31 Dec
2008
31 Dec
'08
9:48 a.m.
hi,
this is no complaint (!) about anything, just to reflect over my
yesterday's exerience with a newly purchased Focusrite Saffire Pro10,
connected via Belkin PCMCIA Firewirecard to my Lenovo R61 - the FW Chip
in this Laptop is one of those crappy Ricoh parts, I tried this, too,
but as expected it just didn't work...
So I compiled the ffado Release Candidate, afterwards compiled the
actual Jack 0.116.1 and finally got things running.
Compared to my Lexicon Omega Studio USB Device the performance of that
FW assembly (don't want to blame a single device of that chain) just was
disappointing, already in idlemode without any clients open the
jackserver showed a workload of roughly 5% (128 frames/3 periods), while
my Lexicon produces between 0.5 to 1% (128 frames/2 periods).
Opening Ardour and just replaying one of my sessions produced glitches
where only a strong hiss could be heard. (The session itself caused a
workload of less than 20%.) - I remember having had those hiss glitches
in my early days in 2005 with the MAudio Transit USB and Kernel 2.6.11 -
I never expected them coming back some day... *lol*
I then had a look on my interrupts and had to notice the FW Card sharing
it's interrupt (no 16) with the Nwidia Graficscard - so I might have
found the culprit - or not? - I don't know... :-(((
I tried to change the pci settings in the bios - no success.
Now my question to those who also tried firewire audio: was it a success
story for you from the beginning or did anybody make similar experiences
and how did you proceed further?
thanks for your attention
Susanne
31 Dec
31 Dec
12:09 p.m.
hi susanne!
Susanne Schneider wrote:
hi,
this is no complaint (!) about anything, just to reflect over my
yesterday's exerience with a newly purchased Focusrite Saffire Pro10,
connected via Belkin PCMCIA Firewirecard to my Lenovo R61 - the FW Chip
in this Laptop is one of those crappy Ricoh parts, I tried this, too,
but as expected it just didn't work...
So I compiled the ffado Release Candidate, afterwards compiled the
actual Jack 0.116.1 and finally got things running.
Compared to my Lexicon Omega Studio USB Device the performance of that
FW assembly (don't want to blame a single device of that chain) just was
disappointing, already in idlemode without any clients open the
jackserver showed a workload of roughly 5% (128 frames/3 periods), while
my Lexicon produces between 0.5 to 1% (128 frames/2 periods).
well, ffado is more cpu-intensive than a kernel driver. there's no way
around that. as to the figured you mention: how many channels does your
lexicon have? in my experience, the cpu load is somewhat proportional to
the number of channels ffado has to heave back and forth. for
comparison, my saffire pro26 (26 channels i/o) uses close to 12% cpu,
and proportionally less if i disable the additional adat i/os.
Opening Ardour and just replaying one of my sessions
produced glitches
where only a strong hiss could be heard. (The session itself caused a
workload of less than 20%.) - I remember having had those hiss glitches
in my early days in 2005 with the MAudio Transit USB and Kernel 2.6.11 -
I never expected them coming back some day... *lol*
I then had a look on my interrupts and had to notice the FW Card sharing
it's interrupt (no 16) with the Nwidia Graficscard - so I might have
found the culprit - or not? - I don't know... :-(((
I tried to change the pci settings in the bios - no success.
Now my question to those who also tried firewire audio: was it a success
story for you from the beginning or did anybody make similar experiences
and how did you proceed further?
yeah, firewire is tricky. you definitely need real-time privileges, and
an rt kernel doesn't hurt either. i'm currently not using one for
various reasons, but i have frequent xruns with large sessions.... when
i'm tracking live, i always use one.
ieee1394 controllers are another hairy issue. both chips i use (a pc
card and the one built into my studio box) behave well, but i've heard
people have had a hard time with theirs.
if you want to dig in deeper, you should probably move this thread to
the ffado-user list.
regarding your interrupt woes: are you running a closed-source nvidia
driver? if so, throw it out if at all possible.
just as a final reassuring remark: despite some rough edges, my saffire
is performing nicely. sometimes ardour kills jackd when loading my
current session, but this is a 20ch monster with 5 nine-channel busses
:) restarting jackd and reconnecting ardour fixes it for me.
best,
jörn
2:05 p.m.
New subject: Saffire Pro26 [was: disappointing experience with firewire]
On 31 December 2008 at 17:09, <nettings(a)folkwang-hochschule.de> wrote:
... my saffire pro26 (26 channels i/o) uses close to
12% cpu,
and proportionally less if i disable the additional adat i/os.
It seems that you like your pro26. Would you buy another one?
What are you using to drive the ADAT channels? Could you let us know
what versions of distribution, kernel, and ffado you are using?
Thanks much....
--
Kevin
3:17 p.m.
New subject: Saffire Pro26 [was: disappointing experience with firewire]
Kevin Cosgrove wrote:
On 31 December 2008 at 17:09,
<nettings(a)folkwang-hochschule.de> wrote:
i can't (this one cost me my soul). but if i were ever granted a spare
soul, i might.
