LWN: Comments on "Can you hear me now?"

Can you hear me now?

muwlgr — Sat, 09 May 2009 04:47:27 +0000

Ok, the mic is fixed now with 2.6.30-2. But anyway, it was a memorable hell.

Can you hear me now?

ariveira — Fri, 08 May 2009 15:56:23 +0000

Not here in my Ubuntu 8.10 system. It shows only one slider for pulseaudio

Can you hear me now?

muwlgr — Fri, 08 May 2009 11:57:21 +0000

I use KUbuntu 9.04 on my laptop (very simple system with snd_hda_intel driver) and I have lost an ability to record sound from its embedded microphone perhaps since the arrival of 8.10. In 8.04, that mic "just worked" in my Skype for some now unknown reason. Under Windows, it still works well. Do I have a reliable way to determine what channel name corresponds to my embedded mic and how to unmute it ? Skype gives me a list of channels with 'hw' or 'plughw' in the name, as well as variances of each with the number '0' or '6' after the comma. Total of 4. As 'Default' choice does not work, I am probably supposed to guess by myself which one of these is correct. And then, every of them could be muted, and I am supposed to know how to really unmute them and control recording gain level. In KMix, I see the proverbial picture very similar to shown above (why-alsa-sucks), only a little shorter. Of course there is no any hint about which is hw/plughw/0/6, but anyway I am able to guess that 'Front Mic' is a suitable input source for my case, as by unmuting it I can hear interference whistle from laptop's speakers, also it really reproduces other sounds heard around of it. I set 'Input Source' to 'Front Mic' in KMix but still have no luck of getting these sounds into my Skype (complete silence is heard in test calls). I turn 'Capture' on, or off, ho changes.

There should really be an understandable and convenient testing tool for sound inputs of the system (mic built-in or mic plugged into socket, as well as line input), and also accessibility of these inputs to the sound-system-of-the-day (oss-alsa-arts-gstreamer-pulseaudio-phonon-qtaudio-whatever). Like Sound Recorder in Windows, for example: if it works and I see waveforms and can hear them back, every other multimedia app should work as well. Else it is going to be useless for me. Soon I am going to reboot into Karmic 2.6.30-2 kernel in the hope this could give me any help. But now, I really really feel myself in the hell, and a prolonged hell.

Can't possibly work

ovitters — Thu, 07 May 2009 11:18:51 +0000

If you cannot reply politely, suggest to refrain from replying. Your comment is not constructive with things like "going right over the heads of the GNOMEs".

Can you hear me now?

obi — Thu, 07 May 2009 03:14:54 +0000

That's actually great! I've always been struggling trying to get maximum volume without distortion by trying to find the "right" percentage for master/pcm etc. I definitely won't miss having to tweak them manually, I look forward to the day this feature hits my distro.

Can you hear me now?

jschrod — Wed, 06 May 2009 10:56:36 +0000

Well, I found it interesting -- being not a Fedora, but a SUSE user, where PA is used by default as well and similar changes might be upcoming. And, as usual on LWN.net, the article was informative, not continuing any flame war.

Just because you are reading fedora-devel and are obviously involved in the development does not mean that we non-Fedora users are not interested to learn what happens in one of the major Linux distributions. This kind of information is what makes LWN.net so attractive.

Can you hear me now?

lacostej — Sun, 03 May 2009 12:00:13 +0000

I like what you said. I've been using Linux for years, and I sometimes have problems getting the sound right, especially with new hardware and programs like skype.

My volume control (ubuntu 9.04) has several playback devices

HTA intel (Alsa mixer)
Analog Devices AD1986A (OSS Mixer)
Playback: HDA Intel - AD198x Analog (PulseAudio mixer).

and 2 capture ones

When I play sound, it seems that all of these device affect it in some way. When I screw up, I never know why.

Would be happy to know how things are layered architecturally so that I can grasp these conceps better.

