Re:  Re: Physical distance performance with Linux?

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

Re:  Re: Physical distance performance with Linux?

Christoph Kuhr
Hi,

we were just playing remote a few minutes ago. We used soundjack with a round trip time of 50ms and less, since we all live in cologne but our ISPs peer in Frankfurt. Thus, you should be able to achieve such latencies from everywhere inside germany. But I don't know about other countries.
We are playing some hard rock stuff with tempo of 120 bpm and more. If you try to be in sync with each other, your tempo goes down. If you want to maintain the original tempo, everybody has to maintain their own tempo and play a litte before the time you hear your mates. Then everybody can hear their remote team mates in sync, only you yourself are too early.
After some practice you get used to this...

Best,
Ck


Am 05.05.20, 15:01 schrieb Giso Grimm <[hidden email]>:
On 05.05.20 14:38, Mac wrote:
> In our current world situation I see a lot of videos, and in some cases
> live, performances by multiple musicians/vocalists in physically
> different locations.
>
> Some are obviously in a MacGyver setup in the closet, others are in
> rather tricked out home studios (like the one I just watched of the
> Doobie Brothers, where they all were in their own studio).
>
> Some are obvious compilations, video effects, etc. added. But, when they
> are live how do they deal with delays and monitoring for the local
> performance (in the Doobie's one, there was some post processing, but,
> the duets seemed really tight, both instruments and vocals.)
>
> My question is, have any Linux folks had any experience setting this,
> with Linux tools obviously..., and how did you deal with delays and
> monitoring when live or (I assume when processed video is involved)
> monitoring and delay for multi-tracking from around the world?

There are several tools out, some of them open source based:

jacktrip ( https://ccrma.stanford.edu/software/jacktrip/)

soundjack ( https://soundjack.eu/, closed source)

jamulus ( http://llcon.sourceforge.net/)

Just to name a few.

We developed our own system based on zita-njbridge and some self-written
network managing tools, which we use almost every day:

https://github.com/gisogrimm/ovbox

The delay which can be achieved (with all of the systems - the
underlying network is the most critical factor) will be in the order of
30-80 ms between musicians (one way). 30-50 ms feels more or less ok, if
it is more you need to play actively before the time.

Best,

Giso


>
> Regards,
> Mac
>
> _______________________________________________
> Linux-audio-user mailing list
> [hidden email]
> https://lists.linuxaudio.org/listinfo/linux-audio-user
>
_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user

_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user
Reply | Threaded
Open this post in threaded view
|

Re: Physical distance performance with Linux?

Paul Davis
The one thing I remember about soundjack from the presentation at tonmeistertage years ago was that inter-continental latencies are MUCH worse than in-country, often to the point of not being very usable. This isn't due to soundjack but the nature of routing across sub-ocean links etc. Just something to be aware of if you're trying to connect and work with people a *long* way away.

On Tue, May 5, 2020 at 11:07 AM Christoph Kuhr <[hidden email]> wrote:
Hi,

we were just playing remote a few minutes ago. We used soundjack with a round trip time of 50ms and less, since we all live in cologne but our ISPs peer in Frankfurt. Thus, you should be able to achieve such latencies from everywhere inside germany. But I don't know about other countries.
We are playing some hard rock stuff with tempo of 120 bpm and more. If you try to be in sync with each other, your tempo goes down. If you want to maintain the original tempo, everybody has to maintain their own tempo and play a litte before the time you hear your mates. Then everybody can hear their remote team mates in sync, only you yourself are too early.
After some practice you get used to this...

Best,
Ck


Am 05.05.20, 15:01 schrieb Giso Grimm <[hidden email]>:
On 05.05.20 14:38, Mac wrote:
> In our current world situation I see a lot of videos, and in some cases
> live, performances by multiple musicians/vocalists in physically
> different locations.
>
> Some are obviously in a MacGyver setup in the closet, others are in
> rather tricked out home studios (like the one I just watched of the
> Doobie Brothers, where they all were in their own studio).
>
> Some are obvious compilations, video effects, etc. added. But, when they
> are live how do they deal with delays and monitoring for the local
> performance (in the Doobie's one, there was some post processing, but,
> the duets seemed really tight, both instruments and vocals.)
>
> My question is, have any Linux folks had any experience setting this,
> with Linux tools obviously..., and how did you deal with delays and
> monitoring when live or (I assume when processed video is involved)
> monitoring and delay for multi-tracking from around the world?

