stuttering in outgoing announcements

classic Classic list List threaded Threaded
17 messages Options
Reply | Threaded
Open this post in threaded view
|

stuttering in outgoing announcements

Rob Janssen
I am trying to investigate an issue that has been present in one of our repeaters
for quite some time.  Before spending even more time on it, I want to ask here if
other people experience the same thing.

We have a 10 meter repeater.  As usual for such repeaters, it is split-site.  At the
receiver site svxlink is running, at the transmitter site remotetrx.  There are practical
reasons why this cannot be swapped easily.

When the repeater makes a timed announcement or plays the CW ID, once every 2 or 3
times there is a small pause about 2 seconds after the start of the transmission.
This makes the announcement sound funny or the CW ID plainly incorrect.  The pause
is maybe 200ms.

There are no other pauses in the playback other than that one at the beginning,
and it does not appear to happen in normal QSO mode.

Our repeaters have a stream on internet, which is connected via a second Tx
feeding a darkice streamer, and it does not occur on the stream (the same announcement
that is garbled on the radio is correct on the stream).

So it seems to happen only on the remotetrx.  The link to the TX site is over 6cm WiFi,
but the link is very stable and can transport lots of data without loss.  I have changed
the CODEC from S16 to OPUS but there is no change (I suspected that there was a
large chunk of data being sent and overflowing some queue, and there would be a
re-try on TCP level, but in that case the timing should be different for S16 and OPUS,
which it isn't).

We have no other installations where the transmitter audio is over remotetrx (there
are other remote transmitters on our other repeaters but they use my co-channel
program), so I have little to compare.   Do others use remote transmitters with a
remotetrx connection and did they ever observe this?

Rob

------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

David Ranch-3

Hello Rob,


> When the repeater makes a timed announcement or plays the CW ID, once every 2 or 3
> times there is a small pause about 2 seconds after the start of the transmission.
> This makes the announcement sound funny or the CW ID plainly incorrect.  The pause
> is maybe 200ms.

Are you saying that there is a 2 second delay between when PTT asserts
and when the announcement is started?  I don't see how this would mess
up the CW ID.  If you switch to a voice ID, do you still have the same
issue?

--David


------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Rob Janssen
David Ranch wrote:

> Hello Rob,
>
>
>> When the repeater makes a timed announcement or plays the CW ID, once every 2 or 3
>> times there is a small pause about 2 seconds after the start of the transmission.
>> This makes the announcement sound funny or the CW ID plainly incorrect.  The pause
>> is maybe 200ms.
> Are you saying that there is a 2 second delay between when PTT asserts
> and when the announcement is started?  I don't see how this would mess
> up the CW ID.  If you switch to a voice ID, do you still have the same
> issue?
>

No, what I mean is the transmitter is keyed, then the announcement plays for 2 seconds or so,
then there is a brief pause and the announcement continues where it paused.
(there is nothing lost)
"papa india six t......ango echo november"

We use both voice and CW ID messages (voice on :00 and :30 and CW every 5 minutes other
than those two), and it happens for both of them.
It also happens when an Echolink station connects and the call is announced.   Or when * is
keyed in DTMF and the repeater description is sent.

Rob



------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Marius Petrescu
Hello,

I am having a similar situation on my Raspberry Pi.
On the timed announcement, there is a gap before the last wave file to be
played.
Something like "yankee oskar two kilowat quebec tango r.....epeater"

I modified the tcl script to announce the time and temperature on each half
hour.
And it always sounds:
"temperature is ten de....grees"

...the gap being always in the last sound file to be played.

In the help, echolink and metar information modules, everything seems ok.

Marius, yo2loj

-----Original Message-----
From: Rob Janssen [mailto:[hidden email]]
Sent: Thursday, April 23, 2015 21:21
To: Discussions about development issues
Subject: Re: [Svxlink-devel] stuttering in outgoing announcements

David Ranch wrote:
> Hello Rob,
>
>
>> When the repeater makes a timed announcement or plays the CW ID, once
every 2 or 3
>> times there is a small pause about 2 seconds after the start of the
transmission.
>> This makes the announcement sound funny or the CW ID plainly incorrect.
The pause
>> is maybe 200ms.
> Are you saying that there is a 2 second delay between when PTT asserts
> and when the announcement is started?  I don't see how this would mess
> up the CW ID.  If you switch to a voice ID, do you still have the same
> issue?
>

No, what I mean is the transmitter is keyed, then the announcement plays for
2 seconds or so,
then there is a brief pause and the announcement continues where it paused.
(there is nothing lost)
"papa india six t......ango echo november"

We use both voice and CW ID messages (voice on :00 and :30 and CW every 5
minutes other
than those two), and it happens for both of them.
It also happens when an Echolink station connects and the call is announced.
Or when * is
keyed in DTMF and the repeater description is sent.