Can you hear me now?

sfink — Sat, 02 May 2009 06:27:49 +0000

In my long and ultimately unsuccessful struggle with Linux audio, I did come to the opinion that Pulse is on the right track technically. My main issue is that I wanted both low-latency triggered audio and zero-dropout music tracks at the same time. Sounds simple, but as far as I could tell, was impossible by design in everything up until Pulse.

In Pulse, it's merely impossible in practice. I get more glitches in its glitch-free playback than I did pre-Pulse (I'm using Fedora 10). But the underlying design seems like it may eventually lead to functional audio, finally.

I've also noticed that the various audio projects seem to attract some of the least empathetic developers around, people who are willfully clueless about usability. That happens in a lot of areas, but some areas -- like audio -- seem to actively chase away the people who could actually make things work nicely for other people. The pack senses the lack of purity and exploits it as a weakness or something.

Can you hear me now?

broonie — Fri, 01 May 2009 13:38:45 +0000

0dB is a generally good reasonable default - where that is on the scale is largely irrelevant. If it's at the top of the scale then you have an attenuation only control and full scale would (typically) not be expected to clip or distort, it's mostly amplification where this is a concern.

Can you hear me now?

broonie — Fri, 01 May 2009 13:30:03 +0000

There's quite a bit of demand for this - it would really rock in the embedded space where we're already collecting the audio routing information and it's is typically *much* more complex than what you see in a PC once get to something like the audio hubs used in mobile phones.

If I get my act together next year I may try posting this as a suggested GSoC project for ALSA or the kernel.

If only if was so simple!

broonie — Fri, 01 May 2009 13:25:45 +0000

This stuff is all published by the hardware vendors (potentially not in public datasheets but still) - the audio devices really aren't usable without this information.

Can you hear me now?

broonie — Fri, 01 May 2009 13:11:22 +0000

That's not true any more - there is now a standard interface for reporting jack status to user space with a helper in kernel for the drivers and support is being added to drivers to use it - 2.6.29 has some that should cover a lot of common laptops.

Note that not all sound devices are going to be able to do jack detection - it does require hardware support and simpler devices will often not have that support.

Can you hear me now?

liljencrantz — Fri, 01 May 2009 11:42:30 +0000

I'm still having occasional problems with Pulse Audio killing all sound on my system on Intrepid. The extra functionality provided by PA is of the «nice to have» variety, but for a lot of people, getting sound out of their computer is critical. For well over a year, we have been experiencing highly reduced reliability in a critical functionality in order to obtain a couple of «nice to have» functions. This is very bad prioritising.

Can you hear me now?

rfunk — Thu, 30 Apr 2009 18:02:11 +0000

In the general case you're probably right. But I just checked on my old
SB Audigy2, and 0dB = 100%. I suspect that it's a common case if not
universal.

BTW, I'm happy to have the dB info; last time I looked at alsamixer I
don't remember that being there.

Can you hear me now?

mezcalero — Thu, 30 Apr 2009 15:44:56 +0000

No, that's not how PA is going to do it.

PA will try to initialize the inner sliders to 0 dB! That is something different than 100%.

Can you hear me now?

mezcalero — Thu, 30 Apr 2009 15:43:26 +0000

First of all, on most drivers we have dB information. Hence setting anything to 100% or so is a bad idea. What makes more sense, as I already tried to make clear is to initialize to 0dB using that information. Forget about the % scale. If we habe the dB scale volume controls actually start to make sense!

And I often made clear that I think that onboard audio is more than good and is what I am focussing on. So if you claim we'd care only about proper external sound cards: that's complete and utter nonsense.

Can you hear me now?

zlynx — Thu, 30 Apr 2009 13:55:51 +0000

Speaking of this, I was quite impressed by the Vista Realtek driver package for my new X58 motherboard.

It uses jack sensing and whenever something is plugged in, it pops up a dialog window *with a connector diagram* and asks you what you just plugged in. Then it lets you test it and set levels.

