of the waveform fits exactly into the FFT transform buffer (Figure 1). With synchronous acquisition, a window-less FFT can be performed resulting in a leakage-free measurement: i.e., for each sine wave present, all the signal energy is contained in one FFT bin, with no leakage to adjacent bins (Figure 2).

Figure 1. A synchronous sine wave.

Figure 2. Example of a multitone spectrum with logarithmically spaced tones.Once a leakage-free FFT has been performed, a number of measurement results can easily be extracted. If the individual tones in the multitone are spread across the spectrum, the amplitude and phase of the FFT bins corresponding to these tones provide the frequency response and phase response of the DUT. The FFT bins between the tones will contain only distortion products and noise. Power-summing the amplitude of these “non-tone” FFT bins yields the total distortion (both harmonic and intermodulation distortion) plus noise.

Multitone testing can also be used to derive crosstalk measurements in a stereo DUT. For this application, a stereo multitone is constructed with tones in one channel at different frequencies from the tones in the second channel. Then, the power at frequencies unique to Channel 1 is measured in Channel 2 and vice versa. This yields a measure of the crosstalk at these frequencies.

Multitone analysis can take advantage of another property of FFT processing: If the FFT transform buffer is made twice the length of the generator buffer, the analyzer will have twice the frequency resolution of the generator. Using this scheme, even numbered FFT bins may contain fundamental tones, distortion products of those tones, and noise. Odd numbered FFT bins, however, can not contain generator-related signals; i.e., the odd numbered bins can only contain noise. This provides a very powerful means of measuring a DUT’s noise in the presence of signal.

Figure 3. Example of multitone measurements available for a signal with a 1.5 kHz fundamental frequency.

While multitone testing often provides significant advantages over traditional test methods, it’s not always the best choice. For example, in an R&D environment, you may need to measure overall distortion performance, or you might want to use the simplest stimulus possible. In this case, traditional single tone sine testing is a must.

So what audio measurements are good candidates for multitone testing? The most obvious ones are those where test time is at a premium - for example, production line environments where a large number of devices need to be tested in a minimum amount of time. Another often time-sensitive application that is well suited to multitone methods is power amplifier testing. Sometimes, an amplifier can only be driven at high power levels for very short periods of time - for example, if one wanted to evaluate a power amplifier’s performance before heat sinks have been installed. In cases like this, multitone testing can be a powerful tool.

Broadcast applications are also well-suited to multitone testing. To minimize program interruption, a multitone stimulus signal can be made very short (less than one second long). And, as mentioned above, multitone stimulus signals are inherently more “program like” than a single tone. A multitone can be made even more program like, by varying the amplitude of the individual tones to more closely match the spectral content of music or speech.

Other applications where multitone methods should be considered are testing of devices whose noise level varies with signal level, such as noise gates, compressors, and some DACs which shut down with zero digital input. For these devices, measuring noise in the presence of signal is a fundamental requirement, and multitone testing provides this important capability.

Audio Precision’s AP2700 series and ATS-2 analyzers provide a rich multitone feature set, with many multitone measurements built-in. AP’s multitone (or Fasttest) analyzers also include a number of features that provide the flexibility to tackle virtually any multitone measurement, including sophisticated multitone triggering and synchronous, windowed or frequency-error-corrected processing. In addition, a powerful multitone creation utility is included to help users build an optimum multitone waveform for their test.

If you haven’t yet used multitone testing for your audio application, we encourage you to consider it. And audio test engineers with production line test needs should be sure to check out HST. It provides a special multitone stimulus bundled with a convenient User Interface and some additional measurement results.

Sound Advice: Audio Test Q&A

Using your AP to do a Power Supply Rail Health Check

by Kendall Castor-Perry, special to Audio.TST

Modern professional-quality audio systems are commonly put together from sub-assemblies which are frequently designed, qualified, manufactured and tested independently. The hierarchical test process, where sub-units are tested separately to a module specification, substantially eradicates expensive whole-system rework caused by building in defective units. The final system test generally concentrates on proving connectivity, full functionality and key audio specifications.

Some potential interactions between sub-assemblies can remain unexplored by this degree of testing, however. These interactions may not be detected in the final, “goods-out” testing of a completed unit. One such interaction to be wary of is that between an audio circuit or module and its power supply. Differences in the characteristics of the supply used for upstream module testing compared to the one installed in the final product can cause distortion and coloration of the audio signal in the channel, as well as crosstalk between channels.

A source of such problems is the difference in output impedance versus frequency characteristics; this depends in detail on the power supply topology and the values of some of the key components employed. The supply current of the audio circuits is almost invariably modulated by the signal being handled (an exception being very carefully designed, single-supply class A amplifying stages). This varying supply current then causes a varying signal voltage to appear on the power supply rails because the output impedance of the power supply is finite. This residual voltage in turn can sometimes be inadequately rejected by the audio circuits themselves. For a detailed series of articles on this subject, showing effects both inside and beyond the audio frequency range, go to http://www.planetanalog.com/features/showArticle.jhtml?articleID=206905337.

Your AP analyzer can help in several ways. Firstly, the effect of finite power supply impedance reacting with signal-varying load current can be monitored directly. This is done by driving the generator output signal through one of the audio channels and monitoring the power supply rail (with a DC value of up to 230 Volts for 2700 Series & APx520/525 or 200V for ATS-2) directly with an analyzer input (note that you can't, however, do this on the APx585/586 analyzers, which have DC-coupled input circuits).

When you use a sine-wave test signal, expect to see the signal on the supply rail contain both the fundamental frequency and a spread of harmonics, with the 2nd harmonic usually being particularly strong. Plotting both the fundamental level and the THD+N as functions of the input signal level as the signal frequency is swept over the audio band delivers a signature for the response of that particular power supply responding to the demands of that particular circuit.

Changing components in either the power supply (a regulator or capacitor, for instance) or the audio circuitry (substituting an op-amp, maybe), can affect this signature. If audio circuits weren't sensitive to what happens on their supply rails, this would be an interesting check of production consistency but wouldn't worry us on audio quality grounds. But they are sensitive to it. Some of this residual nastiness on the power supply rail will find its way into the output signal of the channel you're driving – and into others too. This causes a distorted form of crosstalk - which gives us an ideal method for quantifying the situation using an AP analyzer.

To make use of this, make a crosstalk measurement between two channels in the equipment, paying particular attention to the harmonic components in the 'victim' channel. Now, simple crosstalk caused by signal coupling between cables, circuit board traces and components is generally a linear process. The presence of harmonics in the crosstalk signal is highly likely to be due to coupling through the supplies. It's easy to see this distortion signal on the 'victim' channel because it has no signal of its own, but this signal will also be present on the main channel. As it is likely to have a strong frequency dependent behavior originating in the interaction with the power supply, this can cause detectable timbral changes in the audio signal path, which may be detectable by critical users of the equipment.

Keep a careful eye on these interactions and make sure that subtle changes in module build specifications don't lead to a change in audio performance. Your AP analyzer is a useful partner in keeping this under control!

Guest contributor Kendall Castor-Perry is an audio expert and longtime AP user with 21 years experience as an analog designer and systems engineer. He was also a runner up in AP's recent Test Bench of the Year competition.

Test Results: AP News & Events

UPCOMING EVENTS

IBC 2008 | Amsterdam

Sept 12 -16

Visit the show websiteAES San Francisco 2008 | Moscone Center, San Francisco