[alsa-devel] [PATCH] New ASoC Drivers for ADI AD1938 codec

[宋宝华]

2009/6/22 Mark Brown <broonie at opensource.wolfsonmicro.com>

> On Mon, Jun 22, 2009 at 11:08:27AM +0800, 宋宝华 wrote:
>
> [Please reply in-line, interspersing your new text into the message
> you're replying to - it makes the discussion much easier to follow.]
>
> > Hi Mark,
> > ***For the new DAI format
> > According to I2S spec, it doesn't definite a I2S with TDM as a standard
> I2S.
> > http://www.nxp.com/acrobat_download/various/I2SBUS.pdf
>
> It's a de facto standard - several other vendors implement TDM in
> exactly the same fashion.  If you think about it this is a natural way
> to handle TDM in I2S.
>
> > It looks like you are admitting this kind of timing into I2S DAI too:
> > http://i3.6.cn/cvbnm/8f/3d/08/268a4560e0daa1b41d69b82419da06e1.jpg
> > I think I can follow it too.
>
> The timing you're showing there is essentially a DSP mode with the left
> and right channels alternating rather than an I2S style where the
> polarity of the LRCLK signal indicates if the data transmitted at the
> same time is for the left or right channel.  I'd need to think about it
> in slightly more detail but probably it is actually a DSP mode - with
> the DSP modes only one edge of the frame sync is used so the other edge
> can be anywhere else within the frame.
>
> > Due to my test boards, at present, the AD1938 is working in and
> supporting
> > TDM timing like the diagram:
> > http://i3.6.cn/cvbnm/2f/e2/f2/03ae2b51c4e90749972e70bf887f926f.jpg
> > It looks like DSP mode with TDM, so can I path related codes into
> > SND_SOC_DAIFMT_DSP switch?
>
> Yes, that's DSP mode.


Based on the "de facto ", I think I will merge TDM related codes into
blackfin I2S DAI directly.

>
>
> > ***For volume controls based on stereo pairs
> > Even though DAC1-DAC8 are named as DACL1,DACR1, DACL2,DACR2..., but the
> > DACLx and DACRx are not always in a pair, in fact, they are independent.
> As
> > a codec supporting 8 channels, it can be configed into 2, 2.1, 4.1, 5.1,
> > 6.1, 7.1, how to handle the pairs?
>
> That's fairly standard and not really a practical problem.  Since
> applications can address each channel of a stereo control independantly
> they don't loose any control from grouping the channels together.
> Having the stereo controls just makes it a bit easier when they are used
> that way.
>

So, does it mean I need to support both single and pair since we don't know
how users will use it? If I only create 4 pairs, how could users control
channels not in any pair?