Fullscreen HTML5Med Res (??MB)HTML4

Review: APx525 By ALLAN LINTON-SMITH & NICHOLAS VINEN A s with previous Audio Precision analysers, the APx525 is essentially a computer peripheral. It relies quite heavily on digital signal processing and much of this is done using the host PC, connected via USB 2. This is much more convenient than the dedicated interface cards needed for some of the older analysers and a mid-range laptop is more than enough to drive it. Everything is controlled by the computer and the analysis results are displayed on its screen. The APx525 adds many new features and capabilities compared to its predecessors. Its design was influenced by an internet survey of engineers and Audio Precision users. The results of this survey showed a need for a digital audio analyser with HDMI and Bluetooth interfaces along with a 84  Silicon Chip variety of other wish list items. The result is an unprepossessing instrument 432 x 467 x 129mm, weighing 10.9kg. There are no controls on the unit itself – the front panel consists mainly of input and output connectors to interface the unit with the device under test (DUT). In addition to the XLR, BNC and banana connectors for the balanced and unbalanced stereo audio inputs and outputs, it has TOSLINK, BNC and XLR (AES/EBU) sockets for digital audio input and output. Using the software, you can easily set it up to use any combination of these, to suit the instrument you are testing. The digital audio interface supports sampling rates of 22kHz-192kHz and resolutions of 8-24 bits. As well as linear PCM, the digital output can generate Dolby Digital and DTS signals, to interface with home theatre gear. HDMI, Bluetooth and other interfaces are added with optional modules – there is space for both. The combination of analog and digital inputs and outputs allows a variety of equipment to be tested. You can test analog devices such as preamps, amplifiers and filters but then you can also test DACs (digital-to-analog converters) or digital home theatre receivers using a combination of the digital output and analog inputs. Similarly, to test an ADC (analogto-digital converter) you would use the analog outputs and digital input. Devices containing DSPs (digital signal processors) can be tested using just the digital inputs and output. siliconchip.com.au Audio Precision Audio Analyser SILICON CHIP has a long record of using the latest available instruments to validate our designs – typical of this has been our use of the Audio Precision System One to measure all our high performance audio designs. This month we review the latest such instrument from Audio Precision, the APx525. • • • • • • • Performance The new features of the APx525 make testing audio equipment easier and gives more comprehensive results. But what about the performance? In the digital domain of course there isn’t really any distortion apart from the limitations of the digital format itself. But when testing analog gear you don’t want the analyser’s own noise and distortion to interfere with the measurements you are making. The good news is that the on-board analog signal generators are capable of delivering a very low distortion analog signal, to less than 0.0002% (-114dB) and the input circuitry is equally good. Audio Precision is promoting it as the “second best” audio analyser in the world for THD+N (total harmonic distortion plus noise) measurements, siliconchip.com.au Basic Specifica ti ons: Output frequenc y range: 0.1Hz80.1kHz Output frequenc y accuracy: 2 pp m Maximum outp ut amplitude: 30 V peak Maximum inpu t voltage: 300V RMS (balanced) /160V RMS (unb alanced) Input bandwidth : >90kHz Input/output am plitude accuracy : ±0.05dB Input/output am plitude flatness: ±0.008dB Residual input noise: 1.3µV Input/output re sidual THD+N: -105 Digital samplin g rate: 22kHz-19 dB + 1.4uV 2kHz Digital output fo rmats: PCM, D olby Digital, DTS • • • • the “best one” being their 2700 series! That’s when it’s fitted with the AG52 (Advanced Analog Generator), a roughly $1000 option. The standard unit is rated for a typical residual distortion of 0.0004% (-108dB), which is slightly better than our System One. On the demonstrator unit, we actually measured a THD+N of 0.00025% (-112dB) from the analog loop-back test, despite it not being fitted with the AG52 option. We expect if it were fitted, we could have easily achieved the stated figure of 0.00018% (-115dB). Features The software provided with the unit is easy to use and it quickly became apparent that the APx525 can do many things that our System One can not. It has a number of new test modes which can give important insights into the performance of the DUT and it also provides the ability to automate a large number of tests with the press of a button – something that could come in very handy when comparing the performance of similar units or during R&D when you want to see the precise effect of a design tweak. It’s also a boon for production testing. Because this unit uses digital processing, it can easily distinguish between THD+N and harmonic distortion (ie, THD without the noise component). It can even show you the contribution each individual harmonic makes to the overall distortion of the waveform. Alternatively, you can look at the output of the inbuilt spectrum analyser to see the relative height of the peaks at each harmonic frequency. This March 2012  85 Fig.2: the information at the top of the report shows how the APx525 was configured for the tests and how the DUT was connected. analyser can be left running permanently in the corner of the software while you