... my saffire pro26 (26 channels i/o) uses close
to 12% cpu,
and proportionally less if i disable the additional adat i/os.
It seems that you like your pro26. Would you buy another one? What are you using to drive the ADAT channels? Could
you let us know
what versions of distribution, kernel, and ffado you are using?
when i'm in need of extra analog i/o, i hook up a (gasp!) behringer ada
8000. definitely not the nicest of converters, but does the job (unless
you need decent mic preamps).
my current kernel is (alternating) a .24-rt-something and a .27.
i follow ffado's 2.0 svn branch, jack is also from svn, but the latest
releases should be ok.
4:03 p.m.
New subject: Saffire Pro26 [was: disappointing experience with firewire]
On 31 December 2008 at 20:17, <nettings(a)folkwang-hochschule.de> wrote:
OK, I understand the money thing. I was more curious about whether
you'd buy one if you were rich. My concern is that some day I might
not be able to get a motherboard with two PCI slots for my two
M-Audio Delta 1010s. At that point I'll need to go with something
else, and then firewire must be considered. I need high channel
count inputs - my drumkit really needs 12 mics ;-)
I'd also have to get an ADAT i/o converter of some sort to get to my
desired channel count. Do you know if multiple firewire devices will
sync with each other?
It seems that
you like your pro26. Would you buy another one?
i can't (this one cost me my soul). but if i were ever granted a spare
soul, i might. What are you
using to drive the ADAT channels? Could you let us know
what versions of distribution, kernel, and ffado you are using?
when i'm in need of extra analog i/o, i hook up a (gasp!) behringer ada
8000. definitely not the nicest of converters, but does the job (unless
you need decent mic preamps). my current kernel is (alternating) a .24-rt-something
and a .27.
i follow ffado's 2.0 svn branch, jack is also from svn, but the latest
releases should be ok.
I'm running jack 0.116.1 and haven't branched out (no SVN pun) into
firewire yet.
Thanks....
--
Kevin
12:22 p.m.
i think i read somewhere that the bad performance is related to some
kernel-firewire implementation stuff... (as i understand this will not
change too fast)
most of my problems with my saffire le have been related to irq sharing
... but i had no hiss glitches ...
sorry that i can not help ...
Susanne Schneider schrieb:
hi,
this is no complaint (!) about anything, just to reflect over my
yesterday's exerience with a newly purchased Focusrite Saffire Pro10,
connected via Belkin PCMCIA Firewirecard to my Lenovo R61 - the FW Chip
in this Laptop is one of those crappy Ricoh parts, I tried this, too,
but as expected it just didn't work...
So I compiled the ffado Release Candidate, afterwards compiled the
actual Jack 0.116.1 and finally got things running.
Compared to my Lexicon Omega Studio USB Device the performance of that
FW assembly (don't want to blame a single device of that chain) just was
disappointing, already in idlemode without any clients open the
jackserver showed a workload of roughly 5% (128 frames/3 periods), while
my Lexicon produces between 0.5 to 1% (128 frames/2 periods).
Opening Ardour and just replaying one of my sessions produced glitches
where only a strong hiss could be heard. (The session itself caused a
workload of less than 20%.) - I remember having had those hiss glitches
in my early days in 2005 with the MAudio Transit USB and Kernel 2.6.11 -
I never expected them coming back some day... *lol*
I then had a look on my interrupts and had to notice the FW Card sharing
it's interrupt (no 16) with the Nwidia Graficscard - so I might have
found the culprit - or not? - I don't know... :-(((
I tried to change the pci settings in the bios - no success.
Now my question to those who also tried firewire audio: was it a success
story for you from the beginning or did anybody make similar experiences
and how did you proceed further?
thanks for your attention
Susanne
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
12:29 p.m.
Susanne!
Now my question to those who also tried firewire
audio: was it a success
story for you from the beginning or did anybody make similar experiences
and how did you proceed further?
Wenn Du kuerzestmoegliche Latenz brauchst und mit dem Notebook arbeiten MUSST,
dann wirst Du nicht um etwas wie eine RME Multiface/Digiface herumkommen -
mit PCMCIA-Adapter. Firewire ist leider nicht wirklich brauchbar in der
Hinsicht.
Oder ist Firewire ein MUSS-Kriterium?
Flo
--
Machines can do the work, so people have time to think.
public key 6C002249 x-hkp://wwwkeys.eu.pgp.net
12:46 p.m.
On Wednesday 31 December 2008 17:29:23 Florian Faber wrote:
First: This is still an english-speaking list...
Second: That might be your oppinion, but its not a universal rule. I did use
my firewire-setup with great success for live-foh-mixing and foh-effects. Yes, I
did have low-latency. Firewire is capable of that...
And I don't think buying an expensive RME pcmcia interface will help Susanne
because her pcmcia-slot is sharing its interrupt with the graphics card.
Regardless whether there is a modem, a fw-controller or a rme-interface in
there, its always sharing its interrupt with the (nvidia-)graphics card...
Have fun,
Arnold
Now my
question to those who also tried firewire audio: was it a success
story for you from the beginning or did anybody make similar experiences
and how did you proceed further?