Rob



----------------------------------------------------------------------------
--
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel



------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Rob Janssen
Marius Petrescu wrote:

> Hello,
>
> I am having a similar situation on my Raspberry Pi.
> On the timed announcement, there is a gap before the last wave file to be
> played.
> Something like "yankee oskar two kilowat quebec tango r.....epeater"
>
> I modified the tcl script to announce the time and temperature on each half
> hour.
> And it always sounds:
> "temperature is ten de....grees"
>
> ...the gap being always in the last sound file to be played.
>
> In the help, echolink and metar information modules, everything seems ok.
>
> Marius, yo2loj
>

Interesting information (and also nice to see you here as well :-)

In our case the system is not a Raspberry Pi.  On both ends it is a PC.
And it is not happening in the last wav file, but more like at a certain time offset into
the message (sometimes a single wav file, sometimes a series of short wav files, sometimes
generated tones/pauses like in the CW ID).

We have a single wav file with the entire "papa india six tango echo november" and
also separate wav files with single letters used for the echolink connects, or digits
for the time announcement and in both cases there is a pause somewhere inside
the callsign.

So it appears there is a common issue but it is not yet clear exactly where.
In your case, is the entire repeater running on a single Raspberry Pi or do you have a
remotetrx link as well?
We have other repeaters that run on a single system and we do not observe this issue there.

Rob

------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Tobias Blomberg
In reply to this post by Marius Petrescu
On Thursday 23 April 2015 22:04:36 Marius Petrescu wrote:

> Hello,
>
> I am having a similar situation on my Raspberry Pi.
> On the timed announcement, there is a gap before the last wave file to be
> played.
> Something like "yankee oskar two kilowat quebec tango r.....epeater"
>
> I modified the tcl script to announce the time and temperature on each half
> hour.
> And it always sounds:
> "temperature is ten de....grees"
>
> ...the gap being always in the last sound file to be played.
>
> In the help, echolink and metar information modules, everything seems ok.

This sounds like a problem that have been present in SvxLink from time to
time. I thought it was fixed though. Are you running the latest code from git
master?

The problem then seemed to have something to do when playing silence at the
end of annoncements. As I remember it, there were a filter that misbehaved when
sending only zeros through it. As a test, try to remove the playSilence 500 at
the end of "send_short_ident" to see if that change anything for the short
identifications.

Look at the svxlink CPU usage (using "top") when the stuttering happens. Is it
100%?

However, the problem Rob is having does not sound like this bug. I have never
seen that before. Rob, maybe you could try to set up another RemoteTrx over
another (wired?) link. It could just be a dummy transmitter using a speaker,
just to see if the problem is there as well.

73's de SM0SVX / Tobias


>
> Marius, yo2loj
>
> -----Original Message-----
> From: Rob Janssen [mailto:[hidden email]]
> Sent: Thursday, April 23, 2015 21:21
> To: Discussions about development issues
> Subject: Re: [Svxlink-devel] stuttering in outgoing announcements
>
> David Ranch wrote:
> > Hello Rob,
> >
> >> When the repeater makes a timed announcement or plays the CW ID, once
>
> every 2 or 3
>
> >> times there is a small pause about 2 seconds after the start of the
>
> transmission.
>
> >> This makes the announcement sound funny or the CW ID plainly incorrect.
>
> The pause
>
> >> is maybe 200ms.
> >
> > Are you saying that there is a 2 second delay between when PTT asserts
> > and when the announcement is started?  I don't see how this would mess
> > up the CW ID.  If you switch to a voice ID, do you still have the same
> > issue?
>
> No, what I mean is the transmitter is keyed, then the announcement plays for
> 2 seconds or so,
> then there is a brief pause and the announcement continues where it paused.
> (there is nothing lost)
> "papa india six t......ango echo november"
>
> We use both voice and CW ID messages (voice on :00 and :30 and CW every 5
> minutes other
> than those two), and it happens for both of them.
> It also happens when an Echolink station connects and the call is announced.
> Or when * is
> keyed in DTMF and the repeater description is sent.
>
> Rob


