Audio.TST June 2008 Matlab and when to go off automatic

Notes from the Test Bench

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

Hello addressee_placeholder

On January 1, 2009, AP will discontinue System One service. We anticipated this day when we announced Conditional Service for the System One in March 2007. Several key components have reached end-of-life and the technician time required to keep these older units within specification means Service is no longer cost effective.

System One had a good run. It enjoyed a 17-year active product life from the first unit sold in 1985 to the last one in 2002. I don't believe very many other high technology products can make a similar claim! However time and technology moves on, and so must AP. System One is no longer a good match for today's performance needs including 192k sample rates, new interfaces like HDMI, and multi-channel applications.

The good news is there is a clear migration path from System One. AP’s new analyzers are faster, more accurate and easier to use than any System One.  The  APx525  in particular was designed as a System One replacement and the 2700 Series long ago assumed the mantle of best performance of any audio analyzer in the world.

We will be looking at migration from System One in more depth in the months between now and January. And of course, your local AP Sales Partner is the best qualified person to look at your application and advise on how to update your test procedures without disruption.

An important note: System Two’s Service “Conditional Service” status is unchanged: We will continue to service System Two units for the foreseeable future.

Bruce

Output: Tech tips and new applications from AP

New Updates for Audio Precision Matlab Toolbox

AP has updated the Audio Precision Toolbox for Matlab.  Matlab is a powerful mathmatics application used in a wide variety of industries such as aerospace and finance where developers need fast number crunching. In audio, Matlab is very useful for DSP. It can also be used for generating customized arbitrary waveforms.

The AP Toolbox consists of five .m files that support the transfer of data between Matlab and the 2700 Series, ATS-2, System Two, and System Two Cascade instruments.

As of June 2008, these have been updated to be fully compatible with Matlab r2008a. Users of the previous version of the toolbox should note that some function parameters have changed, so existing code may need to be updated to use the new versions.

Matlab r2008a

The functions are:

• ap_read_wave.m Loads an acquisition file (.aam for mono files, .aas for stereo) saved from AP equipment into Matlab. This is useful for post-processing acquired data in Matlab. The function automatically detects whether the waveform is analog or digital and scales it accordingly.

• ap_write_wave.m Saves Matlab data as an AP generator file (.agm for mono files, .ags for stereo). The primary application for this function is to allow special waveforms generated inside Matlab to be played back on AP equipment. The function automatically creates the appropriate file type and, if desired, normalizes and dithers the waveform for optimal performance.

• ap_read_filter.m Loads an AP DSP filter file (.afh for highpass, .afw for weighting, .afl for lowpass) into Matlab. This allows existing AP DSP filters to be examined or checked for correctness inside Matlab. Note: DSP filter files are not available in System One or System Two.

• ap_write_filter.m Saves Matlab filters as AP DSP filter files. Matlab is a popular tool for filter design. This function allows you to use Matlab to create filters suitable for AP equipment. Filters of up to 18th order can be saved. Note: DSP filter files are not available in System One or System Two.

• ap_eq_wave.m Cyclically filters generator waveforms. Since the AP generator circularly plays the generator buffer, the generator waveform must be cyclically continuous to prevent a glitch each time the end of the waveform is reached and the waveform stats again. This function allows cyclically continuous waveforms to be filtered inside Matlab in such a way that the resulting filtered waveform is itself cyclically continuous.

More information is available in the help text for each function or contact Tech Support.

Related links & downloads

• AP Matlab Toolbox
• Matlab website www.mathworks.com

Sound Advice: Audio Test Q&A

Ranging and Settling: When to Go Off Automatic

Applications Engineer Jim Williams looks at when it's time to disengage auto-ranging and alter settling to optimize measurement accuracy and speed.