Wenn Du kuerzestmoegliche Latenz brauchst und mit
dem Notebook arbeiten
MUSST, dann wirst Du nicht um etwas wie eine RME Multiface/Digiface
herumkommen - mit PCMCIA-Adapter. Firewire ist leider nicht wirklich
brauchbar in der Hinsicht.
1 Jan
1 Jan
12:40 p.m.
Arnold,
First: This is still an english-speaking list...
Yes, evolution defaults to author, kmail to list, I got mixed up.
Second: That might be your oppinion, but its not a
universal rule.
I wanted to point out an alternative if she needs lots of IO and is
stuck on a notebook. I don't see how that classifies as opinion, it's
just another option.
I did use
my firewire-setup with great success for live-foh-mixing and foh-effects. Yes, I
did have low-latency. Firewire is capable of that...
Latency and system load are still worse than with native PCI/PCIe
solutions in current implementations, no matter what Firewire is
theoretically capable of. And since I don't know what exactly she is
doing, ..
And I don't think buying an expensive RME pcmcia
interface will help
Susanne because her pcmcia-slot is sharing its interrupt with the
graphics card.
Yes, that's broken by design. Something that even can be seen on recent
notebooks. Maybe we should start a list of notebooks with proper
interrupt setup to inform people *before* buying a new machine. What
would be a good place for this?
Flo
--
Machines can do the work, so people have time to think.
public key 6C002249 x-hkp://wwwkeys.eu.pgp.net
1 p.m.
Florian Faber wrote:
Arnold,
There are too many places that it could be hosted. Most obvious is a
wiki format on the linuxaudio.org server.
I'll link to it on the LAU-guide when it is up. I'm sure it will be
helpful as netbook adoption rates are gonna keep growing over the next
few years.
Cheers.
--
Patrick Shirkey
Boost Hardware Ltd.
First: This is still an english-speaking list...
Yes, evolution defaults to author, kmail to list, I got mixed up.
Second: That might be your oppinion, but its not
a universal rule.
I wanted to point out an alternative if she needs lots of IO and is
stuck on a notebook. I don't see how that classifies as opinion, it's
just another option.
I did use
my firewire-setup with great success for live-foh-mixing and foh-effects. Yes, I
did have low-latency. Firewire is capable of that...
Latency and system load are still worse than with native PCI/PCIe
solutions in current implementations, no matter what Firewire is
theoretically capable of. And since I don't know what exactly she is
doing, ..
And I don't think buying an expensive RME
pcmcia interface will help
Susanne because her pcmcia-slot is sharing its interrupt with the
graphics card.
Yes, that's broken by design. Something that even can be seen on recent
notebooks. Maybe we should start a list of notebooks with proper
interrupt setup to inform people *before* buying a new machine. What
would be a good place for this?
31 Dec
31 Dec
12:39 p.m.
Hi,
On Wednesday 31 December 2008 14:49:42 Susanne Schneider wrote:
Compared to my Lexicon Omega Studio USB Device the
performance of that
FW assembly (don't want to blame a single device of that chain) just was
disappointing, already in idlemode without any clients open the
jackserver showed a workload of roughly 5% (128 frames/3 periods), while
my Lexicon produces between 0.5 to 1% (128 frames/2 periods).
As Jörn pointed out this is streaming-in-kernel-space vs. streaming-in-
userspace. And probably your fw-device has _lots_ more channels. (USB1 can
only carry four channels and USB2 has no audio-standard.)
Opening Ardour and just replaying one of my sessions
produced glitches
where only a strong hiss could be heard. (The session itself caused a
workload of less than 20%.) - I remember having had those hiss glitches
in my early days in 2005 with the MAudio Transit USB and Kernel 2.6.11 -
I never expected them coming back some day... *lol*
Did you try with higher latencies?
As long as you don't need to do software-monitoring, you don't need low-
latency for recording. It actually becomes safer to use higher latencies for
recording as the hd has more time to store and read the data.
I then had a look on my interrupts and had to notice
the FW Card sharing
it's interrupt (no 16) with the Nwidia Graficscard - so I might have
found the culprit - or not? - I don't know... :-(((
Probably that is the reason. If ardour shows flashing VU-meters, your gui will
update very often and generate lots of interrupts. This is guaranteed to
interfere with the fw-card.
While it is definitely worth trying to use the free driver (and the only
possibility to get support from kernel-devs and near-kernel-devs), I don't
think this will improve your situation. I have the feeling that if you can't
change the interrupt of either the graphics card or the pcmcia-slot, you are
out of luck. :-(
What is the exact model of your internal ricoh fw-controller? ("lspci -nn |
grep Fire" or add your line to http://subversion.ffado.org/wiki/HostControllers
directly)
Have fun and a happy new year,
Arnold
5601
days inactive
5602
days old
linux-audio-user@lists.linuxaudio.org
10 comments
7 participants
participants (7)
-
Arnold Krille -
Florian Faber -
Jörn Nettingsmeier -
Kevin Cosgrove -
Patrick Shirkey -
sh0099 -
Susanne Schneider