------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Rob Janssen
SM0SVX wrote:
> However, the problem Rob is having does not sound like this bug. I have never
> seen that before. Rob, maybe you could try to set up another RemoteTrx over
> another (wired?) link. It could just be a dummy transmitter using a speaker,
> just to see if the problem is there as well.
>

That was already my plan for a next experiment, but I decided to first ask if others have
seen the same thing.   I can setup a remotetrx at home, there is no WiFi inbetween but
of course a lot of other link hops.   I'll see if it shows the same issue.

My first guess was that it is somehow related to a load peak that may be present at :00
because probably other tasks are running then.  But then I also observed it at other moments.

Rob

------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Tobias Blomberg
On Thursday 23 April 2015 21:58:12 Rob Janssen wrote:

> SM0SVX wrote:
> > However, the problem Rob is having does not sound like this bug. I have
> > never seen that before. Rob, maybe you could try to set up another
> > RemoteTrx over another (wired?) link. It could just be a dummy
> > transmitter using a speaker, just to see if the problem is there as well.
>
> That was already my plan for a next experiment, but I decided to first ask
> if others have seen the same thing.   I can setup a remotetrx at home,
> there is no WiFi inbetween but of course a lot of other link hops.   I'll
> see if it shows the same issue.

A recording of the network traffic on both sides using for example
tshark/wireshark would be interesting as well to see if you are having some
kind of window size problem and/or packet loss. You could get a problem like
that if your ping times are high since ACKs are not getting back in time
before the TCP window fills up.

Another classic is if some part of the network does not handle path MTU
discovery properly. That can happen if for example the needed ICMP packets is
not let through a firewall somewhere along the path. That may lead to sudden
stops in network traffic, but I think the stops usually are longer than 200ms.

73's de SM0SVX / Tobias


>
> My first guess was that it is somehow related to a load peak that may be
> present at :00 because probably other tasks are running then.  But then I
> also observed it at other moments.
>
> Rob


------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Rob Janssen
SM0SVX wrote:

> A recording of the network traffic on both sides using for example
> tshark/wireshark would be interesting as well to see if you are having some
> kind of window size problem and/or packet loss. You could get a problem like
> that if your ping times are high since ACKs are not getting back in time
> before the TCP window fills up.
>
> Another classic is if some part of the network does not handle path MTU
> discovery properly. That can happen if for example the needed ICMP packets is
> not let through a firewall somewhere along the path. That may lead to sudden
> stops in network traffic, but I think the stops usually are longer than 200ms.
>

That is why one of my experiments was to change from CODEC=S16 to CODEC=OPUS,
as that should reduce the traffic and the effect should at least change.  However, that did
not happen.   Note that the WiFi link is not very bad:

ping tx.pi6ten.ampr.org
PING tx.pi6ten.ampr.org (44.137.37.10) 56(84) bytes of data.
64 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=1 ttl=64 time=0.864 ms
64 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=2 ttl=64 time=0.866 ms
64 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=3 ttl=64 time=0.843 ms
64 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=4 ttl=64 time=0.850 ms
64 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=5 ttl=64 time=0.842 ms
^C
--- tx.pi6ten.ampr.org ping statistics ---
5 packets transmitted, 5 received, 0% packet loss, time 4005ms
rtt min/avg/max/mdev = 0.842/0.853/0.866/0.010 ms

ping -M do -s 1472 -i 0.1 tx.pi6ten.ampr.org
PING tx.pi6ten.ampr.org (44.137.37.10) 1472(1500) bytes of data.
1480 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=1 ttl=64 time=2.73 ms
1480 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=2 ttl=64 time=2.67 ms
1480 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=3 ttl=64 time=2.67 ms
1480 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=4 ttl=64 time=2.67 ms
1480 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=5 ttl=64 time=2.72 ms
1480 bytes from tx.pi6ten.ampr.org (44.137.37.10): icmp_req=6 ttl=64 time=2.67 ms
^C
--- tx.pi6ten.ampr.org ping statistics ---
6 packets transmitted, 6 received, 0% packet loss, time 502ms
rtt min/avg/max/mdev = 2.670/2.694/2.738/0.026 ms