are running other tests or it can be maximised to see more detail. The FFT can be set up with a very large window (1M) and wide bandwidth (up to 100kHz). With the averaging option enabled, it gives a noise floor around -160dB! You can therefore clearly see all the harmonics and their intermodulation products (see Fig.4). Speaking of averaging, another handy feature of the instrument is the ability to automatically perform a measurement several times and average the result. This allows you to easily ignore artefacts in the test results which may be due to external interference (eg, from nearby mains-powered equipment) or to further reduce measurement noise. Measurements There are six main measurements which show all the most important aspects of a given device. Audio Precision call them “the big six” and these are the default measurements enabled when you start a new project. These are: • Level and Gain • THD+N • Frequency Response • Signal to Noise Ratio • Crosstalk • Inter-channel Phase It can run many more tests though, including CMRR (common mode rejection ratio), IMD (intermodulation distortion), stepped level and frequency sweeps (to give you THD+N vs power, THD+N vs frequency), dynamic range, burst (peak output) tests, etc. Many of the tests can be run in different modes. For example, THD+N vs frequency can be measured using a stepped frequency sweep, continuous sweep, multi-tone burst or impulse mode. One of the brilliant aspects of this unit is that you only need to run the continuous sweep, multi-tone burst or impulse tests once and you get a whole slew of results including THD+N, phase response and crosstalk. We found that the continuous sweep tests also gave better frequency resolution and detail compared to the stepped sweep (see Fig.3) although there are still times when the stepped mode is preferable. Returning to the standard or “streaming” measurements, these can be made with stimulus signals other than sine waves. This includes: • Mixed twin tone Sine wave • Special Sine wave • Square wave • Mixed square wave and sine wave • Arbitrary waveform (loaded from hard disk or CD) and • Diagnostic digital signal Software Fig.3: each test result in the report shows the parameters of the test with the results below. In this case, it is a continuous sweep test of THD. We set it to the maximum sweep time of 2.5s; the default is less than a second and in this time the unit can generate THD, THD+N, frequency response, inter-channel phase and a whole slew of other graphs. This also shows the very low distortion levels this unit can measure at around 0.0004%. 86  Silicon Chip The software for this unit requires Windows XP SP2 or later and fully supports 64-bit Windows which is good since it is becoming very common and is pre-installed on most new computers. The stated requirements are 2GHz+ CPU and 2GB+ of RAM, which means it will run on most computers purchased within the last year or so. A screenshot is shown in Fig.5 and the photo on p84. Being familiar with the Audio Precision System One and audio gear testing in general, despite the plethora of siliconchip.com.au new features, we figured out the APx525 software straight away. It’s intuitive and well-organised although it has a few quirks which threw us off at first, since we are so used to the older APWin software. For any features which are not obvious, you can refer to the 636-page printed manual. It is clearly written and comes with good explanations of each feature and function, including the optional ones. They even go to the trouble of explaining the basic principles of each measurement; just the thing to flip through when you are in the mood for a little light reading! The software provided with the machine was already out of date when we received it but it was a simple matter of downloading the latest version (2.9, 103MB) from the Audio Precision website (http://ap.com). One of the best features of the software is the ability to create your own “project”. The project defines the test set-up and consists of any number of individual tests, each configured as required. It’s then just a matter of pressing a button and the instrument runs through them all. The result is a PDF which includes information on the test set-up (see Fig.2) as well as all the figures and graphs (see Fig.3). Very handy! The tests to run are selected from the tree list at the left side of the screen and the settings for each is displayed on the right when it is selected. Each option for a test (bandwidth, start/stop frequency, sweep time etc) can then be adjusted. You can also control the formatting of the resulting graph: axes, trace styles and so on. It’s pretty easy once you understand how the tests work. For use on a production line, all the test measurements can be preset and locked in. The instrument can also be set up with limits and will simply display a “go” or “no go” to indicate whether the DUT is operating to spec. Options waves. It is also required for the DIM IMD (intermodulation distortion) test modes (but not the SMPTE, MOD or DFD IMD modes which are supported by the base unit). The APx-BW52 (Ultra-high Bandwidth) option extends the bandwidth to 1MHz for one channel or 500kHz for two channels. This is useful for observing very high order harmonics and digital switching noise in audio signals. The DSIO (Digital Serial I/O) option adds serial audio I/Os to the front panel, allowing direct interfacing with DACs, ADCs and DSPs. It enables devices to be more easily tested at the board level and supports I2S and similar formats, with up to eight channels. Other models There are