It's very useful.

Can you hear me now?

zdzichu — Thu, 30 Apr 2009 13:11:31 +0000

Your doing it wrong. Do PCM=100%, Master=50% sound distorted? It's how PA will going ot control the volume, inner sliders to the max and outer to control actual volume.

BTW, if you happen to have soundcard with actual amplifier controlled by Master, you are hurting yourself by setting PCM to 50%. This way you decrease quality by reducing number of possible quantification steps.

Can you hear me now?

cladisch — Thu, 30 Apr 2009 12:31:49 +0000

> does anyone know of "digital-output only" soundcard that just pipes the sounds to my dedicated amplifier with a digital input?

Any cheap onboard sound chip (HDA or AC'97) or CMI8738-based card with a digital output should work just fine for this; there aren't too many routing or amplification choices that could go wrong.

It is, as usual, a driver problem.

cortana — Thu, 30 Apr 2009 12:13:43 +0000

At least with Vista, the manufacturers test their audio device & drivers and make sure it works. That is the problem with Linux. I've had an Intel-based laptop for over a year now, and sound still doesn't work properly. I doubt it ever will at this rate.

Signal vs. noise

stickster — Thu, 30 Apr 2009 12:10:28 +0000

I hope that people who are interested in improving the situation read this message from Adam Williamson on fedora-devel-list:

http://www.redhat.com/archives/fedora-devel-list/2009-Apr...

There are instructions in that message for helping to identify problems with their sound card drivers that are being exposed by PulseAudio's volume control change. Please take the time to download and burn the Fedora 11 Preview on a Live CD, follow these instructions, and file a bug with results.

Can you hear me now?

nhippi — Thu, 30 Apr 2009 10:20:46 +0000

> I am also interested in the way how the new pulse will handle my laptop, where with Master=100%, PCM=50% the sound is distorted, while with Master=75%, PCM=75% it is perfectly OK.

This seems to be a relatively common case. ie. top 10% of master mixer slider is sound distorted and bottom 5% expose hissing. So it seems any generic "one slider" solutions should avoid pushing any of the (analog) mixer controls close the limit.

> I don't think it is feasible to call most of the sound hardware "crap" and refuse to work around it.

This seems to have been the attitude of alsa developers from the beginning.. IE "your integrated soundcard sucks, get a real multichannel one if you want play sounds from many apps at the same time"...

But, crapitalism rules, if the hw is good enough for windows, that's what is going to be put on pc's.

slighly offtopic, does anyone know of "digital-output only" soundcard that just pipes the sounds to my dedicated amplifier with a digital input? And, ofcourse, works without problems in Linux..

Can you hear me now?

aigarius — Thu, 30 Apr 2009 06:01:45 +0000

There are two programs - gnome-volume-control and pavucontrol. The first one is the advanced one with all the knows the hardware provides. It is needed sometimes - to switch to 5.1 sound or to turn the correct microphone input on. Most of the time it is much easer to use pavucontrol which provides not only an easy volume control per-soundcard, but also provides the unique feature of volume control per-application.

Both must be available. Preferably pavucontrol should be the default one with gnome-volume-control or gnome-alsamixer provided when you press the 'Advanced...' button.

alsamixer

tialaramex — Thu, 30 Apr 2009 05:34:10 +0000

alsamixer can be instructed to talk to a specific device instead of your default (which is now PulseAudio)

e.g "alsamixer -c 0" shows the low-level mixer for the first card, or "alsamixer -D any-device-name" if you know the ALSA device name of the mixer you are interested in.

This is obviously not an adequate solution for my mother, but it should get you back where you were before.

Can you hear me now?

alankila — Thu, 30 Apr 2009 01:35:39 +0000

On the other hand, running alsamixer on ubuntu actually addresses the hardware sliders, so this one particular pitfall mentioned on the article is avoided.