The link is from a VLAN at the svxlink machine, via a managable switch to the WiFi AP,
then to the remote AP and a switch to the remotetrx machine.  There are no routers
inbetween, and the iptables fw on the machine allows ICMP:

ping -M do -s 1474 tx.pi6ten.ampr.org
PING tx.pi6ten.ampr.org (44.137.37.10) 1474(1502) bytes of data.
 From gw.pi6ten.ampr.org (44.137.37.1) icmp_seq=1 Frag needed and DF set (mtu = 1500)

So it does not seem to be the obvious error...   but I am still researching it.

Rob

------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Marius Petrescu
In reply to this post by Tobias Blomberg
Hi,
I'm using the 14.08 standard release, not the latest GIT code (it works
untouched on my Pi for almost a year, so I don't want to kill it).
However, removing the trailing "playSilence 250" in the script seems to
solve the issue.

Marius, YO2LOJ


-----Original Message-----
From: SM0SVX [mailto:[hidden email]]
Sent: Thursday, April 23, 2015 22:51
To: Discussions about development issues
Subject: Re: [Svxlink-devel] stuttering in outgoing announcements

On Thursday 23 April 2015 22:04:36 Marius Petrescu wrote:
> Hello,
>
> I am having a similar situation on my Raspberry Pi.
> On the timed announcement, there is a gap before the last wave file to be
> played.
> Something like "yankee oskar two kilowat quebec tango r.....epeater"
>
> I modified the tcl script to announce the time and temperature on each
half
> hour.
> And it always sounds:
> "temperature is ten de....grees"
>
> ...the gap being always in the last sound file to be played.
>
> In the help, echolink and metar information modules, everything seems ok.

This sounds like a problem that have been present in SvxLink from time to
time. I thought it was fixed though. Are you running the latest code from
git
master?

The problem then seemed to have something to do when playing silence at the
end of annoncements. As I remember it, there were a filter that misbehaved
when
sending only zeros through it. As a test, try to remove the playSilence 500
at
the end of "send_short_ident" to see if that change anything for the short
identifications.

Look at the svxlink CPU usage (using "top") when the stuttering happens. Is
it
100%?

However, the problem Rob is having does not sound like this bug. I have
never
seen that before. Rob, maybe you could try to set up another RemoteTrx over
another (wired?) link. It could just be a dummy transmitter using a speaker,

just to see if the problem is there as well.

73's de SM0SVX / Tobias


>
> Marius, yo2loj
>
> -----Original Message-----
> From: Rob Janssen [mailto:[hidden email]]
> Sent: Thursday, April 23, 2015 21:21
> To: Discussions about development issues
> Subject: Re: [Svxlink-devel] stuttering in outgoing announcements
>
> David Ranch wrote:
> > Hello Rob,
> >
> >> When the repeater makes a timed announcement or plays the CW ID, once
>
> every 2 or 3
>
> >> times there is a small pause about 2 seconds after the start of the
>
> transmission.
>
> >> This makes the announcement sound funny or the CW ID plainly incorrect.
>
> The pause
>
> >> is maybe 200ms.
> >
> > Are you saying that there is a 2 second delay between when PTT asserts
> > and when the announcement is started?  I don't see how this would mess
> > up the CW ID.  If you switch to a voice ID, do you still have the same
> > issue?
>
> No, what I mean is the transmitter is keyed, then the announcement plays
for
> 2 seconds or so,
> then there is a brief pause and the announcement continues where it
paused.
> (there is nothing lost)
> "papa india six t......ango echo november"
>
> We use both voice and CW ID messages (voice on :00 and :30 and CW every 5
> minutes other
> than those two), and it happens for both of them.
> It also happens when an Echolink station connects and the call is
announced.
> Or when * is
> keyed in DTMF and the repeater description is sent.
>
> Rob


----------------------------------------------------------------------------
--
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel



------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Tobias Blomberg
On Saturday 25 April 2015 17:11:35 Marius Petrescu wrote:
> Hi,
> I'm using the 14.08 standard release, not the latest GIT code (it works
> untouched on my Pi for almost a year, so I don't want to kill it).
> However, removing the trailing "playSilence 250" in the script seems to
> solve the issue.

