[alsa-devel] What is exact definition of "normal/inverted" FSYNC signal polarity?
Hi
I am trying to configure TegraX1 SoC and Realtek5677 codec. These two drivers have different meaning of FSYNC polarity in TDM mode A format.
* TegraX1 driver thinks that "normal" polarity is when a frame starts with FSYNC falling edge. TegraX1 applies it to all formats (I2S, PCM/TDM mode A). And inverted FSYNC means that frame starts with rising edge of FSYNC.
* RT5677 thinks that "normal" for I2S is "frame starts with falling edge". And for TMD mode A "normal" means "frame starts with rising FSYNC edge".
Thus normal/inverted FSYNC in TDM mode A are opposite at these 2 chips. Which one is correct? What is exact definition of SND_SOC_DAIFMT_xB_yF constants in soc-dai.h?
I tried to google if I2S spec [1] defines FSYNC polarity but no luck.
It worth to clarify definition of BCLK polarity as well. I believe "normal" BCLK is "line sensing happens at rising edge", "inverted" BCLK - "line sensing happens at falling edge". Is it correct?
[1] https://web.archive.org/web/20060821114007/http://www.semiconductors.philips...
On Mon, Aug 31, 2015 at 09:40:04AM -0700, Anatol Pomozov wrote:
I am trying to configure TegraX1 SoC and Realtek5677 codec. These two drivers have different meaning of FSYNC polarity in TDM mode A format.
The simplest thing to do is look at the datasheets for Wolfson parts like the wm8731, they have some very clear diagrams of all the data modes we support.
Hi
Thanks.
On Mon, Aug 31, 2015 at 11:02 AM, Mark Brown broonie@kernel.org wrote:
The simplest thing to do is look at the datasheets for Wolfson parts like the wm8731, they have some very clear diagrams of all the data modes we support.
I looked at http://www.rockbox.org/wiki/pub/Main/DataSheets/WM8731_8731L.pdf spec and it is what I see
"normal" bitclock means data sensing happens at rising edge of BCLK. It is the same as TegraX1 and rt5677 examples above.
As of FSYNC - the wm8731 driver uses Reg0x7[4] to set FSYNC polarity. If FSYNC is inverted then bit set to 1. According to the spec:
DACLRC phase control (in left, right or I2S modes) 1 = Right Channel DAC data when DACLRC high 0 = Right Channel DAC data when DACLRC low (opposite phasing in I2S mode) or DSP mode A/B select (in DSP mode only) 1 = MSB is available on 2nd BCLK rising edge after DACLRC rising edge 0 = MSB is available on 1st BCLK rising edge after DACLRC rising edge
Thus "normal" FSYNC for this driver means that the frame (L channel) starts at rising edge of FSYNC (wm8731 datasheet calls it DACLRC). "inverted" means frame starts at the falling edge. It differs both from TegraX1 and rt5677 above.
For TDM mode A the chip seems always expect to see "normal" polarity FSYNC (frame starts at rising edge of FSYNC).
It looks like each driver uses its own definition of FSYNC polarity. To avoid compatibility problems between different codecs/socs there should be a clean definition of what is FSYNC polarity is. I would propose something simple like: **** "normal" FSYNC means frame starts at rising edge of FSYNC, and "inverted" frame starts at falling FSYNC edge **** i.e. the same as for 8731.
On 08/31/2015 09:05 PM, Anatol Pomozov wrote:
Hi
Thanks.
On Mon, Aug 31, 2015 at 11:02 AM, Mark Brown broonie@kernel.org wrote:
The simplest thing to do is look at the datasheets for Wolfson parts like the wm8731, they have some very clear diagrams of all the data modes we support.
I looked at http://www.rockbox.org/wiki/pub/Main/DataSheets/WM8731_8731L.pdf spec and it is what I see
"normal" bitclock means data sensing happens at rising edge of BCLK. It is the same as TegraX1 and rt5677 examples above.
