[alsa-devel] [PATCH] ASoC: bcm2835: Increase channels_max to 8

Matt Flax flatmax at flatmax.org
Sun Feb 5 14:49:45 PST 2017


On 06/02/17 07:37, Matt Flax wrote:
> On 06/02/17 03:34, Mark Brown wrote:
>> On Sun, Feb 05, 2017 at 12:26:31PM +1100, Matt Flax wrote:
>>> On 05/02/17 02:49, Mark Brown wrote:
>>>> On Thu, Feb 02, 2017 at 10:37:44AM +1100, Matt Flax wrote:
>>>> I can't understand how this works.  We're programming the hardware 
>>>> in an
>>>> identical fashion for both channel counts, that means that what we're
>>> I am not surprised it is confusing, to my knowledge this is the 
>>> first time
>>> that someone has demonstrated this type of I2S signalling.
>> More than stereo I2S?  That is actually a thing, while most people use
>> DSP modes for higher channel counts it's fairly common to have the
>> support especially in CODECs that also support DSP mode since the
>> hardware implementation is fairly straightforward if you abstracted in
>> the bus interface.
>>
>>> In this case BCM2835 is a clock slave and not concerned with 
>>> generating the
>>> I2S clocks, only receiving - think of all clocks and synchrony being
>>> generated/managed in hardware on the sound card.
>>> In short the BCM2835 silicon is only concerned with shifting bits 
>>> into two
>>> hardware registers which (after DMA) are interpreted by ALSA 
>>> according to
>>> the specified channel count.
>> Sure, that's what all audio controllers do.  But there's a few problems.
>> One is that nothing in your code that prevents this running in master
>> mode.  Another is that if this *is* for I2S mode then when doing
>> multichannel I2S you usually still respect the left/right indication in
>> LRCLK so you will end up with the data in the in memory format:
>>
>>    L1 R1 L2 R2 L3 R3 L4 R4
>>
>> but multi channel I2S would usually be rendered with all the left
>> channels and all the right channels grouped together so you get:
>>
>>    L1 L2 L3 L4 R1 R2 R3 R4
>>
>> It'll also require exact clocking with no spare bits in order to sync
>> the right channels with the LRCLK switch but that's not so unusual.
>> Looking at the driver it may be possible for the hardware to do DSP
>> modes at which point this gets much easier but right now it only has I2S
>> support, I've got a feeling that the other end of the link may actually
>> be running in a DSP mode.
> You've got it !
>
> This can also work with the BCM2835 in master mode, no reason at this 
> point to limit it to slave mode only.
>
> I haven't looked into why, but I am getting the correct channel 
> mapping under the hood. To me that is a concern of the higher up 
> levels of the ALSA driver - it was a pleasant surprise when it worked 
> nicely, because to me it indicated a robustly coded system.
>
> My next patch is to allow DSP modes in this I2S driver. I wanted to 
> make sure that this got through first however.
>
> The majority of these other concerns you mention are controlled at the 
> machine driver level.
>
Mark,

On deeper thought around what you were saying about requiring master 
mode (for the sound card), I think you are right. Whilst it is possible 
for the BCM2835 to be master, it would pose significant problems for the 
sound card to regenerate the correct clocks.

Would you like me to resubmit the patch ensuring master mode for 
channels != 2 and also adding DSP mode capability ?

thanks
Matt
p.s. I didn't realise this was a thing already, are there any machine 
drivers for other SoCs which are currently using this paradigm of 
signalling ?




More information about the linux-rpi-kernel mailing list