This was solved once already

drag — Thu, 30 Apr 2009 00:24:13 +0000

Well the difference is that Alsamixer is low-level. It's actually reflecting the capabilities of the card and the drivers. I have a feeling that many of the mixers are proceedurally built simply from hardware features.

With Windows 2000/XP they introduced Kmix, which is a mixer for the Win32 sound system. Thus everything you do in Windows is software mixed, from the beginning. Then, I beleive, it's up to the sound card makers to provide their own applications and mixer controls for situations were they have features that can't be controlled via the bog-standard Windows stuff.

Which is why you end up with people bitching and moaning about creative drivers and their bloated apps and stuff in Windows. They need all that because the functionality they need are not provided for by the operating system.

And on top that you have huge latency issues and other things that affect Windows users. Similar to what people are running into with PulseAudio. (of course PA is not nearly as mature right now, as the Kmix stuff that was introduced with Windows NT-based desktops)

Of course for the needs of psuedo-professional purposes.. like doing high quality recordings, running low-latency recording, multiple I/O streams and such things.. then any sort of functionality provided for Windows is a complete wasteland.

The audio industry for the longest time stuck with Windows 9x systems and Mac OS systems due to the performance latency issues that newer versions of Windows introduced.. but since Microsoft is the dominate force in PCs and everybody uses them people eventually relented and created their own audio and driver stack for Windows XP called ASIO.

In order to use ASIO, of course, you need special hardware that supports is and you need to use special drivers. This, of course, conflicts with the normal operation of Windows audio.

All in all Windows audio in XP is quite a bit of a cluster-fsck. The only reason it works as well as it does is becuase of the huge amount of money and effort poured into it by the industry and individual card manufacturers providing their own sane default settings and so on and so forth.

The nice thing about Alsa is that it avoids all that stuff.. the drivers are suitable for high performance as well, with the addition and growing maturity of Pulse-Audio, is gaining usefulness as the desktop.

I know that the Windows interface may seem simple and attractive, but it's actually quite a bit more confusing then at first blush. Getting high quality mic results out of it is actually kinda confusing and can take a long time to get set just right.

Which is probably why Microsoft scrapped the whole thing for Vista.

Of course that is nothing compared to the current state of Linux audio confusion.. but what can you do?

----------------------------
----------------------------

Probably the most correct approach is a policy-driven or profile-driven way to setup the drivers.

Does the user want... Stereo out? Digital out to a receiver? 5.1 Surround sound? 7.1 surround sound? Quadraphonic sound? Ac3 or DTS pass-thru? etc etc.

The user selects the profile and the UI adjusts accordingly, hiding all the little weird tweaks that needs to be done on Alsa's low-level. Doing that will easily surpass what you can get in Windows in terms of user friendliness.

Having a 'Expert Mode' is just admitting failure and is a fundamentally broken design.. but it may be necessary for the time being.

What is ultimately needed is a 'sound card tweak database' that can be used to answer such questions as "What low-level settings on the card can lead to working digital output?" and "What low-level settings lead to the Mic being useful". Then tie that into udev or hal or devicekit or whatever then hell is going to be used to keep track of driver settings.

Of course the difficult part is gathering that sort of information...

This was solved once already

ikm — Wed, 29 Apr 2009 23:24:26 +0000

Never saw Vista, but XP had it ok. Single click on a tray icon brings you a master slider, double-click shows more sliders -- but not like on that screenshot, of course, usually around 10 at max. You can opt-in to show some more in the options.

I think the root of the problem is in Alsa. It doesn't try to differentiate between different types of controls. Whoever was designing this haven't thought that through, I guess. And also the fact that driver writers don't all eat the same dog food -- if you develop a driver under Windows, you've got that mixer app already and you'll surely check if it looks alright or not. Under Linux, well, no UI's a standard. I think it is some kind of a Linux immaturity stereotype -- there are always too many options, and no one's good enough. Or, returning to the initial topic: either have just one control, or whole hundreds of them. No middle options :)

This was solved once already

jmitchel — Wed, 29 Apr 2009 23:01:19 +0000

Which version of Windows? Vista is just as cussedly bad! The speaker output is almost never loud enough on my new laptop, except when it is, and I can't find any secondary sliders to twiddle to set things right.

