Still dropout with softsqueeze

[Christian Pernegger]

Hi!

I've had problems with dropouts in softsqueeze since I first tried it
(~slimserver 6.0 alpha). Every few weeks I try again just to see if it
has gotten better. Well, no luck yet ...

The music files (90% flac) and slimserver reside on a 2x Athlon MP 1.6
GHz w/ 2 GB RAM and Debian Testing, slimserver 6.1-trunk

The PC I'm trying to use softsqueeze from is a 2x Opteron 240 w/ 2 GB
RAM, Windows XPsp1a and Java 1.5.0_02, softsqueeze in 6.1-trunk
(2.0b9)

Both machines are connected via Gbit (copper).

I have yet to have a dropout on the "real" sb2 (which is wireless) or
any other software player playing the music directly from samba share.

I tried enabling the buffer fullness - it never goes below 100% with
the sb2 save when changing songs. With softsqueeze I sometimes get
negative readings, but when they're sane it fluctuates between 75% and
95% on an otherwise idle machine. I could see it dropping to 0 under
load but then the behaviour is always exactly the same regardless of
configured buffer size.

Is the buffer size setting not used perhaps? If the buffer were always
the same relatively small size that would explain this behaviour.

C.

[max.spicer]

I've had problems with dropouts in softsqueeze since I first tried it
(~slimserver 6.0 alpha).
How do you mean dropouts? I use softsqueeze without many problems, but it does occasionally output nothing but static when I'm playing tracks. Fast forwarding to the next track cures this. I've never had a similar problem with my SqueezeBox2. Does this sound like your problem?

Max

[Christian Pernegger]

> How do you mean dropouts?

Sometimes there's a short period of silence (<.5 secs), then the song
continues ... and so on. It sounds a bit like my old CD player, which
liked to skip in its old age.

It's never the same songs and never the same spots in a song. It might
go hours without, then dropout every 30 secs or so.

I only use softsqueeze for background audio while working at the
computer, and while I don't much care about audio quality in this
case, the short burst of silence in the middle of a song has about the
same effect on me as a ringing telephone - interruption of any and all
useful thoughts I might have had going on :)

C.

[max.spicer]

Sometimes there's a short period of silence (<.5 secs), then the song
continues ... and so on.
Is anything hogging your CPU when this happens, I wonder? Are you using Java 1.5? Could it even be something like the Java garbage collector kicking in? I'm afraid that I'm no expert here - I'm just throwing up a few ideas until someone more knowledgable (probably Richard Titmuss, who wrote SoftSqueeze) appears. ;-)

Max

[Christian Pernegger]

> Is anything hogging your CPU when this happens, I wonder?

It happens even when it is otherwise idle and / or softsqueeze has
higher priority.

> Are you using Java 1.5?

Yes, any known issues?

> Could it even be something like the Java garbage collector kicking in?

Possibly. Richard had the same idea some time ago and I sent him some
logs - apparently that wasn't it.

For the moment my money is on the wierd buffer behaviour. Negative
readings, almost never at 100%, no change with even two orders of
magnitude bigger buffer.

C.

[max.spicer]

>

> Are you using Java 1.5?
Yes, any known issues?
[color=blue]
Not with 1.5, no. Java Sound was bad pre 1.5 - you used to get a sizeable, non predictable delay on audio. I tried writing a bell ringing simulator in Java 1.3/1.4 once (as you do). It required real-time performance, and didn't get close. I keep meaning to dig the code out and see how it fares in 1.5. But I digress... ;-)

Max