I have Ubuntu Studio Karmic with the real-time kernel.
I run Dragon NaturallySpeaking through wine, so I am most concerned with clear, low-latency, incoming ("recording") sound.
I have edited /etc/alsa/client.conf to stop pulseaudio from re-spawning, and killall pulseaudio after I log on.
Under winecfg:
If I select OSS the results are pretty good. It works. Sound quality seems a bit worse than under ALSA (which used to work, pre-pulseaudio) but it's not bad.
If I select ALSA, the results are not good. No mixer shows up unless I install esound (which uninstalls pulseaudio-esound-compat and ubuntu-studio-desktop).
Then a mixer shows up and my program shows an audio system present, but freezes at the first incoming sound.
Sometimes after an update sound does not work.
Updates do not take kindly to tinkering.
This could be because of something to do with esound, it could be because I tried to update without ubuntu-studio-desktop, or it could be because I set the default soundcard using asoundconf.
As you probably know, Ubuntu has dropped asoundconf from Karmic, and I use the one from Jaunty. (I put it under /usr/local/bin.)
To make sound work again, I must re-install the settings from pulseaudio, which I do as follows:
sudo apt-get --purge remove pulseaudio
sudo apt-get install pulseaudio (and, if you want, all the other programs that the above command uses; not all of them are re-installed)
Then it works.
This could be useful in case you sound suddenly goes out after an update, as mine does.
As I said, I'm just documenting this as my experience. Comments? Anyone else need really good incoming sound?
