TAS2563: Echo reference signal is choppy

Hi Prashanth,

Perhaps data is misaligned and that's why after a specific amplitude the data from adjacent slot couples in as MSB/LSB are sampled at the incorrect slot. Maybe 2-bits shifted?
Are you using the EVM or a custom board? Can you share your PPC3 file as well?

Best regards,
-Ivan Salazar
Applications Engineer

Hi Ivan,

We're using our hardware.  I believe that you already have our PPC3 file but please LMK if it needs to be resent through the local TI FAE.

Regards, O.R.

Hi Oren,

Didn't notice it was the same application. I'll get some captures of the effect of data misalignment. However to be completely sure of the problem we need to know the host configuration in terms of I2S data format, can you confirm the word and slot length, slot count and any offset settings from the host side?

Best regards,
-Ivan Salazar
Applications Engineer

Hi Ivan,

It's standard I2S (2 slots, MSB offset by 1 BCLK cycle), configured as follows:

BCLK = 3.072 MHz
FSYNC = 48 kHz
Slot length = 32 bit
Word length = 16 bit

Regards, O.R.

Hi Ivan, these are the settings we're using per our previous discussions.  Please review and advise if any change is needed.

Hi Oren,

Can you please try with the attached config files? I'm using a different version of the software that enables some advanced settings for the output data. If this works I can send further instructions through our TI FAE.

Best regards,
-Ivan Salazar
Applications Engineer

TAS2563_Tuning_v1_ddc.zip

Thanks Ivan.  Can you please expand on what is being changed vs. what's in the standard software?  Regards, O.R.

Hi Oren,

I'll share further information on the differences later today.
Were you able to test this new config file? Is it working better?

Best regards,
-Ivan Salazar
Applications Engineer

Hi Ivan, the EC path is not functional with this new config file.  I shared details by e-mail through the local TI FAE about the differences vs. register settings generated with the standard software.  Regards, O.R.  

Hi Oren,

I'll come back over email thread later today or tomorrow morning.

Best regards,
-Ivan Salazar
Applications Engineer

Hi Ivan, do you have an update?  Also, is there a way to exclude the pilot tone from the EC data?  Thanks, O.R.

Hi Oren,

I've been testing this problem related to the EC data, I'm adding a few specific comments below:

• I noticed the configuration files generated from the latest version I provided of PPC3 are actually incorrect on the slot selection settings. You may read registers 0x0b through 0x0d from both devices and most likely they have the same values, which means the data from both devices may be colliding in the same SDOUT slot.
  • We can change this with some manual I2C commands
  • I'm currently checking if there something we have to do on PPC3 to fix it
• Regarding the pilot tone on EC data, I wouldn't expect the pilot tone to be included in the EC data, I just verified this on the EVM using PPC3 for configuration (not using the config files at this moment)
  • Are you getting the pilot tone on EC data? Or just asking this to make sure the pilot tone is not there? I think you're still having trouble getting EC data back, correct?

Best regards,
-Ivan Salazar
Applications Engineer

Hi Ivan,

Two different issues...

I'm looking at the EC data using your recommended TDM transmitter settings in the standard PPC3.  The 60 Hz pilot tone is present in the SDOUT EC data when the SDIN audio amplitude is high enough to cause it to be activated.  I'm using a -12 dBFS 1 kHz tone, what level are you using?

The EC doesn't work at all when using the new config file that you provided.

Thanks, O.R.

EDIT:  I took this screenshot with the recommended TDM transmitter settings in the standard PPC3.  This is a close-up of SDIN and SDOUT during a left slot. The light blue vertical cursors show where the MSB and LSB I2S data is clocked in. You can see that both SDIN and SDOUT are 16-bit and that the left amplifier stops driving the common SDOUT line after the LSB is clocked. I took this screenshot when the 60 Hz pilot was present but whether it’s active makes no difference in these waveforms.

I believe that the data is aligned correctly for standard I2S so please advise what you are seeking to change vs. what is shown.

This screenshot shows the 60 Hz pilot tone when measuring SDOUT directly.  SDIN is a 1 kHz tone at -12 dBFS.  The -17 dBFS SDOUT level at 1 kHz is expected per tuning.

