High-fidelity test specifications

Microsoft began to force system manufacturers to pass the sound quality specifications SYSFUND-0118 (Premium) and SYSFUND-0127 (Basic) in the Windows Logo Program (WLP3.0X) on June 1, 2007 if they wanted to get the Vista logo.

As a result, the industry had its first sound quality standard for electronic/computer products, and Microsoft's leading position in the industry made it a sound specification that manufacturers were willing to cooperate with and improve product quality.

Figure 1: The eight standard contents of Microsoft specification

Audio Precision

Microsoft specified the use of Audio Precision's SYS-2722 Audio Analyzer as a program test instrument in the Policy, and even the Test Kits were developed for SYS-2722, making this company famous overnight.

This professional sound test instrument company located in Oregon, USA was founded in 1984 by a group of engineers who originally worked for Tektronix. The instruments it developed were also the first to use a computer graphical interface to replace the traditional old-fashioned instrument. Its products have not only been widely specified by various consumer electronics manufacturers for many years, but it is also a long-term partner of Dolby Laboratories in the United States. Currently, most of the certification tests launched by Dolby also specify its instruments as special equipment for signal measurement. Its position in the field of professional sound signal measurement can be said to be a world-recognized leading company.

SYS-2722 is a two-channel multi-functional sound measurement instrument. In addition to sound analysis, it also supports sound generators up to 192KHz and is equipped with most digital analog output and input interfaces. The instrument itself does not have any control function and must be connected to a computer to be controlled through proprietary software. Currently, most consumer entertainment product manufacturers use it to test the distortion, signal-to-noise ratio, separation, etc. of the product's sound output, which are basic tests that everyone is familiar with.

Figure 2: SYS-2722 connection method and AP2700 control interface

WLP stipulates eight tests for sound output quality and tests for sound input, microphone input and headphone output. The Test Kits (DTM) currently provided by Microsoft for automatic testing provide five tests that support sound output.

If developers want to know whether their products meet all the specifications set in the Windows Logo Program, in addition to using the Test Kits control instruments to automatically complete certain items, they also need to use the AP SYS-2722 instrument to manually measure other items that are not currently supported by the Test Kits.

Figure 3: Test items currently supported by DTM

What is Audio Performance?

The purpose of this article is to take you to understand each test one by one. Although you may not be able to get started just by reading the article, at least you can have a basic understanding. The so-called Fidelity Test or Audio Performance on PC, in short, is to test the quality of its D/A, A/D conversion, involving the distortion caused by the capacitor on the board and any other noise-causing causes. Overall, after passing any possible adverse effects, the quality of the converted sound signal output is called the Audio Performance of the product, which is what Microsoft calls Fidelity.

Understanding the test items

Full Scale

Full Scale is also called the maximum output voltage. The entire chart will present a continuous curve. In the field of sound signals, we usually define 0dB as the definition of the maximum volume, and usually 0dB is equal to 1V voltage. Therefore, the specification defines that the maximum output of the product must exceed 0dB / 1V. Although it is the maximum output voltage, strictly speaking it should be the maximum undistorted output voltage, because according to experience, when many products reach the maximum output, the upper and lower edges of their waveforms will be cut off, which is called clipping, resulting in nonlinear distortion. At this time, its maximum output is not recognized, and its maximum undistorted ability can only be determined when the volume is lowered to a non-clipping state.

Figure 4: There cannot be any clipping on the continuous waveform, otherwise it will look square from top to bottom.

In this test, we will output a 0dB 48KHz Sample Rate 1KHz sound from the DUT and then input it into the instrument for analysis. Then adjust the computer volume to the maximum to see if there is any harmonic distortion (harmonic distortion will be discussed later). Then lower the volume until the harmonic distortion rate is less than 0.2. The voltage value displayed on the screen is the Full Scale of this product. This value must be recorded and will be used in subsequent tests.

Another test item [Sampling frequency accuracy] is usually measured at the same time. The accuracy of the sample rate of the test signal before and after entering the D/A and compare the original sample rate of the signal with the actual measured ratio. Usually, the first test signal is used for testing together. The Microsoft specification is 0.02%.

THD+n

Total Harmonic Distortion with Noise is the total harmonic distortion plus noise value.

The principle is that when the signal is converted and output, there will be non-linear distortion in the original signal, that is, additional non-original signal components will be added.
Simply put, when we input a 1KHz signal, it is possible that additional abnormal signals that are not the original signal will appear at its multiples such as 2K or 4K.

Figure 5: Only the 1K portion is the original signal, and the other multiples are harmonics.

This abnormal harmonic will of course affect the reproduction of the original signal. This test is to detect the difference between the gain of the harmonic and the original signal. Assuming that the original signal is 0dB and the maximum harmonic is -60dB, the distance between the harmonic and the original signal is 60dB.