Can you hear me now?

eklitzke — Wed, 29 Apr 2009 22:57:02 +0000

*sigh*

I've been using Ubuntu for a long time on my laptop at work, and had just recently installed the F11 Beta over this, since that's what I've been using at home and I prefer Fedora. I ended up switching back after a couple days to the latest Ubuntu release over this because audio was messed up in Fedora.

In my case, I'm using an older Macbook Pro that has some sort of headphone detection problem. When headphones are plugged in, the volume on the front speakers are reduced, but not all of the way. This means I can't listen to music or do anything else audio-related since it's distracting for everyone else nearby. The solution I've been using up until now is to just mute the front speaker device; this works great because I had headphones, so I don't care about the fact that the speakers don't work. Unfortunately I couldn't figure out how to do this with the new Pulseaudio tool in Fedora. The alsamixer tool was also displaying to me a single output device, without knobs to change the front speaker/headphone volumes indepently.

I know this is kind of a corner case, and if the hardware just worked it wouldn't have been a problem. But as we all know, the hardware doesn't always just work. If the Fedora developers end up changing this, I have Fedora installed on another partition, so I might try switching back...

If only if was so simple!

farnz — Wed, 29 Apr 2009 22:03:44 +0000

For the case of HDA codecs (the majority of modern on-board sound chips), ALSA does has this information. Look at codecgraph to see how it can be drawn - and note that ALSA knows which of the nodes it's sending sound to or controlling with a mixer control.

This was solved once already

ikm — Wed, 29 Apr 2009 21:56:47 +0000

Just do the way Windows does. That's not trolling, seriously - I like the way they did.

If only if was so simple!

JoeBuck — Wed, 29 Apr 2009 21:32:57 +0000

It's true that in many cases we don't have the knowledge to build the right diagram. But in many cases we do: for a standard laptop sold by a reasonably Linux-friendly company, or from a Linux-oriented reseller.

We couldn't expect the schematic to be filled in by asking the sound card, because the sound card doesn't know (or the driver isn't exporting what it does know, like what's plugged in to what). But I still think that it's an ideal to aim for. When the user has built the box herself, she could edit the diagram to match. It would be more useful than an array of ten sliders.

If only if was so simple!

russell — Wed, 29 Apr 2009 21:19:13 +0000

The fact that it's embedded is the Windows driver probably means that it's useful information. Perhaps we ( Linux distros ) should start collecting and using this information too.

Can you hear me now?

russell — Wed, 29 Apr 2009 21:14:15 +0000

That's one of the most sensible ideas I've heard.

Can't possibly work

jmorris42 — Wed, 29 Apr 2009 20:26:14 +0000

One slider can't possibly work. Period end of story.

Problem the First:

Too many cards have poor (none) documentation or worse incorrect documentation. Manual intervention will be required to get those systems making sound. After it is working a single slider could work for the uninteresting use cases, see below.

Problem the Second:

Everyone doesn't have a simple desktop with two speakers (and perhaps a sub) hooked to a single port. Or a laptop with only the internal speakers ever being used. For example my Thinkpad has a seperate slider to control the built in speaker and another to control the line out on the dock. Unless Pulse is going to get smart enough to detect the dock and mute the internal speaker for me I need a second control exposed. Or I can open the lid and fiddle with the buttons on the keyboard I suppose.... NOT.

On my desktop I had to experiment to find the right position of the MASTER, PCM and VIA DXS mixers to give enough output level to drive my speakers without audible amplifier hum yet low enough to prevent horrible clipping in the sound card's analog circuits. Unless Pulse is going to develop an extensive database these issues will always require manual intervention.