Hi Oren,

• Going back to standard release, is the problem that EC data is noisy or chopped? Or that is not happening anymore? (this was originally reported on this thread)
• I'll double check if we can remove the pilot tone from EC data on the standard release, I know this changes in firmware as the PPC3 version changes. I'll update you on this soon.
• Additional question, what configuration file format are you using for your custom system? bin, header or cfg file?

Best regards,
-Ivan Salazar
Applications Engineer

Hi Ivan,

The EC audio is distorted but the behavior is intermittent.  However, when I measure SDOUT directly and inspect the waveforms I don't see it so the issue may be downstream from the amplifiers.  You said that it may be an alignment problem but as far as I can see the data format is correct, please advise about that.

We are using header files.

Thanks, O.R.

Hi Oren,

I can see the pilot tone now on EC data on my side, however I have to enable the pilot tone regardless if the input amplitude; I have identified that for some reason, when using the EVM, the Audio In data is scaled down so the pilot tone is not being added to the output.
Regarding the distortion, given the fact that it shows intermittent behavior, I still think it may be some timing or format mismatch between the TAS2563 and the host.
I'll keep working towards a solution without the pilot tone on EC data, and keep working on the distortion root cause.

Another follow up question regarding custom system configuration: Can you read back register 0x0b, 0x0c and 0x0d from both devices? Are you writing to these registers as part of your initialization?

Best regards,
-Ivan Salazar
Applications Engineer

Hi Ivan,

Yes, these registers are part of the initialization.  I noted a difference between the values in PPC3 and what is read back:

Actual readings:

Device 1 (Left):

0x0b 0x00
0x0c 0x44
0x0d 0x04

Device 2 (Right):

0x0b 0x00
0x0c 0x40
0x0d 0x04

Values in PPC3:

Device 1 (Left):

0x0b 0x02
0x0c 0x40
0x0d 0x04

Device 2 (Right):

0x0b 0x02
0x0c 0x44
0x0d 0x04

Hi Oren,

Thanks for checking these registers, content on register 0x0b doesn't matter as long as bit 6 is 0 (which means that set of data is disabled). Regarding register 0x0c value swap only means the data is swapped on the SDOUT channels, but as long as they're different values with difference of 4 on the LSBs this should be OK.

I noticed on my end that PPC3 configures this correctly but cfg is not, so I was wondering if you could be getting something similar.

Can you also try toggling bit 0 on register 0x07 from book 0 page 0? I wonder if data could be sampling at incorrect edge and it's marginally sampling incorrectly.

Best regards,
-Ivan Salazar
Applications Engineer

Hi Ivan,

I toggled the LSB in register 0x07 and it caused SDOUT to be corrupted on both the audio analyzer and in our hardware.  Makes sense since I2S is sampled on rising BCLK so it was correct.  One of my colleagues advised that the team noticed and compensated for the register 0x0c setting swap.

Please advise about suppressing the pilot tone in the EC data.

Thanks, O.R.

Hi Oren,

Thanks for the notice.
I'm currently debugging this and other projects in parallel, sorry for the late response. I'll come back a soon as possible,

Best regards,
-Ivan Salazar
Applications Engineer

Hello Ivan, 

Any updates on this issue?

Regards,

Prashanth

Hi Prashanth,

I just edit my previous response, please disregard if you get an email notification with my previous reply.
Getting the pilot tone removed from the EC data is not a simple enable/disable operation, so we may need to change the algorithm for that purpose.
The speaker is actively playing 60Hz, can you elaborate on why you need the EC data without the pilot tone even if the speaker is playing that signal? Does it affect the echo cancellation algorithm?

Best regards,
-Ivan Salazar
Applications Engineer

Hello Ivan, 

More than the pilot tone we are more concerned about the discontinuities caused in the echo reference signal due to the VI sense sharing SDOUT. These will definitely affect the Echo cancellation algorithm. The EC capture signal is choppy when the signal is above a certain threshold. 

Hi Prashanth,

In this same thread we mentioned registers 0x0b, 0x0c and 0x0d, these control the enable and location of different sets of data for each device.
Can you please try by enabling only one device data? I mean disable all data coming out of device B and only check data on device A. If you do that, is data still affected?

Best regards,
-Ivan Salazar
Applications Engineer