[alsa-devel] [PATCH] ASoC: ssm2602: Fix ADC powerup sequencing

[Marco Felsch]

Hi Rob,

On 18-05-23 11:53, Rob Herring wrote:
> On Fri, May 18, 2018 at 12:46:49PM +0200, Philipp Zabel wrote:
> > Hi Rob,
> > 
> > On Thu, 2018-05-17 at 11:14 -0500, Rob Herring wrote:
> > > On Thu, May 17, 2018 at 8:30 AM, Marco Felsch <m.felsch at pengutronix.de> wrote:
> > > > From: Philipp Zabel <p.zabel at pengutronix.de>
> > > > 
> > > > According to the ssm2603 data sheet (control register sequencing), the
> > > > digital core should be activated only after all necessary bits in the
> > > > power register are enabled, and a delay determined by the decoupling
> > > > capacitor on the VMID pin has passed. If the digital core is activated
> > > > too early, or even before the ADC is powered up, audible artifacts
> > > > appear at the beginning of the recorded signal.
> > > > 
> > > > The digital core is also needed for playback, so when recording starts
> > > > it may already be enabled. This means we cannot get the power sequence
> > > > correct when we want to be able to start recording after playback.
> > > > 
> > > > As a workaround put the MIC mute switch into the DAPM routes. This
> > > > way we can keep the recording disabled until the MIC Bias has settled
> > > > and thus get rid of audible artifacts.
> > > > 
> > > > The delay we have to wait depends on a board specific capacitor
> > > > connected to the VMID pins, so make the delay configurable in the device
> > > > tree.
> > > > 
> > > > Signed-off-by: Philipp Zabel <p.zabel at pengutronix.de>
> > > > Signed-off-by: Sascha Hauer <s.hauer at pengutronix.de>
> > > > Signed-off-by: Marco Felsch <m.felsch at pengutronix.de>
> > > > ---
> > > >  .../devicetree/bindings/sound/adi,ssm2602.txt |  7 +++++
> > > >  sound/soc/codecs/ssm2602.c                    | 30 +++++++++++++++++--
> > > >  2 files changed, 35 insertions(+), 2 deletions(-)
> > > > 
> > > > diff --git a/Documentation/devicetree/bindings/sound/adi,ssm2602.txt b/Documentation/devicetree/bindings/sound/adi,ssm2602.txt
> > > > index 3b3302fe399b..9334d48c0462 100644
> > > > --- a/Documentation/devicetree/bindings/sound/adi,ssm2602.txt
> > > > +++ b/Documentation/devicetree/bindings/sound/adi,ssm2602.txt
> > > > @@ -11,9 +11,16 @@ Required properties:
> > > >    - reg : the I2C address of the device for I2C, the chip select
> > > >            number for SPI.
> > > > 
> > > > +Optional properties:
> > > > +
> > > > +  - startup-delay-us : delay between powering on and activating the digital
> > > > +                       core, determined by the decoupling capacitor C on the
> > > > +                      VMID pin: delay [µs] = C [µF] * 25000 / 3.5
> > > > +
> > > 
> > > We already have similarly defined property. Please reuse that. See mmc
> > > pwrseq binding.
> > 
> > Do you mean 'post-power-on-delay-ms' from 'mmc-pwseq-simple'?
> > It is documented as:
> > 
> > - post-power-on-delay-ms : Delay in ms after powering the card and
> >         de-asserting the reset-gpios (if any)
> > 
> > The startup delay here is not after powering the whole IC or deasserting
> > resets and before it can be used, but after powering up a specific part
> > of the codec (the ADC) and before unmuting the MIC input to the digital
> > core during start of decoding. With this in mind, do you still think the
> > property should be called the same as the mmc full-chip poweron delay?
> > If so, would it be acceptable to use post-power-on-delay-us to keep the
> > microsecond resolution?
> 
> Okay, then I'd suggest something a bit more specific. Perhaps 
> "pre-unmute-delay-us" and document in a common location.
> 
> Rob

The delay is not just for the line-in/mic path it is also for the out
path. More technical, it is needed to charge the decouple capacity which
provides the voltage bias for the analog input/output frontends.

I found the: "ti,charge-period (sound/ti,tas5086.txt)" binding which
represents nearly the same but it is not common. One opportunity would be to
introduce a common "charge-period-us" binding and change the ti binding the
common binding later.

Would that be okay?

Marco