Audio in Linux becomes annoying again (continued)

Well, it turned out that the problem playing some system sounds in Thunderbird, described in my previous post, was due to PulseAudio.

Despite sounding fine when played by SMPlayer, the audio clips that sounded distorted/scratchy and too loud when played by Thunderbird also sounded that way when played by VLC. Then I discovered several other .wav files on various Web sites that sounded distorted when played by the browser’s Windows Media Player plug-in (Gecko Media Player). So the problem clearly was not caused by Thunderbird itself. I began to wonder if PulseAudio was the cause. So I adjusted PulseAudio’s sampling frequency, number of fragments and fragment size, and all the clips that previously sounded distorted and too loud now play fine. Here is what I did to fix the problem…

1. Check what PulseAudio is configured to use:

$ pulseaudio --dump-conf
### Read from configuration file: /etc/pulse/daemon.conf ###
daemonize = no
fail = yes
high-priority = yes
nice-level = -11
realtime-scheduling = yes
realtime-priority = 5
allow-module-loading = yes
allow-exit = yes
use-pid-file = yes
system-instance = no
local-server-type = user
cpu-limit = no
enable-shm = yes
flat-volumes = no
lock-memory = no
exit-idle-time = 20
scache-idle-time = 20
dl-search-path = /usr/lib64/pulse-5.0/modules
default-script-file = /etc/pulse/default.pa
load-default-script-file = yes
log-target =
log-level = notice
resample-method = auto
enable-remixing = yes
enable-lfe-remixing = no
default-sample-format = s16le
default-sample-rate = 44100
alternate-sample-rate = 48000
default-sample-channels = 2
default-channel-map = front-left,front-right
default-fragments = 4
default-fragment-size-msec = 25
enable-deferred-volume = yes
deferred-volume-safety-margin-usec = 8000
deferred-volume-extra-delay-usec = 0
shm-size-bytes = 0
log-meta = no
log-time = no
log-backtrace = 0
rlimit-fsize = -1
rlimit-data = -1
rlimit-stack = -1
rlimit-core = -1
rlimit-rss = -1
rlimit-as = -1
rlimit-nproc = -1
rlimit-nofile = 256
rlimit-memlock = -1
rlimit-locks = -1
rlimit-sigpending = -1
rlimit-msgqueue = -1
rlimit-nice = 31
rlimit-rtprio = 9
rlimit-rttime = 1000000
$

2. Check PulseAudio’s output sample rate:

$ pacmd list-sinks | grep sample
sample spec: s16le 2ch 44100Hz
$

So the sample rate is 16 bits @ 44100 Hz and there are 2 output channels (Stereo). My laptop does indeed have two built-in stereo speakers. ‘s16le‘ means ‘signed 16-bit little-endian’.

3. Check PulseAudio’s input sample rate:

$ pacmd list-sources | grep sample
sample spec: s16le 2ch 44100Hz
sample spec: s16le 2ch 44100Hz
$

So the sample rate is 16 bits @ 44100 Hz and there are 2 input channels (Stereo). My laptop does indeed have two built-in microphones. ‘s16le‘ means ‘signed 16-bit little-endian’.

4. Find out the audio card’s maximum sample rate (Hz):

$ arecord -f dat -r 60000 -D hw:0,0 -d 5 test.wav
Recording WAVE 'test.wav' : Signed 16 bit Little Endian, Rate 60000 Hz, Stereo
Warning: rate is not accurate (requested = 60000Hz, got = 48000Hz)
please, try the plug plugin
$

5. As the console output shows that my audio card supports a sample rate of 48000 Hz, edit /etc/pulse/daemon.conf and change the sample rate accordingly:

$ su
Password:
# grep sample-rate /etc/pulse/daemon.conf
; default-sample-rate = 44100
; alternate-sample-rate = 48000
# nano /etc/pulse/daemon.conf
# grep sample-rate /etc/pulse/daemon.conf
default-sample-rate = 48000
; alternate-sample-rate = 48000
# exit
exit
$

