Re: ACP7.0 SoundWire: two TAS2783 amps, only one channel reproduced?

On 8/12/26 04:31, Kel Modderman wrote:
> [You don't often get email from [email protected]. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
>
> Hi,
>
> I have an HP OmniBook X Flip 14-kc0xxx (board 8EA1) with AMD ACP 7.0
> (1022:15e2 rev 0x70, subsystem 103c:8ea1), and I am trying to get its
> speakers working with assistance from claude. I have got most of the way
> there but I am stuck on one point, and I would be grateful for your advice
> on whether the layout I am using is the right one.
>
> The machine has:
>
>    link 1, uid 0 - Realtek RT712 SDCA (headset jack and microphones)
>    link 0, uid 9 - TI TAS2783
>    link 0, uid C - TI TAS2783
>
> Two local changes were needed to get a card at all:
>
>    1. The BIOS reports acp-audio-config-flag = 16
>       (FLAG_AMD_LEGACY_ONLY_DMIC), so snd_amd_acp_find_config() returns a
>       value that neither acp70_pci_probe() nor pci-ps.c accepts, and the ACP
>       ends up with no driver bound and nothing in dmesg. Adding the board to
>       acp70_acpi_flag_override_table solves that.
>
>    2. There is no entry in snd_soc_acpi_amd_acp70_sdw_machines[] matching
>       this topology, so I added one.
>
> With those two changes the card is built correctly. The headset jack, the
> microphones and the UCM profile all work, and both amplifiers load their
> firmware from ti/audio/tas2783/8EA1-0-0-0x{9,C}.bin without errors.
>
> My problem is that only one channel of the stream is ever reproduced. I
> measured this with the machine's own microphones rather than by ear, so
> the numbers are repeatable:
>
>    amps on group_position 0 and 1:
>      only the position-0 amp plays, and only stream channel 0
>
>    the same, but swapped over:
>      only the (new) position-0 amp plays, and only stream channel 0
>
>    both amps on group_position 0:
>      both amps play, but both play stream channel 0
>
> So both amplifiers are healthy - either one on its own reproduces the
> whole signal when it sits on position 0 - but stream channel 1 is never
> reproduced, whatever layout I use. Nothing changes if I use the other
> firmware variant (8EA1-1-0-0x{9,C}.bin).
>
> I modelled the two-position layout on acp70_rt721_l1u0_tas2783x2_l1u8b,
> and the shared-position layout on tas2783_2_adr, which pairs two devices
> per position with /* left */ and /* right */ comments.
>
> So my question is really a simple one: is a two-TAS2783 configuration on
> ACP 7.0 expected to work, and if so, what should the machine entry look
> like? I am probably missing something obvious.
>
> I am happy to test patches or collect any further information you would
> find useful - I have the hardware in front of me and can rebuild modules
> quickly. I can also send the two patches I am carrying if that is helpful.
>
> This is on 7.1, and I checked the relevant files against the current master
> and the sound tree's for-next as well.
For better tracking, could you please create a Bugzilla ticket?
>
> Thanks very much for your time,
>
> Thanks, Kel