Loopback latency issue

Loopback latency issue

[Maxim Doumkine]

[-- Attachment #1.1: Type: text/plain, Size: 681 bytes --]

Hello everyone,

We are trying to build a Sox based real-time sound altering app, and we’ve got everything working on our LUbuntu 16.04 stations except for an issue w latency.

When we test it on actual hardware, we are having a latency of about 2 seconds, when we test it on our virtual machine the latency is more like 250ms.

sox -t pusleaudio default -t pulseaudio null pitch n
Whenever we add the null loopback or aloop, we experience a delay. No amount of config tinkering seems to affect this latency on an actual workstation.

We’d appreciate an advice or any input but would be equally glad to hire a consultant for a fee to get this resolved.

Thank you.

[-- Attachment #1.2: Type: text/html, Size: 1023 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]



[-- Attachment #3: Type: text/plain, Size: 158 bytes --]

_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Darryl Lewis]

[-- Attachment #1.1: Type: text/plain, Size: 1099 bytes --]


Is your virtual machine exactly the same (OS version) as your physical PC?
It might be a driver issue with different versions.


From: Maxim Doumkine <maxd@coregears.com>
Reply to: "sox-users@lists.sourceforge.net" <sox-users@lists.sourceforge.net>
Date: Saturday, 11 July 2020 at 3:58 am
To: "sox-users@lists.sourceforge.net" <sox-users@lists.sourceforge.net>
Subject: [SoX-users] Loopback latency issue

Hello everyone,

We are trying to build a Sox based real-time sound altering app, and we’ve got everything working on our LUbuntu 16.04 stations except for an issue w latency.

When we test it on actual hardware, we are having a latency of about 2 seconds, when we test it on our virtual machine the latency is more like 250ms.

sox -t pusleaudio default -t pulseaudio null pitch n
Whenever we add the null loopback or aloop, we experience a delay. No amount of config tinkering seems to affect this latency on an actual workstation.

We’d appreciate an advice or any input but would be equally glad to hire a consultant for a fee to get this resolved.

Thank you.

[-- Attachment #1.2: Type: text/html, Size: 3539 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]



[-- Attachment #3: Type: text/plain, Size: 158 bytes --]

_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Måns Rullgård]

Maxim Doumkine <maxd@coregears.com> writes:

> Hello everyone,
>
> We are trying to build a Sox based real-time sound altering app, and we’ve
> got everything working on our LUbuntu 16.04 stations except for an issue w
> latency.
>
> When we test it on actual hardware, we are having a latency of about 2
> seconds, when we test it on our virtual machine the latency is more like
> 250ms.
>
> sox -t pusleaudio default -t pulseaudio null pitch n
> Whenever we add the null loopback or aloop, we experience a delay. No amount
> of config tinkering seems to affect this latency on an actual workstation.

The first thing I'd do is get rid of pulseaudio and access the ALSA
device directly.  Even then, however, there are several places where
latency is added.  You can find the ALSA buffer sizes currently uses by
examining /proc/asound/card0/pcm0p/sub0/hw_params (or whatever numbers
your system uses).  The size (in bytes) of the main buffer in SoX is set
with the --buffer option.  Some effects have additional internal buffers.

-- 
Måns Rullgård


_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Maxim Doumkine]

[-- Attachment #1.1: Type: text/plain, Size: 648 bytes --]

We’ve tried that with no effect. :-/ it’s curious how it doesn’t work on actual workstation but works fine on virtualized pc.
On 11 Jul 2020, 16:12 +0300, Måns Rullgård <mans@mansr.com>, wrote:
>
> The first thing I'd do is get rid of pulseaudio and access the ALSA
> device directly. Even then, however, there are several places where
> latency is added. You can find the ALSA buffer sizes currently uses by
> examining /proc/asound/card0/pcm0p/sub0/hw_params (or whatever numbers
> your system uses). The size (in bytes) of the main buffer in SoX is set
> with the --buffer option. Some effects have additional internal buffers.