6. Find the buffer size and fragment size for each sound card:

$ echo autospawn = no >> ~/.config/pulse/client.conf
$ pulseaudio --kill
$ LANG=C timeout --foreground -k 10 -s kill 10 pulseaudio -vvvv 2>&1 | grep device.buffering -B 10
I: [pulseaudio] sink.c: alsa.driver_name = "snd_hda_intel"
I: [pulseaudio] sink.c: device.bus_path = "pci-0000:00:1b.0"
I: [pulseaudio] sink.c: sysfs.path = "/devices/pci0000:00/0000:00:1b.0/sound/card0"
I: [pulseaudio] sink.c: device.bus = "pci"
I: [pulseaudio] sink.c: device.vendor.id = "8086"
I: [pulseaudio] sink.c: device.vendor.name = "Intel Corporation"
I: [pulseaudio] sink.c: device.product.id = "3b56"
I: [pulseaudio] sink.c: device.product.name = "5 Series/3400 Series Chipset High Definition Audio"
I: [pulseaudio] sink.c: device.form_factor = "internal"
I: [pulseaudio] sink.c: device.string = "front:0"
I: [pulseaudio] sink.c: device.buffering.buffer_size = "19200"
I: [pulseaudio] sink.c: device.buffering.fragment_size = "4800"
--
I: [pulseaudio] source.c: alsa.driver_name = "snd_hda_intel"
I: [pulseaudio] source.c: device.bus_path = "pci-0000:00:1b.0"
I: [pulseaudio] source.c: sysfs.path = "/devices/pci0000:00/0000:00:1b.0/sound/card0"
I: [pulseaudio] source.c: device.bus = "pci"
I: [pulseaudio] source.c: device.vendor.id = "8086"
I: [pulseaudio] source.c: device.vendor.name = "Intel Corporation"
I: [pulseaudio] source.c: device.product.id = "3b56"
I: [pulseaudio] source.c: device.product.name = "5 Series/3400 Series Chipset High Definition Audio"
I: [pulseaudio] source.c: device.form_factor = "internal"
I: [pulseaudio] source.c: device.string = "front:0"
I: [pulseaudio] source.c: device.buffering.buffer_size = "19200"
I: [pulseaudio] source.c: device.buffering.fragment_size = "4800"
$ sed -i '$d' ~/.config/pulse/client.conf # You can just delete the file instead if it didn't exist in the first place.
$

N.B. Depending on your distribution, the PulseAudio file client.conf (if it exists) may be in a different sub-directory of the user’s home directory.

The console output shows that the buffer size was 19200 bits and the fragment size was 4800 bits.

7. Calculate the number of fragments and the fragment size (msec):

default-fragments = 19200 / 4800 = 4

default-fragments-size-msec

= device.buffering.fragment_size [bits] / (sample rate [Hz] x sample width [bits] x number of channels)

= 4800 / ( 48000 x 16 x 2 )

= 0.003125 seconds = 3.125 msec = 3 msec to the nearest integer

8. Edit /etc/pulse/daemon.conf to set these two parameters to the above-mentioned values:

# grep default-fragment /etc/pulse/daemon.conf
; default-fragments = 4
; default-fragment-size-msec = 25
# nano /etc/pulse/daemon.conf
# grep default-fragment /etc/pulse/daemon.conf
default-fragments = 4
default-fragment-size-msec = 3
# exit
exit
$

9. Restart PulseAudio:

$ pulseaudio --kill
$ pulseaudio --start # Only needed if you have 'autospawn = no' in ~/.config/pulse/client.conf
$

10. Check the PulseAudio configuration:

$ pulseaudio --dump-conf
### Read from configuration file: /etc/pulse/daemon.conf ###
daemonize = no
fail = yes
high-priority = yes
nice-level = -11
realtime-scheduling = yes
realtime-priority = 5
allow-module-loading = yes
allow-exit = yes
use-pid-file = yes
system-instance = no
local-server-type = user
cpu-limit = no
enable-shm = yes
flat-volumes = no
lock-memory = no
exit-idle-time = 20
scache-idle-time = 20
dl-search-path = /usr/lib64/pulse-5.0/modules
default-script-file = /etc/pulse/default.pa
load-default-script-file = yes
log-target =
log-level = notice
resample-method = auto
enable-remixing = yes
enable-lfe-remixing = no
default-sample-format = s16le
default-sample-rate = 48000
alternate-sample-rate = 48000
default-sample-channels = 2
default-channel-map = front-left,front-right
default-fragments = 4
default-fragment-size-msec = 3
enable-deferred-volume = yes
deferred-volume-safety-margin-usec = 8000
deferred-volume-extra-delay-usec = 0
shm-size-bytes = 0
log-meta = no
log-time = no
log-backtrace = 0
rlimit-fsize = -1
rlimit-data = -1
rlimit-stack = -1
rlimit-core = -1
rlimit-rss = -1
rlimit-as = -1
rlimit-nproc = -1
rlimit-nofile = 256
rlimit-memlock = -1
rlimit-locks = -1
rlimit-sigpending = -1
rlimit-msgqueue = -1
rlimit-nice = 31
rlimit-rtprio = 9
rlimit-rttime = 1000000
$

