mirror of https://github.com/thesofproject/sof.git
124 lines
5.0 KiB
Plaintext
124 lines
5.0 KiB
Plaintext
SOF audio quality test scripts
|
|
|
|
1. Introduction
|
|
|
|
The purpose of the script is to test playback (and/or recording
|
|
quality of a SOF platform with help of PC with a high quality USB
|
|
sound card. Or in other words a sound card with sufficient number of
|
|
analog inputs and outputs with sufficient signal levels handling
|
|
capability, and good analog signal quality. The sound card should be
|
|
better than expected quality of the SOF device.
|
|
|
|
+------+ +--------+
|
|
| |<--- Ethernet or WLAN connection ---->| |
|
|
| PC | +-------+ | SOF |
|
|
| |<--- usb --->| Sound |--- line in --->| device |
|
|
+------+ | card |<-- line out ---| |
|
|
+-------+ +--------+
|
|
|
|
The tests can also use digital interfaces like S/PDIF if available in
|
|
the SOF device. That helps to move testing focus to smaller digital
|
|
signal processing issues instead of usually dominant ADC/DAC
|
|
performance (unless there's a major quality issue).
|
|
|
|
The key tests to execute to quickly verify playback/recording audio
|
|
quality are gain, frequency response (FR), and total harmonic
|
|
distortion plus noise (THD+N). The test procedure and results reporting
|
|
tries to follow AES17 recommendations.
|
|
|
|
This work can be taken as a inexpensive introduction to objective
|
|
audio quality parameters testing that should be useful for individual
|
|
developers without access to professional test equipment for relative
|
|
performance indicators testing. Tests like THD+N also work for
|
|
automatic testing needs to quickly flag regressions in playback audio
|
|
integrity.
|
|
|
|
IMPORTANT NOTE: These test scripts are not calibrated and not suitable
|
|
to produce absolute metric. The results will vary depending on the
|
|
used sound card and depend on correctly done preparations on the
|
|
PC. Therefore for professional usage with need for industry comparable
|
|
metric we recommend instead dedicated test systems and services.
|
|
|
|
|
|
2. General preparations
|
|
|
|
The PC should be installed with a Linux distribution that provides
|
|
octave and octave-signal packaces or use commercial Matlab (R)
|
|
software. For Octave it's recommended to use an initialization script
|
|
that loads the needed packages and disables the pager to freely scroll
|
|
the test text printings without frequently stopping to press <space>
|
|
to proceed.
|
|
|
|
$ sudo apt-get install octave octave-signal octave-io
|
|
|
|
$ cat ~/.octaverc
|
|
more off
|
|
pkg load signal
|
|
pkg load io
|
|
|
|
The SOF device should be made accessible via ssh without
|
|
password. Also the test user should belong to group audio in both PC
|
|
and SOF device.
|
|
|
|
To avoid including the PC audio server and it's signal processing into
|
|
the test chain it is recommended to disable audio servers like
|
|
pulseaudio permanently (rename the executable /usr/bin/pulseaudio) on
|
|
the test PC and use the ALSA provided sound devices directly.
|
|
|
|
There is need to go trough the alsamixer settings for the sound card
|
|
and store them persistently. Also on sophisticated sound cards need to
|
|
do internal settings via front panel or dedicated control SW for
|
|
settings those are not exposed to alsamixer.
|
|
|
|
|
|
3. Playback and recording test preparations
|
|
|
|
See script sof_audio_quality_test_config.m. It is a template to
|
|
configure the audio interfaces for test playback and capture of the
|
|
playback. Make a copy of the file for your test setup and edit the
|
|
script sof_audio_quality_test_top.m to use that configuration script
|
|
instead of the default. See line
|
|
|
|
configs = {'sof_audio_quality_test_config.m'};
|
|
|
|
The line initializes a cell array of strings. Multiple test
|
|
configurations are added to cell array with comma as separator.
|
|
|
|
To test playback of SOF device there's need to edit "play.user" to
|
|
correpond to actual test user and hostname or IP address of the
|
|
device. Also edit the "play.dev" to contain the audio playback device
|
|
to test.
|
|
|
|
For playabck capture to PC edit the "rec.dev", "rec.sft", "rec.nch",
|
|
and "rec.ch" fields. The example is set for a 8ch sound card with SOF
|
|
device connected into analog inputs 1 and 2.
|
|
|
|
The setting "test.att_rec_db" is important. There must be a sufficient
|
|
analog signal headroom for the sound card ADC to capture SOF DAC
|
|
playback signal without clipping distortion. The example assumes 3 dB
|
|
attenuation but it can be adjusted to be e.g. within 1 - 6 dB.
|
|
|
|
You can start with 3 dB and run the test script
|
|
sof_audio_quality_test.sh. It won't proceed if the gain test is not
|
|
passing. Based on the reported difference vs. expected 0 dB gain
|
|
adjust the alsamixer settings for you USB sound card capture gain. The
|
|
controls may be also in the front panel.
|
|
|
|
If the result still deviates from 0 dB and it's known that the
|
|
playback level is correct it is possible to to adjust this attenuation
|
|
parameter to get the 0 dB result.
|
|
|
|
4. Tests reporting
|
|
|
|
The script is configured to proceed without opening plot windows to
|
|
desktop. Once the gain test is passed the script executes the FR and
|
|
THD+N tests.
|
|
|
|
The shell window will show a brief test report and the same will
|
|
appear to directory "reports" with name prefix defined in "test.id" in
|
|
the configuration file.
|
|
|
|
The plots in PNG format are exported to directory "plots".
|
|
|
|
That's all, happy testing!!
|