There are several tools out, some of them open source based:

jacktrip ( https://ccrma.stanford.edu/software/jacktrip/)

soundjack ( https://soundjack.eu/, closed source)

jamulus ( http://llcon.sourceforge.net/)

Just to name a few.

We developed our own system based on zita-njbridge and some self-written
network managing tools, which we use almost every day:

https://github.com/gisogrimm/ovbox

The delay which can be achieved (with all of the systems - the
underlying network is the most critical factor) will be in the order of
30-80 ms between musicians (one way). 30-50 ms feels more or less ok, if
it is more you need to play actively before the time.

Best,

Giso


>
> Regards,
> Mac
>
> _______________________________________________
> Linux-audio-user mailing list
> [hidden email]
> https://lists.linuxaudio.org/listinfo/linux-audio-user
>
_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user
_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user

_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user
Reply | Threaded
Open this post in threaded view
|

Re: Physical distance performance with Linux?

Chris Caudle
On Tue, May 5, 2020 1:07 pm, Paul Davis wrote:
> The one thing I remember about soundjack from the presentation at
> tonmeistertage years ago was that inter-continental latencies are MUCH
> worse than in-country, often to the point of not being very usable. This
> isn't due to soundjack but the nature of routing across sub-ocean links
> etc. Just something to be aware of if you're trying to connect and work
> with people a *long* way away.

That should also apply in general to the number of routing devices the
link has to traverse, and general distance.  So same city would be better,
same ISP  in the same city wold probably be better still.  Best would be
same building connecting through a single building Ethernet switch.

I watched that Doobie Brothers performance originally mentioned, and I
don't think that was live.  That is one possibility if you don't mind,
have one person start a session then pass it around, everyone adds their
part.  Not as enjoyable as performing together with other people, but
probably results in a better quality performance overall.

--
Chris Caudle
_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user
Reply | Threaded
Open this post in threaded view
|

Re:  Re: Physical distance performance with Linux?

Christoph Kuhr
In reply to this post by Christoph Kuhr
Well, actually using the same ISP in the same city does not make a difference. My mate and me have both Unitymedia/Vodafone cable network access. We had the same RTT as with the other one at another ISP with ADSL.

It is worthwile noting that we are limited by the speed of light. If you do a signal propagation calculation just for fun, you will notice the boundaries of our reality. ;-)
For cologne <-> frankfurt it is little less then 2ms. frankfurt <-> nyc 60ms already. Not accounting for routing devices etc...
I read a paper a while ago which stated that most delay occurs in the access network, the last mile, due to shared media and signal multiplexing.


Best,
Ck
Am 05.05.20, 21:03 schrieb Chris Caudle <[hidden email]>:
On Tue, May 5, 2020 1:07 pm, Paul Davis wrote:
> The one thing I remember about soundjack from the presentation at
> tonmeistertage years ago was that inter-continental latencies are MUCH
> worse than in-country, often to the point of not being very usable. This
> isn't due to soundjack but the nature of routing across sub-ocean links
> etc. Just something to be aware of if you're trying to connect and work
> with people a *long* way away.

That should also apply in general to the number of routing devices the
link has to traverse, and general distance. So same city would be better,
same ISP in the same city wold probably be better still. Best would be
same building connecting through a single building Ethernet switch.

I watched that Doobie Brothers performance originally mentioned, and I
don't think that was live. That is one possibility if you don't mind,
have one person start a session then pass it around, everyone adds their
part. Not as enjoyable as performing together with other people, but
probably results in a better quality performance overall.

--
Chris Caudle
_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user

_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user
Reply | Threaded
Open this post in threaded view
|

Re: Physical distance performance with Linux?