Notice that PulseAudio is now configured to use new values for default-sample-rate, default-fragments and default-fragment-size-msec.

$ echo autospawn = no >> ~/.config/pulse/client.conf
$ pulseaudio --kill
$ LANG=C timeout --foreground -k 10 -s kill 10 pulseaudio -vvvv 2>&1 | grep device.buffering -B 10
I: [pulseaudio] sink.c: alsa.driver_name = "snd_hda_intel"
I: [pulseaudio] sink.c: device.bus_path = "pci-0000:00:1b.0"
I: [pulseaudio] sink.c: sysfs.path = "/devices/pci0000:00/0000:00:1b.0/sound/card0"
I: [pulseaudio] sink.c: device.bus = "pci"
I: [pulseaudio] sink.c: device.vendor.id = "8086"
I: [pulseaudio] sink.c: device.vendor.name = "Intel Corporation"
I: [pulseaudio] sink.c: device.product.id = "3b56"
I: [pulseaudio] sink.c: device.product.name = "5 Series/3400 Series Chipset High Definition Audio"
I: [pulseaudio] sink.c: device.form_factor = "internal"
I: [pulseaudio] sink.c: device.string = "front:0"
I: [pulseaudio] sink.c: device.buffering.buffer_size = "2304"
I: [pulseaudio] sink.c: device.buffering.fragment_size = "576"
--
I: [pulseaudio] source.c: alsa.driver_name = "snd_hda_intel"
I: [pulseaudio] source.c: device.bus_path = "pci-0000:00:1b.0"
I: [pulseaudio] source.c: sysfs.path = "/devices/pci0000:00/0000:00:1b.0/sound/card0"
I: [pulseaudio] source.c: device.bus = "pci"
I: [pulseaudio] source.c: device.vendor.id = "8086"
I: [pulseaudio] source.c: device.vendor.name = "Intel Corporation"
I: [pulseaudio] source.c: device.product.id = "3b56"
I: [pulseaudio] source.c: device.product.name = "5 Series/3400 Series Chipset High Definition Audio"
I: [pulseaudio] source.c: device.form_factor = "internal"
I: [pulseaudio] source.c: device.string = "front:0"
I: [pulseaudio] source.c: device.buffering.buffer_size = "2304"
I: [pulseaudio] source.c: device.buffering.fragment_size = "576"
$ rm ~/.config/pulse/client.conf # I didn't have a client.conf to begin with, so I just deleted it.
$ pulseaudio --start
$

Notice that PulseAudio is now using new values for device.buffering.buffer_size and device.buffering.fragment_size.

11. Check PulseAudio’s output sample rate:

$ pacmd list-sinks | grep sample
sample spec: s16le 2ch 48000Hz
$

Notice that PulseAudio is now using a new output sample rate.

12. Check PulseAudio’s input sample rate:

$ pacmd list-sources | grep sample
sample spec: s16le 2ch 48000Hz
sample spec: s16le 2ch 48000Hz
$

Notice that PulseAudio is now using a new input sample rate.

Audio in Linux becomes annoying again

At the moment I seem to be having more audio problems than usual. Last month I blogged about having to fix the ALSA Speaker volume level resetting to zero at boot, and recently two other audio problems have cropped up.