Other people have surround sound systems that the tech for is anything but settled, thus frequent manual setup and adjustment.

Basically, it isn't going to "Just Work" and this attitude at GNOME that "We will make all these decisions for you, so don't you worry yer pretty little head none 'bout all that scary tech stuff." is getting a bit tiresome. Yes things should be made as simple as possible. But no more. That second part appears to be going right over the heads of the GNOMEs. If we wanted to trade simplicity and kewl design for usability I think we all know the URL of the Apple Store.

Can you hear me now?

jengelh — Wed, 29 Apr 2009 20:09:27 +0000

Nobody can hear my comments!

>an unbelievable mess of confusing sliders. See this image for an example.

Seriously, if you have that many sliders on one card, it must be a good one. I have seen utter crap on-board soundchips that have even less mixers than pcspkr.ko. Then again, 10+ chorus sliders? What for? Angels won't start singing even if you up them all.

Yeah, I side with "do it Windows style", i.e. display Master, Master Mono, Headphones, PCM, Line, CD, and Mic by default. kthx. Is that really so hard?

If only if was so simple!

khim — Wed, 29 Apr 2009 19:59:28 +0000

Imagine an interface that draws an accurate diagram of the audio paths in the system, with all available volume adjustments depicted at the correct place in the diagram.

Yes, imagine that. Imagine the perfect world where ALSA knows what it's doing. Feel it, love it, then go back to reality. Where 90% of controls exposed by ALSA are doing god knows what.

ALSA exposes what the sound chip is exposing. But this worldview does not include "diagram of the audio paths in the system"!!! There are nothing to visualize!!! Only hardware maker knows what knobs are chaging what - and they are not saying. This knowleadge is embedded in Windows- driver or, even worse, control program for said driver - and does not exist in any other form...

We can try to build such a system - but right now, today it'll support just a few cheap soundcards. Where you just don't have enough knobs to organize complex scheme and so all these simplifications are not needed anyway...

Can you hear me now?

elanthis — Wed, 29 Apr 2009 19:30:30 +0000

"What these extra sliders are exposing is the same kind of thing as the controls under the flap on your VCR. The people too scared to even programme the VCR clearly shouldn't be playing with these sliders, but there was a reason they were provided."

For the odd use cases, sure.

For the common cases where the audio hardware just has 20 different knobs that all need to be set to specific values to get consistent audio volume out of the two or three common outputs on the hardware, I shouldn't have to touch any advanced knobs. It should just work. But ALSA is not making it just work, and ALSA is not exporting enough information for applications like PulseAudio or desktop volume tools to set the mixer values properly.

The applications could create some kind of database mapping sound cards and their respective mixers to a list of standardized labels and attributes, but there's no guarantee that ALSA won't just break those in the future by changing names, because they're non-standardized and not considered part of the API/ABI. The only way such a database would be useful is if it was maintained and shipped as part of ALSA itself.

But if ALSA were to do that, they could just fix the freaking issue at its core and be done with it.

Can you hear me now?

mezcalero — Wed, 29 Apr 2009 19:04:00 +0000

ALSA includes dB information for most its sliders. While it is not always clear what the reference level for those dB values is for the 'inner' sliders such as PCM it usually means amplification relative to the previous element in the pipeline. It is hence a good idea to set them to 0dB.

The % scale most mixers expose is pretty useless.

Can you hear me now?

Yenya — Wed, 29 Apr 2009 18:51:56 +0000

Lennart,

I, for one, am glad that the topic has been reported on LWN. I do not follow a fedora-devel list, but this is clearly interesting even for us, Fedora (and other pulse-based distros) users, who will use the new pulseaudio really soon (in F11).

I am also interested in the way how the new pulse will handle my laptop, where with Master=100%, PCM=50% the sound is distorted, while with Master=75%, PCM=75% it is perfectly OK. I don't think it is feasible to call most of the sound hardware "crap" and refuse to work around it.

-Yenya