Giso Grimm
In reply to this post by Chris Caudle
On 05.05.20 21:02, Chris Caudle wrote:

> On Tue, May 5, 2020 1:07 pm, Paul Davis wrote:
>> The one thing I remember about soundjack from the presentation at
>> tonmeistertage years ago was that inter-continental latencies are MUCH
>> worse than in-country, often to the point of not being very usable. This
>> isn't due to soundjack but the nature of routing across sub-ocean links
>> etc. Just something to be aware of if you're trying to connect and work
>> with people a *long* way away.
>
> That should also apply in general to the number of routing devices the
> link has to traverse, and general distance.  So same city would be better,
> same ISP  in the same city wold probably be better still.  Best would be
> same building connecting through a single building Ethernet switch.

in our setup we are five participants, all in North Germany (Hamburg,
Bremen, Oldenburg and two villages not very far away). Interestingly the
peer-to-peer ping times are not much shorter than the added ping times
from the peers to the server (in Frankfurt) and back to the other peers.

>
> I watched that Doobie Brothers performance originally mentioned, and I
> don't think that was live.  That is one possibility if you don't mind,
> have one person start a session then pass it around, everyone adds their
> part.  Not as enjoyable as performing together with other people, but
> probably results in a better quality performance overall.
>
_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user
Reply | Threaded
Open this post in threaded view
|

Re: Physical distance performance with Linux?

Chris Caudle
In reply to this post by Christoph Kuhr
On Tue, May 5, 2020 2:43 pm, Christoph Kuhr wrote:
> Well, actually using the same ISP in the same city does not
> make a difference. My mate and me have both
> Unitymedia/Vodafone cable network access. We had the same RTT
> as with the other one at another ISP with ADSL.

That is a really interesting observation.  Not what I expected, so I am
curious why there is not more added latency from the routing out of your
ISP and  into the other ISP.
Perhaps the latency on the modems is high enough that the customer end
equipment dominates over the routing delays.

> It is worthwile noting that we are limited by the speed of light.

300km/ms speed of light, the lowest ping time I have ever seen from my
computer to a machine outside my house is about 20ms.  I know not all of
those machines were 6000km away, so I think speed of light is pretty low
on the list of factors influencing delay.  Yes for large distances, but
should not have been a factor in your connection inside the same city
through the same ISP.


> I read a paper a while ago which stated that most delay occurs in the
> access network, the last mile, due to shared media and signal
> multiplexing.

Also both DSL  and cable modem involve lots of pretty sophisticated DSP
operations, that all takes time.  I have not seen any information on time
delay between ethernet packet in and cable or DSL packet out, or the
reverse.  That may be a significant portion of the delay.

And for anyone not familiar with the term "bufferbloat" or "buffer bloat"
you should verify that you are using appropriate active queue management
at your home router.  There have been many papers, blogs, presentations,
etc. on how poorly most equipment manages network queues at the points
where network bandwidth changes (such as from your home 1Gb connection to
the usually much slower upstream connection).  If the queues are too deep
then many packets can be buffered in the network equipment waiting for
transmission, especially if you have multiple applications active and
sharing network bandwidth simultaneously.

--
Chris Caudle
_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user
Reply | Threaded
Open this post in threaded view
|

Re: Physical distance performance with Linux?

Chris Caudle
In reply to this post by Christoph Kuhr
On Tue, May 5, 2020 2:43 pm, Christoph Kuhr wrote:
> Well, actually using the same ISP in the same city does not
> make a difference. My mate and me have both
> Unitymedia/Vodafone cable network access. We had the same RTT
> as with the other one at another ISP with ADSL.

I have been looking for information on the latency associated with the
cable modem.  So far this is the only information I found, but it is not
clear whether this refers just to the latency on the cable side, or also
includes the latency of converting from the Ethernet side to the cable
side:

"The average DOCSIS upstream latency for best effort traffic has been
measured to be 11-15 ms with the potential of a significantly higher
maximum latency up to 50 ms under medium to heavy channel utilization."

So already a significant latency, and potentially high latency variation,
before even getting to the first router.

It is impressive that you have found a way to play live together with that
connection.

--
Chris Caudle
_______________________________________________
Linux-audio-user mailing list
[hidden email]
https://lists.linuxaudio.org/listinfo/linux-audio-user