[-- Attachment #1.2: Type: text/html, Size: 1067 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]



[-- Attachment #3: Type: text/plain, Size: 158 bytes --]

_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Maxim Doumkine]

[-- Attachment #1.1: Type: text/plain, Size: 1411 bytes --]

Yes it is an exact clone.
On 10 Jul 2020, 21:09 +0300, Darryl Lewis <darryl.lewis@unsw.edu.au>, wrote:
>
> Is your virtual machine exactly the same (OS version) as your physical PC?
> It might be a driver issue with different versions.
>
>
> From: Maxim Doumkine <maxd@coregears.com>
> Reply to: "sox-users@lists.sourceforge.net" <sox-users@lists.sourceforge.net>
> Date: Saturday, 11 July 2020 at 3:58 am
> To: "sox-users@lists.sourceforge.net" <sox-users@lists.sourceforge.net>
> Subject: [SoX-users] Loopback latency issue
>
> Hello everyone,
>
> We are trying to build a Sox based real-time sound altering app, and we’ve got everything working on our LUbuntu 16.04 stations except for an issue w latency.
>
> When we test it on actual hardware, we are having a latency of about 2 seconds, when we test it on our virtual machine the latency is more like 250ms.
>
> sox -t pusleaudio default -t pulseaudio null pitch n
> Whenever we add the null loopback or aloop, we experience a delay. No amount of config tinkering seems to affect this latency on an actual workstation.
>
> We’d appreciate an advice or any input but would be equally glad to hire a consultant for a fee to get this resolved.
>
> Thank you.
> _______________________________________________
> Sox-users mailing list
> Sox-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/sox-users

[-- Attachment #1.2: Type: text/html, Size: 3008 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]



[-- Attachment #3: Type: text/plain, Size: 158 bytes --]

_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Steve Masters]

[-- Attachment #1.1: Type: text/plain, Size: 1679 bytes --]

I wonder if it's because the virtual pc doesn't have the same amount of
processes running on it, nor physical data traffic, that your actual
workstation has to deal with. Added to that, the memory available to it -
not being used elsewhere - would improve perfoemance. Just my thinking.

On Sat, 11 Jul 2020, 14:31 Måns Rullgård, <mans@mansr.com> wrote:

> Maxim Doumkine <maxd@coregears.com> writes:
>
> > Hello everyone,
> >
> > We are trying to build a Sox based real-time sound altering app, and
> we’ve
> > got everything working on our LUbuntu 16.04 stations except for an issue
> w
> > latency.
> >
> > When we test it on actual hardware, we are having a latency of about 2
> > seconds, when we test it on our virtual machine the latency is more like
> > 250ms.
> >
> > sox -t pusleaudio default -t pulseaudio null pitch n
> > Whenever we add the null loopback or aloop, we experience a delay. No
> amount
> > of config tinkering seems to affect this latency on an actual
> workstation.
>
> The first thing I'd do is get rid of pulseaudio and access the ALSA
> device directly.  Even then, however, there are several places where
> latency is added.  You can find the ALSA buffer sizes currently uses by
> examining /proc/asound/card0/pcm0p/sub0/hw_params (or whatever numbers
> your system uses).  The size (in bytes) of the main buffer in SoX is set
> with the --buffer option.  Some effects have additional internal buffers.
>
> --
> Måns Rullgård
>
>
> _______________________________________________
> Sox-users mailing list
> Sox-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/sox-users
>

[-- Attachment #1.2: Type: text/html, Size: 2316 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]



[-- Attachment #3: Type: text/plain, Size: 158 bytes --]

_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Jan Stary]

On Jul 10 20:40:09, maxd@coregears.com wrote:
> We are trying to build a Sox based real-time sound altering app, and we’ve got everything working on our LUbuntu 16.04 stations except for an issue w latency.
> 
> When we test it on actual hardware, we are having a latency of about 2 seconds, when we test it on our virtual machine the latency is more like 250ms.
> 
> sox -t pusleaudio default -t pulseaudio null pitch n

I don't use pulseaudio, but I assume that "default"
and "null" are names of pulseaudio (pseudo)devices.

How can you tell the delay when writing to a null device?
(If that means what I think it means: dropping the pulseaudio,
would it be what sox -d -n does?) I just tried

	sox -d -d pitch 1000

an OpenBSD current/amd64, and there is indeed
about a half second of delay.

> Whenever we add the null loopback or aloop, we experience a delay.

I don't understand: what does your application do
that you need to "add a null loopback"?


