Latency in Audio

[Karl MacMillan]

On Tue, 6 Mar 2001, sfogar at libero.it wrote:

> Hi Karl,
>
> this is only partially true, I'm using Pd with standard (Rme) audio
> drivers, of course !
>

Hmmm - I was thinking that RME would simply put the MME layer over the
same low-level drivers they use for ASIO.  I guess the way to really
determine this is to try to open the card twice with two applications - if
both apps can use the hardware at the same time then they are using the
standard Windows driver model, if not then they are circumventing the
parts of that cause large latencies (kmixer).

> And Karl, what about a pd porting to OS X when it will be ready, aren't
> you working on Apple hardware ?
>

I do run apple hardware and had the MacOS X public beta for a while - it
looks really nice.  As long as tcl/tk is available I think the port should
be pretty easy (and if nothing else there are X servers available that run
transparently with the native gui so it would be possible to use the unix
tcl/tk).  I did a partial FreeBSD port a while ago and that was basically
just a matter of changing some includes (the thing I didn't finish was the
audio I/O and then I needed the disk space again and had to get rid of
FreeBSD).  The MaxOS X port should be similar.  One thing that would make
all of this easier would be to replace the soundcard i/o portion with
something like PortAudio (www.portaudio.com).  That looks to be gaining
some popularity and is really nice to use (I have used it on Windows and
Mac).

Karl

> Cheers
>
> Alessandro Fogar
>
> > ---------- Initial message -----------
> >
> > From    : Karl MacMillan <karlmac at peabody.jhu.edu>
> > To      : "sfogar at libero.it" <sfogar at libero.it>
> > Cc      : <pd-list at iem.kug.ac.at>
> > Date    : Mon, 5 Mar 2001 11:44:11 -0500 (EST)
> > Subject : Re: Latency in Audio
> >
> > On Fri, 2 Mar 2001, sfogar at libero.it wrote:
> >
> > > Hi all,
> > >
> > > in my experience Pd works much better on Win2k than on Win98.
> > >
> >
> > This is not surprising since you are using ASIO - the scheduler is
> > supposed to be significantly more stable under 2000, but the standard
> > audio drivers have additional buffering (in the range of 30
> milliseconds).
> > ASIO circumvents these additional buffers allowing you to take
> advantage
> > of the better overall performance of 2000.
> >
> > Karl
> >
> > > Using Asio and Rme Hammerfall I could obtain good latency (under 50
> ms)
> > > with Cubase.
> > >
> > > Under Win2k the best should be to use a multiprocessor Pc.
> > >
> > > And remember you have to carefully look at the interrupts (perhaps
> > > installing the standard PC Hal helps).
> > >
> > > Consider that you'll not be able to achieve good results using the
> SB
> > > Live with standard drivers and Pd (I tried).
> > >
> > > With The Hammerfall and Pd the latency is acceptable but not ideal.
> > >
> > > Don't know about the Terratec.
> > >
> > > We'd need the Asio interfacing but is it worth ?
> > >
> > > Wouldn't it be better to use Linux or (possibly in the future) Apple
> > > Mac OS X ?
> > >
> > > Cheers
> > >
> > > Alessandro Fogar
> > >
> > > > ---------- Initial message -----------
> > > >
> > > > From    : Karl MacMillan <karlmac at peabody.jhu.edu>
> > > > To      : claudio <c.scozzafava at tin.it>
> > > > Cc      : <pd-list at iem.kug.ac.at>
> > > > Date    : Thu, 1 Mar 2001 22:15:40 -0500 (EST)
> > > > Subject : Re: Latency in Audio
> > > >
> > > > Claudio,
> > > >
> > > > You are not going to do very well trying to get better results
> out of
> > > > Windows ME / 2000 until PD supports something like ASIO.  Even
> then, I
> > > > don't think that the results are going to be that good.  In some
> tests
> > > > that I did recently the best full-duplex performance I could get
> out
> > > of
> > > > Windows ME was 73ms and Windows 2000 was 120ms - this was using a
> > > minimal
> > > > 'ideal' application with the MME interface.  This was on a 700mhz
> > > PIII and
> > > > a 933mhz PIII respectively.
> > > >
> > > > Many companies have been advertising much better results under
> > > Windows and
> > > > I have to say it seems that either a) these results represent
> > > something
> > > > other than real world performance or b) the app runs almost
> entirely
> > > as a
> > > > kernel driver (gigasampler being the best example of this).  If
> you
> > > really
> > > > want good latency you should try Linux or Irix.  For the same test
> > > that I
> > > > mentioned above I got latencies below 3ms using Linux 2.4.1 and
> Alsa -
> > > > this is using standard programming interfaces and the same
> computer
> > > as the
> > > > Windows 2000 test.  If you have to use windows, I would suggest
> > > > downgrading to windows 98 as it is slightly better.
> > > >
> > > > Karl
> > > >
> > > >
> > > > > Hi,
> > > > >        I've some problems with the latency in PD's Audio in/out.
> > > > >     I use PD under Windows ME and Win2000 and works very well
> (in
> > > Win2K midi timing is more stable), but I've some problems to use it
> for
> > > my works in real time. I must use it for very little patch
> (filtering,
> > > ring modulation ...), but the latency is about 300ms.
> > > > >     I run it on a Athlon Th. 800 Mhz, 256 Ram, HD IBM 30Gb,
> > > audioCard SBLive.
> > > > >     How can I reduce the latency? I will buy a Terratec 24/96,
> so I
> > > hope to reduce the latency time.
> > > > >     Can someone give me some councils?
> > > > > With my best regards
> > > > > Claudio Scozzafava
> > > > >
> > > >
> > > > _____________________________________________________
> > > > | Karl W. MacMillan                                 |
> > > > | Computer Music Department                         |
> > > > | Peabody Institute of the Johns Hopkins University |
> > > > | karlmac at peabody.jhu.edu                           |
> > > > | www.peabody.jhu.edu/~karlmac                      |
> > > > -----------------------------------------------------
> > > >
> > > >
> > > >
> > >
> >
> > _____________________________________________________
> > | Karl W. MacMillan                                 |
> > | Computer Music Department                         |
> > | Peabody Institute of the Johns Hopkins University |
> > | karlmac at peabody.jhu.edu                           |
> > | www.peabody.jhu.edu/~karlmac                      |
> > -----------------------------------------------------
> >
> >
> >
>

_____________________________________________________
| Karl W. MacMillan                                 |
| Computer Music Department                         |
| Peabody Institute of the Johns Hopkins University |
| karlmac at peabody.jhu.edu                           |
| www.peabody.jhu.edu/~karlmac                      |
-----------------------------------------------------