OK. Then the problem will probably be solved in the next release. Until then,
you know how to work around the issue.

73's de SM0SVX / Tobias


>
> Marius, YO2LOJ
>
>
> -----Original Message-----
> From: SM0SVX [mailto:[hidden email]]
> Sent: Thursday, April 23, 2015 22:51
> To: Discussions about development issues
> Subject: Re: [Svxlink-devel] stuttering in outgoing announcements
>
> On Thursday 23 April 2015 22:04:36 Marius Petrescu wrote:
> > Hello,
> >
> > I am having a similar situation on my Raspberry Pi.
> > On the timed announcement, there is a gap before the last wave file to be
> > played.
> > Something like "yankee oskar two kilowat quebec tango r.....epeater"
> >
> > I modified the tcl script to announce the time and temperature on each
>
> half
>
> > hour.
> > And it always sounds:
> > "temperature is ten de....grees"
> >
> > ...the gap being always in the last sound file to be played.
> >
> > In the help, echolink and metar information modules, everything seems ok.
>
> This sounds like a problem that have been present in SvxLink from time to
> time. I thought it was fixed though. Are you running the latest code from
> git
> master?
>
> The problem then seemed to have something to do when playing silence at the
> end of annoncements. As I remember it, there were a filter that misbehaved
> when
> sending only zeros through it. As a test, try to remove the playSilence 500
> at
> the end of "send_short_ident" to see if that change anything for the short
> identifications.
>
> Look at the svxlink CPU usage (using "top") when the stuttering happens. Is
> it
> 100%?
>
> However, the problem Rob is having does not sound like this bug. I have
> never
> seen that before. Rob, maybe you could try to set up another RemoteTrx over
> another (wired?) link. It could just be a dummy transmitter using a speaker,
>
> just to see if the problem is there as well.
>
> 73's de SM0SVX / Tobias
>
> > Marius, yo2loj
> >
> > -----Original Message-----
> > From: Rob Janssen [mailto:[hidden email]]
> > Sent: Thursday, April 23, 2015 21:21
> > To: Discussions about development issues
> > Subject: Re: [Svxlink-devel] stuttering in outgoing announcements
> >
> > David Ranch wrote:
> > > Hello Rob,
> > >
> > >> When the repeater makes a timed announcement or plays the CW ID, once
> >
> > every 2 or 3
> >
> > >> times there is a small pause about 2 seconds after the start of the
> >
> > transmission.
> >
> > >> This makes the announcement sound funny or the CW ID plainly incorrect.
> >
> > The pause
> >
> > >> is maybe 200ms.
> > >
> > > Are you saying that there is a 2 second delay between when PTT asserts
> > > and when the announcement is started?  I don't see how this would mess
> > > up the CW ID.  If you switch to a voice ID, do you still have the same
> > > issue?
> >
> > No, what I mean is the transmitter is keyed, then the announcement plays
>
> for
>
> > 2 seconds or so,
> > then there is a brief pause and the announcement continues where it
>
> paused.
>
> > (there is nothing lost)
> > "papa india six t......ango echo november"
> >
> > We use both voice and CW ID messages (voice on :00 and :30 and CW every 5
> > minutes other
> > than those two), and it happens for both of them.
> > It also happens when an Echolink station connects and the call is
>
> announced.
>
> > Or when * is
> > keyed in DTMF and the repeater description is sent.
> >
> > Rob


------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Rob Janssen
In reply to this post by Tobias Blomberg
SM0SVX wrote:
>
> However, the problem Rob is having does not sound like this bug. I have never
> seen that before. Rob, maybe you could try to set up another RemoteTrx over
> another (wired?) link. It could just be a dummy transmitter using a speaker,
> just to see if the problem is there as well.
>

This weekend I have reconfigured my home network connection so it is easier to connect
to a remotetrx here at home from the repeater, and I have tested the above setup for some
time.  It looks like it does not reproduce.  The sound was clean every time.

As I don't believe there is a network problem in the link to the transmitter, I now want to
test the PC at the tx site.  Maybe there is a problem with the sound driver that causes that
hickup shortly after opening the device.  Maybe a missed interrupt that is recovered with
a timeout in the driver.