On Jul 11 14:12:13, mans@mansr.com wrote:
> Some effects have additional internal buffers.

It seems there is a delay even when just playing
default input to default output; even without the effects:

	sox -d -d

(So I don't think it's the effect processing
that introduces the delay.)

Jan



_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Jan Stary]

On Jul 13 13:49:27, hans@stare.cz wrote:
> I just tried
> 
> 	sox -d -d pitch 1000
> 
> on OpenBSD current/amd64, and there is indeed
> about a half second of delay.

For comparison,

	 aucat -c 0:0 -o - | aucat -c 0:0 -i -

has about the same delay.



_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Maxim Doumkine]

[-- Attachment #1.1: Type: text/plain, Size: 1782 bytes --]

	It’s definitely not the effect processing…  we are using this to modify voice in realtime, make someone sound deeper/older for instance. Without any effects there’s still a delay.
On Jul 13, 2020, 4:16 PM +0400, Jan Stary <hans@stare.cz>, wrote:
> On Jul 10 20:40:09, maxd@coregears.com wrote:
> > We are trying to build a Sox based real-time sound altering app, and we’ve got everything working on our LUbuntu 16.04 stations except for an issue w latency.
> >
> > When we test it on actual hardware, we are having a latency of about 2 seconds, when we test it on our virtual machine the latency is more like 250ms.
> >
> > sox -t pusleaudio default -t pulseaudio null pitch n
>
> I don't use pulseaudio, but I assume that "default"
> and "null" are names of pulseaudio (pseudo)devices.
>
> How can you tell the delay when writing to a null device?
> (If that means what I think it means: dropping the pulseaudio,
> would it be what sox -d -n does?) I just tried
>
> sox -d -d pitch 1000
>
> an OpenBSD current/amd64, and there is indeed
> about a half second of delay.
>
> > Whenever we add the null loopback or aloop, we experience a delay.
>
> I don't understand: what does your application do
> that you need to "add a null loopback"?
>
>
> On Jul 11 14:12:13, mans@mansr.com wrote:
> > Some effects have additional internal buffers.
>
> It seems there is a delay even when just playing
> default input to default output; even without the effects:
>
> sox -d -d
>
> (So I don't think it's the effect processing
> that introduces the delay.)
>
> Jan
>
>
>
> _______________________________________________
> Sox-users mailing list
> Sox-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/sox-users

[-- Attachment #1.2: Type: text/html, Size: 2586 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]



[-- Attachment #3: Type: text/plain, Size: 158 bytes --]

_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Måns Rullgård]

Jan Stary <hans@stare.cz> writes:

> On Jul 11 14:12:13, mans@mansr.com wrote:
>> Some effects have additional internal buffers.
>
> It seems there is a delay even when just playing
> default input to default output; even without the effects:
>
> 	sox -d -d
>
> (So I don't think it's the effect processing that introduces the
> delay.)

No, then it's just the main buffer that is set with the --buffer option.

-- 
Måns Rullgård


_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users

Re: Loopback latency issue

[Rafal Maszkowski]

On Mon, Jul 13, 2020 at 09:52:00PM +0100, Måns Rullgård wrote:
> Jan Stary <hans@stare.cz> writes:
> > On Jul 11 14:12:13, mans@mansr.com wrote:
> >> Some effects have additional internal buffers.
> > It seems there is a delay even when just playing
> > default input to default output; even without the effects:
> > 	sox -d -d
> > (So I don't think it's the effect processing that introduces the
> > delay.)
> No, then it's just the main buffer that is set with the --buffer option.

There are some buffers in the kernel and large default buffers in
pulseaudio. I have used only small part of the advice in
https://wiki.archlinux.org/index.php/PulseAudio/Troubleshooting
and that decreased the delay from some 1/2 s to few tens of ms if I can
judge by ear. There is more to read there though.

    Modify PulseAudio's configuration file (2/4)

    /etc/pulse/daemon.conf

    default-fragments = 3
    default-fragment-size-msec = 5

    Restart the PulseAudio daemon (3/4)


R.
-- 
„Walczy on z całym zapamiętaniem przeciwko intelektowi” - z akt personalnych prof. A. Baeumlera


_______________________________________________
Sox-users mailing list
Sox-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-users