When doing this test, the value measured at Full Scale will be used as the reference value, that is, as the maximum output signal.

The signal used in this test is generally a -3dB 1KHz signal as the actual measurement signal, and then the measured harmonic value is compared with the previous Full Scale value.

The test required by Microsoft is more stringent than the original manufacturer's requirements. It requires not only 1KHz signal measurement but also full-frequency signal measurement, that is, 20Hz - 20KHz is divided into 31 frequency bands, and there will be 31 values ​​obtained, and then the worst value is taken, so it is called THD+N VS Frequency.

The specification of this test is that it must be greater than or equal to 80dB relative to Full Scale.

Figure 6: Full-frequency THD+n test results

Dynamic Range

refers to the difference between the maximum output signal and the minimum signal (noise). It can also be used to measure the difference between the size of the signal noise and the normal signal, so it can generally be used to measure the signal to noise ratio (SNR).

This test is measured by inputting a -60dB 1KHz signal. In principle, the -60dB value is used to measure its noise when the signal is present, and -60dB can also avoid the uncertainty of harmonic distortion. This test currently

uses a 1KHz single frequency as the test signal. The reason for not doing full frequency is that the signal is too small to be locked by the instrument, so it must be fixed at 1KHz for testing. The full frequency test requires 31 sound files of different frequency bands, and then each frequency band is fixed for testing.

The requirement of this specification for this test is greater than or equal to 90dB.

This item can be discussed together with another test item "Noise Level during system activity". The biggest difference from Dynamic Range is that although the input signal of "Noise Level during system activity" is also -60dB, data will be stored in the LSB Bit 0 of the signal so that the Codec will always be in working state.

Interchannel Phase Delay

tests the phase delay between the left and right channels at full frequency.

Phase describes the position of the signal waveform change in degrees. Whether the two channels will be out of sync with each other when the signal enters and outputs is the focus of this test.

This measurement uses -20dB Multitone as the input signal (Multition is full-frequency simultaneous sounding rather than step-by-step sounding) and
directly uses the instrument's Phase Meter to obtain the angle, which can then be converted into time through a formula.

Figure 7: The angle value can be converted into time

Microsoft's standard for this test is more flexible. It uses whichever value is better, 30 degrees or 12.5 microseconds under full frequency.

Cross-talk

is testing crosstalk, which means testing the separation of channels. For example, the sound of the left channel should not go to the right channel, and the right channel should not go to the left channel. However, it seems impossible to completely avoid interference with each other on the terminals used for computer sound output.

For this test, we should prepare two different Mono files, one is full frequency 31-order -20dB but only the left channel, and the other is only the right channel. When measuring, the Cross-talk test function of the instrument itself will check the difference between the channel with sound and the other channel without sound, and the ratio obtained is the Cross-talk value we want. Therefore, this test needs to be run twice, once for the left channel to compare with the right channel, and the other time is the opposite.

Figure 8: Cross-talk result chart

Microsoft's standard for this test is greater than or equal to 60dB. The

Magnitude Response

test is the frequency response that everyone is familiar with before, which tests the gain difference of the signal in each frequency band.

Figure 9: Magnitude Response result chart diagram

From the above figure, we can see that the frequency band of the entire test result is divided into two parts, one is the transition band, and the other is the outside transition band. The sky blue line represents the signal curve, and the red part is the limit range, which is used to indicate whether the curve value is correct or not. We can see that the signal becomes unstable in the high and low frequency parts and runs out of the red line range. We call that the beginning of a transition band. There are three points to observe in this test: (1) Is it smooth in the pass band (outside transition band) and whether its ripple value is within the red line range; (2) Is it in compliance with the standard at the end point of the transition band; (3) After the signal enters the transition band, does it continue to extend outward in a single direction and cannot return to the red line curve range? The industry calls this "Monotonic".

Microsoft's standard for this test is that the ripple value must be within plus or minus 0.25dB, that is, within the red line range, and the difference cannot be greater than 1dB at the highest frequency and greater than 3dB at the low frequency.

Third-party testing labs

The above are the eight most important tests. These eight tests can show the fidelity of your product, which is what China calls fidelity and what we call high fidelity, meaning the degree of restoration of the original signal.

Overall, although the test does not take much time, the construction of the test environment and the required investment of funds and manpower are not small. Developers also need to face the time cost of error correction.

Now the environment is ready, and international manufacturers are also ready. I wonder if your product is ready? If not, you may refer to the third-party information that supports Audio Fidelity testing provided on the Microsoft website to choose the most effective test method.

Figure 10: Excerpt from Microsoft's official introduction to the Third Party Lab