Thunderbird

I have been having trouble with Thunderbird’s ‘system sound’ that announces the arrival of a new e-mail. Lately, Thunderbird has started playing too loud and with significant distortion the audio clip it had been playing perfectly for the last four years. This is especially strange because I created that audio clip with Audacity from another audio clip that sounded too loud when Thunderbird played it. Ironically, the work-around for this latest problem was to switch to the original, much louder sound clip alert.wav instead of the quieter alert_quiet.wav. Not only does Thunderbird now play alert.wav at a lower volume than alert_quiet.wav, but the sound of alert.wav is not distorted when Thunderbird plays it. Yet if I play alert.wav and alert_quiet.wav using SMPlayer, the former is much louder than the latter and neither is distorted. Figure that one out.

The event notification sound that Thunderbird uses to remind me about an impending meeting scheduled in the calendar has now started sounding very distorted too. I still have not found a work-around for that. Event sounds played by the desktop environment I use (KDE) are not distorted, so what is Thunderbird doing? Perhaps the problem is not Thunderbird itself but the audio library it uses, so I need to investigate further.

Skype

Yet another audio problem cropped up this morning when I connected my laptop to an external monitor and keyboard (and thus I left the laptop’s lid almost closed) in an open-plan office and booted the laptop. I entered my username and password on the KDM log-in screen, and the KDE splash screen appeared as usual. After a few seconds the laptop’s speakers suddenly emitted a piercing, continuous howl; the well-known sound of audio feedback from speakers to microphone. It was LOUD. The volume control buttons on the keyboard made no difference, and the sound was so loud that everyone in the office noticed and several people came over to tell me to reset the BIOS (apparently that had fixed the problem for their laptops running Windows).

I kept my finger on the laptop’s power switch and, after several seconds, the laptop powered off. My laptop dual boots Windows 7 and Gentoo Linux, and the audio feedback did not occur when I booted Windows 7. After booting Linux again a couple of times and annoying everyone in the office even more, I discovered I could open the laptop’s lid far enough back to reduce the feedback to a low whine, so I could let KDE finish launching and display the Desktop. I then launched ALSAMixer and discovered that ‘Internal Mic Boost’ was set to 100%. So I immediately lowered it to zero. Then the penny dropped: I had used Skype the previous night without bothering to connect my headphones and external microphone, and Skype had automatically raised ‘Internal Mic Boost’ all the way up to 100%. So I immediately launched Skype, selected ‘Options’ > ‘Sound Devices’ and unticked ‘Allow Skype to automatically adjust my mixer levels’. The next thing I did was add the following lines to the script /etc/local.d/20set_alsa_volume.start mentioned in my previous blog post Fix for ALSA Speaker volume level resetting to zero at boot:

su -c "amixer -c 0 -- sset 'Internal Mic Boost' 0%" -s /bin/sh fitzcarraldo
su -c "amixer -c 0 -- sset 'Internal Mic' 0%" -s /bin/sh fitzcarraldo
su -c "amixer -c 0 -- sset 'Mic Boost' 0%" -s /bin/sh fitzcarraldo
su -c "amixer -c 0 -- sset 'Mic' 0%" -s /bin/sh fitzcarraldo

From now on, only I am allowed to adjust microphone settings! To avoid any possibility of feedback loops in future, the above-mentioned script sets all the microphone channels to zero and I will have to adjust them myself before use. I already have ALSAMixerGUI in the KDE Launcher menu, so it won’t be a big deal to do that.

This fiasco with Skype got me thinking: if Skype is set to automatically adjust mixer levels when you are in a conversation, when you exit Skype why doesn’t it automatically set mixer levels back to the way they were when Skype was launched? It could be done easily and would be more user-friendly than the current way Skype works.

Interrelationship between PulseAudio and ALSA

The final thing I did (yet again) was to adjust all the various ALSA channels and PulseAudio channels to try and get the resulting audio input and output sounding reasonable. This is easier said than done. I often have to mess around with ALSAMixer and PulseAudio Volume Control in order to get audio input and output working satisfactorily in all applications that use audio. It is actually more difficult than it sounds (ouch!) to get ALSA and PulseAudio ‘balanced’ (for want of a better word). In the days before PulseAudio existed, one only had to adjust ALSA. Now, with two agents controlling audio, the task turns out to be surprisingly awkward sometimes.

To sum up, boo to Thunderbird (or whatever it uses to play sounds), boo to Skype and boo to PulseAudio. I’m fed up with audio issues in Linux at the moment. 😡

Update (January 19, 2015): It turns out that the problem in Thunderbird was due to PulseAudio. See my next post for details of how I fixed it.