Hi!
v2 -> v3 changes: - update the preparatory atmel_ssc_dai patch with error checking and fallback handling of the CMR_DIV divider in addition to the new fallback handling of TCMR_PERIOD/RCMR_PERIOD from v2.
v1 -> v2 changes: - new subject for the bindings patch to make it easier to find - move the driver to the sound/soc/atmel directory - add cached values for add/loop1/loop2 to avoid gets from output gpios - use _cansleep when updating gpios - add comment on how the regulator voltage is mapped to the ana enum - drop the .hw_params hook that did set the cpu dai divider and... - ...add a preparatory patch that does this in the cpu dai driver instead - drop .init and set the dapm routes directly in the card struct instead
The TSE-850 is an FM Transmitter Station Equipment, designed to generate baseband signals for FM, mainly the DARC subcarrier, but other signals are also possible.
This adds a driver for the "sound" bits of the device (quoted since it is normally not used for normal sound output, but that works too of course).
I have not provided a patch to add axentia as a devicetree vendor prefix, since such a patch is already pending in an IIO series [1] that seems close to being accepted.
However, there are a couple of points that I'm not 100% satisfied with for this driver.
First, I do not know how to describe the relays that control if the IN1/IN2 signals are directly routed towards OUT1/OUT2 or if they are routed to the "add" switch. The dapm routing treats this as if the IN1/IN2 signals are always routed to both the "add" switch and to the muxes feeding OUT1/OUT2. This is fine with me since nothing is powered in those sections anyway, so what dapm thinks does not really matter. But it is a wart all the same.
Second, there's my comment in tse850_put_mix() when the "add" switch is updated. I believe this update should really happen as a side effect of the call to snd_soc_dapm_mixer_update_power(), so that it happens at the right point compared to other stuff that is powered. But I do not know how to hook that up and instead I flip the switch before the call since it doesn't really matter. I.e., any noise resulting from this badness is negligeble in practice.
Cheers, Peter
[1] http://lwn.net/Articles/705931
Peter Rosin (3): ASoC: atmel_ssc_dai: if not provided, default to sensible dividers ASoC: tse850: document axentia,tse850-pcm5142 bindings ASoC: atmel: tse850: add ASoC driver for the Axentia TSE-850
.../bindings/sound/axentia,tse850-pcm5142.txt | 88 ++++ MAINTAINERS | 7 + sound/soc/atmel/Kconfig | 10 + sound/soc/atmel/Makefile | 2 + sound/soc/atmel/atmel_ssc_dai.c | 83 +++- sound/soc/atmel/atmel_ssc_dai.h | 1 + sound/soc/atmel/tse850-pcm5142.c | 472 +++++++++++++++++++++ 7 files changed, 656 insertions(+), 7 deletions(-) create mode 100644 Documentation/devicetree/bindings/sound/axentia,tse850-pcm5142.txt create mode 100644 sound/soc/atmel/tse850-pcm5142.c