Tomorrow is king's day so again some time to test and debug.

Rob

------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Rob Janssen
Rob Janssen wrote:
>
> As I don't believe there is a network problem in the link to the transmitter, I now want to
> test the PC at the tx site.  Maybe there is a problem with the sound driver that causes that
> hickup shortly after opening the device.  Maybe a missed interrupt that is recovered with
> a timeout in the driver.


More testing reveals that it probably is a network problem, but it is still a guess why it manifests
itself like this.

My testing with ping was not representing the real situation. Svxlink uses TCP to transfer the
voice data, and WiFi handles TCP differently from datagram protocols like UDP and ICMP.

To avoid retransmitting data that is retransmitted at the higher level, WiFi does not re-try lost
TCP packets and leaves the retransmission to the network stack.  But UDP and ICMP are
retransmitted by the access points themselves.  This retransmit happens very quickly, so
it is not noticed.  Linux retransmits TCP packets when 2-3 more packets have arrived with
higher sequence numbers, so it does not have to wait for a timeout (Fast Retransmit).
However, apparently it is still too late.

What I still find puzzling is why it affects announcements, some time after the start, and not
the normal QSO traffic.  Maybe in the latter case there is more buffering and the time to
retransmit is no problem, but I also notice that during a longer transmission there are little
or no retransmissions, and during the beginning of an announcement there often are some.

This could be caused by the startup of the TCP transmit sending some packets when they
become available, while later the packets are sent as replies to the ACK traffic.  That could
mean on the initial packets there are collisions on the (half-duplex) WiFi link, while this does
not happen that much later in a sequence.

What would probably help in this case is a UDP mode for transferring the voice.  Maybe I can
find some trick to encapsulate the TCP packets in UDP (e.g. a VPN).

Rob

------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Rob Janssen
Rob Janssen wrote:
>
> What would probably help in this case is a UDP mode for transferring the voice.

I have used a UDP output device on the svxlink side and a socat|aplay pipe on the
transmitter site, controlling the PTT using a remotetrx with CODEC=NULL.

Of course this sends way too much data (192000 bytes/s) but the link should be able
to handle that.   However, the problem is similar.

Apparently it really is a problem with the link.   Funny that it is not detected when
pinging (even at high rate) or when downloading files.  This will be difficult to solve, I fear.

Rob

------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Rob Janssen
Rob Janssen wrote:
>
> Apparently it really is a problem with the link.   Funny that it is not detected when
> pinging (even at high rate) or when downloading files.  This will be difficult to solve, I fear.

As a coincidence I found this problem is caused by the switch at the server end of the connection.
It caused a small packet loss on some (but not all) connections.  It is unclear why, but it could
be fixed by a configuration change.

Rob

------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

David Ranch-3
Hello Rob,

Curious, what is the change that fixed it?  Since you say it's a configuration change, you're talking about a "web" class switch or a full blown smart switch but I'm curious what was happening.

--David


On 05/04/2015 06:32 AM, Rob Janssen wrote:
Rob Janssen wrote:
Apparently it really is a problem with the link.   Funny that it is not detected when
pinging (even at high rate) or when downloading files.  This will be difficult to solve, I fear.
As a coincidence I found this problem is caused by the switch at the server end of the connection.
It caused a small packet loss on some (but not all) connections.  It is unclear why, but it could
be fixed by a configuration change.

Rob

------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud 
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel


------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel
Reply | Threaded
Open this post in threaded view
|

Re: stuttering in outgoing announcements

Rob Janssen
David Ranch wrote:
Hello Rob,

Curious, what is the change that fixed it?  Since you say it's a configuration change, you're talking about a "web" class switch or a full blown smart switch but I'm curious what was happening.

--David


It is a Cisco 3550 L3 switch.  The configuration has QOS and policing, and it is routing as well.  It all worked ok until
the switch was powercycled due to a widespread powerfailure.  Then it had a small packetloss on some paths but
not on all.  It was not really clear which paths were affected and which were not.   Removing a "service-policy input" on
another port fixed it.  I still need to debug the thing further, probably when I apply the policy again it still works, as it
worked before as well.   It could be a bug in the switch, or maybe I misunderstand how the policing works.
(the path we are talking about has only priority queuing based on QOS and no policing)

Rob

------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Svxlink-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/svxlink-devel