18 Apr
2011
18 Apr
'11
10:21 p.m.
Hello all,
A new release of jack_delay is available at
<http://kokkinizita.linuxaudio.org/linuxaudio/downloads>
From the README:
jack_delay 0.4.0 - 18/04/2011
-------------------------------
Jack_delay can be used to measure the round-trip latency of a soundcard.
To do this, start the program and connect like this:
jack_delay -> playback_port -> cable from soundcard output to input -> capture
port -> jack_delay
Jack_delay generates a signal consisting of 13 sine waves, measures the
phase difference between the input and output for each of these, and
computes the delay from those phase differences. The algorithm used is
one developed originally for satellite ranging - that is measuring the
distance between a satellite and a ground station.
With a good sound card jack_delay will measure the round-trip latency
with an accuracy of around 1/1000 of a sample. The assumption is that
the delay is more or less independent of frequency. The actual value
displayed is the one for a frequency of 1/16 of the sample rate. The
phase measurement for this frequency of course only provides a result
in the range of 0..16 samples. The other frequencies are used to extend
this interval to 4096 * 16 samples, more than a second at 48 kHz.
This release should be much less sensitive to frequency-dependent delay
than the previous ones.
The following options are avaiable (use jack_delay -h to see them):
-O playback port connect output to named port.
-I capture port connect input to named port.
-E show excess latency instead of full latency.
Using -E requires -O and -I, as the the computation depends on
the latency values reported by jack for the ports used.
The excess latency is the measured value minus the expected one,
taking into account any corrections set by jack's -I and -O options.
That is, if you have the right values for these options, then the
value displayed with -E will be at most +/- half a sample.
To determine the correct values for jack's -I and -O, set both
of them to zero ('default' in qjackctl) and measure the latency
using the -E option. Then set each of the -I and -O options to
half the value displayed.
Ciao,
--
FA
18 Apr
18 Apr
10:50 p.m.
Thanks,
Builds and runs fine on Debian Wheezy AMD64. What latency would be
considered average? What numbers should one expect?
Cheers,
S.M.
On Mon, Apr 18, 2011 at 08:21:26PM +0000, Fons Adriaensen wrote:
Hello all,
A new release of jack_delay is available at
<http://kokkinizita.linuxaudio.org/linuxaudio/downloads>
--
From the README:
jack_delay 0.4.0 - 18/04/2011
-------------------------------
Jack_delay can be used to measure the round-trip latency of a soundcard.
To do this, start the program and connect like this:
jack_delay -> playback_port -> cable from soundcard output to input -> capture
port -> jack_delay
Jack_delay generates a signal consisting of 13 sine waves, measures the
phase difference between the input and output for each of these, and
computes the delay from those phase differences. The algorithm used is
one developed originally for satellite ranging - that is measuring the
distance between a satellite and a ground station.
With a good sound card jack_delay will measure the round-trip latency
with an accuracy of around 1/1000 of a sample. The assumption is that
the delay is more or less independent of frequency. The actual value
displayed is the one for a frequency of 1/16 of the sample rate. The
phase measurement for this frequency of course only provides a result
in the range of 0..16 samples. The other frequencies are used to extend
this interval to 4096 * 16 samples, more than a second at 48 kHz.
This release should be much less sensitive to frequency-dependent delay
than the previous ones.
The following options are avaiable (use jack_delay -h to see them):
-O playback port connect output to named port.
-I capture port connect input to named port.
-E show excess latency instead of full latency.
Using -E requires -O and -I, as the the computation depends on
the latency values reported by jack for the ports used.
The excess latency is the measured value minus the expected one,
taking into account any corrections set by jack's -I and -O options.
That is, if you have the right values for these options, then the
value displayed with -E will be at most +/- half a sample.
To determine the correct values for jack's -I and -O, set both
of them to zero ('default' in qjackctl) and measure the latency
using the -E option. Then set each of the -I and -O options to
half the value displayed.
Ciao,
--
FA
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-user 
11:21 p.m.
On Mon, Apr 18, 2011 at 4:52 PM, S. Massy <lists(a)wolfdream.ca> wrote:
Thanks,
Builds and runs fine on Debian Wheezy AMD64. What latency would be
considered average? What numbers should one expect?
the "excess" value is generally on the order of several tens of
samples in each direction. it can reach as high as the low hundreds.
with USB audio (and potentially some firewire devices, the excess can
even get up into the thousands, but its dependent on a variety of
factors. this assumes you are doing the obvious thing and measuring a
loopback via a D/A and A/D converter pair.
for reference, my RME HDSP system, connected to a Frontier Designs
Tango24 converter, meaures 68 samples of total excess latency, so
about 34 in each direction.
19 Apr
19 Apr
1:05 a.m.
On Mon, Apr 18, 2011 at 05:21:46PM -0400, Paul Davis wrote:
On Mon, Apr 18, 2011 at 4:52 PM, S. Massy
<lists(a)wolfdream.ca> wrote:
That's a very typical value.
If you have an HDSP or any other card with ADAT (or MADI) interfaces
it can be interesting to measure the excess latency with a digital
loopback as well.
If this is much less than the value obtained when going through
the DA/AD converters, then the difference is an indication of the
type and lenght of the anti-aliasing filters used in the converters.
If these filters are linear-phase (i.e. a symmetric FIR), then
each of them will add a delay equal to half the length of the
filter, so the excess delay is a measure of the filter length.
Values I've seen so far appear to be much less than e.g. the filter
lengths used by libsrc or zita-resampler at high quality settings.
Ciao,
--
FA
Thanks,
Builds and runs fine on Debian Wheezy AMD64. What latency would be
considered average? What numbers should one expect?
the "excess" value is generally on the order of several tens of
samples in each direction. it can reach as high as the low hundreds.
with USB audio (and potentially some firewire devices, the excess can
even get up into the thousands, but its dependent on a variety of
factors. this assumes you are doing the obvious thing and measuring a
loopback via a D/A and A/D converter pair.
for reference, my RME HDSP system, connected to a Frontier Designs
Tango24 converter, meaures 68 samples of total excess latency, so
about 34 in each direction. 
18 Apr
18 Apr
11:24 p.m.
On 04/18/2011 10:52 PM, S. Massy wrote:
Thanks,
Builds and runs fine on Debian Wheezy AMD64. What latency would be
considered average? What numbers should one expect?
If you run it with '-E': something around 50-100 frames for PCI devices,
and for USB-devices it depends on the period-size.
http://wiki.linuxaudio.org/wiki/jack_latency_tests#interpretation_and_analy…
ciao,
robin
5101
days inactive
5101
days old
linux-audio-user@lists.linuxaudio.org
4 comments
4 participants
participants (4)
-
Fons Adriaensen -
Paul Davis -
Robin Gareus -
S. Massy