Ranging

When measuring frequency response, THD+N, crosstalk, and other characteristics of audio circuits, Audio Precision analyzers automatically auto-range the input signal and set measurement settling parameters in order to achieve the most accurate results. There are some types of signals, however, where it is advantageous or even necessary to modify the default settings for both the auto-ranging and settling functions.

Auto-ranging of the inputs allows the analyzer to accept a wide range of signals and to automatically set the input range to achieve optimum dynamic range. From the residual input noise of a microphone preamplifier to the full output of a 1.5KW concert power amplifier (a range of a few microvolts to hundreds of volts), the auto-range circuitry adds the appropriate gain or attenuation to the signal path. 

When auto-ranging the input, peak-sensitive detectors select the proper range for the analyzer input signals. This feature can save valuable time.

Test signals with low duty cycles or high crest factors, such as audio program material, cause the ranging relays to chatter as the detectors struggle to find the correct range. One common application that requires fixing a particular range is power amplifier testing. Many modern amplifier designs will go into protection if a continuous full scale sine wave is applied to the input. When testing this type of amp, it becomes necessary to use a sine burst with a low duty cycle (a few cycles on and many cycles off). To set the range properly, the procedure is to first determine the amplitude of the burst portion of the signal, and then set the analyzer to use the first maximum range setting that is larger than this amplitude. For example, if the burst portion of the signal is 8.2 V, then select the 10 V range.

Two extremes must be balanced when selecting a fixed range on the analyzer input. First, if too low a range is chosen, signal peaks may exceed the full scale value of the detector, resulting in clipping and measurement errors.If the range selected is too high and the value to be measured is smaller then the dynamic range of the detector, (on the order of 50 dB), then the detector cannot make a measurement and the level will be shown as -999 dB.

One other reason to select a fixed range, in addition to optimizing measurement accuracy, is to speed up measurement time. It takes time for the detector to select the right range every time the test is run. The signal level in this case, however, must be within a predictable range.

Settling

In audio testing, just like any other test and measurement application, it is desirable to make measurements under steady state conditions, when readings are stable. When a stimulus signal is first connected, or when a sweep changes steps, transients can get introduced into the generator, the DUT, and the analyzer. These transients can result in unsettled readings. Fortunately, AP audio analyzers include a number of settling parameters to help ensure that stable, accurate measurements can be recorded under a wide variety of test conditions. These features, accessible on the Settling Panel, include several settling algorithms and a number of other parameters that work together to provide a great deal of flexibility.

On first glance, the Settling Panel looks complicated. However, the analyzers have default settings for the various measurements which are optimized to work well in most cases. For example, the Amplitude function defaults to the Flat settling algorithm. This setting mimics the behavior a test engineer would intuitively use when recording fluctuating levels from a real-time meter - it ignores the initial portion of data immediately following the transient, and then waits until a few successive readings agree within “reasonable” limits. However, unlike the manual recording process, the analyzer uses a strict mathematical algorithm that ensures repeatable results every time.

While the default settling parameters work well for most measurement applications, some situations require that they be changed. For example, when measuring a very noisy signal, with the Flat algorithm, readings might not converge to within the specified tolerance, resulting in a “Timeout” condition. This is denoted by a white T above the plotted point on the graph, or a T beside the data point in the Data Editor. When this occurs, a good troubleshooting technique is to turn settling off (switch the settling Algorithm to None) and observe the results. This can help in choosing more appropriate settings for the various settling parameters. For a very noisy signal, the Average settling algorithm is a better choice. This algorithm ignores the initial portion of the reading specified by the Delay parameter, and then computes the average of the specified number of points (i.e., a moving average).

A 2700 Series graph in a timeout condition

Although ranging and settling parameters are usually already optimized and require no user intervention, knowing how and when to alter them is a great tool that audio test engineers can use to help ensure stable and accurate readings and optimize test times.

Test Results: AP News & Events

UPCOMING EVENTS

IBC 2008 | Amsterdam

Sept 12 -16

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

Oct 2-5

Visit the show website

back to top >