Strange SCO problem

Strange SCO problem

[Yaroslav Isakov]

Hello! I'm currently trying to modify pulseaudio to be able to use HFP
profile with mSBC codec on native backend, but found that, while
working de-jure, real sound is completely ruined, so I can barely hear
what is sent over SCO link. I've tried two different Intel adapters
(included in 8260 and 8265 chipset) and KillerWifi (ath10k driver), on
two different PCs with two different headsets, results are mostly the
same, the only difference is that on ath10k words are more or less
distinguishable.

I've even tested with pulse just for profile setup, and SCO is sent
via modified scotest from bluez, which sends some test sound, manually
encoded via sbcenc (with added h2, of course). I've tried also msbc
branch from bluez-alsa, same result.

I'm sure that every time socket is connected with setsockopt(fd,
SOL_BLUETOOTH, BT_VOICE, &voice, sizeof(voice));

I've also tried with this patch
https://www.spinics.net/lists/linux-bluetooth/msg76982.html, with no
luck.

Any ideas?

Re: Strange SCO problem

[Sathish Narasimman]

Hi,

Intel adapter 9260 and 9265(with the latest Firmware - from
Bluetooth-next) supports USB ALT 6 settings and was verified.

Along with backend ofono
https://patchwork.freedesktop.org/patch/245272/

https://www.spinics.net/lists/linux-bluetooth/msg76982.html

Regards,
Sathish N


On Wed, Mar 6, 2019 at 5:15 AM Yaroslav Isakov
<yaroslav.isakov@gmail.com> wrote:
>
> Hello! I'm currently trying to modify pulseaudio to be able to use HFP
> profile with mSBC codec on native backend, but found that, while
> working de-jure, real sound is completely ruined, so I can barely hear
> what is sent over SCO link. I've tried two different Intel adapters
> (included in 8260 and 8265 chipset) and KillerWifi (ath10k driver), on
> two different PCs with two different headsets, results are mostly the
> same, the only difference is that on ath10k words are more or less
> distinguishable.
>
> I've even tested with pulse just for profile setup, and SCO is sent
> via modified scotest from bluez, which sends some test sound, manually
> encoded via sbcenc (with added h2, of course). I've tried also msbc
> branch from bluez-alsa, same result.
>
> I'm sure that every time socket is connected with setsockopt(fd,
> SOL_BLUETOOTH, BT_VOICE, &voice, sizeof(voice));
>
> I've also tried with this patch
> https://www.spinics.net/lists/linux-bluetooth/msg76982.html, with no
> luck.
>
> Any ideas?

Re: Strange SCO problem

[Yaroslav Isakov]

ср, 6 мар. 2019 г. в 09:28, Sathish Narasimman <nsathish41@gmail.com>:
>
> Hi,
>
Hello, Sathish
> Intel adapter 9260 and 9265(with the latest Firmware - from
> Bluetooth-next) supports USB ALT 6 settings and was verified.
>
I have only 8260 and 8265. How I can check if they supports ALT 6
setting? I've checked via lsusb
lsusb -d 8087:0a2b -v | grep bAlt
      bAlternateSetting       0
      bAlternateSetting       0
      bAlternateSetting       1
      bAlternateSetting       2
      bAlternateSetting       3
      bAlternateSetting       4
      bAlternateSetting       5
can't get device qualifier: Resource temporarily unavailable
can't get debug descriptor: Resource temporarily unavailable
> Along with backend ofono
> https://patchwork.freedesktop.org/patch/245272/
>
I've based my patches for native backend on your work. Unfortunately,
I cannot use ofono, as it needs phonesim which is not working on my
system (and is abandoned long ago...)

> https://www.spinics.net/lists/linux-bluetooth/msg76982.html
>
> Regards,
> Sathish N
>
>
> On Wed, Mar 6, 2019 at 5:15 AM Yaroslav Isakov
> <yaroslav.isakov@gmail.com> wrote:
> >
> > Hello! I'm currently trying to modify pulseaudio to be able to use HFP
> > profile with mSBC codec on native backend, but found that, while
> > working de-jure, real sound is completely ruined, so I can barely hear
> > what is sent over SCO link. I've tried two different Intel adapters
> > (included in 8260 and 8265 chipset) and KillerWifi (ath10k driver), on
> > two different PCs with two different headsets, results are mostly the
> > same, the only difference is that on ath10k words are more or less
> > distinguishable.
> >
> > I've even tested with pulse just for profile setup, and SCO is sent
> > via modified scotest from bluez, which sends some test sound, manually
> > encoded via sbcenc (with added h2, of course). I've tried also msbc
> > branch from bluez-alsa, same result.
> >
> > I'm sure that every time socket is connected with setsockopt(fd,
> > SOL_BLUETOOTH, BT_VOICE, &voice, sizeof(voice));
> >
> > I've also tried with this patch
> > https://www.spinics.net/lists/linux-bluetooth/msg76982.html, with no
> > luck.
> >
> > Any ideas?