four models in this range: • APx520 – two channels, analog I/O only • APx521 – four channels, analog I/O only • APx525 – two channels, analog and digital I/O (reviewed Fig.4: this shows the APx525’s FFT display with the maximum window size (1M) and 8x averaging. The noise floor is extremely low (-160dB). Ch1 shows the residual distortion from our Audio Precision System One’s analog signal generator at 1kHz while the mauve display underneath shows the output from the APx525’s own signal generator. The APx525 has less 50Hz hum and less spread around the 1kHz fundamental but more high-frequency noise. As well as the aforementioned HDMI and Bluetooth options, you can also get a separate input/output switching unit that plugs into the Aux Control sockets on the back of the APx525. This is then controlled by software and allows an even greater degree of testing automation, suitable for use in production lines and so on. The HDMI option interfaces with the digital audio I/O on the unit and is features the same PCM, Dolby Digital and DTS signal handling. Since HDMI carries audio and video, the video stream can either be an internally generated test pattern or supplied by external equipment. The APx525 and its siblings are currently the only distortion analysers we know of with HDMI support. For Bluetooth, the unit supports HFP and HSP profiles for voice and A2DP for music. With A2DP, it can operate as either an audio source or sink (not both) while the other profiles are bidirectional (full duplex). Fig.5: another screenshot, this time showing more test options at the left. The The APx-AG52 option, mentioned instrument is running an acoustic response test and is set up to average the earlier as giving lower distortion re- result of eight runs. Between the list of tests at the left and the result graph at the siduals, also allows the signal genera- right are the settings for the current test. The status bar at the bottom shows how tor to produce fast rise/fall time square the instrument has been configured. siliconchip.com.au March 2012  87 This close-up of the front of the APx525 shows that it is not designed to win any beauty pageant . . . but it will win when it comes to function and operation! here) • APx526 – four channels, analog and digital I/O There are several other current Audio Precision models, both cheaper and more expensive but those are outside the scope of this review. Drawbacks While this new unit has a lot more features than our System One, it has unfortunately lost some too. One of the biggest drawbacks that we noticed straight away is that there’s no good way to monitor the distortion residual waveform (ie, the actual distortion waveform being introduced) during a test. With our System One, we monitor the residual whenever we are using it. It gives important clues as to the behaviour of the DUT and the nature of the distortion. Unlike the System One, the APx525 has no output sockets to hook up to a scope. It does have the facility to view the test waveform on the computer screen but this does not update very fast and there’s no way to view the residual synchronised to the test signal itself. You could argue that having a spectrum analyser means you don’t need to see the residual in this manner but while the spectrum analyser is an excellent feature, we don’t think it really replaces being able to see the distortion waveform itself. For example, it’s easy to see whether an amplifier is suffering from crossover distortion by looking at the residual as the spikes occur near the zero crossing of the test signal. How do you make the same diagnosis from a frequency spectrum display? It’s possible but nowhere near as easy. We also found it a bit odd that when you generate a PDF of your test results, the embedded graphs are just bitmaps rather than vector plots. As a result, they look quite grainy when printed. When we used the APx525 to test our Crystal DAC project (February 2012), which we happened to be working on at the time, we ended up tracing the curves to get graphs we could publish. This could obviously be fixed with a simple software upgrade or alternatively, you can export test results as a CSV file and plot them yourself using spreadsheet or graphing software. The analog outputs of the APx525 are generated using internal DACs. This means that while they have extremely low distortion with a 22Hz-22kHz bandwidth, if the bandwidth is opened up (say, to 80kHz) there is a measurable increase in noise. In fact, with a 22Hz-80kHz bandwidth, the analog signal generator in our System One gives a lower THD+N figure. This is not a major problem but it goes to show just how good the System One was for its time (see Fig.4). Conclusion Overall the instrument is easy to use, has a slew of useful automated tests and we found ourselves using it in preference to our System One most of the time, especially due to the ease of automating tests and the digital audio I/O. We hardly had to refer to the manual at all. The APx525 is available from VICOM Australia Pty Ltd, Level 4, 71-79 Parramatta Rd Silverwater NSW 2128. Contact them at (02) 9648 4757. The cost at the time of writing is approximately $13,360 and the AG52 option is $1,380. The HDMI + Bluetooth option is approx $10,000. SC And if you thought the front panel was rather spartan, here’s the back of the machine to prove it too is very utilitarian! Apart from the ubiquitous IEC mains input connector, all you have is a couple of D9 connectors for aux control in and out, another for software options and a miniUSB for computer interface. 88  Silicon Chip siliconchip.com.au