Page 1 of 1

DSD recording

Posted: Thu Jun 07, 2018 2:13 pm
by JPS64
Hello,

today more ADC with DSD interface are available on the market (AKM, TI). I know that XMOS multichannel solutions exists (DIYINHK) but they´ve only PCM recording capabilities.

What is to do to enable XMOS silicon DSD recording?

Jean-Paul

Re: DSD recording

Posted: Thu Jun 07, 2018 5:03 pm
by MaximLiadov
You just need to pay about 1 mln dollars to some dev team to make it real on XMOS or you could send your mail to Santa and be good whole year.

Re: DSD recording

Posted: Fri Jun 08, 2018 10:48 am
by AlexAdvice
JPS64 wrote:Hello,

today more ADC with DSD interface are available on the market (AKM, TI). I know that XMOS multichannel solutions exists (DIYINHK) but they´ve only PCM recording capabilities.

What is to do to enable XMOS silicon DSD recording?

Jean-Paul
Normal way: to change a firmware, to enable I2S port operate in DSD input (it is simple multi data SPI). But I don't know if a specific software on PC exists to capture the data to dff/dsf?

Crazy way: use additional small CPU (STM32, for example) with the necessary quantity of SPI and I2S ports, which will be transformed DSD signal to I2S DoP.
Then any software, dedicated to PCM can be used to capture the data.
You need to write some simple PC utility, to convert WAV file in PCM-DoP to DSD dff or dsf file.

Re: DSD recording

Posted: Mon Jun 11, 2018 9:21 am
by infiniteimprobability
Hmm, this is an interesting one.

As far as I am aware, DSD native is only available on Windows platforms and requires the Thesycon driver. I am not sure though whether this driver can support input interfaces for DSD (using UAC_FORMAT_TYPEI_RAW_DATA instead of the normal UAC_FORMAT_TYPEI_PCM format). It's certainly doable from the FW point of view.

What will definitely be possible though is DoP. For example, a 176.4kHz 24b audio stream could easily contain 16b DSD data with a DoP header. This could comfortably be done in the main "I2S" loop of USB audio. You certainly wouldn't need an external chip for this - it's precisely the sort of thing XMOS is really good at.

As far as the host is concerned it would just be a 24b PCM stream so the host recording software must be told to record in the correct format and unpack the DSD data. I am not sure if such a software exists but I'm sure google search will tell us. It wouldn't be hard to write a script to extract out the DSD from the DoP.
You just need to pay about 1 mln dollars to some dev team to make it real
- This would just about cover my fees ;)