Sound cards (Was: Re: no sound

[Ralf]

On Sat, 2011-05-21 at 00:30 +0200, Thomas Orgis wrote:
> Am Fri, 20 May 2011 15:33:39 +0200
> schrieb Ralf Mardorf <ralf.mardorf at alice-dsl.net>: 
> 
> > No! But I suspect a bug for Jack and btw. there already was a rounding
> > bug that was fixed, perhaps years ago.
> 
> Ah, OK. I'll ignore the discussion about different sound cards influencing this as a heap of confusion and settle for this: If JACK would not have bugs, by design it would do bit-exact copies, or at least copies with floating point computation errors in non-audible amplitude. I don't claim that this ideal case is indeed the reality.
> 
> When you find those bugs, I hope they'll get reported and fixed, for the benefit to us all. In the meantime, those of us with some time on their hands can just try to verify the issue of bit-exact copying of data through the pipeline ... I just don't want to go down that path right now, since I'm already not able to do my actual work because of debugging some gcc 4.6 breakage on code of mine.
> 
> 
> Alrighty then,
> 
> Thomas.
> 
> 

Is there a way of monitoring the data that is captured and read by Jack
and Jack clients in a human readable way, e.g. to see if there would be
rounding errors? I always used Audacity's spectral view (to show
recordings of square and sine waves), when writing with Rui. For
Qtractor there really was and is an issue when connecting it's outputs
to the inputs, that is caused by Jack. 

Phew, a lot of OT blah blah, you don't need to read does follow, pardon:

(That's why I wish to stop the discussion about the audio quality loss
issue.)

I guess it's because a vector points to the original buffer, instead of
coping it, but I might be completely wrong. Anyway, if a client's input
is connected to it's output this only should work correctly, if there's
a special order.
Here I did avoid to do this and there still is loss, similar to cheap
4-Track cassette recorders. If I directly connect Hydrogen or Yoshimi to
Qtractor or Ardour2, that does mean one client to _another_ client,
there's this kind of audible loss and this shouldn't happen :(.

I can't help with reading the source code. When I programmed in C 20
years ago, I just did it for one program and then switched back to
Assembler. On Linux I tried to learn C again, but I wasn't able to write
even simplest make files, IIRC 20 years ago it was the work, the
compiler had to do ;), I could take a look on my Atari ST's 80286
hardware emulation, there still should be the old editor and compiler.
So, I'm not a coder ;). I'm an audio/video engineer and the computers I
privately programmed have less in common with current PCs, e.g. the
Atari's TOS and the emulation running DR DOS, are sharing a 42MB hard
disk. Especially regarding to music software, there was the advantage
that those machines don't do real multitasking and the hardware, e.g.
for the C64 MIDI interface, was directly accessible, that's why there's
no MIDI jitter for old hardware. Turn of the interrupt, check the
ACIA/UART and send in real-time, that's how it did work years ago, today
we do have USB protocols etc. that at all events do cause MIDI jitter.
Some smart guys, I guess Stéphane is one of them, do work on this issue
and they/he already had success, at least for my machine. If I use PCI
MIDI and

edubuntu at edubuntu:~$ jackd -V
jackdmp 1.9.8

using the -Xalsarawmidi switch + a2jmidi_bridge I'm able to use external
MIDI equipment without audible jitter, unfortunately I need to disable
-Xalsarawmidi, if I wish to record MIDI events with a Linux sequencer,
because there's no bridge vice versa and as far as I know no good
sequencer using Jack MIDI, resp. I guess Ardour3 might use Jack MIDI.
Btw. if I run latency tests I also get best results for Alsa, but at
least I'm able to hear that there's jitter, a lot of people with less
good results aren't able to hear jitter.

There're a lot of issues for Linux audio, some are already solved by svn
versions, other issues aren't solved. It makes me wonder that there are
so less people in the Linux community who notice those issues.

I can't program myself, but I could run tests and report issues.

Linux audio and MIDI IMO do need a lot of upgrades and some isolation
from the non-audio community.