Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1400733 > unrolled thread

Re: [PATCH v3 2/2] ASoC: codecs: add support for TAS5720 digital amplifier

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH v3 2/2] ASoC: codecs: add support for TAS5720 digital  amplifier Mark Brown <broonie@kernel.org> - 2016-05-13 14:00 +0200
    Re: [PATCH v3 2/2] ASoC: codecs: add support for TAS5720 digital  amplifier Andreas Dannenberg <dannenberg@ti.com> - 2016-05-13 16:20 +0200

#1400733 — Re: [PATCH v3 2/2] ASoC: codecs: add support for TAS5720 digital amplifier

[Multipart message — attachments visible in raw view] — view raw

On Tue, Apr 26, 2016 at 01:01:05PM -0500, Andreas Dannenberg wrote:
> On Tue, Apr 26, 2016 at 06:29:36PM +0100, Mark Brown wrote:

> > Is the device actually going to mess up if someone sends it something
> > else or is it just going to ignore the extra bits (given that it's doing
> > autodetection anyway).

> well in any of the left-justified modes (which are the only ones the
> driver supports) the device takes and processes as many bits as it can
> given the clock and divider settings. Any extra bits provided will get
> ignored, and the next sync happens on the frame sync signal and not by
> counting bits so there is no downside also as confirmed by some bench
> testing I did feeding in 32-bit long frames for one channel. This seems
> like a case of preferring tolerance over strictly enforcing
> datasheet-advertised bit-widths. Will take out the check code.

OK, accepting extra bits is fine.  You should set sig_bits in the DAI so
userspace can see what's going on if it cares.

> Along these lines, earlier as I was rummaging through the existing
> drivers looking for a solution I could model after I noticed that
> most(?) ASoC codec drivers don't have any type of HW fault checking, at
> at least whatever drivers I looked at. Not sure why this is but given
> this discussion this seems like a general opportunity to make
> improvements.

There are some with over temperature handling (eg, wm8962) but it's
relatively uncommon for observable protection features to be implemented
in silicon and even rarer for the interrupts to be hooked up (and hence
useful to support in software) unless there is also accessory detection
in the device.  On older devices the required digital logic was often
excessively expensive and realistically only relatively high power
speaker drivers have much risk of something going wrong - things like
headphone outputs or smaller speaker drivers end up with protection from
their supplies collapsing well before the device is in physical danger.

[toc] | [next] | [standalone]

#1400823

Mark, please see below...

On Fri, May 13, 2016 at 12:55:40PM +0100, Mark Brown wrote:
> On Tue, Apr 26, 2016 at 01:01:05PM -0500, Andreas Dannenberg wrote:
> > On Tue, Apr 26, 2016 at 06:29:36PM +0100, Mark Brown wrote:
> 
> > > Is the device actually going to mess up if someone sends it something
> > > else or is it just going to ignore the extra bits (given that it's doing
> > > autodetection anyway).
> 
> > well in any of the left-justified modes (which are the only ones the
> > driver supports) the device takes and processes as many bits as it can
> > given the clock and divider settings. Any extra bits provided will get
> > ignored, and the next sync happens on the frame sync signal and not by
> > counting bits so there is no downside also as confirmed by some bench
> > testing I did feeding in 32-bit long frames for one channel. This seems
> > like a case of preferring tolerance over strictly enforcing
> > datasheet-advertised bit-widths. Will take out the check code.
> 
> OK, accepting extra bits is fine.  You should set sig_bits in the DAI so
> userspace can see what's going on if it cares.

I just had a quick look to see what sig_bits does, yes this is a good
addition. Will post a follow-up patch based on the driver that you
already accepted into your tree.

(Btw I'm working on another codec driver and I don't stop getting
positively surprised how flexible and powerful the overall ASoC
framework is).

Regards,
Andreas

> > Along these lines, earlier as I was rummaging through the existing
> > drivers looking for a solution I could model after I noticed that
> > most(?) ASoC codec drivers don't have any type of HW fault checking, at
> > at least whatever drivers I looked at. Not sure why this is but given
> > this discussion this seems like a general opportunity to make
> > improvements.
> 
> There are some with over temperature handling (eg, wm8962) but it's
> relatively uncommon for observable protection features to be implemented
> in silicon and even rarer for the interrupts to be hooked up (and hence
> useful to support in software) unless there is also accessory detection
> in the device.  On older devices the required digital logic was often
> excessively expensive and realistically only relatively high power
> speaker drivers have much risk of something going wrong - things like
> headphone outputs or smaller speaker drivers end up with protection from
> their supplies collapsing well before the device is in physical danger.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web