As of FSYNC - the wm8731 driver uses Reg0x7[4] to set FSYNC polarity. If FSYNC is inverted then bit set to 1. According to the spec:
DACLRC phase control (in left, right or I2S modes) 1 = Right Channel DAC data when DACLRC high 0 = Right Channel DAC data when DACLRC low (opposite phasing in I2S mode) or DSP mode A/B select (in DSP mode only) 1 = MSB is available on 2nd BCLK rising edge after DACLRC rising edge 0 = MSB is available on 1st BCLK rising edge after DACLRC rising edge
Thus "normal" FSYNC for this driver means that the frame (L channel) starts at rising edge of FSYNC (wm8731 datasheet calls it DACLRC). "inverted" means frame starts at the falling edge. It differs both from TegraX1 and rt5677 above.
For TDM mode A the chip seems always expect to see "normal" polarity FSYNC (frame starts at rising edge of FSYNC).
It looks like each driver uses its own definition of FSYNC polarity. To avoid compatibility problems between different codecs/socs there should be a clean definition of what is FSYNC polarity is. I would propose something simple like: **** "normal" FSYNC means frame starts at rising edge of FSYNC, and "inverted" frame starts at falling FSYNC edge **** i.e. the same as for 8731.
The modes supported by the wm8731 is DSP A (framesync precedes first data by 1 bit clock cycle) and DSP B(framesync happens during the first bit clock cycle).
But otherwise agreed rising edge = normal framesync, falling edge = inverted framesync
On Mon, Aug 31, 2015 at 12:05:51PM -0700, Anatol Pomozov wrote:
Thus "normal" FSYNC for this driver means that the frame (L channel) starts at rising edge of FSYNC (wm8731 datasheet calls it DACLRC). "inverted" means frame starts at the falling edge. It differs both from TegraX1 and rt5677 above.
It looks like each driver uses its own definition of FSYNC polarity. To avoid compatibility problems between different codecs/socs there should be a clean definition of what is FSYNC polarity is. I would propose something simple like: **** "normal" FSYNC means frame starts at rising edge of FSYNC, and "inverted" frame starts at falling FSYNC edge **** i.e. the same as for 8731.
Right, the Tegra and Realtek drivers are buggy here - the wm8731 definition is the one that essentially everything is using.
Hi Mark
On Tue, Sep 1, 2015 at 5:14 AM, Mark Brown broonie@kernel.org wrote:
On Mon, Aug 31, 2015 at 12:05:51PM -0700, Anatol Pomozov wrote:
Thus "normal" FSYNC for this driver means that the frame (L channel) starts at rising edge of FSYNC (wm8731 datasheet calls it DACLRC). "inverted" means frame starts at the falling edge. It differs both from TegraX1 and rt5677 above.
It looks like each driver uses its own definition of FSYNC polarity. To avoid compatibility problems between different codecs/socs there should be a clean definition of what is FSYNC polarity is. I would propose something simple like: **** "normal" FSYNC means frame starts at rising edge of FSYNC, and "inverted" frame starts at falling FSYNC edge **** i.e. the same as for 8731.
Right, the Tegra and Realtek drivers are buggy here - the wm8731 definition is the one that essentially everything is using.
Are you OK to accept the patch http://mailman.alsa-project.org/pipermail/alsa-devel/2015-August/097126.html ?
If yes then I will work with NVidia/Realtek on updating their drivers.
On Fri, Sep 11, 2015 at 10:48:05PM -0700, Anatol Pomozov wrote:
Are you OK to accept the patch http://mailman.alsa-project.org/pipermail/alsa-devel/2015-August/097126.html ?
If yes then I will work with NVidia/Realtek on updating their drivers.
Please allow a reasonable time for review, especially during times like the merge window.
participants (3)
-
Anatol Pomozov -
Lars-Peter Clausen -
Mark Brown